University of Southern California Dissertations and Theses

Delivering better care for children with special health care needs: analyses of patient-centered medical home and types of insurance

(USC Thesis Other)

Delivering better care for children with special health care needs: analyses of patient-centered medical home and types of insurance

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
1

Delivering Better Care for Children with Special Health Care Needs:
Analyses of Patient-Centered Medical Home and Types of Insurance

By
Chia-Wei Lin

A Dissertation Presented to The
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Fulfillment of the
Requirements for the Degree of
DOCTOR OF PHILOSOPHY
PHARMACEUTICAL ECONOMICS AND POLICY
December 2017

2
TABLE OF CONTENTS

CHAPTER 1. INTRODUCTION ............................................................................................................................. 4
CHAPTER 2. THE RELATIONSHIP BETWEEN THE PATIENT-CENTERED MEDICAL HOMES,
HEAL THCARE EXPENDITURES, AND QUALITY OF CARE AMONG CHILDREN WITH SPECIAL
HEAL TH CARE NEEDS ............................................................................................................................................ 9
ABSTRACT ................................................................................................................................................................................. 9
INTRODUCTION ...................................................................................................................................................................... 11
METHODS ............................................................................................................................................................................... 12
RESULTS .................................................................................................................................................................................. 18
DISCUSSION ............................................................................................................................................................................ 21
CONCLUSIONS ........................................................................................................................................................................ 23
TABLES & FIGURES ............................................................................................................................................................... 24
CHAPTER 3. HEAL TH CARE UTILIZATION AND EXPENDITURES FOR CHILDREN WITH SPECIAL
HEAL TH CARE NEEDS: A CROSS-SECTIONAL ANAL YSIS FOR MEDICAID AND PRIVATE
INSURANCE ENROLLEES ................................................................................................................................... 34
ABSTRACT ............................................................................................................................................................................... 34
INTRODUCTION ...................................................................................................................................................................... 36
METHODS ............................................................................................................................................................................... 38
RESULTS .................................................................................................................................................................................. 44
DISCUSSION ............................................................................................................................................................................ 47
CONCLUSION .......................................................................................................................................................................... 51
TABLES & FIGURES ............................................................................................................................................................... 52
CHAPTER 4. THE EFFECTS OF THE HEAL TH CARE HOME ON COSTS, UTILIZATION, AND
QUALITY FOR CHILDREN WITH SPECIAL HEAL TH CARE NEEDS IN MINNESOTA: A
LONGITUDINAL ANAL YSIS OF INSURANCE CLAIMS ................................................................................. 58
ABSTRACT ............................................................................................................................................................................... 58
INTRODUCTION ...................................................................................................................................................................... 60
METHODS ............................................................................................................................................................................... 64

3
RESULTS .................................................................................................................................................................................. 76
DISCUSSION ............................................................................................................................................................................ 80
CONCLUSIONS ........................................................................................................................................................................ 84
TABLES & FIGURES ............................................................................................................................................................... 85
CHAPTER 5. SUMMARY & FUTURE RESEARCH DIRECTIONS ............................................................... 95
ACKNOWLEDGEMENTS ..................................................................................................................................... 98
APPENDIX .............................................................................................................................................................. 99
APPENDIX A. APPLICATION AND VALIDATION OF THE CSHCN SCREENER TO INSURANCE CLAIMS ........................ 99
APPENDIX B. QUALITY OF CARE OUTCOMES CONSTRUCTION .................................................................................... 109
APPENDIX C. CSHCN ATTRIBUTION AND HEALTH CARE HOME INDICATOR CONSTRUCTION ............................. 113
APPENDIX D. SUPPLEMENT TABLES FOR “THE EFFECTS OF THE HEALTH CARE HOME ON COSTS, UTILIZATION,
AND QUALITY FOR CHILDREN WITH SPECIAL HEALTH CARE NEEDS IN MINNESOTA: A LONGITUDINAL ANALYSIS
OF INSURANCE CLAIMS ” .................................................................................................................................................... 118
REFERENCES ....................................................................................................................................................... 125

4
Chapter 1. Introduction
Children with special health care needs
The Maternal and Child Health Bureau have defined children with special health care needs
(CSHCN) as “children who have or are at increased risk for a chronic physical, developmental,
behavioral, or emotional condition, and who also require health and related services of a type or
amount beyond that required by children generally.”
1
In the United States, 15.1% of all children,
or approximately 11.2 million people, were identified as CSHCN.
2
Common conditions of
CSHCN include allergy, asthma, attention deficit hyperactivity disorder, developmental delay,
depression, and anxiety.
3
The CSHCN screener is the gold standard for CSHCN identification.
4

There are five non-condition specific, health consequence-based, non-mutually exclusive items
in the screener. It assesses whether a child: (1) has prescriptions; (2) has higher than usual
medical services compared to other children of the same age; (3) has a functional limitation; (4)
receives special therapy, such as physical, occupational, or speech therapy; and (5) receives
counseling or mental health services.
4

CSHCN have relatively high health care utilization and expenditures.
5-9
They were estimated
as incurring medical expenses at a rate three times higher than healthy children and accounted
for 42% of all health care expenditures among US children.
7
Caring for CSHCN is likely to
impose financial burdens on families and caregivers.
10-13
CSHCN were also more likely to report
having unmet medical needs, especially in mental health services.
14,15
The majority of CSHCN
studies either use parent-reported survey data, such as the National Survey of CSHCN and the

5
National Survey of Child Health, or operationalize using selected chronic conditions in claims
analysis. The former may suffer recall and under-reporting biases, while the latter does not
comprehensively include children with less severe conditions that require long term use of health
services.

The Patient-Centered Medical Home
The Patient-Centered Medical Home (PCMH) is a model of primary care that aims to improve
the patterns of health care utilization and quality of care in the ongoing healthcare reform in the
US.
16-18
The concept of the medical home was developed for CSHCN by the American Academy
of Pediatrics (AAP) in 1967.
19
In 2007, the AAP, along with the American College of Physicians
(ACP), the American Academy of Family Physicians (AAFP), and the American Osteopathic
Association (AOA), published a set of joint principles, which aimed to characterize an ideal
PCMH. They specified that each patient should have an ongoing relationship with a personal
physician who leads a medical team providing whole-person oriented care.
18
This care should be
coordinated and integrated across various elements in the health care system and community.
18

The joint principles also emphasized quality and safety, with enhanced access and value-based
payments.
18
As of 2014, there were over 114 existing payment incentives and demonstration
PCMH programs, serving more than 21 million patients in 44 states.
17,20
The payers, who were
either the State government or a multi-payer collaboration employing a mix of private and public
sectors, incentivized PCMH clinics to improve both quality of care and health outcomes, by

6
providing enhanced per-member-per-month (PMPM) payments.
PCMH was expected to reduce costs and improve quality of care, by delivering primary care
that met the aforementioned principles. In general, PCMHs have demonstrated some improved
outcomes in cost and utilization, though not all of differences were statistically significant and the
directions of effects have not been uniform.
17,21-25
Increases in primary care service utilization for
patients receiving care from a PCMH was a common significant utilization outcome amongst the
publications reviewed.
21
In a recently-published systematic review of the literature, that
synthesizes evidence of PCMH effects, significant associations were found between having a
PCMH and total medical spending (-4.2%), use of specialty services (-1.5%), and cervical cancer
screenings (-1.2%).
26
However, the review also found no overall effect of a PCMH on primary
care, emergency department (ED), and inpatient visits, or on quality measures for diabetes care
and preventive screenings.
26
It is worth to note that these evaluation studies typically examined the
results over short periods, after one or two years after PCMH transformation.
21
PCMH were also
related to a trend towards positive results or no change in quality of care, including outcomes in
health care quality measures, such as well-child visits and colorectal cancer screening, and patient
satisfaction.
21

Compared to numerous studies looking at impacts of PCMHs on adults and the overall
pediatric population, when looking at the CSHCN group only a handful of studies have
investigated relationships between the PCMH (or its components), and outcomes in cost,
utilization, and quality. Among CSHCN, Homer’s review article showed the medical home to be

7
associated with positive health outcomes, such as decreased ED utilization and hospitalization
rates, though this effect is not consistent across all studies.
27
Family-centeredness and having a
usual source of care, typically defined as having a family physician, have been two of the most
common PCMH features in the literature addressing CSHCN.
28-30
In general, these interventions
were associated with some positive outcomes in utilization and quality of care.
27-30

CSHCN is a particularly relevant population given that PCMHs were originally developed
for this group. However, the effects of these recent efforts in the PCMH programs have not been
comprehensively evaluated for CSHCN. The first study in this dissertation aims to update the
literature and evaluate whether the effects of the PCMH evolved with time at the population level.
We followed previous research
6
definitions to investigate the impacts of receiving care from a
parent-reported medical home on health care expenditures, utilization, and quality of care for
CSHCN, using more recent data in the Medical Expenditure Panel Survey (2008-2012). Health
care quality outcomes, which were not reported in previous studies, were also included in our
study to better inform the impacts of receiving care from a PCMH for CSHCN.
The second study in the dissertation aims to describe the differences in health care
expenditures and patterns of care by type of insurance coverage, using the recent development
and validation of the claim-based CSHCN identification algorithm by Arim et al.
31
This
algorithm allows us to better capture CSHCN in insurance claims and to estimate and categorize
health care services with greater accuracy, compared to parent-reported survey data. We utilized
insurance claims from a payer that offered managed Medicaid and commercial coverage in the

8
Upper Midwest. With the shift from private insurance to public coverage for CSHCN and
potential changes to Medicaid policy, the cross-sectional study aims to improve our
understanding the differences in health service use and quality of care using real-world data for
publicly and privately insured CHSCN.
Finally, the third study of the dissertation utilized panel data from the same insurance claims
data source and the same sample specification as the second study, to investigate the effects of
PCMH for CSHCN. A difference-in-differences framework was applied to infer the causal
relationship between receiving care from a well-executed PCMH certification program, the
Health Care Home in Minnesota, and health care expenditure, utilization, and quality of care.
The studies presented in this dissertation collectively aim to inform the unmet needs of this
vulnerable population, to generate evidence in supporting the value of the PCMH, and to provide
insights into improving primary care delivery for CSHCN.

9
Chapter 2. The Relationship between the Patient-Centered Medical
Homes, Healthcare Expenditures, and Quality of Care among Children
with Special Health Care Needs

Abstract
OBJECTIVE To examine the association between having a patient-centered medical home
(PCMH) and healthcare expenditures and quality of care for children with special health care
needs (CSHCN).
METHODS We conducted a cross-sectional analysis of 8,802 CSHCN using the 2008-2012
Medical Expenditure Panel Survey. A PCMH indicator was constructed from survey responses.
Inverse probability treatment weighting was applied to balance the cohort. CSHCN’s annual
health care expenditures and quality were analyzed using two-part and logistic models,
respectively.
RESULTS Covariate-adjusted annual total expenditures were similar between CSHCN with and
without a PCMH ($4,267 vs. $3,957, p=0.285). CSHCN with a PCMH had higher odds of
incurring expenditure (OR=1.66, 95% CI:1.22-2.25)—in particular, office-based services and
prescriptions (OR=1.46 and 1.36, 95% CI:1.24-1.72 and 1.17-1.58, respectively)—compared to
those without a PCMH, without shifting expenditures. When examined in detail, PCMH was
associated with lower odds of accessing office-based mental health services (OR=0.80, 95%
CI:0.66-0.96), leading to lower expenditures ($106 vs. $137; p=0.046). PCMH was associated

10
with higher odds of post-operation and immunization visits (OR=1.23 and 1.22, 95%
CI:1.05-1.45 and 1.004-1.48, respectively) without changing expenditures. Parents of CSHCN
with a PCMH were more likely to report having the best health care quality (OR:2.33, p< 0.001).
CONCLUSIONS CSHCN who had a PCMH experienced better health care quality and were
more likely to access preventive services, with unchanged expenditures in the study period.
However, they were less likely to access mental health services. As the effects of PCMH varied
across services for CSHCN, more research is warranted.

11
Introduction
Children with special health care needs (CSHCN) have been defined as “children who have
or are at increased risk for a chronic physical, developmental, behavioral, or emotional condition,
and who also require health and related services of a type or amount beyond that required by
children generally.”
1
In the United States, 15.1% of all children (approximately 11.2 million
people) were identified as CSHCN.
32
These children were estimated as incurring medical
expenses at a rate three times higher than healthy children and accounting for 42% of health care
expenditures among all US children.
7

The concept of the patient-centered medical home (PCMH) has gained popularity in recent
years as a desirable primary care model. An ideal PCMH, as described by the American Academy
of Pediatrics, is accessible, family-centered, continuous, comprehensive, coordinated,
compassionate, and culturally effective.
19
Approximately 43% of CSHCN have reported
receiving care at clinics with medical home features.
32
CSHCN who received care from a PCMH
had an increased probability of having medical needs met, enjoyed improved health outcomes,
and had an increased probability of using preventive care than CSHCN without a PCMH.
14,27,33-35

For health care expenditures, Romaire et al. is one of the key studies that examined the effect
of PCMHs at population level for CSHCN using the 2003-2008 Medical Expenditure Panel
Survey (MEPS) data.
6
They found an increase in probability of accessing hospital outpatient and
office-based visits, and prescriptions, without significant changes in overall expenditures, for
CSHCN with a PCMH relative to those without one.
6
Han et al. investigated the annual health care

12
expenditure and quality among all children, rather than CSHCN, and a longer time horizon
(2004-2012).
36
For all children, no significant effects of PCMH on health care expenditures were
found, but a significant increase in parent-reported health care quality resulted from the presence
of a PCMH.
36

This study aimed to use more recent data to examine the current relationship of having a
PCMH on annual health care expenditures and health care quality for CSHCN, the population for
whom the model was originally designed. We investigated the association between having a
PCMH and annual health care expenditures and quality of care. The PCMH model is intended to
lower health care expenditures while improving health care quality.

Methods
Data Source
We used the Agency for Healthcare Research and Quality MEPS Household Component
2008-2012 full-year and event data, an in-person interview survey of US non-institutionalized
families and individuals. These public-use files include data on individual demographics, health
conditions, healthcare expenditures and utilization, insurance coverage, and health status. MEPS
also collected information surrounding individuals’ perception of health care quality using
Consumer Assessment of Healthcare Providers and Systems (CAHPS), a survey instrument
measuring consumer-perceived healthcare quality during the past 12 months.
37
By accounting for
complex multiple stages and sampling survey designs, we were able to obtain national estimates

13
from a representative non-institutionalized population using MEPS.
38
The Institutional Review
Board of the University of Southern California determined this study was exempt from review
because MEPS data are publicly available and de-identified.

Study Population
Children under 18 years of age in 2008-2012 and identified as CSHCN by the responses to
the CSHCN Screener were included as the study sample (n= 8,802), representing 14.6 million
CSHCN in the US. The screener is a non-condition-specific, health consequence-based tool for
CSHCN identification in several national surveys, including MEPS.
4
We excluded infants
younger than 30 days old, a population unlikely to have immediate exposure to a medical home
and whose healthcare expenditures were unrelated to medical home.
6
Since the health
maintenance organizations (HMOs) were linked to primary care provider based care
management and referral, we conducted sensitivity analyses for CSHCN who were not covered
by HMO plans to examine independent impacts of PCMH. The findings were directionally but
lost statistical power due to exclusion of 47% CSHCN with HMO coverage (eTables 1 and 2).

Exposure/ Independent Variable
We followed the MEPS-based algorithm described and validated in Romaire et al
39
to
construct a binary PCMH indicator. The algorithm maps 22 survey items to one of the following
medical home domains: usual source of care, accessibility, family centeredness,

14
comprehensiveness, and compassionate care.
6
A child must have a usual source of care and a
minimum score in each domain to be considered as having a medical home.
6
This approach
allowed us to capture parent-reported use of a PCMH for CSHCN.

Outcome Measures/ Dependent Variables
The outcomes of interest in this study were health care expenditures and a quality of care
indicator. The expenditures in MEPS were estimated using an algorithm that incorporates
information on payments from the respondents and medical providers with an imputation for
missing expenditures using payments for similar services.
37
Expenditures in MEPS represent the
costs from all sources, including out-of-pocket payments, Medicare, Medicaid, and private
insurance.
37
Direct payments for services reported by respondents were summed up as the annual
total expenditure including payments to hospitals, physicians, other health care providers, and
pharmacies.
37
Our study examined annual expenditures in total and by type of service.
Since office-based care is one of the most common settings for primary care, we studied
utilization and expenditures by type of care for in-person, office-based visits. Respondents
assigned each of their office-based visits to one of following mutually exclusive categories:
general checkup, well-child exam, diagnosis or treatment, psychotherapy/mental health
counseling, follow-up/ post-operation visit, immunizations/ shots, and others.
40
Following the
literature,
41
we combined visits for general checkups and well-child exams into a “well-child
care” category. We included the same types of visits in hospital outpatient settings as a sensitivity

15
analysis. We combined visits with sparse utilization, including emergency, vision exam, and
pregnancy-related visits, into the “other” category. All expenditures were converted into 2012 US
dollars using the Personal Health Care Expenditures Price index.
42

Timely well-child visits represent high-quality health care as they provide children with
preventive services, developmental evaluations, and ensure timely immunizations.
43
Thus, our
study used receipt of a well-child visit as our first quality of care measure. We conducted
sensitivity analyses applying a broader definition of well-child visit that included visits for
immunization or shots, as used in the literature.
41,44
To further assess quality, we use
parent-reported rating of health care quality from the CAHPS as another quality of care
measure.
36
This survey item asked parents to rate the health care that their children received from
0 (the worst possible) to 10 (the best possible).
37
Because responses were heavily weighted
toward higher values, we defined perception of high-quality health care as a rating of 10, with
sensitivity testing using cut-offs of higher than 9 or 8.

Covariates
We controlled for a range of predictors and covariates that were likely to affect outcomes.
We adopted Anderson’s behavioral model
45
of health services use as a framework to categorize
our predictors and covariates into: 1) predisposing factors, including age range, sex, and
race/ethnicity; 2) enabling factors, including poverty level, maternal educational level, insurance

16
coverage, geographic region, whether the child resided in a metropolitan area, number of family
members, and language spoken at home; and 3) needs factors, including whether the child had
functional or sensory limitations, and parent-reported children’s health status. A set of dummy
variables was included to adjust for systemic differences between cohort years. Mode imputation
was applied for missing predictors, including age groups (1.6%), language (0.1%), and region
(0.1%), number of family members (2.7%), and mother’s education level (13.6%). The
sensitivity analysis excluding children with missing covariates yielded directionally similar
effects of slightly larger size (eTable 3 and 4). Thus, we reported conservative findings from
mode imputations.

Statistical Analysis
This study is a cross-sectional analysis of observational survey data. Survey design weights
in MEPS were applied to all findings in order to generate national representative estimates. We
first reported and compared summary statistics for independent (Table 1) and dependent
variables (Table 2) by whether CSHCN were receiving care from a PCMH during the study
period. These tables show statistically significant differences between the populations using
bivariate 𝜒 2
tests for categorical variables and t-tests for continuous variables. To balance
observed characteristics that were significantly different between CSHCN with a PCMH and
those without a PCMH, propensity score weighting, also called inverse probability treatment
weighting (IPTW),
46
was applied. Specifically, to adjust complex survey design with the IPTW

17
following DuGoff et al.’s recommendation,
47
we first calculated the propensity score of having a
PCMH using a logistic regression model as a function of the observable patient characteristics
and the survey weights. The final weights in the survey-based IPTW were the products of the
survey weights and the inverse propensity score.
47

Annual health care expenditures were also compared between CSHCN with and without a
PCMH using a multivariate two-part model.
48
All multivariate results used IPTW to balance the
populations on observable characteristics, in addition to adjusting for the covariates using the
multivariate model, to produce doubly robust results. The two-part model was designed to
address patients with zero expenditures.
48
Part 1 modeled the likelihood of incurring
expenditures using a logistic regression; the finding can be interpreted as the probability of the
child accessing care. Part 2 estimated the association between the PCMH and health care
expenditures, conditional on accessing care. The general linear model in Part 2 accounts for the
right-skewed distribution of the expenditure data by using a log-link with gamma distribution.
The mean differences between CSHCN with and without a PCMH were estimated using both
parts in the two-part model, and the 95% confidence intervals (CI) were estimated from 1,000
bootstrapped samples.
49
For the parent-reported health care quality rating, we assessed the
association of PCMH with the quality rating by estimating the probability parents reported
high-quality health care using a multivariate logistic regression model. All statistical significance
was judged at a two-sided 𝛼 = 0.05 level. Data preparation and analyses were performed using
STATA MP version 14.0 (StataCorp).

18

Results
Characteristics of CSHCN
The observable characteristics of the CSHCN are summarized in Table 1. Before
implementing IPTW, 48.5% of CSHCN reported receiving care from a PCMH. They were more
likely to be male, have private insurance, to be in higher income categories, speak English at
home, report better health statuses, reside in metropolitan areas, and have higher maternal
education, while being less likely to have functional or sensory limitations. Geographic
differences were found between CSHCN with and without a PCMH. Aforementioned CSHCN
characteristics were balanced after IPTW, but the age distribution became significantly different
between PCMH groups. CSHCN who reported having a PCMH were younger on average when
IPTW was applied.

Health Care Expenditures
During the study period, the majority of CSHCN had incurred healthcare expenditures
(Table 2). In descriptive statistics with IPTW, CSHCN with a PCMH were more likely to use
office-based services and prescriptions. No significant differences were found in health care
expenditures conditional on accessing care.
Consistent with the summary statistics in Table 2, Part 1 of the multivariate two-part model
(Table 3) showed that having a PCMH was associated with significantly higher odds of accessing

19
healthcare (odds ratio [OR]: 1.66, 95% CI: 1.22-2.25). CSHCN with a PCMH had higher odds of
using office-based services and prescriptions after adjusting for covariates (OR: 1.46 and 1.36,
95% CI: 1.24-1.72 and 1.17-1.58, respectively). Part 2 of the model estimated the percent change
in dollars, conditional on accessing care. Expenditures for office-based services were
significantly lower for CSHCN with a PCMH (-11.2% change, 95% CI: -20.7% to -0.6%)
conditional on accessing care. When both parts were considered collectively, annual health care
expenditures for CSHCN with a PCMH were estimated at $4,267, while those without a PCMH
were estimated at $3,957, a statistically insignificant difference. No significant differences in
annual expenditures were found for any type of services when both parts of the model were taken
into account. In other words, an increase in accessing care was not associated with significantly
greater expenditures for CSHCN with a PCMH.
We then examined office-based services by type of visit to identify the source of the
increase in accessing office-based care and the decrease in office-based expenditures, conditional
on accessing care (Table 4). When examining Part 1, we found an increase in accessing a
follow-up/post-operation visit (OR: 1.23, 95% CI: 1.05-1.45) and an immunizations/shots visit
(OR: 1.22, 95% CI: 1.004-1.48) for CSHCN with a PCMH. Having a PCMH was also associated
with a 20% decrease in odds of accessing psychotherapy/ mental health counseling (OR: 0.80,
95% CI: 0.66-0.96). In Part 2, only follow-up/ post-operation visits demonstrated a decreased
expenditure conditional on accessing care (-25.8% change, 95% CI: -39.1% to -9.6%).
Collectively, the follow-up/post-operation visits did not show a significant increase in annual

20
expenditures because the increase in accessing the services was offset by the decrease in
conditional expenditure. We only found that PCMH was associated with lower annual
expenditures in psychotherapy/mental health counseling visits (-$31, 95% CI: -61 to -1), and
they were driven by changes in the likelihood of accessing the services.

Quality outcomes
We evaluated quality of care based on parental rating of health care quality as well as receipt
of well-child care. For parent-reported high-quality health care, CSHCN who received care from
a PCMH were more likely to report having best possible health care quality after adjusting for
covariates (OR: 2.33, p < 0.001; Figure 1). When we lowered cutoffs in sensitivity testing, these
odds ratios increased (i.e., for rating >= 8, OR: 3.74, p < 0.001).
Receipt of well-child care corresponds to Part 1 of the two-part model. As shown in Table 4,
having a PCMH was not associated with improved access to well-child care. However, when the
immunizations/shots category was included for the broader definition of a well-child visit,
PCMH became significantly associated with improved access to well-child visit (OR: 1.16, 95%
CI: 1.03-1.31, p=0.014). The addition of visits in the hospital outpatient setting showed similar
results. This suggested that the immunizations visit was the key contributor to the
PCMH-associated improvement in accessing preventive services.

21
Discussion
Receiving care from a PCMH was found to have no association with overall health care
expenditures, but was significantly associated with better parent-reported health care quality in a
nationally representative sample of CSHCN in 2008-2012. Our study advanced evidence
surrounding the impact of PCMH on different types of office-based services, where most
primary care services are delivered, and applied a model that distinguished the mechanisms of
the changes. On the one hand, CSHCN with a PCMH were more likely to access prescriptions
and office-based services for follow-up/ post-operation and immunizations. On the other hand,
we observed a decrease in accessing psychotherapy/ mental health counseling in office-based
visits, resulting in a reduction in expenditures for this type of service for CSHCN with a PCMH.
These findings suggested the effects of PCMH varied across types of visits in the office-based
setting.
Our findings support the prior literature on the impact of PCMHs on children, but focusing on
CSHCN through 2012.
6,36
We advance the literature by examining hospital outpatient and
office-based services separately, finding that PCMH affected them differently. For office-based
services, the significance in both parts of the two-part model suggested an offset between the
increase in CSHCN accessing the care and a reduction in expenditures when using office-based
services. These findings suggested that PCMH addressed some access issues,
14
and potentially
lowered the expenditures for individual CSHCN. However, a decrease in mental health services
access was found for CSHCN with a PCMH, suggesting that the improvement in accessing care

22
may not be consistent across types of primary care visits. While there is an apparent rise in
unmet mental health care needs and potentially forgone mental health services among
CSHCN,
50,51
more research is needed to understand how the PCMH model affects individuals’
access to mental health services separate from other types of office-based services.
Consistent with the findings in health care quality for PCMHs in the overall pediatric
population in Han et al.,
36
a PCMH was significantly associated with better health care quality
for CSHCN without increase in expenditures. Applying the broad definition of well-child visit,
an increase in access to preventative care was associated with PCMH, consistent with the
literature.
33,34
In particular, our findings suggested that seeking immunizations/ shots drove the
increasing probability of well-child visits.
This study had several limitations. First, the PCMH indicator was constructed by
parent-reported survey information. This distinguished our study from facility-based PCMH
assignment, which identified whether a person received care from a qualified PCMH practice.
Since family-centered care is a key element in the PCMH model and the care burden for CSHCN
was likely to fall on family members’ shoulders, parent/familial perceptions of the PCMH have
been viewed as relevant when evaluating for CSHCN.
19
Second, expenditures and types for
office-based visits were subject to recall bias and to misreporting of health care use in the survey
setting. However, because the extent of underreporting was substantial to all socioeconomic
groups and the relative bias between groups was small, the behavioral analysis in comparing
between CSHCN with and without a PCMH was likely unaffected.
52
Moreover, we only included

23
a functional and sensory limitation indicator as a covariate to stratify CSHCN. As the effects of
PCMH may differ between health conditions, more research is needed to confirm that resources
are allocated to CSHCN who benefit the most from the model. Lastly, due to the cross-sectional
nature of this observational study, we cannot establish causal inference because of the potential
existence of unobserved factors that affect having a PCMH and the expenditures. Applying IPTW,
which relied on observed CSHCN characteristics, may not fully balance unobserved
confounders.

Conclusions
A PCMH was associated with an increased access to prescriptions and office-based services in
follow-up/post-operation visits and immunizations for CSHCN, without significantly shifting
overall expenditures and lowering the quality of care. However, we found that a PCMH did not
enhance access to mental health services in the primary care setting resulting in decreased
expenditures for these services, suggesting the effects of PCMH varied across different types of
office-based services. As gaps in mental health services were profound for CSHCN, channels
other than the PCMH should be considered when addressing their needs. More research is
warranted to capture the changes and causal relationships among health services in various settings
of PCMHs for CSHCN.

24
Tables & Figures
Table 1. CSHCN Characteristics by Patient-Centered Medical Home Status
a

Without IPTW

With IPTW
Characteristics, %
n= 8,802

PCMH P-

PCMH P-
All

Yes No Value

All

Yes No Value
Reported had a PCMH 48.5

51.3

Age, year

0.548

<0.001
0 to 5 18.0

17.6 18.4

18.6

21.1 15.8

6 to 11 38.7

38.3 39.1

38.3

36.2 40.5

12 to 17 43.3

44.1 42.5

43.2

42.7 43.7

Male 59.6

57.9 61.3 0.013

59.7

59.3 60.1 0.547
Race/ethnicity

0.208

0.733
White, non-Hispanic 25.8

26.5 25.1

25.8

25.9 25.7

Black, non-Hispanic 4.2

4.2 4.2

4.2

4.4 4.1

Hispanic 63.6

63.6 63.6

63.5

63.7 63.4

Other, non-Hispanic 6.4

5.7 7.1

6.4

6.1 6.8

Insurance

<0.001

0.991
Any private 59.6

64.9 54.5

59.0

57.3 60.8

Public only 37.4

32.7 41.9

34.5

38.2 36.7

Uninsured 3.0

2.5 3.6

3.6

4.6 2.5

Family size, n 4.2

4.2 4.2 0.159

4.2

4.2 4.2 0.344
Family income (% FPL)

<0.001

0.264
< 100% 21.5

17.7 25.1

21.7

22.2 21.2

100-199% 20.3

19.1 21.5

20.4

21.6 19.1

200-400% 30.4

30.9 29.8

30.3

29.6 31.1

>400% 27.8

32.3 23.6

27.6

26.6 28.6

Language spoke at home

<0.001

0.791
English 93.0

95.0 91.0

92.9

92.9 92.8

Spanish 5.9

4.1 7.6

6.0

6.1 5.9

Other 1.2

0.9 1.2

1.2

1.1 1.3

Health status

<0.001

0.637
Excellent/very good 67.5

70.2 65.0

67.4

67.7 67.1

Good 25.5

24.2 26.8

25.7

25.7 25.6

Fair/poor 7.0

5.7 8.2

6.9

6.5 7.3
Had functional or sensory
limitation 9.3

8.3 10.2 0.042

9.3

8.6 10.0 0.120
Resided in metropolitan area 83.8

85.7 81.9 0.033

83.5

84.2 82.7 0.403
Region

0.001

0.205
Northeast 18.9

21.2 16.6

18.9

18.0 19.9

Midwest 24.6

26.2 23.1

24.3

23.9 24.7

25
South 38.3

37.2 39.3

38.7

40.9 36.4

West 18.2

15.3 21.0

18.1

17.3 19.0

Year

0.240

0.221
2008 18.5

17.7 19.3

18.6

18.4 18.8

2009 19.3

19.8 18.8

19.1

20.5 17.6

2010 19.3

19.3 19.3

19.2

19.3 19.1

2011 21.5

20.6 22.3

21.8

20.9 22.7

2012 21.4

22.6 20.3

21.4

21.0 21.8

Maternal education

<0.001

0.110
Less than high school 9.9

7.1 12.5

10.0

9.9 10.1

High school/ GED degree 48.4

47.3 49.6

48.4

50.4 46.3

College degree 31.8

33.9 29.8

31.6

30.3 33.0

Post-college degree 9.9

11.8 8.2

9.9

9.3 10.5

CSHCN, children with special health care needs; GED, general education diploma; FPL, federal
poverty level; IPTW, inverse probability treatment weighting; PCMH, patient-centered medical
home.
a
Survey weights and designs were applied.

26
Table 2. Summary of Annual Healthcare Expenditures for CSHCN by Patient-Centered
Medical Home Status
a

CSHCN, children with special health care needs; exp, expenditure; IPTW, inverse probability
treatment weighting; PCMH, patient-centered medical home; SD, standard deviation.
a
Survey designs and weights were applied.

Without IPTW With IPTW
Type of Expenditures

PCMH
P-
Value
PCMH
P-
Value

All Yes No All Yes No
Total

Incurring exp, % 96.5 97.6 95.5 <0.001 96.6 97.2 95.9 0.011
Exp among incurring, mean 4,183 4,476 3,901 0.204 4,311 4,588 4,016 0.304
(SD) (15,880) (19,268) (11,135)

(17,701) (21,237) (11,252)

Inpatient

Incurring exp, % 4.5 4.6 4.3 0.604 4.5 4.9 4.0 0.085
Exp among incurring, mean 17,495 20,639 14,331 0.148 19,056 22,629 14,394 0.164
(SD) (39,653) (48,488) (24,409)

(50,611) (57,664) (25,317)

Emergency room

Incurring exp, % 16.9 16.6 17.1 0.595 16.9 17.2 16.5 0.468
Exp among incurring, mean 896 943 853 0.296 908 929 886 0.643
(SD) (1,359) (1,457) (1,241)

(1,410) (1,438) (1,320)

Hospital outpatient

Incurring exp, % 12.3 12.3 12.2 0.913 12.3 12.2 12.4 0.869
Exp among incurring, mean 2,468 3,025 1,937 0.161 2,448 2,939 1,937 0.202
(SD) (10,977) (14,855) (4,209)

(11,172) (14,576) (4,015)

Office-based services

Incurring exp, % 87.0 90.0 84.1 <0.001 87.0 88.7 83.3 <0.001
Exp among incurring, mean 1,071 1,013 1,129 0.225 1,068 978 1,167 0.055
(SD) (2,401) (1,948) (2,813)

(2,372) (1,894) (2,859)

Prescriptions

Incurring exp, % 81.6 84.2 79.1 <0.001 81.5 83.5 79.3 <0.001
Exp among incurring, mean 1,419 1,531 1,307 0.494 1,508 1,582 1,425 0.722
(SD) (8,890) (11,246) (5,145)

(9,929) (11,972) (6,025)

Dental services

Incurring exp, % 54.0 56.6 51.6 0.004 53.8 53.5 54.2 0.708
Exp among incurring, mean 737 734 740 0.951 740 670 813 0.172
(SD) (1,974) (2,023) (1,907)

(1,942) (1,778) (2,109)

27
Table 3. Multivariate Two-Part Model of Annual Healthcare Use and Expenditures
a

Types of Expenditure
Part 1:
Adjusted Odds Ratio
of Accessing Care
Associated with
PCMH (95% CI)
Part 2:
% Change in
Conditional
Expenditures Associated
with PCMH (95% CI)
Estimated Annual Expenditure from
Combined Model ($)
With
PCMH
Without
PCMH
Differ-
ence
(95% CI of
Difference)
Total 1.660
***
(1.224, 2.253) 6.4% (-0.071, 0.218) 4,267 3,957 309 (-258, 875)
Inpatient 1.236 (0.943, 1.620) 7.9% (-0.229, 0.509) 945 727 218 (-163, 599)
Emergency room 1.016 (0.877, 1.178) 0.4% (-0.155, 0.192) 154 152 3 (-27, 32)
Hospital outpatient 1.028 (0.855, 1.236) 26.7% (-0.031, 0.656) 308 238 70 (-21, 161)
Office-based services 1.461
***
(1.244, 1.715) -11.2%
*
(-0.207, -0.006) 887 964 -76 (-171, 19)
Prescription 1.358
***
(1.166, 1.580) 8.2% (-0.086, 0.281) 1,256 1,104 153 (-56, 361)
Dental 1.092 (0.933, 1.279) -8.6% (-0.209, 0.057) 375 397 -22 (-86, 41)
CI, confidence interval; CSHCN, children with special health care needs; PCMH,
patient-centered medical home.
a
Survey designs and the inverse probability treatment weighting were applied.
*
p-value < 0.05,
**
p-value < 0.01,
***
p-value < 0.001

28
Table 4. Multivariate Two-Part Model of Office-Based Services Utilization and
Expenditures by Types of Visits
a

Types of
Office-Based Visits
Part 1:
Adjusted Odds Ratio
of Accessing Care
Associated with
PCMH (95% CI)
Part 2:
% Change in
Conditional
Expenditures Associated
with PCMH (95% CI)
Estimated Annual Expenditure
from
Combined Model ($)
With
PCMH
Without
PCMH
Differ
-ence
(95% CI of
Difference)
Well-child care
b
1.119 (0.993, 1.261) 3.5% (-0.090, 0.177) 167 154 13 (-7, 33)
Diagnosis or treatment 1.257 (0.967, 1.635) 17.2% (-0.059, 0.459) 422 451 -29 (-93, 35)
Psychotherapy/ mental
health counseling
0.797
*
(0.664, 0.956) -6.8% (-0.237, 0.138) 106
*
137 -31 (-61, -1)
Follow-up/ post-operation
visit
1.229
*
(1.045, 1.445) -25.8%
b
(-0.391, -0.096) 55 63 -8 (-23, 7)
Immunizations/shots 1.219
*
(1.004, 1.481) -2.8% (-0.263, 0.281) 38 33 5 (-5, 16)
Other 1.110 (0.950, 1.297) -18.1% (-0.367, 0.061) 96 108 -12 (-41, 17)
CI, confidence interval; CSHCN, children with special health care needs; PCMH,
patient-centered medical home.
a
Survey designs and the inverse probability treatment weighting were applied.
b
Well-child care included visits for general checkup and well-child exam;
*
p-value < 0.05,
**
p-value < 0.01,
***
p-value < 0.001

29
Figure 1. Multivariate-Adjusted Parent-Reported Health Care Quality Rating for CSHCN
a

CSHCN, children with special health care needs; PCMH, patient centered-medical home; SE,
standard error.
a
Health care quality rating range= 0 (worst) to 10 (best); survey designs and the inverse
probability treatment weighting were applied.
49.4%
30.0%
0%
10%
20%
30%
40%
50%
60%
PCMH No PCMH
% of Parents reported best possible
health care quality
Average rating (SE)
= 9.1 (0.03)
Average rating (SE)
= 8.4 (0.04)

30
eTable 1. Multivariate Two-Part Model of Annual Healthcare Use and Expenditures for
CSHCN with Non-HMO Coverage
a
Types of Expenditure
Part 1:
Adjusted Odds Ratio of
Accessing Care
Associated with PCMH
(95% CI)
Part 2:
% Change in Conditional
Expenditures Associated
with PCMH (95% CI)
Estimated Annual Expenditure from
Combined Model ($)
With
PCMH
Without
PCMH
Differ-
ence
(95% CI of
Difference)
Total 1.644
*
(1.081, 2.499) 2.9% (-0.168, 0.248) 4,379 4,232 148 (-718, 1014)
Inpatient 1.024 (0.712, 1.472) -10.8% (-0.424, 0.381) 1,004 1,104 -99 (-873, 674)
Emergency room 0.861 (0.689, 1.075) 6.5% (-0.115, 0.282) 141 149 -8 (-42, 26)
Hospital outpatient 1.028 (0.789, 1.338) 3.7% (-0.255, 0.445) 238 224 13 (-89, 116)
Office-based services 1.340
*
(1.070, 1.677) -14.5%
*
(-0.259, -0.014) 750 851 -101

(-209, 8)
Prescription 1.146 (0.926, 1.420) 20.7% (-0.033, 0.506) 1,581

1,278 302

(-17, 622)
Dental 1.094 (0.880, 1.361) -3.7% (-0.174, 0.122) 289 290 -1 (-56, 55)
CI, confidence interval; CSHCN, children with special health care needs; HMO, health maintenance organizations;
PCMH, patient-centered medical home.
a
n=4,641 (52.7% of sample) CSHCN were included in this subgroup analysis; survey designs and the inverse
probability treatment weighting were applied.
*
p-value < 0.05,
**
p-value < 0.01,
***
p-value < 0.001

31
eTable 2. Multivariate Two-Part Model of Office-Based Services Utilization and
Expenditures by Types of Visits for CSHCN with Non-HMO Coverage
a
Types of
Office-Based Visits
Part 1:
Adjusted Odds Ratio of
Accessing Care
Associated with PCMH
(95% CI)
Part 2:
% Change in Conditional
Expenditures Associated
with PCMH (95% CI)
Estimated Annual Expenditure from
Combined Model ($)
With
PCMH
Without
PCMH
Differ-
ence
(95% CI of
Difference)
Well-child care
b
1.159 (0.987, 1.361) 9.9% (-0.074, 0.305) 166 142 24 (-2, 50)
Diagnosis or treatment 1.069
*
(1.069, 1.294) -6.9% (-0.218, 0.109) 338 354 -16 (-89, 56)
Psychotherapy/ mental health
counseling
0.738
*
(0.553, 0.984) -14.8% (-0.341, 0.101) 101 151 -50
**
(-94, -7)
Follow-up/ post-operation
visit
1.027 (0.832, 1.268) -17.2% (-0.335, 0.032) 52 62 -8 (-26, 7)
Immunizations/shots 1.145 (0.877, 1.496) -29.8%
**
(-0.441, -0.118) 17 22 -5 (-11, 2)
Other 1.207 (0.956, 1.524) -42.6%
***
(-0.571, -0.233) 71 106 -35
*
(-71, 0.3)
CI, confidence interval; CSHCN, children with special health care needs; HMO, health maintenance organizations;
PCMH, patient-centered medical home.
a
n=4,641 (52.7% of sample) CSHCN were included in this subgroup analysis; survey designs and the inverse
probability treatment weighting were applied.
b
Well-child care included visits for general checkup and well-child exam.
*
p-value < 0.05,
**
p-value < 0.01,
***
p-value < 0.001

32
eTable 3. Multivariate Two-Part Model of Annual Healthcare Use and Expenditures for
CSHCN with No Missing Covariates
a

Types of Expenditure
Part 1:
Adjusted Odds Ratio of
Accessing Care
Associated with PCMH
(95% CI)
Part 2:
% Change in Conditional
Expenditures Associated
with PCMH (95% CI)
Estimated Annual Expenditure from
Combined Model ($)
With
PCMH
Without
PCMH
Differ-
ence
(95% CI of
Difference)
Total 1.665
*
(1.183, 2.343) 4.6% (-0.087, 0.199) 4,299 4,057 242 (-390, 874)
Inpatient 1.175 (0.876, 1.577) 10.2% (-0.241 0.598) 985 776 208 (-242, 659)
Emergency room 1.008 (0.853, 1.192) 0.1% (-0.156, 0.207) 159 157 3 (-30, 35)
Hospital outpatient 0.988 (0.797, 1.223) 14.8% (-0.113, 0.485) 238 224 13 (-58, 129)
Office-based services 1.420
***
(1.201, 1.679) -15.4%
***
(-0.248, -0.048) 871 998 -126
*

(-234, -19)
Prescription 1.391
***
(1.180, 1.640) 13.8% (-0.043, 0.353) 1,241

1,033 208
*
(0.1, 416)
Dental 1.076 (0.914, 1.267) -9.8% (-0.228, 0.054) 387 418 -31 (-103, 41)
CI, confidence interval; CSHCN, children with special health care needs; PCMH, patient-centered medical home.
a
n=7,124 (80.9% of sample) CSHCN were included in this sensitivity analysis; survey designs and the inverse
probability treatment weighting were applied.
*
p-value < 0.05,
**
p-value < 0.01,
***
p-value < 0.001

33
eTable 4. Multivariate Two-Part Model of Office-Based Services Utilization and
Expenditures by Types of Visits for CSHCN with No Missing Covariates
a
Types of
Office-Based Visits
Part 1:
Adjusted Odds Ratio of
Accessing Care
Associated with PCMH
(95% CI)
Part 2:
% Change in Conditional
Expenditures Associated
with PCMH (95% CI)
Estimated Annual Expenditure from
Combined Model ($)
With
PCMH
Without
PCMH
Differ-
ence
(95% CI of
Difference)
Well-child care
b
1.128 (0.990, 1.287) -0.9% (-0.131, 0.129) 168 161 13 (-7, 33)
Diagnosis or treatment 1.152
***
(1.012, 1.312) -12.3% (-0.251, 0.027) 436 477 -29 (-93, 35)
Psychotherapy/ mental health
counseling
0.810
***
(0.668, 0.984) -10.6% (-0.286, 0.119) 99 132 -31
***
(-61, -1)
Follow-up/ post-operation
visit
1.131 (0.943, 1.358) -32.5%
**
(-0.451, -0.170) 52 69 -8
*
(-23, 7)
Immunizations/shots 1.222 (0.991, 1.506) 5.9% (-0.207, 0.415) 38 32 5 (-5, 16)
Other 1.087 (0.913, 1.294) -24.4%
***
(-0.420, -0.015) 96 105 -12 (-41, 17)
CI, confidence interval; CSHCN, children with special health care needs; PCMH, patient-centered medical home.
a
n=7,124 (80.9% of sample) CSHCN were included in this sensitivity analysis; survey designs and the inverse
probability treatment weighting were applied.
b
Well-child care included visits for general checkup and well-child exam.
*
p-value < 0.05,
**
p-value < 0.01,
***
p-value < 0.001

34
Chapter 3. Health Care Utilization and Expenditures for Children with
Special Health Care Needs: A Cross-Sectional Analysis for Medicaid
and Private Insurance Enrollees

Abstract
OBJECTIVES: To describe the differences as the first claims-based study in health care
utilization, expenditures, and quality of care between Medicaid and privately insured children
with special health care needs (CSHCN).
METHODS: We conducted a cross-sectional analysis using 2013 insurance claims from a payer
offering commercial and managed Medicaid coverages in the Upper Midwest. CSHCN were
identified using a validated algorithm that applied the CSHCN screener to claims. Inverse
probability treatment weighting (IPTW) was applied to balance Medicaid and private insurance
cohorts. A standardized fee schedule was used to account for differences in reimbursement rates.
We examined expenditures and quality of care indicators using a two-part model and a logistic
regression model to adjust for CSHCN characteristics, neighborhood socioeconomic status, and
health conditions.
RESULTS: A total of 17,775 CSHCN were included in our sample. 31% of these CSHCN had
Medicaid coverage. After weighting and adjusted for covariates, CSHCN with Medicaid were
more likely than privately insured CSHCN to use care in a hospital outpatient setting (OR=1.192,
95% CI:1.085-1.309) and in places other than hospital and office-based settings, including

35
emergency departments (ED: OR=1.794, 95% CI:1.628-1.977), urgent care facilities (UC:
OR=1.221, 95% CI:1.089-1.370), and other settings (OR=1.638, 95% CI:1.495-1.796). Medicaid
coverage was associated with a decreased use of office-based services (OR=0.247, 95% CI:
0.178-0.341) relative to private coverage, especially visits for evaluation and management.
Conditional on accessing care, places other than hospital and office-based settings were
associated with significantly higher average expenditures for Medicaid CSHCN (ED=+27.3%,
p<0.001; UC=+9.3%, p=0.027; and other=+92.4%, p<0.001), and lower expenditures for
hospital inpatient (-36.6%, p<0.001) and outpatient care (18.9 %, p<0.001). Quality of care
appeared to be similar between types of insurance, but Medicaid CSHCN were more likely to
have avoidable ED visits in comparison to private coverage (OR=1.673, 95% CI: 1.305-2.145).
CONCLUSIONS: Medicaid CSHCN were more likely than private insured CHSCH to rely on
ED, UC, and other non-hospital and non-office-based settings as sources of care. They also had
significantly higher average expenditures in these settings when accessing this care. This pattern
is consistent with existing evidence raising concerns about optimal delivery of care to the
Medicaid population. The CSHCN population is particularly vulnerable, and its access to care is
particularly important.

36
Introduction
Children with special health care needs (CSHCN) have been defined as “children who have
or are at increased risk for a chronic physical, developmental, behavioral, or emotional condition,
and who also require health and related services of a type or amount beyond that required by
children generally.”
1
Common conditions of CSHCN include allergies, asthma, attention deficit
hyperactivity disorder (ADHD), developmental delay, depression, and anxiety. In the United
States, 15.1% of all children, approximately 11.2 million people, have been identified as CSHCN
(14.3% in Minnesota).
2
It is well known that CSHCN have relatively high health care utilization
and expenditures.
5-7
They were estimated as incurring medical expenses at a rate three times
higher than healthy children and accounted for 42% of health care expenditures among all US
children.
7
Caring for CSHCN is likely to impose financial burdens on families and caregivers.
10-13

Having insurance (public or private) has been proven to protect families from financial hardship,
to improve access to care (including both usual sources of care and specialty care), and to
address unmet health care needs and delays in care for CSHCN.
53-56

The impact of insurance type (public or private) on access and utilization are mixed for
CSHCN. Couple studies have found that public insurance is associated with less unmet needs,
lower likelihood of plan-related access issues, and higher utilization rates of various services
(including mental health, ED, inpatient and outpatient care), compared to privately insured
CSHCN.
55-57
One potential explanation is that Medicaid provides Early and Periodic Screening,
Diagnostic, and Treatment (EPSDT) benefits for children, which is important for CSHCN since

37
private insurance provision in special health services may be inadequate.
58
Several other studies
have found that private insurance provides a better usual source of care, typically defined as
having a family doctor, while some studies concluded no differences were found for delayed care,
specialty use, and unmet needs, when compared to public insurance.
55,59
When it comes to health
care expenditures, to our knowledge, only out-of-pocket were compared between public and
privately insured CSHCN. CSHCN with private insurance were more likely to have high
out-of-pocket costs compared to public insurance.
55,60

CSHCN were more likely to be covered by public insurance and it was estimated 1 in 4
children in Minnesota were covered by Medicaid, a public insurance program for people with
disability or low income.
61
Between 2001-2010, the proportion of CSHCN covered by public
insurance increased from 21.7% to 34.7%, while the proportion of privately insured CSHCN
declined from 64.7% to 50.7%.
57,62,63
Potential reasons for this shift includes the Medicaid and
State Children’s Health Insurance Program (SCHIP) changes in enrollment eligibility, parents
loss of employment, and crowd-out due to increased private insurance premiums.
57
As a result, it
is even more important to understand how type of insurance interacts with health care utilization
and quality of care for this population.
Given that the CSCHN screener is the gold standard to identify CSHCN, information on
health care expenditures, utilization, and quality of care mostly comes from parent-reported
survey data, such as the Medical Expenditures Panel Survey and National Survey of CSHCN.
Arim et al and colleagues developed and validated an algorithm to apply the CSHCN screener
4

38
(which is the gold standard of CSHCN identification) to insurance claims.
31
We adopted the
approach of Arim et al. to better capture CSHCN in insurance claims, since previous claims
analysis studies usually operationalized CSHCN by selected chronic conditions.
31
In addition,
insurance claims also allowed us to estimate and categorize health care expenditures with greater
accuracy in comparison to parent-reported survey data.
64

With the shift from private to public coverage among CSHCN and the potential policy
changes in the Affordable Care Act (ACA) and its repeal bills, this article aims to improve our
understanding of real-world health care utilization, expenditures, and quality of care for CSHNC
with managed Medicaid and private insurance using insurance claims data. The findings can help
researchers and policymakers to identify opportunities and challenges for better care for CSHCN
in public and private sectors.

Methods
Data and sample
We conducted a cross-sectional analysis using 2012-2013 insurance claims from a payer
offering commercial and managed Medicaid coverage in the Upper Midwest. Enrollees who
were: 1) 18 years old or younger in 2013, 2) resided in Minnesota, 3) had full-year medical and
pharmacy coverage in 2013, and 4) had any medical coverage in 2012, were included for
CSHCN identification. Pharmacy coverage was imposed because pharmacy coverage is required
for managed Medicaid in Minnesota. The Institutional Review Board of the University of

39
Southern California determined this study was qualified for exemption from review because the
protected health information was removed and/or de-identified in the claims data for this study in
concordant to the Health Insurance Portability and Accountability Act.
The CSHCN screener is the gold standard used to identify CSHCN in multiple national
surveys, including the National Survey of Children’s Health (NSCH) and the National Survey of
Children with Special Health Care Needs (NS-CSHCN).
4
There are five non-condition specific,
health consequence-based, non-mutually exclusive items in the screener. It assesses: 1) need for
or use of medicine prescribed by a doctor (other than vitamins); 2) need for or use of more
medical care, mental health care, or educational services than is usual for children of the same
age; 3) being limited or unable to do the things most children of the same age can do; 4) need for
or receipt of special therapy, such as physical, occupational, or speech therapy; and 5) emotional,
developmental, or behavioral problems for which treatment or counseling is needed or received.
A “yes” in any of the questions triggers a follow-up question regarding duration and whether the
consequence is attributable to medical, behavioral, or other health conditions (except in question
5). The screener is designed to capture both health services needs and health services use if the
health consequences last or are expected to last for at least 12 months.
4

CSHCN were identified by applying a validated algorithm that maps the CSHCN screener
onto claims.
31
The algorithm was developed and validated by Arim et al.
31
It demonstrated
similar overall prevalence and demographic and socioeconomic distributions compared to
findings in Canadian surveys for children age 6-10.
31
We followed Arim et al. specification to

40
use a total of 3 years (including 2014, if available) of medical and pharmacy claims at the most
were used to determine whether a child belonged to CSHCN in 2013.
31
A brief description of
criteria for each screener item is shown in Figure 1. To our knowledge, no claims-screener
mapping for CSHCN has been published in the U.S. The Arim algorithm has demonstrated good
validity when compared to the Pediatric Medical Complexity Algorithm (PMCA)
65
and also with
resource utilization bands (RUBs) in the Johns Hopkins ACG system.
66
Detailed algorithm
applications and a validity check are in Appendix A. After imposing the enrollment criteria, we
had 17,775 CSHCN in the final sample (Figure 2).

Exposure / Independent Variable
The variable of main interest was type of insurance. CSHCN with full-year Medicaid
coverage were assigned to the Medicaid group (n=5,518), while those covered by preferred
provider organization (PPO) private insurance plans were in the private insurance group (n=
12,257). It is noted that 38 CSHCN who had both private and public insurance coverage in 2013
were assigned to the private insurance group.

Predictors & Covariates
We controlled for a range of predictors and covariates that were likely to affect outcomes.
We adopted Anderson’s behavioral model
45
of health services use as a framework to categorize
our predictors and covariates into: 1) predisposing factors, including age and sex from the

41
enrollment data; 2) enabling factors, including residence in an urban area and neighborhood
effects, which captured race and ethnicity, language, education, and income and were matched at
census-tract level to the enrollee’s address using the 2010 US Census and the 2011 American
Community Survey (ACS) 5-year average data; and 3) needs factors, including the prior-year
(2012) comorbidity burden estimated by the resource utilization bands (RUBs) from the Johns
Hopkins ACG system
66
, and selected common condition indicators for CSHCN.

Outcome measures: health care expenditures, utilization, and quality of care
To study the patterns of health care expenditures and utilization, we categorized claims by
place of service (POS), including hospital outpatient, inpatient, office-based visits, emergency
department (ED) visits, urgent care (UC) facilities, and other POS. Other POS include health
care services provided at home, independent laboratories, community mental health centers,
schools, comprehensive rehabilitation center, etc. We also categorized expenses by type of
services using CPT codes. Common types of services include inpatient facilities, evaluation and
management visits, laboratories, radiology units, and surgical units. Other diagnosis and
treatments included biopsy, diagnostics, preventive treatment, and mental health services. Other
services with relatively smaller shares, including anesthesia, dental, vision and hearing,
medication/infusion, transportation, and equipment supplies, were combined into the “other
types” category in our analysis. A standardized fee schedule was created using available
insurance claims by averaging the dollar amounts for professional and facility medical claims

42
within each CPT and DRG code, and by NDC for prescriptions, across all types of insurance
plans and years. Applying the standardized fee schedule to account for different reimbursement
rates allows us to interpret the magnitude of the differences as utilization. All costs estimates
were reported at per-member-per-month (PMPM) level.
We selected five claims-based National Committee for Quality Assurance
(NCQA)-endorsed measures from the core set of child health quality measures for Medicaid and
CHIP programs
67
: follow-up visits for mental health hospitalization (FUH7, FUH30), medication
management for people with asthma (MMA), appropriate testing for children with pharyngitis
(CWP), appropriate treatment for children with upper respiratory infection (URI), and follow-up
care for children prescribed ADHD medication (ADD). We constructed each indicator following
the 2016 Healthcare Effectiveness Data and Information Set (HEDIS) technical specifications,
with minor modification to accommodate data availability. Pediatric quality of care measures
that are highly correlated with primary care visits, such as well-child visits and immunization,
were not included because the number of office-based visits is one of our sample selection
criteria. We compared the percentage of CSHCN that met the HEDIS goals to the 2013 national
average rates, which were calculated using the NCQA health plan records that covered 54% of
US population (including adults and elderly).
68
Note that the national average rates for the FUH7
and FUH30 were calculated using patients aged 6 years or older while other measures in our
study only included children.
68
In addition, an indicator of having any avoidable ED visits was
also included as a quality measure. ED visits that did not result in hospital admission were

43
classified into non-emergent and emergent levels following rules in Ballard et al. and the NYU
ED algorithm.
69,70
Detailed information on the construction of these quality of care measures is
in Appendix B.

Statistical analysis
We conducted a cross-sectional analysis of observational data (2013 insurance claims). We
first reported and compared summary statistics for independent and dependent variables by type
of insurance for CSHCN. We assessed statistically significant differences between CSHCN
covered by Medicaid and private insurance using bivariate 𝜒 2
tests for categorical variables and
t-tests for continuous variables. To balance observed characteristics that were significantly
different between CSHCN with different types of insurance, propensity score weighting, also
called inverse probability treatment weighting (IPTW)
46
, was applied. The likelihood of
Medicaid coverage was estimated using logistic regression with the aforementioned predicators.
Health care expenditures were compared between CSHCN covered by Medicaid and private
insurance using a multivariate two-part model.
48
All multivariate results used IPTW to balance
the populations on observable characteristics, in addition to adjusting for the covariates using the
multivariate model, to produce robust results. The two-part model addressed patients with zero
expenditures.
48
Part 1 modeled the likelihood of incurring expenditures using a logistic
regression; the finding can be interpreted as the probability of the child accessing care. Part 2
estimated the relationship between Medicaid coverage and health care expenditures, conditional

44
on accessing care. The general linear model in Part 2 accounts for the right-skewed distribution
of the expenditure data by using a log-link with gamma distribution. Note that we applied only
Part 2 to medical services, since almost all CSHCN in our sample incurred expenses. We
assessed the association between positive quality of care outcomes and types of insurance for
CSCHN using a multivariate logistic regression model. All statistical significance was judged at
a two-sided α= 0.05 level. Data were prepared using SAS software, version 9.4 (SAS Institute,
Inc.) and were analyzed using STATA MP version 14.0 (StataCorp).

Results
A total of 17,775 CSHCN were included in our study sample; 31% had Medicaid coverage.
Their characteristics are summarized in Table 1. Before implementing IPTW, Medicaid CSHCN
were more likely to be male, younger in age, had higher comorbidity burdens, and lived in
neighborhoods with race/ethnicity other than white, with lower incomes and educational levels
and less English-only speaking homes (Table 1). They were also more likely to have selected
conditions, including anxiety, asthma, ADHD, behavioral / conduct problems, depression, and
developmental delays, while they were less likely to be diagnosed with autism compared to
privately insured CSHCN (Table 1). IPTW balanced all demographic factors and neighborhood
characteristics between Medicaid and privately insured CSHCN (Table 1). For selected
conditions, IPTW balanced only the prevalence of autism, but Medicaid CSHCN became more

45
likely to have migraine/headache (Table 1).

Health care expenditures & utilization
Unadjusted and adjusted health care expenditures predicted PMPM expenditures for all
medical services, and prescriptions were similar between Medicaid and privately insured
CSHCN (unadjusted: $626 vs. $590, Table 2; adjusted: $681 vs. $642, Table 3).
Medicaid CSHCN were significantly more likely to incur expenditures in all places of
service other than office-based settings, but had higher-than-average spending in UC and other
POS (Table 2). After adjusting for covariates, CSHCN with Medicaid sustained their higher
PMPM costs in ED, UC, and other POS setting when examined by place of service (Table 3).
Medicaid coverage was associated with both larger odds of accessing care and higher average
expenses when accessing care in ED, UC, and in other POS (Table 3). Meanwhile, lower PMPM
costs were found to be associated with Medicaid coverage in inpatient, outpatient hospital, and
office-based expenditures for CSHCN (Table 3). The former two were driven by decreased
average expense when accessing care, while the office-based services were driven by decreased
odds of accessing care (Table 3). These findings suggest that Medicaid CSHCN could use ED,
UC, and other places of service as their sources of care, instead of office-based settings.
Increases in ED, UC, and other POS were offset by reductions in inpatient, hospital outpatient,
and office-based expenses for Medicaid CSHCN. In other words, in relation to Medicaid,
privately insured CSHCN experienced higher use of inpatient care.

46
When examined by type of service, the differences in inpatient facility and other diagnostic
and treatment utilization between Medicaid and privately insured CSHCN becomes insignificant
with multivariate adjustment (Table 2 and Table 3). Access to other types of services remained
significantly higher for Medicaid CSHCN in the model. Higher overall expenditures for other
diagnostic and treatment and other types of services were found for Medicaid CSHCN, driven by
both parts of the model, when adjusted for covariates (Table 3). Medicaid CSCHN also had
lower average expenses when accessing care in inpatient facility and surgeries, leading to
significant reductions in overall inpatient facility and surgery services spending (Table 3).
Medicaid CSHCN seemed to use a broader variety of services, but had less costly hospital stays
compared to privately insured CSHCN.

Quality measures
Medicaid and privately insured CSHCN demonstrated comparable quality of care outcomes
in all five claims-based quality measures for unadjusted and covariate-adjusted models (Table 4).
Compared to national average HEDIS rates for commercial PPO and managed Medicaid, larger
portions of CSHCN met the HEDIS goals in MMA, ADD, CWP, and URI regardless of type of
coverage (Table 4). For privately insured CSHCN, the rates of early follow-up after mental
health hospitalization (FUH7) was lower than the national average but the difference was smaller
when examined the 30-day follow-ups (FUH30) (Table 4). These rates were similar for Medicaid
CSHCN and the national average (Table 4). When it came to reasons of ED use, Medicaid

47
CSHCN were associated with significantly higher odds of having avoidable ED visits, which
were for non-emergent or primary care-treatable diagnoses (Table 5).

Discussion
Medicaid and privately insured CSHCN in Minnesota had similar overall health utilization,
expenses and quality of care in 2013. However, the access patterns were significantly different.
Medicaid CSHCN relied on ED, UC, and other non-hospital and non-office-based settings as the
sources of care, and were more likely to visit ED for non-emergent and primary care-treatable
causes. Their access pattern was not optimal, because the office-based setting is the ideal place
for primary care delivery and serves as the usual source of care. Moreover, ED use for
non-urgent conditions worsens the overcrowding, reduces quality of care, and is more costly than
care delivered in other settings.
71
Our findings were consistent with literature suggesting that
adults and children in the general population with Medicaid enrollment had higher ED utilization
rates and were more likely to use ED for non-emergent reasons, implying that these CSHCN
might consider ED as a regular source of care.
72-74
From Medicaid patients’ perspective, evidence
suggests that seeking care in ED was likely a decision based on not trusting a primary care
practitioner to be able to treat complex conditions, or of considering ED as providing more
comprehensive care.
75
This situation seemed no different for managed Medicaid CSHCN in our
study sample.
Limited numbers of primary care providers accepting Medicaid could be a plausible

48
explanation of lower office-based visits in comparison to private insured CSHCN. Decker et al.
concluded that Medicaid children with significant health conditions or developmental delays
were more likely to report problems in accessing physicians due to insurance source prior to the
implementation of ACA.
76
Minnesota is one of the states that is committed to the ACA Medicaid
expansion in 2014, providing coverage for childless non-elderly adults up to 138% Federal
poverty level.
61
It was estimated that 222,900 (18%) Medicaid enrollees gained coverage due to
the ACA expansion.
61
Even though the expansion did not directly affect the eligibility for
children, the rates of enrollment among eligible Medicaid/SCHIP children in Minnesota rose
from 84.9% in 2013 to 93.0% in 2014.
77
This was a relatively large increase compared to
non-expansion states and could be attributed to the ACA publicity effects on take-up rates.
77
With
the increase in Medicaid population, the access to primary care provider problems
78
could be
aggravated for CSHCN, as our pre-expansion findings suggested already significantly lower
odds of accessing office-based services compared to private coverage prior to the expansion. In
contrast, previous research suggested that expanding public insurance program eligibility, such
as SCHIP, increased the physician participation in the program in the short-run.
79
Whether this
supply-side effect compensates the access problem for Medicaid CSCHN warrants more
research.
To our knowledge, this is the first claims-based study examining the health care
expenditures, utilization, and quality of care for CSHCN between different insurance types with
the application of a standardized fee schedule. The fee schedule accounts for the variation in

49
reimbursement rates across product types. This feature strengthened our ability to interpret the
observed differences in expenditures as the intensity of utilization. Previous studies only
examined the differences in these outcomes between Medicaid and private insurance with
selected conditions, such as asthma and autism (11% and 3% of CSHCN, respectively), using
insurance claims.
80,81
For asthma, Chang et al. found that public-insured asthma patients were
significantly more likely to have inpatient and ED visits, with fewer outpatient visits and with
25% higher expenditures in comparison with private coverage.
81
Chang and colleagues then
concluded that Medicaid was less comprehensive than private plans.
81
When it comes to autism,
children received four times higher levels of expenditure compared to private coverage, and
received more specialized therapies, including speech therapy, behavioral modification, and
occupational therapy.
80
Autistic children covered by Medicaid appeared to receive better care,
though this was accompanied with higher expenditures, compared to private coverage.
80
These
findings suggest that the impact of insurance type likely varies between children with different
medical needs. Further research is warranted to ascertain the impacts for subgroups in CSHCN.
Our findings suggested that Medicaid CSHCN had lower expenditure when using inpatient
care compared to privately insured CSHCN. This finding could be because Medicaid CSHCN,
on average, had less severe condition when using inpatient services, or these CSHCN received
different care, such as less likely to have surgical services, when comparing to those privately
insured. CSHCN in Minnesota seemed to be receiving good quality of care despite type of
insurance in comparison to national rates. However, the quality of care measures in this study

50
focused on appropriateness of care and proper follow-ups/outpatient care. Detailing the services
provided in hospital inpatient settings and including inpatient-based quality of care measures
could improve our understanding of the lower costs when using inpatient care for Medicaid and
its impact on quality of care.
We applied a valid mapping algorithm
31
of the CSHCN screener on insurance claims for
CSHCN identification. In the past, CSHCN were operationalized as having selected chronic
conditions in claims analysis. Condition-based CSHCN selection may leave out children with
elevated use of health services due to prolonged periods of diseases of less severity, such as
headache/migraine, and may not meet the true definition of CSHCN. In addition, with regard to
health utilization and expenditures studies, insurance claims have been considered as a more
reliable data source in comparison to parent-reported surveys, because they are less likely to
suffer from underestimation and underreporting.
64
The expenditures can be categorized by place
of service and type of service, which provide a broader picture compared to out-of-pocket costs,
the most common cost outcomes examined in CSHCN literature due to data availability. To sum
up, our findings and estimates are more comprehensive, and depict with greater accuracy, the
utilization patterns for both privately insured and Medicaid CSHCN.
This study had several limitations. Our findings from an Upper Midwest state may not be
generalizable to CSHCN in other states or nationwide, due to variations in Medicaid eligibility,
Medicaid reimbursement rates, and physician supply. In addition, due to the cross-sectional
nature of this observational study, we cannot establish causal inference because of the potential

51
existence of unobserved factors that affect the types of insurance held by CSHCN. Applying
IPTW, which relies on observed CSHCN characteristics, may not fully balance unobserved
confounders between Medicaid and private insured CSHCN. Lastly, a considerable number of
health care encounters and expenditures were categorized into ‘other POS’ and ‘other types’ for
CSHCN. Many of the forms of care in these two categories seemed relevant to specialty
care/services, such as service claims for respiratory, mental health, and nutrition care. A finer
categorization might be more informative of different patterns of care for Medicaid and privately
insured CSHCN.

Conclusion
Even though the overall expenditures and quality of care were similar between Medicaid
and privately insured CSHCN, Medicaid CSHCN relied on ED, UC, and other non-hospital and
non-office-based settings as the sources of care and less likely to access office-based services.
They also had significantly higher average expenditures when accessing this care. This access
pattern is consistent with the Medicaid literature for other populations, and suggests that care for
Medicaid CSHCN is not optimal, especially when CSHCN were more likely to use ED for
non-emergent and primary care treatable reasons. As Medicaid policy is in a process of change,
this vulnerable population requires extra attention in order to develop efficient patterns of care.

52
Tables & Figures
Table 1. CSHCN characteristics by types of insurance, with/without IPTW

Without IPTW With IPTW
CSHCN characteristics
Medicaid
n= 5,518
(31%)
Private
n= 12,257
(69%)
P-value Medicaid Private P-value
Male, %
53.5 51.4 0.008 52.5 52.0 0.678
Resided in urban area, %
84.0 84.8 0.181 83.8 84.5 0.370
Comorbidity burden, %

<0.001

0.704
Healthy/ no user
7.0 9.3

8.0 8.3

Low
27.3 31.3

30.2 30.0

Moderate
50.1 44.9

46.2 46.1

High
13.3 11.9

12.7 13.2

Very high
2.3 2.7

2.9 2.4

Age in year, mean (SD)
9.1 (5.0) 11.0 (5.0) <0.001 10.4 (4.0) 10.4 (6.0) 0.889
Neighborhood effects, mean (SD)

% of race/ethnic

Hispanic
7.6 (12.3) 3.5 (5.8) <0.001 4.8 (7.1) 4.8 (9.3) 0.977
White, non-Hispanic
72.9 (27.3) 87.3 (13.2) <0.001 82.4 (16.5) 80.6 (26.3) 0.064
Black, non-Hispanic
11.1 (16.7) 3.2 (6.7) <0.001 5.9 (9.3) 7.3 (17.4) 0.072
Others, non-Hispanic
8.5 (9.5) 6.0 (6.9) <0.001 6.9 (6.5) 7.3 (10.1) 0.130
% of education level

Less than high school
12.4 (10.9) 6.0 (5.6) <0.001 8.3 (6.5) 9.1 (11.1) 0.076
High school/GED
28.1 (10.6) 24.8 (11.3) <0.001 26.1 (9.1) 26.1 (12.9) 0.934
Some colleges
32.7 (9.3) 32.6 (8.6) 0.576 32.7 (7.2) 32.4 (10.1) 0.170
4-year/ graduate degree
26.7 (15.5) 36.6 (18.9) <0.001 32.8 (14.7) 32.5 (21.5) 0.469
% of speaking English only
85.3 (16.5) 91.3 (8.6) <0.001 89.2 (10.1) 88.4 (15.9) 0.254
% of < FPL
17.5 (13.1) 7.8 (6.6) <0.001 11.4 (7.9) 12.4 (14.3) 0.076
Selected conditions, %

Anxiety
18.5 15.3 <0.001 20.2 15.7 <0.001
Asthma
15.4 8.5 <0.001 12.7 9.2 <0.001
ADHD
24.5 19.6 <0.001 26.5 20.3 <0.001
Behavioral/conduct problems
9.2 2.6 <0.001 7.6 3.2 <0.001
Autism
1.6 3.0 <0.001 2.8 2.6 0.601
Depression
13.7 10.3 <0.001 15.3 10.9 <0.001
Developmental delay
10.1 7.2 <0.001 9.8 7.4 0.001
Intellectual disability
0.4 0.4 0.789 0.3 0.4 0.583
Migraine/headache
2.1 1.7 0.056 2.5 1.6 0.001
Congenital disorders
4.6 5.9 0.168 6.3 6.2 0.847
Abbreviation: ADHD, attention deficit hyperactivity disorder; CSHCN, children with special health care needs; FPL,
federal poverty level; GED, general educational diploma; IPTW, inverse probability weighting; SD, standard
deviation.

53

Table 2. Unadjusted cost outcomes for CSHCN by types of insurance
PMPM Expenditures, mean (SD)
Medicaid Private P-value
Medical services
626 (1,510) 590 (2,743) 0.366
Incurring expenditure, %
99.8% 99.5% 0.064
Prescriptions
111 (461) 95 (394) 0.190
Incurring expenditure, %
89.2% 88.5% 0.328
By place of service

Inpatient
122 (900) 165 (1,714) 0.047
Incurring expenditure, %
8.3% 6.4% 0.005
Hospital outpatient
173 (801) 177 (1,008) 0.851
Incurring expenditure, %
48.8% 43.4% <0.001
Emergency department
45 (91) 23 (97) <0.001
Incurring expenditure, %
34.0% 22.3% <0.001
Urgent care
4 (9) 3 (11) <0.001
Incurring expenditure, %
20.1% 17.1% 0.008
Office-based visits
139 (195) 168 (742) <0.001
Incurring expenditure, %
94.5% 98.4% <0.001
Other
143 (409) 54 (927) <0.001
Incurring expenditure, %
45.2% 31.4% <0.001
By type of service

Inpatient facility
110 (856) 148 (1,604) 0.059
Incurring expenditure, %
6.9% 5.5% 0.024
Evaluation and management
56 (49) 52 (64) 0.006
Incurring expenditure, %
95.5% 96.3% 0.086
Radiology
28 (154) 23 (136) 0.167
Incurring expenditure, %
38.2% 37.0% 0.284
Laboratory
20 (57) 20 (90) 0.680
Incurring expenditure, %
64.9% 62.8% 0.070
Surgery
26 (76) 35 (155) <0.001
Incurring expenditure, %
22.1% 24.0% 0.052
Other diagnostic & treatment
195 (507) 179 (821) 0.170
Incurring expenditure, %
93.7% 91.8% 0.002
Other types
190 (694) 132 (1,396) 0.010
Incurring expenditure, %
83.5% 57.6% <0.001
Abbreviation: CSHCN, children with special health care needs; PMPM, per-member-per-month; SD, standard
deviation.
Note: The inverse probability treatment weighting (IPTW) was applied to balance Medicaid and privately insured
CSHCN cohorts. A standardized fee schedule was applied to account for different reimbursement rates.

54
Table 3. Multivariate adjusted two-part analysis of health care expenditures for CSHCN

Types of Expenditure
Part 1:
Adjusted Odds Ratio
of Accessing Care
Associated with
Medicaid (95% CI)
Part 2:
% Change in
Conditional
Expenditures Associated
with Medicaid (95% CI)
Estimated Expenditure from
Combined Model (PMPM, $)
Medi-
caid
Private
insurance
Differ
-ence
(95% CI of
Difference)
Medical services N/A

6.0% (-3.7, 16.6) 681 642 38 (-12, 102)
Prescriptions 0.972 (0.843, 1.120) 0.2% (-11.8, 13.9) 106 106 0.1 (-14, 14)
By place of service
Inpatient 1.017 (0.850, 1.216) -36.6%*** (-48.4, -22.0) 114 177 -63*** (-97, -29)
Hospital outpatient 1.192*** (1.085, 1.309) -18.9%*** (-27.7, -9.0) 163 190 -26* (16, 23)
Emergency room 1.794*** (1.628, 1.977) 27.3%*** (16.1, 39.6) 43 24 20*** (-47, -6)
Urgent care 1.221** (1.089, 1.370) 9.3%* (1.0, 18.3) 4 3 1*** (0.4, 1)
Office-based visits 0.247*** (0.178, 0.341) -2.8% (-7.4, 2.0) 152 163 -10** (-18, -3)
Other 1.638*** (1.495, 1.796) 92.4%*** (49.0, 148.4) 154 64 90*** (67, 113)
By type of service
Inpatient facility 0.938 (0.770, 1.142) -31.5%*** (-43.9, -16.3) 103 157 -55*** (-85, -24)
Evaluation &
management
0.725** (0.584, 0.900) 1.9% (-1.7, 5.5) 54 54 0.5 (-1, 2)
Radiology 1.029 (0.937, 1.129) 7.7% (-6.7, 24.2) 28 25 2 (-2, 6)
Laboratory 1.077 (0.981, 1.183) -2.4% (-11.9, 8.1) 20 21 -0.1 (-2, 2)
Surgery 0.901 (0.812, 1.000) -13.3%* (-23.6, -1.5) 27 34 -7** (-11, -2)
Other diagnostic &
treatments
1.162 (0.973, 1.387) 15.8%*** (8.2, 23.9) 226 194 32*** (16, 47)
Other types 3.777*** (3.370, 4.233) -5.1% (-20.2, 12.7) 181 149 32* (6, 59)
Abbreviation: CI, confidence interval; CSHCN, children with special health care needs; N/A, not applicable; OR,
odds ratio; PMPM, per-member-per-month.
Note: The inverse probability treatment weighting (IPTW) was applied to balance Medicaid and privately insured
CSHCN cohorts. Part 1 of overall medical services was not applicable because all CSHCN incurred medical
expenses. Part 2 of medical services was modeled by general linear model with gamma distribution and log-link.
Privately insured CSHCN was the reference group. A standardized fee schedule was applied to account for different
reimbursement rates.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001

55
Table 4. Receipt of quality care among Medicaid and privately insured CSHCN and
published NCQA goals in 2013
Abbreviation: ADHD, attention deficit hyperactivity disorder; CSHCN, children with special health care needs;
HEDIS, Healthcare Effectiveness Data and Information Set; HMO, health maintenance organization; NCQA,
National Committee for Quality Assurance; PPO, preferred provider organization.
Note: The inverse probability treatment weighting (IPTW) was applied to balance Medicaid and privately insured
CSHCN cohorts. No covariate adjustment was made.
a
National average rates were calculated used patients older than 6 years old, which included adults.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001 for Medicaid and private insurance comparison.
% Meeting NCQA HEDIS Goals
CSHCN in
Minnesota
2013 National Average
68

Medicaid Private
Medicaid
HMO
Commercial
PPO
Follow-up for mental health hospitalization within 7 days
a

(FUH7)
43.5% 26.7% 42.0% 49.8%
Follow-up for mental health hospitalization within 30 days
a

(FUH30)
60.2% 62.5% 60.9% 69.0%
Appropriate testing for pharyngitis (CWP) 87.4% 87.1% 66.5% 78.4%
Appropriate antibiotics use for upper respiratory tract
infection (URI)
92.2% 92.7% 85.2% 83.2%
Follow-up care for children with ADHD prescription (ADD)
- Initial phase 54.7% 50.4% 25.6% 38.2%
- Continued & maintenance phase 94.1% 89.5% 46.4% 45.2%
Medication management for people with persistent asthma
(MMA)

- 5-11 years old 41.3% 44.5% 27.6% 42.8%
-12-18 years old 34.0% 39.7% 25.6% 38.8%
Had avoidable emergency department visits *** 81.8% 73.6% N/A N/A

56
Table 5. Multivariate-adjusted logistical analysis of quality of care for CSHCN

Abbreviation: ADHD, attention deficit hyperactivity disorder; CI, confidence interval; CSHCN, children with
special health care needs; n, number of eligible CSHCN.
Note: The inverse probability treatment weighting (IPTW) was applied to balance Medicaid and privately insured
CSHCN cohorts. Privately insured CSHCN was the reference group.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001

Quality of Care Indicators n Odds Ratio 95% CI
Follow-up for mental health hospitalization within 7 days (FUH7) 277 1.079 (0.570, 2.040)
Follow-up for mental health hospitalization within 30 days (FUH30) 274 0.603 (0.316, 1.151)
Appropriate testing for pharyngitis (CWP) 584 1.315 (0.723, 2.392)
Appropriate antibiotics use for upper respiratory tract infection (URI) 594 0.754 (0.368, 1.544)
Follow-up care for children with ADHD prescription (ADD)
- Initial phase 764 1.100 (0.733, 1.651)
- Continued & maintenance phase 297 1.540 (0.565, 4.197)
Medication management for people with persistent asthma (MMA)
- 5-11 years old 304 0.985 (0.509, 1.908)
-12-18 years old 265 0.514 (0.225, 1.176)
Had avoidable emergency department visits 3,118 1.673*** (1.305, 2.145)

57
Figure 1. Application of the children with special health care needs (CSHCN) screener to
insurance claims

Figure 2. Children with special health care needs (CSHCN) sample selection

• Had > 9 months of prescrip on
possession in a 3-year span.
Q1. Need or use medicine
prescribed by a doctor
• Had more office-based visits
than 95% of the same age-sex
children.
Q2. Need or use more medical
care, mental health, or
educa onal services than
average children
• Had been diagnosed with
selected disabled condi ons, such
as epilepsy and spina bifida.
Q3. Func onal limita on or
disability
• Had > 1 claim for physical,
occupa onal, speech, or
respiratory therapy, that were >
6 months apart in a 3-year span.
Q4. Need or receive special
therapy, such as physical,
occupa onal, or speech
therapy
• Had > 1 claim for mental health
services that were > 6 months
apart in a 3-year span.
Q5. Need or receive
treatment/counseling for
emo onal, developmental, or
behavioral problems
CSHCN iden fied in 2013
n= 22,272 (100%)
No missing demographic & neighborhood
effects variables
n= 22,248 (99.9%)
Had any medical coverage in 2012
n=20,837 (93.6%)
Final Sample
Had full-year pharmacy coverage in 2013
n= 17,775 (79.8%)
Medicaid
n= 5,518
(31.0% in final sample)
Privately insured
n= 12,257
(69.0% in final sample)

58
Chapter 4. The Effects of the Health Care Home on Costs, Utilization,
and Quality for Children with Special Health Care Needs in Minnesota:
A Longitudinal Analysis of Insurance Claims

Abstract
Objective: To quantify the effects of the Health Care Homes (HCH) program, a patient-centered
medical home (PCMH) model introduced in Minnesota in 2010, on health care utilization,
spending and quality of care among children with special health care needs (CSHCN).
Methods: We conducted a difference-in-difference analysis using 2007-2014 insurance claims
from a payer offering commercial and managed Medicaid coverages in the Upper Midwest.
CSHCN were identified using a validated algorithm that applied the CSHCN screener to claims.
Identification of CSHCN with a HCH was determined by the clinic attribution using primary
care visits in the prior year. A standardized fee schedule was used to account for variation in
reimbursement rates. We examined expenditures and quality of care indicators using a two-part
model and a logistic regression model with individual-clinic random effects to adjust for CSHCN
characteristics, neighborhood socioeconomic status, and health conditions.
Results: A total of 87,172 patient-year observations were included in our sample, with 46% in
the HCH group. After adjusting for covariates, CSHCN receiving care from an HCH had similar
overall health expenditure and quality of care compared to those not in an HCH. Being in the
HCH group significantly increased the percentage of CSHCN using care in hospital inpatient,

59
outpatient, and other place of service settings (0.72%, p=0.016; 2.64%, p=0.003; and 2.03%,
p=0.019, respectively) with a decrease in urgent care use (-0.88%, p=0.039). Average
expenditures were also similar between the HCH and non-HCH groups, with the exceptions of a
19% reduction in urgent care costs (p=0.002) and a 2% increase in evaluation and management
service costs (p=0.046). When HCH status lasted two or more years, it significantly increased
total medical expenditure and expenditures in other POS settings, as well as evaluation and
management expenditures and other types of services, whereas a non-significant increase was
found in the first year of certification.
CONCLUSIONS: HCH did not appear to either effectively reduce cost or improve quality of
care for these high-risk children. More research is needed to identify any subgroups of CSHCN
that would benefit from the PCMH model to improve primary care delivery.

60
Introduction
Background for the patient-centered medical home
The Patient-Centered Medical Home (PCMH) is a model of care that aims to improve US
primary care in the healthcare reform.
16-18
The concept of the medical home was first developed
for children with special health care needs (CSHCN) by the American Academy of Pediatrics
(AAP).
19
In 2007, the AAP, along with the American College of Physicians (ACP), the American
Academy of Family Physicians (AAFP), and the American Osteopathic Association (AOA),
published a set of joint principles, which aimed to characterize an ideal PCMH. They specified that
each patient should have an ongoing relationship with a personal physician who led a medical team
providing whole-person oriented care.
18
This care should be coordinated and integrated across
various elements in the health care system and community.
18
The principles also emphasized
quality and safety, with enhanced access and value-based payments.
18

In 2008, the Minnesota legislature established a set of standardized criteria and processes for
multi-payer PCMH certification, known as the Health Care Home (HCH) initiative, which
included a comprehensive set of resources and services. Five categories of HCH certification
administrative rules based on community and stakeholders’ inputs were developed. These were
intended to assess access and communication, participant registration and the tracking of
participants’ care activity, care coordination, care plans, and performance reporting and quality
improvement.
82
The final rules for HCH certification were published on January 11, 2010, and
the first HCH certification was issued in July 2010. These rules were set to meet the highest level

61
(Level 3) of the National Committee for Quality Assurance (NCQA) PCMH recognition program,
which has been the most prevalent PCMH recognition program.
83
Every primary care provider in
a certified HCH clinic is required to become certified. HCHs were required to renew their
certification annually, and to participate in the HCH learning collaborative and data collection.
The application guide and assessment tools are available through the Minnesota Department of
Health (MDH) website.
84
The MDH verify whether the primary care delivered in the HCH meet
the criteria through onsite inspections, annual reporting, and regular tracking. This systemic
implementation approach is one of the distinctive features of Minnesota’s HCH development.
83

As of December 31, 2014, 359 (53%) clinics in Minnesota have been certified as HCH,
supporting care for 3.64 million people.
85
The MDH evaluation report found that HCH
introduced a 9% reduction in per-member-per-year costs, that led to an estimated saving of 1
billion dollars during 2010-2014 for Medicare, Medicaid, and dual eligible patients.
85
When
examined the costs by types, HCH decreased inpatient, hospital outpatient, skilled nursing
facility, and pharmacy costs, while increasing ED and ambulatory surgery costs.
85
For health care
utilization, HCHs use fewer services in categories other than ED visits and skilled nursing home
admissions.
85
The same report also found better quality of care for diabetes, vascular conditions,
asthma, depression and colorectal cancer screening for adult patients and asthma for children
who received care in a HCH during the study period.
85
Even though the report applied a set of
advanced analysis methods, it did not take into account the differences in Medicare and Medicaid
reimbursement rates that could mix the effects of HCH in the evaluation. In addition,

62
this government evaluation did not include privately insured population and the majority of the
enrollee in the report were adults. The literature gaps in the HCH effects on children and CSHCN
remained unfilled.

Backgrounds for children with special health care needs
The Maternal and Child Health Bureau have defined CSHCN as “children who have or are at
increased risk for a chronic physical, developmental, behavioral, or emotional condition, and
who also require health and related services of a type or amount beyond that required by children
generally.”
1
In the United States, 15.1% of all children, or approximately 11.2 million people,
were identified as CSHCN (14.3% in Minnesota).
2
Common conditions of CSHCN include
allergy, asthma, attention deficit hyperactivity disorder (ADHD), developmental delay,
depression, and anxiety.
3
CSHCN were estimated as incurring medical expenses at a rate three
times higher than healthy children and accounted for 42% of health care expenditures among all
US children.
7
Caring for CSHCN is likely to impose additional financial burdens on families and
caregivers.
10,11,13

The majority of literature on health care expenditures, utilization, and quality of care for
CSHCN is based on parent-reported survey data, such as the National Survey of CSHCN and the
National Survey of Child Health, as result of applying the non-categorical and health
consequence-based CSCHN screener to identify CSHCN.
4
Arim et al and colleagues developed
and validated an algorithm to apply the CSHCN screener
4
(known as the gold standard of

63
CSHCN identification) to insurance claims.
31
We adopted the approach of Arim et al. to better
capture CSHCN in insurance claims, since previous claims analysis studies usually
operationalized CSHCN by selected chronic conditions.
31

Impacts of PCMH on costs, utilization, and quality of care for CSHCN
In Minnesota, it was estimated that 52% of CSHCN had been receiving care at clinics with
medical home features in 2009/2010.
2
Compared to studies for the impact of PCMH on adults
and the overall pediatric population, when looking at the CSHCN group only a handful of studies
have investigated relationships between the PCMH or its components, and cost, utilization, and
quality outcomes. Among CSHCN, Homer’s review article showed the medical home to be
associated with positive health outcomes such as decreased ED utilization and hospitalization
rates, though this effect is not consistent across all studies.
27
An additional caveat is that many of
the studies reviewed in Homer et al. may only possess some features of a PCMH.
Family-centeredness and having a usual source of care have been two of the most common
PCMH features in the literature addressing CSHCN.
28-30
In general, these interventions were
associated with some positive outcomes in utilization and quality of care.
27-30
Annual health care
expenditures were shown to be similar between CSHCN with and without a PCMH in studies
using the Medical Expenditure Panel Survey (MEPS).
6,86
Lin et al. did find some evidence in
increased access to care in office-based settings and prescriptions, but the costs were not shifted.
86

64
Han et al. also found significant improvements in parent-reported quality ratings without
significantly changing the costs when a PCMH was present among all children.
36

As the PCMH model has gained in popularity and more research findings have reinforced
the benefits from the model in the adult population, we decided to investigate the effects of the
Minnesotan multi-payer HCH model on CSHCN. To our knowledge, there is no previous study
that has evaluated the effects of PCMH focusing on the CSHCN (non-condition based)
population using insurance claims data. Moreover, the HCH annual recertification process in
Minnesota has assured more homogeneous and stable PCMH delivery system characteristics.
The objective of this study is to quantify the impacts of HCH on health care expenditures and
utilization, as well as on quality of care, for CSHCN in Minnesota. We utilized a sophisticated
econometric model for causal inference. The findings will broaden our knowledge in managing
CSHCN by introducing real-world cost data, will generate additional evidence to support the
social value of the PCMH model, and will provide insights into improving primary care delivery
for CSHCN, a particularly vulnerable population.

Methods
Data
We analyzed HCH impacts using 2006-2014 longitudinal insurance claims from a payer
offering commercial and managed Medicaid coverage in the Upper Midwest. The time frame

65
was selected to capture pre-HCH certification using multiple years observations as the first HCH
were certified in mid-July 2010. Enrollees who were: 1) 18 years old or younger in the calendar
year, 2) resided in Minnesota, 3) had a full-year medical in the calendar year, and 4) had at least
6 months of medical coverage in the year prior to the calendar year, were included for CSHCN
identification. The last requirement was included to assure the feasibility of retrospective clinic
attribution and of estimating the prior year comorbidity burden.
The Institutional Review Board of the University of Southern California determined this
study was qualified for exemption from review because the protected health information was
removed and/or de-identified in the claims data for this study, in concordance with the Health
Insurance Portability and Accountability Act.

CSHCN identification
The CSHCN screener is the gold standard used to identify CSHCN in multiple national
surveys, including the National Survey of Children’s Health (NSCH) and the National Survey of
Children with Special Health Care Needs (NS-CSHCN).
4
There are five non-condition specific,
health consequence-based, non-mutually exclusive items in the screener. It assesses: 1) need for
or use of medicine prescribed by a doctor (other than vitamins); 2) need for or use of more
medical care, mental health care, or educational services than is usual for children of the same
age; 3) being limited or unable to do the things most children of the same age can do; 4) need for
or receipt of special therapy, such as physical, occupational, or speech therapy; and 5) emotional,

66
developmental, or behavioral problems for which treatment or counseling is needed or received.
A “yes” in any of the questions triggers a follow-up question regarding duration and whether the
consequence is attributable to medical, behavioral, or other health conditions (except in question
5). The screener is designed to capture both health services needs and health services use if the
health consequences last or are expected to last for at least 12 months.
4

CSHCN were identified by applying a validated algorithm that maps the CSHCN screener
onto claims.
31
The algorithm was developed and validated by Arim et al.
31
It demonstrated
similar overall prevalence in demographic and socioeconomic distributions compared to findings
in Canadian surveys for children age 6-10.
31
We followed Arim et al. specification, a total of 3
years (1 year before/after the calendar year, if available) of medical and pharmacy claims at the
most were used to determine whether a child belonged to CSHCN in that calendar year. A brief
description of criteria for each screener item is shown in Figure 1. To our knowledge, no
claims-screener mapping CSHCN has been published for CSHCN in the US. The Arim algorithm
has demonstrated good validity when compared to the Pediatric Medical Complexity Algorithm
(PMCA)
65
and also with the resource utilization bands (RUBs) in the Johns Hopkins ACG
system.
66
Detailed algorithm applications and a validity check are in Appendix A. In order to
control for individual/clinical differences, individuals and clinics were required to have at least
two observations during the study period (at least one observation in pre- and post-HCH
certification if the child was in the HCH group).
Medicaid programs in Minnesota are required to provide pharmacy coverage, while about

67
15% of preferred provider organization (PPO) plan enrollees did not have pharmacy enrollment
for unknown reasons (this included not offering pharmacy coverage, having pharmacy coverage
by other insurance companies, etc.) Selection problems may exist if the child was included in the
CSHCN sample solely by prescription possession criteria. To determine the extent of this
potential selection bias, we conducted a sensitivity analysis, excluding CSHCN who qualified by
the Q1 item only. As the significance and magnitude of the HCH effects were similar to the main
analysis, we did not consider this as a threat (Appendix D, Table D1).

Explanatory variable: HCH indicator
The HCH indicator was constructed for each child by calendar year in 2007-2014 according
to the clinic attribution results. We performed a retrospective attribution for CSHCN who met the
full-year medical enrollment requirement and had no missing demographic variables. CSHCN
were attributed to a clinic based on the plurality of primary care visits in the year prior to the
calendar year, with the most recent visit breaking ties. A primary care visit was defined by
selected CPT codes, place of service (POS) codes, and specialty codes. Only clinics that
represented more than 35% of the child’s primary care visits were considered in the attribution.
The cutoff was arbitrarily selected to resemble the idea of continuity in care and to guarantee that
only two clinics at the most would be compared in the clinic assignments. Close to 80% of
CSHCN who were identified by the mapping algorithm were successfully attributed to a clinic.
We imputed the attribution results for those who were attributed to a clinic in the prior year but

68
did not have any primary care visits in the current year, using the
last-observation-carried-forward method. The year-to-year patient retention rates were very high,
with only 2.6% of children switching between clinics between years. The HCH status of the
clinic to which a child was attributed determined whether this child was in the HCH group. The
date of the HCH certification was set as the earliest health care provider-certified date in the
HCH-provider list on the MDH website.
87
The year of certification was rounded to the nearest
January 1
st
to the certification date. HCH status remained for all proceeding years after
certification, since almost all of the HCH evaluated completed annual recertification in
2011-2014.
88

Outcomes: health care expenditure, utilization & quality of care
To study the patterns of health care expenditures and utilization, we categorized claims by
POS, including hospital inpatient and outpatient, office-based visits, ED visits, urgent care (UC)
facilities, and other POS. Other POS included health care services provided at home,
independent laboratories, community mental health centers, schools, etc. We also categorized
expenses by type of services using CPT codes. Common types of services included inpatient
facilities, evaluation and management visits, laboratories, radiology units, and surgical units.
Other diagnosis and treatments included biopsy, diagnostics, preventive treatment, and mental
health services. Other services with relatively smaller shares, including anesthesia, dentistry,
vision and hearing-related services, medication/infusion, transportation, and equipment supplies,

69
were combined into the “other types” category in our analysis. A standardized fee schedule was
created using available insurance claims, by averaging the dollar amounts for each CPT and
DRG codes (and NDC level for prescriptions) across all types of insurance plans and years.
Applying the standardized fee schedule to account for different reimbursement rates allows us to
interpret the magnitude of the differences as utilization. To address potential selection bias in
CSHCN with public and commercial plans, we only analyzed prescription costs for CSHCN with
full-year pharmacy benefits. All cost estimates were estimated and reported at per-member
per-month (PMPM) level.
We selected five claims-based NCQA-endorsed measures from the core set of child health
quality measures for Medicaid and CHIP programs
67
: follow-up visits for mental health
hospitalization (FUH7, FUH30), medication management for people with asthma (MMA),
appropriate testing for children with pharyngitis (CWP), appropriate treatment for children with
upper respiratory infection (URI), and follow-up care for children prescribed ADHD medication
(ADD). We constructed each indicator following the 2016 Healthcare Effectiveness Data and
Information Set (HEDIS) technical specifications, with minor modifications to accommodate
data availability. Pediatric quality of care measures that are highly correlated with primary care
visits, such as well-child visits and immunization, were not included because the number of
office-based visits is one of our sample selection criteria. We compared the percentage of
CSHCN that met the HEDIS goals to the 2013 national average rates, which were calculated
using all eligible children.
68
In addition, an indicator of having any avoidable ED visits was also

70
included as a quality measure. ED visits that did not result in hospital admission were classified
into non-emergent and emergent levels following rules set in Ballard et al. and the NYU ED
algorithm.
69,70
Detailed information on the construction of these quality of care measures is in
Appendix B.

Covariates
We controlled for a range of covariates that were likely to affect outcomes. We adopted
Anderson’s behavioral model
45
of health services use as a framework to categorize our predictors
and covariates into: (1) predisposing factors, including age and sex from the enrollment data; (2)
enabling factors, including residence in an urban area, race/ethnicity, education level, spoken
language other than English, and poverty level, estimated by neighborhood socioeconomic
effects derived from the 2010 US Census and the 2011 American Community Survey (ACS)
5-year average data at Census tract level; and (3) needs factors, including the prior-year
comorbidity burden estimated by the resource utilization bands (RUBs) from the Johns Hopkins
ACG system
66
, as well as selected common condition indicators for CSHCN. CSCHN
demographics, neighborhood effects, unbalanced condition indicators, and year dummies (to
control for trend) were adjusted in the model for outcomes.

Statistical analysis
Quantifying the effects of HCH is challenging in the context of this study, because CSHCN

71
were not randomly assigned to a HCH. There are two sources of unobserved preference: (1) the
choice of clinics to become HCH certified, and (2) the choice of a child in getting care in an
HCH clinic. For example, the former might be related to unmeasured organizational policy and
structure that otherwise influence the outcomes of interest while the later might be a result of
parents’ confidence and trust of the clinic. The presence of unobserved variables leads to a biased
coefficient due to the correlation between the explanatory variable (which is the HCH status in
our study), and the error term (omitted variables bias). This endogeneity problem is a threat to
internal validity.
A difference-in-difference (DD) framework in a two-part model with individual and clinic
random effects was applied to examine the impacts of HCH on outcomes. This design allows us
to account for unmeasured differences between HCH and non-HCH groups, in a manner that
enables estimation of HCH effects in a nonrandomized setting. We assumed that individual and
clinic random effects mitigated individual/clinic differences and that they should be able to
capture time-invariant unobserved factors, such as changes to parents’ preference of clinics and
clinic/organization management styles usually take a long period of time. Note that though fixed
effect is considered more reliable for controlling time-invariant confounders, we used random
effects, which also include differences between individuals, to accommodate the two-part model
variance structures.
89

The idea of DD is that changes in outcomes among the HCH clinics following the
certification process cannot be explained by observed changes in outcomes among non-HCH

72
clinics. Therefore, the changes in outcomes were attributed to the HCH. The approach removes
biases in post-HCH period comparisons between HCH and non-HCH that could be the result
from differences in the pre-HCH period, as well as biases from comparisons over time in the
HCH group that could be the result of trends due to other unobserved causes of the outcome. The
basic DD model for child i in clinic j and time period t outcome (𝑌 𝑖𝑗𝑡 ) can be described as
follows:
𝑌 𝑖𝑗𝑡 = 𝛽 0
+ 𝛽 1
𝐻𝐶𝐻 𝑗 + 𝛽 2
(𝐻𝐶𝐻 𝑗 ∗ 𝑝𝑜𝑠𝑡 𝑗𝑡
) + 𝛽 3
𝑦𝑒𝑎𝑟 𝑡 + 𝛽 4
𝑋 𝑖𝑡
+ 𝛼 𝑗 + 𝜑 𝑖 + 𝜀 𝑖𝑗𝑡
where i indexes individual children, j indexes clinics, and t indexes calendar year. The utilization,
cost, or quality outcomes (Y) were estimated as a function of an interaction between a variable
indicating that a clinic was an HCH at some time (𝐻𝐶𝐻 𝑗 ) and whether the clinic was certified in
a particular year (𝑝𝑜𝑠𝑡 𝑗𝑡
), year dummy variables (𝑦𝑒𝑎𝑟 𝑡 ), a vector of time-variant
individual-level characteristics (𝑋 𝑖𝑡
) including age and comorbidities indexes (covariates), a
random effect for clinic j (𝛼 𝑗 ), a random individual effect over years (𝜑 𝑖 ), and a random error
component (𝜀 𝑖𝑗𝑡 ). The coefficient of the interaction term between the HCH indicator and
certification status (𝛽 2
) is the HCH effect of our interest. While analyzing the data, we combined
individual and clinic data because we assumed that these two effects nested within each other.
The assumption was based on the fact that only 2.6% of the sample CSHCN were attributed to
different clinics over the study period (Appendix C). To assure the nested effects were solid, we
excluded these switchers as a sensitivity analysis, and the results were similar (Appendix D:
Table D2).

73
The key assumption of the DD model is that in the absence of HCH certification, the
outcomes would have been the same regardless of whether a child received care from a HCH
clinic or not. To verify this common (parallel) trend assumption between CSHCN in HCH and
non-HCH groups, we added interaction terms for 1 year before HCH certification (HCH
j
∗
pre1
jt
) and 2 or more years before HCH certification (HCH
j
∗ pre2
jt
) in the DD model. The
model is described as follows with the same notations:
𝑌 𝑖𝑗𝑡 = 𝛽 0
+ 𝛽 1
𝐻𝐶𝐻 𝑗 + 𝜃 𝑝𝑟𝑒 2
(𝐻𝐶𝐻 𝑗 ∗ 𝑝𝑟𝑒 2
𝑗𝑡
) + 𝜃 𝑝𝑟𝑒 1
(𝐻𝐶𝐻 𝑗 ∗ 𝑝𝑟𝑒 1
𝑗𝑡
)
+ 𝜃 𝑝𝑜𝑠𝑡 (𝐻𝐶𝐻 𝑗 ∗ 𝑝𝑜𝑠𝑡 𝑗𝑡
) + ∑ 𝛽 3
𝑦𝑒𝑎𝑟 𝑡 + 𝛽 4
𝑋 𝑖𝑡
+ 𝛼 𝑗 + 𝜑 𝑖 + 𝜀 𝑖𝑗𝑡
All notations are the same as the previous model. The coefficients of the interaction terms
(𝜃 𝑝𝑟𝑒 1
, 𝜃 𝑝𝑟𝑒 2
) allowed us to formally test whether HCH and non-HCH groups were significantly
different (against 0). When examined using expenditures outcomes, the parallel assumptions hold
well in general (Appendix D: Table D3).
Health care expenditures and utilization were modeled using a two-part model, in which the
DD framework was applied to both Part 1 and Part 2. The model was designed to address
patients with zero expenditures.
48
Part 1 modeled the likelihood of incurring expenditures using a
probit regression; the finding can be interpreted as the probability of the child accessing care.
Part 2 estimated the impacts of HCH on health care expenditures, conditional on accessing care.
The log-transformed cost in Part 2 accounts for the right-skewed distribution of the expenditures.
HCH effects on the dichotomous quality indicators were estimated using a random effect logistic
regression. All statistical significance was judged at a two-sided α= 0.05 level. Data were

74
prepared using SAS software, version 9.4 (SAS Institute, Inc.) and were analyzed using STATA
MP, version 14.0 (StataCorp).
In addition, we also included interaction terms for the years after HCH certification and the
HCH status in the model, in order to investigate lagged or diminishing effects of the HCH on
outcomes for CSHCN. Specifically, 𝑝𝑜𝑠𝑡 1 and 𝑝𝑜𝑠𝑡 2 are indicators for the year of HCH
certification and the second and more years after HCH certification, respectively. We only
examined the second and more years due to small sample size for observations with three or
more years after HCH certification. The DD model is described as follows with similar features
to the aforementioned one:
𝑌 𝑖𝑗𝑡 = 𝛽 0
+ 𝛽 1
𝐻𝐶𝐻 𝑗 + 𝜃 𝑙𝑎𝑔 1
(𝐻𝐶𝐻 𝑗 ∗ 𝑝𝑜𝑠𝑡 1
𝑗𝑡
) + 𝜃 𝑙𝑎𝑔 2
(𝐻𝐶𝐻 𝑗 ∗ 𝑝𝑜𝑠𝑡 2
𝑗𝑡
) +
∑ 𝛽 3
𝑦𝑒𝑎𝑟 𝑡 + 𝛽 4
𝑋 𝑖𝑡
+ 𝛼 𝑗 + 𝜑 𝑖 + 𝜀 𝑖𝑗𝑡
All notations are the same as the previous model. The coefficients of the interaction terms
(𝜃 𝑙𝑎𝑔 1
, 𝜃 𝑙𝑎𝑔 2
,) are the effects of HCH intervention with one and two or more years after HCH
certification. By comparing these coefficients, we will be able to investigate if any lagged or
diminishing effects exist.
In our analytical framework, CSHCN composition before and after the HCH
implementation (i.e., pre- and post-period) could vary and could be the source of observed
differences in outcomes. In addition, the diffusion of the HCH effects (i.e., some features in the
PCMH may already exist prior to formal certification or may take time to be effective after
certification) around the year of certification should also be taken into account when quantifying

75
the size of HCH effects. We analyzed CSCHN characteristics in relation to HCH status using a
balanced cohort of CSHCN, defined as CSHCN who had 4 consecutive years of observation
between 2009 and 2014, to address the potential compositional changes. When comparing the
study sample and the balanced cohort, the differences met our expectation (Appendix D: Table
D4). For example, we expected that CSHCN were older in age and less likely to have Medicaid
coverage in the balanced cohort in comparison to the study sample (Appendix D: Table D4).
When comparing the health care expenditure outcomes in 1-year post-HCH to 1-year pre-HCH,
the results were similar, but unlikely to be significant due to lack of power under small in sample
size (Appendix D: Table D5).
To investigate whether the effects of HCH varied across subgroups in CSHCN, we
conducted two subgroup analyses: (1) CSHCN with complex needs, defined as CSHCN met
more than three items in the CSHCN screener; and (2) CSHCN with emotional, behavioral, or
developmental conditions (EDB), defined as CSHCN who met Q5 and at least one other item.
90

These two groups of children have been reported to have higher medical expenditures than
averaged CSHCN and are thus where the effort of reducing expenditures should be focused on.
90

Note that we only examined cost outcomes for subgroups due to small eligible sample sizes,
which were expected to limit our powers to interpret the findings.

76
Results
A total of 87,172 person-year observations were included in our sample, with 46% in the
HCH group. Table 1 shows the CSHCN characteristics for the HCH group and the non-HCH
group. Compared to CSHCN not receiving care from a HCH, those who had a HCH were more
likely to : (1) be female; (2) have public insurance and pharmacy coverage; (3) reside in urban
areas; (4) have higher comorbidity burdens in the prior year; (5) be younger in age; and (6) live
in less educated, more Hispanic and Black neighborhoods (Table 1). ADHD was the most
common condition in our study sample and was more prevalent among CSHCN without a HCH
(Table 1). Anxiety, asthma, and autism were also common conditions for CSHCN and these
children were more likely to be in the HCH group (Table 1).

Health care utilization & expenditures
Almost all CSHCN incurred medical expenses (HCH vs. non-HCH group: 99.6% vs. 99.7%,
p=0.005; Table 2). CSHCN in HCH group were more likely to access care in ED and UC and to
use radiology services, whereas non-HCH CSHCN were more likely to access care in other POS
settings and to use laboratory and diagnostic and treatment services (Table 2). After accounting
for baseline characteristic differences and covariates (as shown in Part 1 results in Table 3),
CSHCN who received care from a HCH were more likely to access care in inpatient (increased
by 0.72%, p=0.016), outpatient hospital (2.64%, p=0.003), and other POS settings (2.03%,
p=0.019), whilst ED and UC utilization rates were comparable or lower than those without a

77
HCH (ED: 0.42%, p=0.508; UC: -0.88%, p=0.039; Table 3).
HCH effects did not significantly shift total medical expenditures in both unadjusted (HCH
vs. non-HCH: $535 vs. 524, p=0.455) and multivariate adjusted estimates (1.5% increase for
HCH group, p=0.223). When examined by place and type of services, CSHCN in the HCH group
had higher unadjusted PMPM average expenditures in hospital outpatient and ED settings, as
well as for evaluation and management and other types of health care services, compared to
those in non-HCH groups (Table 2). These differences disappeared in the marginal effects (i.e.,
the combined Part 1 and Part 2 data) of the multivariate two-part model, which controlled for
covariates and differences in CSHCN characteristics, with the exception of expenditures in UC
settings and for evaluation and management services (Table 3). The significant reduction in UC
expenditure (-18.64%, p< 0.001) was driven by a decrease in accessing care (-0.88%, p=0.039)
and a lower average expense for UC services (-4.38%, p=0.048) in CSHCN with a HCH (Table
3). However, UC expenditure only accounted for roughly 0.6% of overall medical expenditures.
For evaluation and management expenses, the increase in overall (i.e., marginal effect)
expenditure became close to significant for CSHCN receiving care from HCH. The increase in
average costs when accessing this service seemed to be the driver (Part 1: -0.01%, p=0.930; Part
2: 2.06%, p=0.090; overall: 2.17%, p=0.046; Table 3). For other POS categories, the increase in
accessing care was offset by a decrease in average expenditure when accessing care, and resulted
in similar overall expenditures for HCH and non-HCH groups, although none of these effects
was significant (Table 3). Note that the bimodal distribution in prescription expenses and the

78
relatively small number of observations meant inpatient expenses failed the two-part model with
random effects. Therefore, if no convergence was achieved, we did not report adjusted
prescription expenditures and the marginal effects of inpatient expenditures.
Table 4 shows the DD coefficient estimates in the year of HCH certification (Year 1) and
the second and later years (Year 2+) of certification. It seemed to be an increasing trend to
accessing care in hospital outpatient and other POS, since the Part 1 coefficient continued to
grow in magnitude from Year 1 to Year 2+ (Table 4). This increasing trend was also found for
overall expenditures in other POS settings, as well as for evaluation and management services,
whilst expenditures in UC continued to decrease. Total medical expenses and expenditure for
other type of health care services decreased in Year 1, then bounced back with a significant
increase in Year 2+ (Table 4). In addition, an initial increase in the probability of accessing
office-based care was found in Year 1, but the effects did not sustain to Year 2+ (Table 4).
Several points of significance emerged in Year 2+, implying the existence of lagged HCH effects
on health care utilization and expenditures.

Quality of care
CSHCN performed similarly in quality of care measures regardless of whether they
received care from HCH or not (Table 5). The only significant difference was that CSHCN with
HCH had lower odds of having proper follow-up visits after ADHD medication was prescribed
(OR= 0.679, 95% CI: 0.459-0.996). The HCH effects on this indicator were only in Year 1; HCH

79
showed no effects in Year 2+ when analyzing the lagged HCH effects (Table 6). As a point of
caution, the eligible sample sizes for the quality measures were small, so their power to detect
differences and provide meaningful estimates was questionable. Therefore, we did not conduct
sensitivity and subgroup analyses on the quality outcomes.

Subgroup analysis: CSHCN with complex needs & emotional, behavioral, or developmental
conditions
We found a total of 19,198 person-year observations (22% of CSHCN sample) belonging to
CSHCN with complex needs. After controlling for CSHCN characteristics, none of the
probabilities of accessing care were significant, which differed from the findings in the CSHCN
sample (Table 7). In addition, the HCH impacts on elevated evaluation and management
expenditure and on the reduction in UC costs were larger in magnitude (Table 7). This suggested
that the access patterns for complex CSHCN were not affected by HCH. Their unmet medical
needs may thus persist with the presence of HCH.
HCH seemed to provide similar access to care in all POS, except for an increase in
accessing hospital outpatient care (Table 8), compared to regular clinics for EDB CSHCN (n=
22,053, 25% of CSHCN sample). HCH impacts on the increased evaluation and management
expenditure and on reduction in UC costs were larger in magnitude compared to CSHCN sample
(Table 8). CSHCN with EDB had significantly higher overall expenditures among various
service types, including the use of inpatient facilities, evaluation and management, radiology, and

80
surgery. For this subgroup of CSHCN, HCH seemed to improve their exposures to different
types of services and led to an increase in expenditures faced by individuals (Table 8).

Discussion
We did not find evidence to support that HCH reduce costs and improve quality of care for
CSHCN in Minnesota. CSCHN who received care from a HCH had similar exposures to various
types of health care services, but had increased probabilities of accessing care in hospital
inpatient and outpatient and other POS settings without significantly shifting costs, compared to
those not receiving care from HCH. Whilst a large body of literature focusing on the adult
population has reported some levels of positive outcomes
21,26,85
, our findings for CSHCN
suggested that the PCMH impacted this vulnerable population differently from general adults,
with the implication that the current PCMH model may not benefit this vulnerable population.
This study leveraged the consistency in the HCH initiative with an advanced econometric
model with sufficient pre-HCH certification years of observation to quantify the impacts of
PCMH on health services for CSHCN, one of the most vulnerable populations that the concept
behind the model was originally designed for. The HCH certification process provides researches
with the ability to identify PCMH with confidence in the level of “medical home-ness”. Both the
certification and annual recertification process require clinics to gradually meet criteria of
comprehensive care and to satisfy quality measures.
88
The examination of the lagged HCH

81
effects resembles the stages of the HCH implementation. When HCH status lasted two or more
years, it significantly increased total medical expenditure and expenditures in other POS settings,
as well as evaluation and management expenditures and other types of services, whereas a
non-significant increase was found in the first year of certification. This is consistent with the
findings that suggest fully implemented PCHM introduced a larger effect than a partially
implemented PCMH and the existent of the lagged effects of the PCMH programs in Iowa and
North Pennsylvania.
24,25,91
However, we found that CSCHN experienced higher medical
expenses years after HCH certification, in contrast to no effects on cost for the pediatric
population described in Paustian et al.
91
More research is needed to examine whether the HCH
effects for CSHCN sustain and further translate into desired utilization patterns, such as efficient
inpatient and ED use, or this is a result of pent-up demand when long-term data becomes
available.
We applied a valid mapping algorithm
31
of the CSHCN screener on insurance claims for
CSHCN identification. In the past, CSHCN were operationalized as having selected chronic
conditions in claims analysis. Condition-based CSHCN selection may leave out children with
elevated use of health services due to prolonged periods of diseases of less severity, such as
headache / migraine, and may not meet the true definition of CSHCN. However, this
improvement in CSHCN identification could introduce greater heterogeneity to the study sample
and provide a more inclusive definition of CSHCN. We approached this issue by analyzing a
subgroup of CSHCN who had EDB conditions. The EDB subgroup in our study had increased

82
expenditures and utilization in various types of services, which suggested a different utilization
pattern compared to overall CSHCN. This reinforces the view that PCMH effects vary across
disease areas
92
(mental and behavioral health services in our case), and warrants the collection of
more evidence to design a model to help those who would be benefit the most.
In addition to differential effects by disease areas, several research projects have suggested
that the PCMH model benefits high-risk / resource use patients the most.
21,93-97
CSHCN are
essentially the high resource use pediatric population, whereas complex CSHCN in our subgroup
analysis represent those with even higher level of health resources use. One randomized clinical
trial of PCMH (i.e., comprehensive care), for pediatric patients with chronic illness who had high
health care use in prior year (>2 ED visits, >1 hospitalization, or >1 pediatric intensive care unit
admission), concluded that those children who received comprehensive care had significant
improvements in inpatient and ED use with a reduction in total cost, driven by a decrease in
inpatient costs.
94
Our findings did not reinforce the argument for the pediatric population using the
real-world claims data, as the HCH did not introduce better health care utilization or reduce costs
for both CSHCN and complex CSHCN. In fact, the increases in accessing care in certain settings
for CSHCN disappeared for complex CSHCN. The sample size of complex CSHCN was relatively
small, so the power of detecting difference using advanced econometric models was limited.
Additional research is warranted to evaluate whether risk stratification is as meaningful as it
appears to be in adult populations, when it is CSHCN who are being considered.
This study had several limitations. First, CSHCN identification was conditional on having

83
health care utilization. Those who had school-based services, such as intellectual disabled
children who attended special educational programs and were only required to be followed up by
health care providers annually or less, could be excluded in our sample. Second, we only
included neighborhood socioeconomic effects for CSHCN and their households. This would
introduce measurement errors into an important covariate. Since the actual socioeconomic status
was not available in claims, this was the best proxy that we could apply. Third, only
cross-sectional HCH-provider relationships were publically available, and were used to link
insurance claims to a clinic and its HCH status for clinic-HCH status determination. The changes
in the HCH-provider linkage could be different (i.e., the physician left a clinic and started work
for the other clinic) during the time prior to which we obtained the list. CSHCN were attributed
to a clinic using the insurance claims-provider-clinic linkage. When minimal switching between
clinics was found, this issue is considered less concerning to us. In addition, a clinic may have
already provided care that met some dimensions of PCMH before being HCH-certified. Some
non-HCH certified practices may embed some features of PCMH in their services, but did not
formally apply for the certification. HCH transformation takes time to prepare and several
months’ effort is needed before the formal application to be certified is made. No information in
the “medical home-ness” for a clinic in the pre-HCH period, or of those clinics without
certification was available. We considered this as a measurement error in the effects of
intervention, which typically attenuates the effects and biases our findings toward null. In
addition, we relied only on the DD model to mitigate time-invariant unobserved confounders.

84
The common trend assumption was not violated but this cannot be formally tested. Individual
random effects may be unable to fully balance unobserved confounders, especially regarding
between-individual factors, in our model specification.
89
Due to the nature of the observational
data, these unobserved confounders could bias the estimations and pose potential threats in
establishing causality. Last but not least, the generalizability of our findings in PCMH impacts on
CSHCN is limited because the HCH is a state-specific initiative.

Conclusions
There is little evidence that HCH resulted in better health care utilization, reduced costs, or
improved quality of care for CSHCN in Minnesota. We found a significant decrease in UC
expenditures immediately and an increase in evaluation and management expenditure, mainly
driven by the lagged HCH effects (i.e., in the second and later years with HCH certification). Our
findings imply that the PCMH impact on CSHCN could be different from adults, since a large
body of literature provides evidence of some outcomes improvement. It is worth noting that the
EDB patients who received care from a HCH had higher inpatient, radiology and surgery costs,
suggesting that the effects of HCH might differ from condition to condition. Current PCMH may
not benefit this vulnerable population the most. Further research with a longer time horizon and a
focus on disease areas or risk stratification could help to better target efforts to optimize primary
care delivery for CSHCN.

85
Tables & Figures
Table 1. Children with special health care needs (CSHCN) characteristics, by health care
home status
CSHCN characteristics
HCH
n= 40,110
(46.0%)
No HCH
n= 47,026
(54.0%)
P-value
Male, % 53.1% 54.8% <0.001
Resided in urban area, % 91.2% 83.0% <0.001
Public insurance, % 25.6% 25.4% <0.001
Had full-year pharmacy coverage, % 90.6% 88.1% <0.001
Comorbidity burden, % <0.001
Healthy/ no user 4.4% 4.5%
Low 28.4% 30.1%
Moderate 50.9% 49.9%
High 13.6% 12.9%
Very high 2.8% 2.7%
Age groups, % <0.001
0-5 years old 18.4% 17.1%
6-11 years old 33.1% 34.0%
12-17 years old 48.4% 48.9%
Age in year, mean (SD) 10.5 (4.7) 10.6 (4.6) 0.005
Neighborhood effects, mean (SD)
% of race/ethnicity
Hispanic 4.3 (7.7) 3.9 (6.6) <0.001
White, non-Hispanic 83.9 (18.7) 85.1 (16.1) <0.001
Black, non-Hispanic 5.1 (10.5) 4.2 (8.7) <0.001
Others, non-Hispanic 6.7 (7.9) 6.8 (7.8) 0.004
% of education level
Less than high school 7.5 (7.7) 6.6 (7.0) <0.001
High school/GED 26.3 (10.8) 23.9 (12.1) <0.001
Some colleges 33.6 (8.6) 31.3 (9.0) <0.001
4-year/ graduate degree 32.6 (17.4) 38.3 (20.5) <0.001
% of speaking English only 89.8 (11.4) 90.3 (10.2) <0.001
% of < FPL 9.8 (9.3) 9.5 (9.1) <0.001
Selected conditions, %
Anxiety 15.0% 13.9% <0.001
Asthma 12.8% 10.8% <0.001
ADHD 24.1% 24.9% 0.006
Behavioral/conduct problems 4.4% 4.2.% 0.268
Autism 11.1% 10.3% <0.001

86
Depression 7.6% 7.8% 0.306
Developmental delay 0.5% 0.5% 0.886
Intellectual disability 6.0% 6.0% 0.782
Migraine/headache 1.6% 1.6% 0.886
Congenital disorders 6.0% 6.0% 0.782
Abbreviation: ADHD, attention deficit hyperactivity disorder; CSHCN, children with special health care
needs; FPL, federal poverty level; GED, general educational diploma; HCH, health care home; n, number
of person-year observation; SD, standard deviation.

87
Table 2. Unadjusted costs for CSHCN by HCH groups
PMPM Expenditures, mean (SD)
HCH No HCH P-value
Medical services
535 (2,244) 524 (2,116) 0.455
Incurring expenditure, %
99.6% 99.7% 0.002
Prescriptions
113 (333) 127 (397) <0.001
Incurring expenditure, %
91.0% 90.8% 0.508
By place of service

Inpatient
122 (1,373) 131 (1,448) 0.344
Incurring expenditure, %
6.7% 6.4% 0.077
Hospital outpatient
153 (869) 143 (785) 0.063
Incurring expenditure, %
41.5% 41.6% 0.687
Emergency department
27 (81) 24 (71) <0.001
Incurring expenditure, %
25.4% 24.7% 0.011
Urgent care
3 (10) 3 (9) <0.001
Incurring expenditure, %
16.2% 15.0% <0.001
Office-based visits
153 (585) 153 (473) 0.873
Incurring expenditure, %
98.5% 98.5% 0.373
Other
78 (830) 71 (655) 0.215
Incurring expenditure, %
34.4% 36.1% <0.001
By type of service

Inpatient facility
109 (1,311) 117 (1,342) 0.351
Incurring expenditure, %
5.6% 5.4% 0.135
Evaluation and management
51 (56) 50 (54) 0.027
Incurring expenditure, %
95.9% 96.1% 0.058
Radiology
23 (109) 23 (120) 0.955
Incurring expenditure, %
40.5% 38.6% <0.001
Laboratory
18 (56) 18 (62) 0.213
Incurring expenditure, %
65.2% 66.3% 0.001
Surgery
27 (106) 27 (111) 0.199
Incurring expenditure, %
23.3% 23.6% 0.176
Other diagnostic & treatment
159 (535) 161 (596) 0.720
Incurring expenditure, %
92.5% 93.0% 0.018
Other types
148 (1,327) 129 (1,017) 0.019
Incurring expenditure, %
64.2% 64.0% 0.417
Abbreviation: CSHCN, children with special health care needs; PMPM, per-member-per-month; SD, standard
deviation.
Note: A standardized fee schedule was applied to account for different reimbursement rates.

88
Table 3. Multivariate adjusted difference-in-difference estimators in two-part model of
health care expenditures for CSHCN

Types of Expenditure
Part 1:
% Change in Accessing
Care with HCH (95%
CI)
Part 2:
% Change in Conditional
Expenditures with HCH
(95% CI)
% Changes in
Expenditure from
Combined Model with
HCH (95%CI)
Medical services 0.06% (-0.01, 0.12) 1.41% (-2.40, 5.23) 1.48% (-2.41, 5.27)
By place of service
Inpatient 0.72%* (0.13, 1.31) -14.96% (-33.44, 3.52) N/A
Hospital outpatient 2.64%** (0.87, 4.41) -1.23% (-8.29, 5.83) 3.59% (-4.35, 10.70)
Emergency room 0.42% (-0.83, 1.67) -0.64% (-5.65, 4.37) 2.11% (-6.20, 11.40)
Urgent care -0.88%* (-1.71, -0.05) -4.38%* (-8.74, -0.03) -18.64%** (-30.68, -3.96)
Office-based visits 0.02% (-0.02, 0.06) -0.54% (-3.54, 2.46) -0.50% (-3.52, 2.27)
Other 2.03%* (0.34, 3.72) -1.66% (-9.74, 6.42) 7.70% (-2.37, 17.18)
By type of service
Inpatient facility 0.29% (-0.13, 0.71) 0.84% (-8.59, 10.28) 20.19% (-5.43, 50.52)
Evaluation &
management
-0.01% (-0.28, 0.25) 2.06% (-0.32, 4.44) 2.17%* (-0.42, 4.79)
Radiology -0.25% (-1.80, 1.21) -1.06% (-7.0, 4.93) -1.64% (-8.08, 5.20)
Laboratory 0.64% (-0.80, 2.09) 0.79% (-3.69, 5.28) 1.30% (-3.04, 5.67)
Surgery -0.01% (-1.15, 1.13) 2.74% (-4.35, 9.83) 2.99% (-5.72, 12.65)
Other diagnostic &
treatments
0.05% (-0.41, 0.51) -1.50% (-5.96, 2.96) -1.44% (-5.73, 2.86)
Other types 1.10% (-0.35, 2.55) 0.87% (-4.99, 6.73) 1.76% (-4.00, 7.65)
Abbreviation: CI, confidence interval; CSHCN, children with special health care needs; HCH, health care home;
N/A, not applicable.
Note: A standardized fee schedule was applied to account for different reimbursement rates. Part 1 of overall
medical services was not applicable because all CSHCN incurred medical expenses. Prescription expenditure and
combined model of inpatient expenditure did not achieve convergence in the estimation model.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001

89
Table 4. Multivariate adjusted difference-in-difference estimators for lagged effects in a two-part model of health care
expenditures for CSHCN

Abbreviation: CI, confidence interval; CSHCN, children with special health care needs; HCH, health care home; N/A, not applicable/not estimated.
Note: CSHCN not receiving care from HCH was the reference group. A standardized fee schedule was applied to account for different reimbursement rates.
Inpatient expenditure did not achieve convergence in the estimation model.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001
Types of Expenditure
Part 1: % Change in Accessing Care
with HCH (95% CI)
Part 2: % Change in Conditional
Expenditures with HCH (95% CI)
% Changes in Expenditure from
Combined Model with HCH (95%CI)
Year 1 Year 2+ Year 1 Year 2+ Year 1 Year 2+
Medical services 0.07% (-0.01, 0.15) 0.05% (-0.02, 0.12) -0.50% (-5.47, 4.47) 3.59% (-0.88, 8.06) -0.56% (-5.60, 4.86) 3.75%* (-0.54, 8.27)
By place of service
Inpatient N/A N/A N/A N/A

N/A N/A
Hospital outpatient -1.09% (-1.20, 3.38) 4.00%*** (1.35, 6.64) -1.39% (-10.65, 7.87) -0.95% (-9.24, 7.33) 0.61% (-9.27, 11.93) 7.28% (-1.95, 18.31)
Emergency room 0.10% (-1.51, 1.71) 0.72% (-0.76, 2.20) -1.84% (-8.42, 4.73) 0.39% (-5.54, 6.32) -1.03% (-11.22, 10.86) 4.73% (-5.29, 16.33)
Urgent care -0.95% (-2.03, 0.13) -0.80% (-1.77, 0.17) -5.42% (-11.24, 0.40) -3.58% (-8.65, 1.48) -21.68%* (-37.71, -5.36) -17.65%** (-32.22, -1.41)
Office-based visits 0.05%* (0.01, 0.09) -0.02% (-0.07, 0.03) -0.76% (-4.67, 3.15) -0.40% (-3.92, 3.13) -0.87% (-4.35, 2.78) -0.42% (-4.04, 2.87)
Other 1.60% (-0.59, 3.79) 2.58%* (0.59, 4.58) -5.36% (-15.85, 5.13) 2.39% (-7.08, 11.87) 1.80% (-10.73, 16.66) 15.40%* (2.92, 30.37)
By type of service
Inpatient facility 0.06% (-0.46, 0.59) 0.50% (-0.03, 1.02) -0.04% (-12,42, 14.62) 1.63% (-9.50, 12.76) 4.24% (-25.38, 40.06) 32.70% (-7.17, 81.82)
Evaluation &
management
0.01% (-0.34, 0.35) -0.03% (-0.34, 0.29) 0.67% (-2.43, 3.76) 3.53%* (0.74, 6.32) 0.59% (-2.42, 3.63) 3.63%** (0.62, 6.70)
Radiology -0.24% (-2.26, 1.79) -2.42% (-2.07, 1.59) -2.55% (-10.33, 5.22) 0.25% (-6.86, 7.37) -2.63% (-10.94, 6.32) -0.38% (-8.85, 8.20)
Laboratory 0.48% (-1.41, 2.36) 0.81% (-0.87, 2.49) 0.04% (-5.78, 5.86) 1.54% (-3.76, 6.85) 0.33% (-5.36, 6.08) 2.22% (-2.34, 7.95)
Surgery 0.26% (-1.22, 1.75) -0.24% (-1.58, 1.10) 0.75% (-8.42, 9.92) 4.46% (-4.03, 12.95) 2.36% (-9.03, 15.58) 3.61% (-6.68, 15.49)
Other diagnostic &
treatments
0.05% (-0.55, 0.64) 0.05% (-0.48, 0.58) -1.92% (-7.74, 3.89) -1.05% (-6.29, 4.19) -1.92% (-7.70, 3.73) -1.05% (-5.92, 4.15)
Other types 1.09% (-0.81, 2.98) 1.14% (-0.56, 2.84) -3.76% (-11.37, 3.84) 5.64% (-1.29, 12.56) -2.90% (-9.80, 4.72) 6.93%* (-0.57, 14.68)

90
Table 5. Unadjusted and multivariate adjusted difference-in-difference estimation of HCH
impacts on quality of care outcomes for CSHCN

Abbreviation: ADHD, attention-deficit/hyperactivity disorder; CI, confidence interval; CSHCN, children with
special health care needs; HCH, health care home; n, number of person-year observations.
Note: HCH and no HCH reported unadjusted percentage of meeting the Healthcare Effectiveness Data and
Information Set (HEDIS) quality measures. CSHCN without a HCH was the reference group for the odds ratio.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001

Quality of Care Indicators n HCH
No
HCH
Odds
Ratio 95% CI
Follow-up for mental health hospitalization within 7 days
(FUH7)
346 48.6% 56.3% 0.659 (0.280, 1.549)
Follow-up for mental health hospitalization within 30
days (FUH30)
346 73.5% 78.4% 1.278 (0.396, 4.128)
Appropriate testing for pharyngitis (CWP) 1,017 88.9% 87.7% 2.023 (0.693, 5.905)
Appropriate antibiotics use for upper respiratory tract
infection (URI)
861 90.5% 87.5% 0.812 (0.262, 2.522)
Follow-up for children with ADHD prescription (ADD)
- Initial phase 1,818 41.4% 42.5% 0.676* (0.459, 0.996)
- Continued & maintenance phase 374 28.8% 31.4% 0.602 (0.085, 4.251)
Medication management for people with persistent asthma (MMA)

- 5-11 years old 1,446 49.5% 47.5% 0.607 (0.317, 1.160)
-12-18 years old 1,317 42.5% 44.8% 1.138 (0.547, 2.367)
Had avoidable emergency department visits 7,364 77.8% 77.0% 1.202 (0.944, 1.531)

91
Table 6. Multivariate adjusted difference-in-difference estimators of lagged HCH impacts
on quality of care outcomes for CSHCN
Abbreviation: ADHD, attention-deficit/hyperactivity disorder; CI, confidence interval; CSHCN, children with
special health care needs; HCH, health care home; OR, adjusted odds ratio.
Note: Year 1 is the year that HCH got certified; Year 2+ is the second and more years after HCH certified. CSHCN
without a HCH was the reference group for the odds ratio.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001

Quality of Care Indicators
HCH: Year 1 HCH: Year 2+
OR 95% CI OR 95% CI
Follow-up for mental health hospitalization within
7 days (FUH7)
0.561 (0.199, 1.581)

0.789 (0.271, 2.293)
Follow-up for mental health hospitalization within
30 days (FUH30)
0.951 (0.251, 3.600)

1.988 (0.432, 9.147)
Appropriate testing for pharyngitis (CWP) 2.312 (0.605, 8.835) 1.724 (0.423, 7.027)
Appropriate antibiotics use for upper respiratory
tract infection (URI)
0.729 (0.185, 2.870)

0.914 (0.219, 3.811)
Follow-up for children with ADHD prescription (ADD)
- Initial phase 0.601* (0.365, 0.990) 0.746 (0.468, 1.187)
- Continued & maintenance phase 0.162 (0.016,1.634) 3.156 (0.229, 43.580)
Medication management for people with persistent asthma (MMA)
- 5-11 years old 0.625 (0.298, 1.311) 0.583 (0.473, 1.469)
-12-18 years old 1.383 (0.574, 3.328) 0.941 (0.259, 2.312)
Had avoidable emergency department visits 1.155 (0.847, 1.577) 1.239 (0.391, 1.649)

92
Table 7. Multivariate adjusted difference-in-difference estimators in a two-part model of
health care expenditures for CSHCN with complex needs

Types of Expenditure
Part 1:
% Change in Accessing
Care with HCH (95%
CI)
Part 2:
% Change in Conditional
Expenditures with HCH
(95% CI)
% Changes in
Expenditure from
Combined Model with
HCH (95%CI)
Medical services N/A 0.51% (-5.26, 6.29) N/A
By place of service
Inpatient 1.68% (-0.10, 3.46) -9.42% (-34.62, 15.77) 14.48% (-9.76, 42.74)
Hospital outpatient 3.20% (-0.60, 7.01) -1.71% (-14.91, 11.48) 1.15% (-11.87, 14.27)
Emergency room 0.71% (-2.30, 3.71) 1.09% (-8.51, 10.68) 3.95% (-9.59, 17.81)
Urgent care -1.57% (-3.62, 0.48) -3.04% (-11.56, 5.49) -21.96%* (-45.34, 0.72)
Office-based visits 0.00% (-0.01, 0.01) -1.97% (-9.78, 5.78) N/A
Other 0.78% (-3.62, 5.18) 4.65% (-27.39, 36.70) N/A
By type of service
Inpatient facility 1.22% (-0.40, 2.84) 0.50% (-9.68, 19.65) 36.44% (-0.80, 83.97)
Evaluation &
management
0.20% (-0.05, 0.46) 4.67% (-0.55, 9.90) 4.91%* (-0.47, 10.13)
Radiology 1.23% (-2.09, 4.54) 6.24% (-6.03, 18.51) 9.11% (-4.63, 22.46)
Laboratory 1.87% (-0.78, 4.51) 0.47% (-9.14, 10.08) 1.51% (-8.06, 12.74)
Surgery 0.56% (-2.07, 3.19) 5.93% (-8.49, 20.34) 9.29% (-8.47, 28.55)
Other diagnostic &
treatments
-0.01% (-0.07, 0.04) -4.24% (-12.27, 3.79) -4.31% (-12.35, 4.06)
Other types -1.65% (-4.30, 1.00) 9.08% (-3.78, 21.93) 8.82% (-3.64, 23.43)
Abbreviation: CI, confidence interval; CSHCN, children with special health care needs; HCH, health care home; n,
number of person-year observation; N/A, not applicable/not estimated.
Note: n=19,198 person-year observations were used in sensitivity analysis. A standardized fee schedule was applied
to account for different reimbursement rates. Part 1 of overall medical services was not applicable because all
CSHCN incurred medical expenses. Combined parts for office-based visits and other place of services expenditure
did not achieve convergence in the estimation model.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001

93
Table 8. Multivariate adjusted difference-in-difference estimators in a two-part model of
health care expenditures for CSHCN with emotional, developmental, or behavioral
conditions
Types of Expenditure
Part 1:
% Change in Accessing
Care with HCH (95%
CI)
Part 2:
% Change in Conditional
Expenditures with HCH
(95% CI)
% Changes in
Expenditure from
Combined Model with
HCH (95%CI)
Medical services N/A 1.12% (-4.23, 6.47) N/A
By place of service
Inpatient 1.38% (-0.02,2.78) -7.45% (-31.65, 16.76) N/A
Hospital outpatient 3.97%* (0.44, 7.49) 2.66% (-10.17, 15.49) 10.61% (-4.05, 26.61)
Emergency room 1.71% (-0.93, 4.36) -3.25% (-12.25, 5.75) 3.20% (-9.10, 15.57)
Urgent care -1.52% (-3.29, 0.25) -4.73% (-12.85, 3.39) -23.42%* (-43.73, 1.34)
Office-based visits -0.02% (-0.09, 0.05) -1.17% (-7.17, 4.84) -0.91% (-7.72, 5.50)
Other -1.83% (-5.52, 1.86) 8.26% (-5.20, 21.72) 5.00% (-9.40, -2.65)
By type of service
Inpatient facility 0.92% (-0.34, 2.18) 8.59% (-5.18, 22.36) 41.12%* (1.61, 94.42)
Evaluation &
management
-0.12% (0.15, -0.40) 5.39%* (0.48, 10.30) 5.50%* (0.72, 10.88)
Radiology 1.59% (-1.25, 4.43) 7.16% (-4.50, 18.82) 12.84%* (-1.81, 27.57)
Laboratory 0.22% (-2.47, 2.92) 2.47% (-6.08, 11.02) 2.88% (-6.38, 12.29)
Surgery 1.40% (-0.71, 3.52) 7.37% (-5.97, 20.71) 19.27%* (-2.58, 43.32)
Other diagnostic &
treatments
-0.04% (-0.14, 0.06) -2.89% (-10.26, 4.48) -2.76% (-9.86, 4.59)
Other types 0.25% (-2.36, 2.85) 4.50% (-6.98, 15.99) 4.81% (-6.15, 17.14)
Abbreviation: CI, confidence interval; CSHCN, children with special health care needs; HCH, health care home; n,
number of person-year observation; N/A, not applicable/not estimated.
Note: n=22,053 person-year observations were used in sensitivity analysis. A standardized fee schedule was applied
to account for different reimbursement rates. Part 1 of overall medical services was not applicable because all
CSHCN incurred medical expenses. Inpatient expenditure did not achieve convergence in the estimation model.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001

94
Figure 1. Application of the children with special health care needs (CSHCN) screener to
insurance claims

• Had > 9 months of prescrip on
possession in a 3-year span.
Q1. Need or use medicine
prescribed by a doctor
• Had more office-based visits
than 95% of the same age-sex
children.
Q2. Need or use more medical
care, mental health, or
educa onal services than
average children
• Had been diagnosed with
selected disabled condi ons, such
as epilepsy and spina bifida.
Q3. Func onal limita on or
disability
• Had > 1 claim for physical,
occupa onal, speech, or
respiratory therapy, that were >
6 months apart in a 3-year span.
Q4. Need or receive special
therapy, such as physical,
occupa onal, or speech
therapy
• Had > 1 claim for mental health
services that were > 6 months
apart in a 3-year span.
Q5. Need or receive
treatment/counseling for
emo onal, developmental, or
behavioral problems

95
Chapter 5. Summary & Future Research Directions
The pattern of health care utilization was not optimal for CSHCN with managed Medicaid
in comparison to privately insured CSHCN in Minnesota, even though health care expenditures
were similar. Medicaid CSHCN were more likely to seek health care in ED, UC, and other
non-hospital settings compared to privately insured CSHCN, while a decreased use in
office-based settings was found for CSHCN with managed Medicaid coverage. Even though
these were not surprising findings, we urge policymakers and researchers not to overlook this
group of children, who are in need of better care as Medicaid policy has been changing under the
ACA and its appealing bills.
The PCMH has been gaining popularity in recent years as a possible model to improve
primary care delivery in the United States. The PCMH model aims to reduce health care costs
and improve quality of care, by providing medical team-based, coordinated, and whole-person
oriented primary care. Though the effects of the PCMH have been studied in adults in recent
years, since PCHM were first endorsed by clinical professional organizations CSHCN only have
been the focal point of a few research. In addition to updating the impacts of the PCMH by using
MEPS data, this dissertation has leveraged the newly available mapping algorithm for the
CSHCN screener to generate evidence for this particularly relevant population.
Although a large body of literature suggests that PCMHs improve some utilization and/or
quality outcomes in the adult population
21,26,85
, we did not find evidence to support that PCMHs
reduced health care expenditures for CSHCN in either MEPS or claims analysis (the first study

96
and the third study in this dissertation). There were some statistically significant increases in the
probability of accessing care, but no shift in overall expenditures when examined by place of
service, with the exception of decreased urgent care use and associated expenditures. However,
we were unable to find a conclusive PCMH effect on patterns of care using the population-based
MEPS findings and the evaluation of state-specific PCMH program results for CSHCN. When it
comes to quality of care, PCMHs were associated with improved parent-reported health care
quality, while claims-based quality of care measures did not reflect such improvements. In
addition, our findings did not support the conclusion that PCMHs benefit high-risk/resource use
patients the most, as previous studies have found in the adult population.
21,93-96
This discrepancy
suggests that PCMH effects for CSHCN likely differed from the adult population. More research
with a focus on CSHCN is warranted, to help improve primary care provision for this vulnerable
population.
We leveraged the consistency and recertification process in the Minnesota HCH initiative to
identify the PCMH with confidence in the level of “medical home-ness”. Examination of lagged
HCH effects resembles the stages of the HCH implementation, which was considered to be
related to the magnitude of the PCMH effects.
91
Our findings in the third study of the dissertation
suggest that a greater time spent being HCH certified was associated with a significant increase
in total medical expenditure and expenditures in other places of services settings, as well as
evaluation and management expenditures and other types of services, whereas a non-significant
shift was found in the first year of certification. More research is needed to examine whether the

97
HCH effects for CSHCN sustain or further translate into desired utilization patterns, such as
efficient inpatient and ED use, when long-term data becomes available.
For CSHCN who had emotional, developmental, and behavioral conditions, the
expenditures associated with inpatient facilities, evaluation and management, radiology, and
surgery were greater for those receiving care in a HCH compared to those who were not. These
shifts were not found in the overall sample. Moreover, having a PCMH was associated with less
use of mental health services in office-based visits in the MEPS study. Collectively, we urge
future research to focus on CSHCN with mental health conditions, since the PCMH affected
these children (who typically bared greater levels of unmet needs
51
) differently, and did not
improve their use of primary care.
When it comes to improvement in quality of care, we found that CSHCN with a PCMH
experienced better health care quality from parents’ perspective compared to those without a
PCMH. However, when examined the claims-based quality of care measures, no significant
differences were found for CSHCN with and without a PCMH. This discrepancy implied that
claims-based and survey-based quality measures differentially capture the PCMH effects on
quality of care for CSHCN. Future research should consider including claims-based quality of
care measures that are highly correlated with parent/patient satisfaction and survey-based quality
of measures that detailed type and place of health care services in order to comprehensively
investigate the effects of PCMH on quality of care.

98
Acknowledgements
I would like to convey my sincerest gratitude to my advisor and dissertation chair Dr. John A.
Romley, for the support and guidance he has provided me. It is through his dedication, and
generosity in his time and resources, that I have been able to complete this project and achieve
my academic and professional goals.
I am infinitely grateful to my committee member and collaborator from the Medica
Research Institute graduate fellowship, Dr. Caroline Carlin, for giving me valuable insights and
being a constant source of encouragement and guidance throughout the fellowship project. Her
passion and expertise in the field has, and forever will, embolden me to be better.
I would like to express my gratitude to my committee members, Dr. Geoffrey Joyce, Dr.
Jeffrey McCombs, and Dr. Neeraj Sood, for the knowledge, advice, and support they have
offered me.
I would also like to thank the Medica Research Institute, the USC Schaeffer Center, and the
USC Provost’s Office, for offering me fellowships to support my doctoral research works.
Without their generous financial supports, I would not have been able to complete this
dissertation project.
Finally, I would like to give a special thank you to my husband, Jung Yu, for always
believing in me and being the joy of my life. Also, I thank all my classmates, friends, and family
here in the US and in Taiwan. With your help and support, this journey couldn’t have been
better.

99
Appendix
Appendix A. Application and validation of the CSHCN screener to insurance
claims
We applied a valid algorithm that mapped the CSHCN screener, the gold standard for
CSHCN identification, to administrative claims to identify CSHCN. This algorithm was
developed and validated by Arim et al. using 6-10 years old children who were covered by
British Columbia’s universal health insurance program administrative data holdings.
31
We
slightly modified the criteria in the mapping algorithm to accommodate the differences between
different administrative claims systems, as described below. We constructed the CSHCN cohorts
for each calendar year in 2007-2013 by examining medical and/or pharmacy claims in a
three-year span that included the cohort calendar year and +/- 1 year for children under the age of
18. Due to data availability, only two-year claims were used for the 2014 cohort CSHCN
identification. Only the 2013 CSHCN cohort was used for the association between types of
insurance and health care expenditure, utilization, and quality of care. All CSHCN identified in
2007-2014 were included to impose other inclusion criteria to investigate the effects of
patient-centered medical home. Each screener item and its respective application are described as
follows:
Q1: Does your child currently need or use medicine prescribed by a doctor (other than
vitamins)?

100
We examined all pharmacy claims for each cohort calendar year. ‘Days of medicine use’
were approximated by total days of prescription possession. ‘Possessing prescription on a date’
was examined using the filled date and day supplies of pharmacy claims. We assumed that the
child started taking his or her medicine on the same day that the prescription was filled and
continued taking the medicine when possessing positive days supplied. As a child can possess
multiple prescriptions on a given date, we did not make adjustments for overlapped claims.
Following the criterion outlined in Arim et al.,
31
children with prescription possession >= 274
days (9 months) in any rolling 365-day period within the three-year span (two-year span for 2014
cohort) met the criteria for Q1 in the calendar year and were coded as Q1=1.

Q2: Does your child need or use more medical, mental health, or educational services than is
usual for most children of the same age?
Counts of physician office visits in the calendar year were used to determine whether the
child had more than usual care. Each unique service date and specialty was considered as a
separate visit. Following the definition in Arim et al., children who had physician visit counts
more than their age-sex specific 95-percentile cutoff numbers (within our claims) of visits were
considered as having more than usual services use.
31
To accelerate data preparation, we used
outpatient visit counts generated by the Johns Hopkins Adjusted Clinical Group (ACG) system
66

to identify the age-sex specific 95-pecentile cutoff visit count for each calendar year. The
outpatient visit count generated by the ACG system is an integer count of encounters with place

101
of service (POS) codes in physician offices (11), outpatient hospitals (22), and others (24, 25, 26,
50, 53, 60, 62, 65, 71, 72).
66
Table A1 shows each age-sex specific 95-percentile number of
outpatient visits, along with the number of children in each age-sex group. Children with more
than the 95-percentile number of physician visits in their age-sex group in the calendar year were
coded as 1 for this item.

Table A1. Age-sex specific 95-percentile number of outpatient visits
2007 2008 2009 2010 2011 2012 2013 2014
Age M F M F M F M F M F M F M F M F
0 16 17 16 16 17 17 17 17 18 17 19 19 19 18 18 19
1 17 17 16 16 17 17 19 18 17 19 17 18 18 17 19 19
2 17 18 16 16 17 18 18 17 19 18 17 19 19 17 19 19
3 17 17 15 17 17 16 17 18 16 19 17 17 20 19 19 20
4 17 16 17 16 17 17 18 17 17 17 17 19 17 17 19 19
5 16 17 16 16 18 16 17 17 19 17 17 19 18 19 18 20
6 17 17 16 16 18 17 17 17 18 17 17 18 17 18 19 19
7 17 17 16 16 16 16 18 17 18 18 18 17 19 20 18 21
8 17 17 15 16 17 16 18 18 17 18 19 17 18 18 18 19
9 17 15 16 16 18 18 18 16 18 17 18 18 18 18 21 18
10 17 17 17 16 17 17 18 17 18 16 18 18 17 18 18 17
11 17 16 15 16 16 16 18 17 18 17 19 18 18 18 18 18
12 16 17 16 16 16 16 16 17 18 17 16 18 22 20 18 19
13 16 17 16 16 16 18 16 17 17 17 17 18 17 18 21 19
14 17 16 16 17 17 18 19 17 17 18 19 17 19 17 17 18
15 16 16 16 16 17 16 16 17 18 18 18 18 17 19 20 19
16 16 18 17 15 16 17 17 17 17 18 19 19 19 19 18 20
17 16 16 16 16 17 17 17 16 18 16 17 17 17 18 19 19
Age, age in year; M, male; F, female.

Q3: Is your child limited or prevented in any way in his or her ability to do the things most
children of the same age can do?
The functional limitation variable was constructed by examining diagnosis codes listed in

102
Table A2 using medical claims for each calendar year. We first flagged patients with any listed
ICD-9 codes in the calendar year. For each date with the ICD-9 code, we examined whether the
patient had the same ICD-9 code in the following or previous six to 18 months (183-548 days). If
the same ICD-9 code was found within the interval, the patient was coded as Q3=1. That is, if
any pair of identical ICD-9 codes was observed with the dates of services six to 18 months apart,
the patient had function limitations that met the CSHCN criteria.

Table A2. Functional limitation diagnostic codes
ICD-9 codes Descriptions
277.x, 277.xx Metabolism disorders- unspecified
296.2x Major depressive disorder single episode
299.xx Pervasive developmental disorders
300.x, 300.xx Anxiety, dissociative and somatoform disorders
307.x, 307.xx Special symptoms or syndromes not elsewhere classified
311 Depressive disorder, not elsewhere classified
312.x, 312.xx Disturbance of conduct not elsewhere classified
313.x, 313.xx Disturbance of emotions specific to childhood and adolescence
314.x, 314.xx Hyperkinetic syndrome of childhood
315.x, 315.xx Specific delays in development
319 Unspecified intellectual disabilities
343.x Infantile cerebral palsy
345.xx Epilepsy and recurrent seizures
359.x, 359.xx Muscular dystrophies and other myopathies
369.x, 369.xx Blindness and low vision
389.x, 389.xx Hearing loss
741.xx Spina bifida
742.x, 742.xx Other congenital anomalies of nervous system
748.x, 748.xx Congenital anomalies of respiratory system
758.x, 758.xx Chromosomal anomalies
760.71 Alcohol affecting fetus or newborn via placenta or breast milk
781.x, 781.9x Symptoms involving nervous and musculoskeletal systems
784.3 Aphasia
784.5x Other speech disturbance
784.6x Other symbolic dysfunction
799.5x Signs and symptoms involving cognition
V40.x, V40.3x Mental and behavioral problems
V41.x Problems with special senses and other special functions

103
Q4: Does your child need or receive special therapy, such as physical, occupational, or speech
therapy?
The special therapy use indicator was constructed using medical claims for each calendar
year. Children who had any medical claims with provider specialty codes from physical medicine
and rehabilitation specialists, physical therapists, occupational therapists, or respiratory therapists
in the calendar year were selected first. For each date of a flagged specialty claim, we examined
whether the patient had services from the same type of providers in the following or previous six
to 18 months (183-548 days). If the same type of provider was found within the interval, the
patient was coded as Q4=1. That is, if any pair of claims from the same type of specialty
provider was six to 18 months apart, the patient experienced long-term use of specialty services
that met the CSHCN criteria.

104
Table A3. Codes to identify special therapies
Physical therapy: Claims with CPT codes or provided by specialty, group specialty, or provider
credential codes
CPT codes Specialty codes Group specialty codes Credential codes
97001, 97002, 97005, 97006,
97010, 97012, 97014, 97016,
97018, 97022, 97024, 97026,
97028, 97032-97036, 97039,
97100, 97110-97113, 97116,
97124, 97139, 97140, 97150,
97530, 97532, 97533, 97535,
97537, 97542, 97545, 97546,
97750, 97760-97762, 97799,
98925-98929, 98940-98943,
4018F, G0151, G0157, G0159,
G8978-G8992, S8990, S9131
0023, 0064 "PMR" "PT","REHAB”
Occupational therapy: Claims with CPT codes or provided by credential codes
CPT codes Specialty codes Group specialty codes Credential codes
97001, 97002, 97005, 97006,
97010, 97012, 97014, 97016,
97018, 97022, 97024, 97026,
97028, 97032-97036, 97039,
97110, 97112, 97113, 97116,
97124, 97139, 97140, 97150,
97530, 97532, 97533, 97535,
97537, 97542, 97545, 97546,
97750, 97760-97762, 97799,
98925-98929, 4018F,
G8978-G8992, 97003, 97004,
G0129, G0152, G0158, G0160,
S9129
"OTR"
Speech therapy: Claims with CPT codes or provided by specialty or provider credential codes
CPT codes Specialty codes Group specialty codes Credential codes
92506-92508, 92521-92524,
92597, 92626, 92627, 92630,
92633, S9152
0046 "ST"
Respiratory therapy: Claims with CPT codes
CPT codes Specialty codes Group specialty codes Credential codes
94640, 94667, 94668, 95783,
95811, 99503, 99504,
G0237-G0239, S5180, S5181

Q5: Does your child have any kind of emotional, developmental, or behavioral problem for
which he or she needs or receives treatment or counseling?
The counseling service use indicator was constructed using available medical claims for

105
each calendar year. Children who had any medical claims with provider specialty/degree codes
from a psychiatrist or counselor/psychologist, in the calendar year were first selected. For each
date of a flagged counseling claim, we examined whether the patient had services from the same
type of providers in the following or previous six to 18 months (183-548 days). If the same type
of provider is found within the interval, the patient was be coded as Q5=1. That is, if any pair of
counseling claims was six to 18 months apart, the patient experienced long-term use of
counseling services that met the CSHCN criteria.

Table A4. Codes to identify mental health and counseling services
Mental health/counseling services: Claims with CPT codes or provided by specialty, group
specialty, or provider credential codes
CPT codes Specialty codes Group specialty codes Credential codes
90785, 90792, 90832-90834,
90836-90840, 90846, 90847,
90849, 90853, 90875, 90876,
90882, 90887, 90899, 90791,
96101-96103, 96116,
96118-96120, 96150- 96154,
99499, H0031, H0032, H0034,
H0038, H2012, H2014, H2015,
H2017-H2019, T1017, T2023,
S9484
0015, 0061,
0062, 0084
"PSYC" "LPC", "MHSA",
"MSN", "PHD",
"PSYD"

CSHCN identification
For children under the age of 18 who had full-year medical coverage in the calendar year,
criteria for each screener items were applied to identify CSHCN. If a child had any of the five
screener items coded as 1, the child belonged to the CSHCN sample. Results are shown by
screener item and by calendar year in Table A5. Fewer CSHCN were identified in 2014 because

106
only two years of claims were used. Close to half (48%) of CSHCN met Q3: functional
limitations criteria followed by Q1: need/use of medicine criteria (41%).

Table A5. CSHCN identification results by the CSHCN screener items

CSHCN, children with special health care needs; n, number of person-year observations.
Note: The percentage represented the portion among all eligible children.

Validation of CSHCN identification
To assure validity of the mapping algorithm, we reviewed the CSHCN identification results
to the distribution of the Pediatric Medical Complexity Algorithm (PMCA)
65
complexity levels
and the distribution of the resource utilization bands (RUBs) from the ACG system.
66

PMCA is a published claims-based algorithm that is used to identify children with chronic
and/or complex conditions. The algorithm categorizes children into three complexity levels:
children with complex chronic disease (C-CD), children with non-complex chronic disease
(NC-CD), and children without chronic disease (without CD).
65
It has been tested and validated
n (%) Overall 2007 2008 2009 2010 2011 2012 2013 2014
CSHCN 179,047
(17.2%)
24,870
(17.0%)
23,149
(16.1%)
22,454
(17.2%)
23,213
(17.6%)
23,832
(17.9%)
22,304
(18.1%)
22,272
(18.0%)
16,953
(15.5%)
Q1 73,876
(7.1%)
10,844
(7.4%)
10,268
(7.2%)
9,302
(7.1%)
9,513
(7.2%)
9,617
(7.2%)
9,177
(7.5%)
8,984
(7.3%)
6,171
(5.6%)
Q2 62,455
(6.0%)
9,260
(6.3%)
8,730
(6.1%)
7,920
(6.1%)
8,021
(6.1%)
8,125
(6.1%)
7,139
(5.8%)
7,230
(5.8%)
6,030
(5.5%)
Q3 86,309
(8.3%)
10,951
(7.5%)
11,074
(7.7%)
10,323
(7.9%)
11,029
(8.4%)
11,621
(8.7%)
11,118
(9.0%)
11,400
(9.2%)
8,793
(8.0%)
Q4 48,714
(4.7%)
6,258
(4.3%)
6,279
(4.4%)
6,095
(4.7%)
6,428
(4.9%)
6,687
(5.0%)
6,540
(5.3%)
6,314
(5.1%)
4,113
(3.8%)
Q5 52,580
(5.1%)
6,652
(4.5%)
6,679
(4.7%)
6,352
(4.9%)
6,809
(5.2%)
7,064
(5.3%)
6,510
(5.3%)
6,891
(5.6%)
5,623
(5.1%)

107
using 0- to 18-year-old children who were insured by Washington State Medicaid and seen at
Seattle Children’s Hospital for more than one emergency department (ED) visit and/or inpatient
stay in 2010.
65
The PMCA has demonstrated good sensitivity and specificity for C-CD and
without CD children in WA-Medicaid claims (sensitivity: 89% and 80%; specificity: 85% and
91%, respectively), and is less valid in identifying NC-CD (sensitivity: 45%; specificity: 91%).
65

The majority of CSHCN (68%) was in the NC-CD and C-CD groups (22% of all eligible
children) under the conservative (i.e., less inclusive) PMCA specification. The overall agreement
rate between the mapping algorithm and the NC-CD and C-CD groups from the PMCA was 81%.
This pattern suggested that the mapping algorithm successfully captured children with chronic
conditions.
The RUBs in the ACG system represent different levels of resource usage, regardless of the
conditions a child may have had. The software assigns six RUB classes: 0= no or invalid
diagnosis, 1= healthy users, 2= low, 3= moderate, 4= high, and 5= very high resources users. As
expected, the distribution of the CSHCN was skewed toward the high resources end (Figure A1),
supporting the mapping algorithm’s accurate identification of CSHCN.

108

Figure A1. CSHCN distribution in each resource utilization bands

Note: CSHCN sample was identified by the CSHCN screener-mapping algorithm. The denominators of the reported
percentage were eligible child-year observations in the same RUBs.
0.9% 2.1% 15.6% 34.5% 65.5% 91.1%
99.1% 97.9% 84.4% 65.5% 34.5% 8.9%
0: No users 1: Healthy
users
2: Low 3: Moderate 4: High 5: Very high
Resource Utilization Bands (RUBs) CSHCN Non-CSHCN

109
Appendix B. Quality of Care Outcomes Construction
We reviewed the core set of children’s health care quality measures for Medicaid and
CHIP
67
and selected applicable claims-based measures in our study (Table B1). Quality measures
that were based on primary care visits, such as well-child visits and immunization status, were
excluded to avoid an endogeneity problem. To be clear, the number of primary care visits was
used as a criterion in the CSHCN screener (Q2) for CSHCN identification and in the health care
home (HCH) attribution assignment. Therefore, we only analyzed the quality measures that do
not solely rely on primary care visits as listed in Table B1. The National Committee for Quality
Assurance (NCQA) has endorsed these measures. We followed the 2016 Healthcare
Effectiveness Data and Information Set (HEDIS) specifications
98
with minor modifications to
construct a dichotomous indicator for each quality measure. Details of our modifications to
accommodate differences in data sources are described in Table B1.

110
Table B1. Medicaid recommended/endorsed claims-based child health quality measures
67,99

Measure Description Modification
Medication
management for
people with asthma
(MMA)
Percentage of children ages 5 to 20
who were identified as having
persistent asthma and were dispensed
appropriate medications that they
remained on during the treatment
period. Percentage of children who
remained on an asthma controller
medication for at least 75% of their
treatment period. Age groups: 5-11,
12-18, 19-20, years old and overall.
Only age groups 5-11 and 12-18 years
old were reported. Instead of a 45-day
allowable gap in pre-calendar year
enrollment, we only allowed a
one-month gap since our enrollment
data are monthly indicators.
Follow-up after
hospitalization for
mental illness (FUH)
Percentage of discharges for children
ages 6 to 20 who were hospitalized
for treatment of selected mental
health disorders and who had an
outpatient visit, and intensive
outpatient encounter, or partial
hospitalization with a mental health
practitioner within seven and 30 days
of discharge.
CSHCN had any mental health
hospitalizations were included as the
denominators. If any of these
discharges had a qualified follow-up
visit, we coded this child as 1. Only
CSHCN up to 18 years old were
included in our sample. HCPCS codes
were not applied when identifying
office visits and non-acute care. UB
types of bill were not applied in
identifying non-acute care.
Follow-up care for
children prescribed
ADHD medication
(ADD)
Percentage of children newly
prescribed ADHD medication who
had at least three follow-up visits
within a 10-month period, one of
which was within 30 days from the
first ADHD medication. Two rates
are reported: the initiation phase (1 in
30 days) and the continuation and
maintenance phase (3 in 10 months).
HCPCS codes were not applied when
identifying office visits. CSHCN who
met the age criteria within the calendar
year, instead of using the intake period
March 1
st
in prior year to February 28
th

of the measurement year, were
included in the analysis.
Appropriate testing
for children with
pharyngitis (CWP)
Percentage of children 2-18 years old
who were diagnosed with
pharyngitis, ordered an antibiotic,
and received a group A Streptococcus
(Strep) test for the episode.
CSHCN who met the age criteria
within the calendar year, instead of
using the intake period July 1
st
in prior
year to June 30
th
of the measurement
year, were included in the analysis.
LOINC codes were not applied to
identify strep tests.

111
Appropriate
treatment for
children with upper
respiratory track
infection (URI)
Percentage of children 3 months- 18
years of age who were diagnosed
with URI and were not dispensed an
antibiotic prescription on or three
days after the episode.
CSHCN who met the age criteria
within the calendar year, instead of
using the intake period July 1
st
in prior
year to June 30
th
of the measurement
year, were included in the analysis.
ADHD, attention deficit/hyperactivity disorder; CPT, Current Procedural Terminology; ED, emergency department;
HCPCS, Healthcare Common Procedure Coding System; LOINC, Logical Observations Identifiers, Names, and
Codes; NCQA, National Committee for Quality Assurance; POS, plaice of services codes; URI, upper respiratory
infection.
In addition to the previously mentioned quality measures, we also examined avoidable
emergency department (ED) utilization as a quality of care indicator. We classified the severity
of an ED visit by applying the ICD-9 code-based NYU algorithm and Ballard et al.
69,70
The NYU
algorithm assigns a set of probabilities that represent the chances of being a non-emergent,
primary care treatable, emergent and preventable, or emergent and non-preventable visit to each
discharge diagnosis.
69
Non-emergent visits are defined as the patient not requiring immediate
care within 12 hours (e.g., toothache). In primary care treatable visits, care was needed within 12
hours but could be taken care of in a primary care setting (e.g. a lumbar sprain). Emergent,
preventable visits involved patients with events that could have been avoided with proper
primary or ambulatory care (e.g. an asthma attack). Emergent, non-preventable visits involved
patients with events that could not have been avoided with primary care (e.g. a heart attack). The
probabilities were constructed by ED and primary care physicians reviewing 6,000 full ED visit
records in detail and mapping the assignment to discharge ICD-9.
69
Mental health, substance
abuse related diagnoses, and uncommon diagnoses are in separate groups and no probabilities are
assigned.
69
In our study, we used the primary ICD-9 code as the proxy of the discharge diagnosis.

112
Even though notable discrepancies between primary and discharge diagnoses for primary care
treatable groups were found
100
, this proxy is still the proxy we can use in our claims data.
We further followed rules established in Ballard et al. to determine whether an ED visit was
in emergent level or non-emergent level. For each ICD-9 code associated with a ED visit, we
combined the probabilities of being non-emergent and primary care treatable to yield the
probability of non-emergent level, and the probabilities of being emergent, both preventable and
non-preventable, to yield the probability of emergent level. The most emergent ICD-9 code was
used to represent the severity status of the ED visit. If the probability of being non-emergent
level was > 50%, the ED visit was considered avoidable.
70
If the child had any avoidable ED
visit in the calendar year, we coded him/her as 1. Having less or no avoidable ED visits is the
desired quality outcome.

113
Appendix C. CSHCN Attribution and Health Care Home Indicator Construction

CSHCN attribution to primary care clinics
The HCH indicator was constructed for each child by calendar year according to the clinic
attribution results. We performed the attribution for CSHCN who met the full-year medical
enrollment, had no missing demographic variables, and had at least six months of enrollment in
the prior calendar year. CSHCN were attributed to a clinic based on the plurality of primary care
visits in the year prior to the calendar year, with the most recent visit breaking ties. If a physician
visit was billed with CPT codes, POS codes and specialty codes listed in Table C1, we
considered the visit as a primary care visit. To ensure that primary care provided by the
designated clinic was meaningful, we only considered clinics that the child had utilized for more
than 35% of his or her primary care visits. The cutoff was selected to represent the idea of
continuity in care and to guarantee only two clinics would be compared, at the most, in the clinic
assignments.

114
Table C1. Criteria for identifying primary care visits
CPT codes POS codes Specialty codes
90000
-99999
3 School 0000
0001
Family practice
5 Indian health services free standing
facility
0004 Internist
6 Indian health services provider based
facility
0012 Pediatrician
11 Office 0057 Convenience care
12 Home 0059 Home health care
13 Assisted living facility 0066 After hours clinic/urgicenter
14 Group home 0073 Family practice specialist
15 Mobile unit 0075 Pediatric specialist
26 Military treatment facility 0076 Internal medicine specialist
49 Independent clinic
50 Federally qualified health center
71 State or local public health clinic
72 Rural health clinic
Among CSHCN identified by the CSHCN screener mapping algorithm, 78% of them were
successfully attributed to a clinic (Figure C1). We imputed the attribution results for CSHCN
who were attributed to a clinic in the prior year but did not have any primary care visits in the
current year (3.8%), using the last-observation-carried-forward method. This carry-forward was
based on the assumption of strong “stickiness” in patient-provider relationships. The year-to-year
patient retention rates for our sample were very high. Only 2.6% of CSHCN in our final sample
were attributed to different clinics in different years (Table C2). This pattern improved our
confidence in the imputation method. Moreover, this pattern implied that the individual fixed
effects were largely nested within the clinic fixed effects.

115
Figure C1. CSHCN sample attrition

CSHCN, children with special health care needs; HCH, health care home; n, number of person-year observations.
Note: Percentages in the graph were calculated using the number of CSHCN identified (the first box, n= 179,074) as
denominators. We applied last-observation-carried-forward for 6,791 (3.8%) of observations in the successful
attribution phase (the 5
th
box).

CSHCN iden fied
n= 179,047 (100%)
No missing in demographic &
neighborhood effects variables
n= 178,806 (99.9%)
Had any primary care visit in prior year
n=144,482 (80.7%)
[Con nuity in care]
The most visited clinic accounted for >=
35% of primary care visits
n= 142,328 (79.5%)
[Successful a ribu on & imputa on]
Had >= 6 months of medical coverage
in prior year
n= 146,382 (81.8%)
[Fixed effect model requirement]
CSHCN had > 1 year of observa on
n= 87,172 (48.7%)
HCH group
n= 17,143 (19.7%)
Non-HCH group
n= 70,029 (29.0%)

116
Table C2. CSHCN who switched clinics in the final sample

CSHCN, children with special health care needs; n, number of CSHCN.

Clinic HCH status determination
The HCH status of the clinic to which a child was attributed determined whether this child
was in the HCH group. Children who were attributed to an HCH clinic were coded as HCH=1,
while those who were attributed to a non-HCH clinic were coded as HCH=0. Year of the HCH
certification can be found on the MDH website.
87
We derived the clinic certified date from the
health care provider certified dates because only the provider certified dates are available in the
HCH search tool on the MDH website.
87
Each provider in a clinic could be certified on different
dates, depending on whether providers left or joined a clinic. As every provider in the HCH
clinic is required to be certified, it is reasonable to assume that the earliest certified date among
all providers within the clinic is the clinic first certified date. The year of certification was
rounded to the nearest January 1
st
of the actual certification date. Years after HCH certification
were coded as HCH=1 since almost all clinics recertified annually. To be specific, all certified
clinics were recertified in 2010-2013 while only four clinics (1% of all certified clinics) were not
recertified in 2014.
88
Information for clinics that were HCH but no longer certified is not
available for public use. The exact certification issued and terminated dates for these four clinics
Overall 2008 2009 2010 2011 2012 2013 2014
Switched clinics from
prior year, n (%)
2,274
(2.6%)
79
(0.7%)
299
(2.6%)
416
(3.4%)
519
(4.2%)
412
(3.5%)
441
(3.8%)
108
(1.5%)

117
were not publically available. Therefore, we selected July 1, 2011 as their certified date because
they were likely to be certified from any time between late 2010-2013.
The final analytical CSHCN sample and their HCH status were summarized in Table C3. A
total of 87,172 person-year observations were included in the sample and close to 20% of them
had HCH=1. We found a similar growing trend of HCH groups over years compared to the HCH
evaluation reports.
101

Table C3. CSHCN in the final sample by HCH status and years
n Overall 2007 2008 2009 2010 2011 2012 2013 2014
CSHCN
sample
87,172 8,403 11,826 11,703 12,281 12,497 11,786 11,462 7,214
HCH
group
17,143
(19.7%)
0 0 0 0 3,731
(29.9%)
4,046
(34.3%)
5,377
(46.9%)
3,989
(55.3%)
CSHCN, children with special health care needs; HCH, health care home; n, number of person-year observations.

118
Appendix D. Supplement tables for “The Effects of the Health Care Home on
Costs, Utilization, and Quality for Children with Special Health Care Needs in
Minnesota: A Longitudinal Analysis of Insurance Claims”

119
Table D1. Multivariate adjusted difference-in-difference estimators in a two-part model of
health care expenditures for CSHCN who identified by criteria other than prescription
coverage (Q1)
Abbreviation: CI, confidence interval; CSHCN, children with special health care needs; HCH, health care home; n,
number of person-year observation; N/A, not applicable/not estimated.
Note: n= 70,369 person-year observations were used in sensitivity analysis. A standardized fee schedule was applied
to account for different reimbursement rates. Part 1 of overall medical services was not applicable because all
CSHCN incurred medical expenses. Inpatient expenditure did not achieve convergence in the estimation model.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001
Types of Expenditure
Part 1:
% Change in Accessing
Care with HCH (95%
CI)
Part 2:
% Change in Conditional
Expenditures with HCH
(95% CI)
% Changes in
Expenditure from
Combined Model with
HCH (95%CI)
Medical services N/A 1.39% (-1.89, 4.66) N/A
By place of services
Inpatient N/A

N/A N/A
Hospital outpatient 2.88%** (0.91, 4.85) 0.57% (-7.07, 8.21) 5.30% (-2.82, 13.84)
Emergency room 0.16% (-1.28, 1.59) 0.66% (-4.77, 6.09) 1.57% (-7.53, 9.97)
Urgent care -0.90%* (-1.75, -0.06) -5.05%* (-9.85, -0.24) -18.36%** (-32.36, -2.29)
Office-based visits -0.01% (-0.04, 0.02) -0.29% (-3.67, 3.10) -0.02% (-4.09, 3.47)
Other 2.12%* (0.22, 4.03) -0.81% (-9.66, 8.03) 8.03% (-2.79, 19.41)
By type of services
Inpatient facility 0.27% (-0.16, 0.70) -2.43% (-12.30, 7.44) 15.21% (-10.07, 45.55)
Evaluation &
management
-0.12% (-0.41, 0.16) 3.64%** (0.95, 6.32) 3.67%** (1.07, 6.61)
Radiology -0.19% (-1.93, 1.55) -1.91% (-8.48, 4.67) -2.31% (-9.65, 4.85)
Laboratory 0.28% (-1.31, 1.87) 3.98% (-1.07, 9.03) 4.37% (-1.19, 9.90)
Surgery 0.17% (-1.12, 1.47) 2.41% (-5.35, 10.16) 3.44% (-6.01, 13.05)
Other diagnostic &
treatments
-0.12% (-0.44, 0.19) -0.84% (-5.63, 3.95) -0.78% (-5.34, 3.92)
Other types 0.80% (-0.78, 2.38) 2.80% (-3.71, 9.30) 3.43% (-3.01, 10.28)

120
Table D2. Multivariate adjusted difference-in-difference estimators in a two-part model of
health care expenditures for CSHCN who were attributed to a clinic across years

Abbreviation: CI, confidence interval; CSHCN, children with special health care needs; HCH, health care home; n,
number of person-year observation; N/A, not applicable/not estimated.
Note: n= 79,113 person-year observations were used in sensitivity analysis. A standardized fee schedule was applied
to account for different reimbursement rates. Inpatient expenditure did not achieve convergence in the estimation
model.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001
Types of Expenditure
Part 1:
% Change in Accessing
Care with HCH (95%
CI)
Part 2:
% Change in Conditional
Expenditures with HCH
(95% CI)
% Changes in
Expenditure from
Combined Model with
HCH (95%CI)
Medical services 0.03% (-0.01, 0.07) 1.34% (-2.56, 5.24) 1.40% (-2.56, 5.26)
By place of services
Inpatient N/A

N/A N/A
Hospital outpatient 2.34%* (0.51, 4.16) -1.20% (-8.56, 7.51) 3.22% (-5.05, 10.76)
Emergency room 0.5% (-0.80, 1.80) -0.27% (-5.49, 4.95) 2.97% (-5.71, 12.75)
Urgent care -0.90%* (-1.75, -0.06) -4.39% (-8.60, 0.17) -19.22%** (-32.38, -5.44)
Office-based visits 0.01% (-0.03, 0.05) 0.15% (-2.93, 3.24) 0.19% (-2.93, 3.07)
Other 2.18%* (0.42, 3.93) -2.84% (-11.21, 5.52) 7.36% (-3.14, 17.27)
By type of services
Inpatient facility 0.27% (-0.16, 0.70) -2.43% (-12.30, 7.44) 15.21% (-10.07, 45.55)
Evaluation &
management
-0.09% (-0.37, 0.20) 2.06% (-0.40, 4.51) 2.17% (-0.50, 4.87)
Radiology -0.06% (-1.68, 1.55) -1.89% (-8.11, 4.33) -1.91% (-8.59, 5.24)
Laboratory 0.41% (-1.09, 1.91) 0.32% (-4.13, 4.95) 0.63% (-3.82, 5.09)
Surgery -0.01% (-1.19, 1.17) 2.02% (-5.37, 9.40) 2.27% (-6.76, 12.35)
Other diagnostic &
treatments
0.16% (-0.31, 0.63) -1.14% (-5.74, 3.46) -0.94% (-5.63, 4.03)
Other types 1.26% (-0.23, 2.76) 0.36% (-5.67, 6.38) 1.33% (-5.16, 7.94)

121
Table D3. Multivariate adjusted difference-in-difference estimators of pre-HCH trends in a two-part model of health care
expenditures for CSHCN

Abbreviation: CI, confidence interval; CSHCN, children with special health care needs; HCH, health care home certified; N/A, not applicable/not estimated.
Note: CSHCN not receiving care from HCH was the reference group. The percent changes reported are compared to the year of HCH certification. A
standardized fee schedule was applied to account for different reimbursement rates. Inpatient expenditure did not achieve convergence in the estimation model.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001

Types of Expenditure
Part 1: % Change in Accessing Care
with HCH (95% CI)
Part 2: % Change in Conditional
Expenditures with HCH (95% CI)
% Changes in Expenditure from
Combined Model with HCH (95%CI)
1 year before HCH 2+ year before HCH 1 year before HCH 2+ year before HCH 1 year before HCH 2+ year before HCH
Medical services -0.08% (-0.18, 0.01) -0.06% (-0.14, 0.03) 0.69% (-5.16, 9.63) 0.35% (-4.77, 5.46) 0.53% (-5.49, 6.86) 0.36% (-4.29, 5.82)
By place of service
Inpatient N/A N/A N/A N/A

N/A N/A
Hospital outpatient -0.40% (-3.09, 2.28) -1.46% (-3.80, 0.89) -3.93% (-14.82, 6.97) 4.24% (-5.29, 13.76) -4.53% (-14.91, 7.31) 1.52% (-8.12, 11.94)
Emergency room 0.41% (-1.49, 2.30) -0.37% (-2.03, 1.28) 2.55% (-5.14, 10.24) 1.48% (-5.28, 8.24) 5.69% (-6.22, 21.61) 0.09% (-10.64, 12.53)
Urgent care 1.32%* (0.03, 2.61) 0.73% (-0.37, 1.84) 2.19% (-4.76, 9.13) 6.75%* (0.89, 12.72) 32.56%* (1.52, 67.30) 24.54% (-2.43, -55.84)
Office-based visits -0.06%* (-0.11, -0.01) -0.04%* (-0.09, -0.0001) 1.06% (-3.55, 5.66) 0.56% (-3.47, 4.58) 1.04% (-0.32, 5.28) 0.47% (-3.61, 4.43)
Other -0.57% (-3.24, 19.97) -2.26%* (-4.5, -0.03) 5.11% (-7.33, 17.55) 4.96% (-5.87, 15.80) 3.22% (-10.29, 20.27) -4.34% (-16.09, 9.86)
By type of service
Inpatient facility -0.15% (-0.77, 0.47) -0.02% (-0.57, 0.52) -3.53% (-18.19, 11.13) 1.53% (-11.15,14.21) -10.27% (-39.53, 23.05) 1.80% (-28.17, 36.68)
Evaluation &
management
0.03% (-3.80, 0.43) -0.03% (-0.39, 0.33) 1.41% (-2.23, 5.06) -2.05% (-5.23, 1.24) 1.42% (-1.92, 4.78) -2.10% (-5.21, 0.93)
Radiology 1.31% (-1.06, 3.68) -0.32% (-2.39, 1.76) -2.39% (-11.51, 0.67) 4.91% (-3.10, 12.92) 1.83% (-7.95, 14.14) 4.53% (-4.77, 15.87)
Laboratory 0.003% (-2.21, 2.22) -0.80% (-3.75, 1.15) 2.38% (-4.51, 9.27) -1.39% (-7.37, 4.60) 2.27% (-4.63, 9.56) -1.97% (-7.25, 4.00)
Surgery -0.49% (-2.24, 1.26) -0.16% (-1.69, 1.38) -1.29% (-12.09, 9.50) -0.50% (-9.92, 8.92) -3.98% (-16.39, 9.70) -1.34% (-12.33, 10.55)
Other diagnostic &
treatments
-0.07% (-0.76, 0.63) -0.04% (-0.04, 0.58) 2.29% (-5.59, 7.38) 1.69% (-4.29, 7.68) 2.29% (-3.98, 9.46) 1.69% (-4.42, 7.97)
Other types -0.41% (-2.65, 1.83) -1.47% (-3.43, 0.49) 2.23% (-6.63, 11.16) 4.60% (-3.22, 12.42) 1.68% (-7.02, 10.71) 3.28% (-4.45, 10.61)

122
Table D4. CSHCN characteristics in the balanced cohort, by health care home status
CSHCN characteristics
HCH
n= 3,020
(46.0%)
No HCH
n= 6,232
(54.0%)
P-value
Male, % 56.9% 56.7% 0.928
Resided in urban area, % 15.1% 16.6% 0.141
Public insurance, %
88.4% 81.6% <0.001
Had full-year pharmacy coverage, %
91.6% 86.3% <0.001
Comorbidity burden, %

<0.001
Healthy/ no user
1.5% 2.7%

Low
24.2% 28.9%

Moderate
51.1% 50.4%

High
17.7% 14.3%

Very high
5.5% 3.8%

Age groups, %

0.313
0-5 years old
11.0% 12.1%

6-11 years old
38.2% 38.8%

12-17 years old
50.9% 49.9%

Age in year, mean (SD) 11.0 (4.7) 10.8 (4.0) 0.045
Neighborhood effects, mean (SD)
% of race/ethnicity
Hispanic
3.4 (5.9) 2.6 (5.8)
0.209
White, non-Hispanic
87.1 (15.2) 85.8 (14.4)
0.001
Black, non-Hispanic
3.5 (7.8) 3.6 (7.1)
0.629
Others, non-Hispanic
6.0 (7.0) 7.1 (7.5)
<0.001
% of education level
Less than high school
6.9 (6.4) 5.5 (5.6)
<0.001
High school/GED
26.8 (10.6) 22.4 (12.0)
<0.001
Some colleges
34.7 (8.6) 30.9 (8.9)
<0.001
4-year/ graduate degree
31.7 (17.0) 41.1 (20.1)
<0.001
% of speaking English only
91.2 (9.3) 90.3 (9.1)
<0.001
% of < FPL 8.6 (8.5) 8.3 (8.1) 0.285
Selected conditions, %
Anxiety
20.0% 16.6% 0.002
Asthma
13.1% 10.8% 0.010
ADHD 34.5% 33.4% 0.412
Behavioral/conduct problems
5.1% 4.6% 0.362
Autism
5.0% 4.3% 0.231
Depression
11.8% 10.7% 0.216
Developmental delay
9.7% 9.7% 0.988
Intellectual disability
1.1% 0.7% 0.206

123
Migraine/headache 1.9% 1.6% 0.384
Congenital disorders 9.6% 7.6% 0.011
Abbreviation: ADHD, attention deficit hyperactivity disorder; CSHCN, children with special health care
needs; FPL, federal poverty level; GED, general educational diploma; HCH, health care home; n, number
of person-year observation; SD, standard deviation.

124
Table D5. Multivariate adjusted difference-in-difference estimators in a two-part model of
health care expenditures for CSHCN, balanced cohort analysis

Abbreviation: CI, confidence interval; CSHCN, children with special health care needs; HCH, health care home; n,
number of person-year observation; N/A, not applicable/not estimated.
Note: n= 9,252 person-year observations (n= 3,020 with HCH) were used in the balanced cohort analysis. A
standardized fee schedule was applied to account for different reimbursement rates. Inpatient expenditure did not
achieve convergence in the estimation model. Part 1 was not estimated for medical and office-based services since
almost all CSHCN had expenditure.
* p-value < 0.05; ** p-value <0.01; *** p-value <0.001
Types of Expenditure
Part 1:
% Change in Accessing
Care with HCH (95%
CI)
Part 2:
% Change in Conditional
Expenditures with HCH
(95% CI)
% Changes in
Expenditure from
Combined Model with
HCH (95%CI)
Medical services N/A

2.02% (-9.03, 13.08) N/A
By place of services
Inpatient 4.44%*** (2.91, 5.98) N/A N/A
Hospital outpatient 4.74% (-4.44, 13.93) -8.24% (-42.93, 26.45) 4.69% (-27.46, 48.41)
Emergency room -0.34% (-6.07, 5.40) 7.79% (-20.39, 35.97) 9.77% (-38.47, 76.74)
Urgent care -1.4% (-4.96, 2.15) 4.50% (-16.48, 25.47) -14.82% (-65.76, 69.60)
Office-based visits N/A -3.50% (-12.75, 5.76) N/A
Other 2.05% (-7.11, 11.21) 10.61% (-22.66, 48.90) 23.94% (-24.71, 79.08)
By type of services
Inpatient facility 0.47% (-1.40, 2.33) 7.72% (-41.70, 57.15) 122.24% (-65.15, 624.58)
Evaluation &
management
0.37% (-0.53, 1.27) -0.64% (-13.15, 11.85) -0.19% (-12.00, 13.49)
Radiology -3.97% (-11.55, 3.60) 9.14% (-21.31, 11.81) -1.68% (-34.65, 40.17)
Laboratory 1.74% (-5.66, 9.14) 2.39% (-20.72, 25.50) 4.18% (-17.39, 31.79)
Surgery 3.26% (-2.64, 9.15) -4.74% (-39.70, 30.20) 13.48% (-27.66, 77.23)
Other diagnostic &
treatments
0.43% (-1.49, 2.35) -1.50% (-5.96, 2.95) -7.98% (-27.72, 16.35)
Other types 1.08% (-6.91, 9.07) 12.95% (-18.98, 44.88) 15.32% (-17.45, 59.13)

125
References
1. McPherson M, Arango P, Fox H, et al. A new definition of children with special health
care needs. Pediatrics. 1998;102(1 Pt 1):137-140.
2. Minnesota Report from the 2009/10 National Survey of Children with Special Health
Care Needs.: Child and Adolescent Health Measurement Initiative, Data Resource Center
for Child and Adolescent Health website.
3. Lindley LC, Lyon ME. A profile of children with complex chronic conditions at end of
life among Medicaid beneficiaries: implications for health care reform. Journal of
palliative medicine. 2013;16(11):1388-1393.
4. THE CHILDREN WITH SPECIAL HEALTH CARE NEEDS (CSHCN) SCREENER.
1998;
http://www.cahmi.org/wp-content/uploads/2014/06/CSHCNS-Survey-and-scoring
.pdf. Accessed April 10, 2015.
5. Berry JG, Toomey SL, Zaslavsky AM, et al. Pediatric readmission prevalence and
variability across hospitals. JAMA. 2013;309(4):372-380.
6. Romaire MA, Bell JF, Grossman DC. Medical home access and health care use and
expenditures among children with special health care needs. Archives of pediatrics &
adolescent medicine. 2012;166(4):323-330.
7. Newacheck PW, Kim SE. A national profile of health care utilization and expenditures for
children with special health care needs. Archives of pediatrics & adolescent medicine.
2005;159(1):10-17.
8. Machado JC, Raszynski A, Totapally BR. Children with Special Health Care Needs:
Resource Utilization in a Tertiary-Care Pediatric Intensive Care Unit. 2017.
9. Ralston SL, Harrison W, Wasserman J, Goodman DC. Hospital Variation in Health Care
Utilization by Children With Medical Complexity. Pediatrics. 2015;136(5):860-867.
10. Lindley LC, Mark BA. Children with special health care needs: Impact of health care
expenditures on family financial burden. Journal of child and family studies.
2010;19(1):79-89.
11. Ghandour RM, Hirai AH, Blumberg SJ, Strickland BB, Kogan MD. Financial and

126
nonfinancial burden among families of CSHCN: changes between 2001 and 2009–2010.
Academic pediatrics. 2014;14(1):92-100.
12. Yotani N, Ishiguro A, Sakai H, Ohfuji S, Fukushima W, Hirota Y . Factor-associated
caregiver burden in medically complex patients with special health-care needs. Pediatrics
International. 2014;56(5):742-747.
13. Romley JA, Shah AK, Chung PJ, Elliott MN, Vestal KD, Schuster MA. Family-Provided
Health Care for Children With Special Health Care Needs. Pediatrics.
2017;139(1):e20161287.
14. Miller JE, Nugent CN, Gaboda D, Russell LB. Reasons for unmet need for child and
family health services among children with special health care needs with and without
medical homes. PloS one. 2013;8(12):e82570.
15. An R. Unmet mental health care needs in U.S. children with medical complexity,
2005-2010. J Psychosom Res. 2016;82:1-3.
16. Hoff T, Weller W, DePuccio M. The Patient-Centered Medical Home A Review of Recent
Research. Medical Care Research and Review. 2012;69(6):619-644.
17. Nielson M OJ, Grundy P, Grumbach K. The Patient-Centered Medical Home's Impact on
Cost and Quality. Patient-Centered Primary Care Collaborative;2015.
18. Physicians AAoF. Joint principles of the Patient-Centered Medical Home. Delaware
Medical Journal. 2008;80(1):21.
19. The medical home. Pediatrics. 2002;110(1 Pt 1):184-186.
20. Edwards ST, Bitton A, Hong J, Landon BE. Patient-centered medical home initiatives
expanded in 2009–13: providers, patients, and payment incentives increased. Health
Affairs. 2014;33(10):1823-1831.
21. Jabbarpour Y DE, Bazemore A, Grundy P. The Impact of Primary Care Practice
Transformation on Cost, Quality, and Utilization: A Systematic Review of Research
Published in 2016. Washington, DC: Patient-Centerd Primary Care Collaborative; July
2017.
22. Rosenthal MB, Alidina S, Friedberg MW, et al. A difference-in-difference analysis of
changes in quality, utilization and cost following the Colorado multi-payer
patient-centered medical home pilot. Journal of general internal medicine.

127
2016;31(3):289-296.
23. Rosenthal MB, Alidina S, Friedberg MW, et al. Impact of the Cincinnati Aligning Forces
for Quality Multi-Payer Patient Centered Medical Home pilot on health care quality,
utilization, and costs. Medical Care Research and Review. 2016;73(5):532-545.
24. Friedberg MW, Rosenthal MB, Werner RM, V olpp KG, Schneider EC. Effects of a
medical home and shared savings intervention on quality and utilization of care. JAMA
internal medicine. 2015.
25. Shane DM, Nguyen-Hoang P, Bentler SE, Damiano PC, Momany ET. Medicaid Health
Home Reducing Costs and Reliance on Emergency Department Evidence From Iowa.
Medical Care. 2016;54(8):752-757.
26. Sinaiko AD, Landrum MB, Meyers DJ, et al. Synthesis Of Research On Patient-Centered
Medical Homes Brings Systematic Differences Into Relief. Health affairs (Project Hope).
2017;36(3):500-508.
27. Homer CJ, Klatka K, Romm D, et al. A review of the evidence for the medical home for
children with special health care needs. Pediatrics. 2008;122(4):e922-e937.
28. Raphael JL, Cooley WC, Vega A, et al. Outcomes for Children with Chronic Conditions
Associated with Parent-and Provider-reported Measures of the Medical Home. Journal of
health care for the poor and underserved. 2015;26(2):358-376.
29. Kuhlthau KA, Bloom S, Van Cleave J, et al. Evidence for family-centered care for
children with special health care needs: a systematic review. Academic pediatrics.
2011;11(2):136-143. e138.
30. Witt WP, Fullerton CA, Chow C, et al. Effect of Having a Usual Source of Care on
Health Care Outcomes Among Children With Serious Emotional Disturbance. Academic
pediatrics. 2017;17(1):45-52.
31. Arim RG, Kohen DE, Brehaut JC, et al. Developing a non-categorical measure of child
health using administrative data. Health reports. 2015;26(2):9-16.
32. Nationwide Report from 2009/10 National Survey of Children with Special Health Care
Needs. Child and Adolescent Health Measurement Initiative, Data Resource Center for
Child and Adolescent Health website.
33. Strickland BB, Jones JR, Ghandour RM, Kogan MD, Newacheck PW. The medical home:

128
health care access and impact for children and youth in the United States. Pediatrics.
2011;127(4):604-611.
34. Garcia-Huidobro D, Shippee N, Joseph-DiCaprio J, O’Brien JM, Svetaz MV . Effect of
Patient-Centered Medical Home on Preventive Services for Adolescents and Young
Adults. Pediatrics. 2016:e20153813.
35. Cooley WC, McAllister JW, Sherrieb K, Kuhlthau K. Improved Outcomes Associated
With Medical Home Implementation in Pediatric Primary Care. Pediatrics.
2009;124(1):358-364.
36. Han B, Yu H, Friedberg MW. Evaluating the Impact of Parent‐ Reported Medical Home
Status on Children's Health Care Utilization, Expenditures, and Quality: A Difference‐
in‐ Differences Analysis with Causal Inference Methods. Health services research.
2017;52(2):786-806.
37. MEPS HC-155: 2012 Full Year Consolidated Data File. Rockville, MD: Agency for
Healthcare Research and Quality;2014.
38. Agency for Healthcare Research and Quality R, MD. MEPS-HC Sample Design and
Collection Process. http://www.meps.ahrq.gov/survey_comp_data_collection.jsp
Accessed 04/04, 2016.
39. Romaire MA, Bell JF. The medical home, preventive care screenings, and counseling for
children: evidence from the Medical Expenditure Panel Survey. Academic pediatrics.
2010;10(5):338-345.
40. MEPS HC-152G: 2012 Office-Based Medical Provider Visits. Rockville, MD: Agency for
Healthcare Research and Quality;2014.
41. Abdus S, Selden TM. Adherence with recommended well-child visits has grown, but
large gaps persist among various socioeconomic groups. Health Affairs.
2013;32(3):508-515.
42. Using Appropriate Price Indices for Analyses of Health Care Expenditures or Income
across Multiple Years. https://meps.ahrq.gov/about_meps/Price_Index.shtml - a3.
Accessed August 10, 2016.
43. Olson LM, Tanner JL, Stein MT, Radecki L. Well-child care: looking back, looking
forward. Pediatric annals. 2008;37(3).

129
44. Caldwell JB, T. Trends in Well-Child Visits: United States, 2002–2009. Rockville, MD.:
Agency for Healthcare Research and Quality.; August 2013.
45. Andersen R. Revisiting the behavioral model and access to medical care: does it matter?
Journal of health and social behavior. 1995;36(1):1-10.
46. Lunceford JK, Davidian M. Stratification and weighting via the propensity score in
estimation of causal treatment effects: a comparative study. Statistics in medicine.
2004;23(19):2937-2960.
47. DuGoff EH, Schuler M, Stuart EA. Generalizing observational study results: applying
propensity score methods to complex surveys. Health services research.
2014;49(1):284-303.
48. Deb P, Manning W, Norton E. Modeling health care costs and counts. Paper presented at:
8th World Congress on Health Economics2006.
49. Belotti F, Deb P, Manning WG, Norton EC. twopm: Two-part models. Stata J.
2015;15(1):3-20.
50. Sannicandro T, Parish SL, Son E, Powell RM. Health Care Changes for Children with
Special Health Care Needs, 2005–2011. Maternal and child health journal. 2016:1-7.
51. An R. Unmet Mental Health Care Needs in US Children with Medical Complexity,
2005–2010. Journal of Psychosomatic Research. 2016;82:1-3.
52. Zuvekas SH, Olin GL. Validating Household Reports of Health Care Use in the Medical
Expenditure Panel Survey. Health services research. 2009;44(5 Pt 1):1679-1700.
53. Chen AY , Newacheck PW. Insurance coverage and financial burden for families of
children with special health care needs. Ambulatory Pediatrics. 2006;6(4):204-209.
54. Busch SH, Barry CL. Does private insurance adequately protect families of children with
mental health disorders? Pediatrics. 2009;124 Suppl 4:S399-406.
55. Jeffrey AE, Newacheck PW. Role of insurance for children with special health care needs:
a synthesis of the evidence. Pediatrics. 2006;118(4):e1027-e1038.
56. Newacheck PW, Houtrow AJ, Romm DL, et al. The future of health insurance for
children with special health care needs. Pediatrics. 2009;123(5):e940-e947.
57. Ghandour RM, Comeau M, Tobias C, et al. Assuring adequate health insurance for
children with special health care needs: progress from 2001 to 2009–2010. Academic

130
pediatrics. 2015;15(4):451-460.
58. Foundation KF. Medicaid Pocket Primer. January 2017.
59. Sarkar M, Earley ER, Asti L, Chisolm DJ. Differences in Health Care Needs, Health Care
Utilization, and Health Care Outcomes Among Children With Special Health Care Needs
in Ohio: A Comparative Analysis Between Medicaid and Private Insurance. Journal of
Public Health Management and Practice. 2017;23(1):e1-e9.
60. Ghandour RM, Hirai AH, Blumberg SJ, Strickland BB, Kogan MD. Financial and
Nonfinancial burden among families of CSHCN: changes Between 2001 and 2009-2010.
Academic pediatrics. 2014;14(1):92-100.
61. Medicaid in Minnesota. Kaiser Family Foundation; June 2017.
62. Bethell CD, Kogan MD, Strickland BB, Schor EL, Robertson J, Newacheck PW. A
national and state profile of leading health problems and health care quality for US
children: key insurance disparities and across-state variations. Academic pediatrics.
2011;11(3 Suppl):S22-33.
63. Davidoff AJ. Insurance for children with special health care needs: patterns of coverage
and burden on families to provide adequate insurance. Pediatrics. 2004;114(2):394-403.
64. Aizcorbe A, Liebman E, Pack S, Cutler DM, Chernew ME, Rosen AB. Measuring health
care costs of individuals with employer-sponsored health insurance in the U.S.: A
comparison of survey and claims data. Statistical journal of the IAOS.
2012;28(1-2):43-51.
65. Simon TD, Cawthon ML, Stanford S, et al. Pediatric Medical Complexity Algorithm: A
New Method to Stratify Children by Medical Complexity. Pediatrics.
2014;133(6):e1647-e1654.
66. The Johns Hopkins ACG System, Technical Reference Guide. Version 10.0.: Johns
Hopkins Bloomberg School of Public Health; December 2011.
67. 2014 Clinical Quality Measures (CQMs) Pediatric Recommendation Core Measures.
Centers of Medicare & Medicaid Services;2014.
68. NCQA. The State of Health Care Quality 2014. Washington, DC: National Committee
for Quality Assurance. ; 2014 2014.
69. Billings J, Parikh N, Mijanovich T. Emergency department use in New York City: a

131
substitute for primary care? Issue brief (Commonwealth Fund). 2000(433):1-5.
70. Ballard DW, Price M, Fung V , et al. Validation of an algorithm for categorizing the
severity of hospital emergency department visits. Medical care. 2010;48(1).
71. Niska R, Bhuiya F, Xu J. National hospital ambulatory medical care survey: 2007
emergency department summary. Natl Health Stat Report. 2010;26(26):1-31.
72. Taubman SL, Allen HL, Wright BJ, Baicker K, Finkelstein AN. Medicaid increases
emergency-department use: evidence from Oregon's Health Insurance Experiment.
Science. 2014;343(6168):263-268.
73. McCormack LA, Jones SG, Coulter SL. Demographic factors influencing nonurgent
emergency department utilization among a Medicaid population. Health care
management science. 2016:1-8.
74. Ben-Isaac E, Schrager SM, Keefer M, Chen AY . National profile of nonemergent
pediatric emergency department visits. Pediatrics. 2010;125(3):454-459.
75. Sieck CJ, Hefner JL, Wexler R, Taylor CA, McAlearney AS. Why do they do that?:
Looking beyond typical reasons for non-urgent ED use among Medicaid patients. Patient
Experience Journal. 2016;3(2):22-30.
76. Decker SL. Acceptance of New Medicaid Patients by Primary Care Physicians and
Experiences with Physician Availability among Children on Medicaid or the Children's
Health Insurance Program. Health services research. 2015;50(5):1508-1527.
77. Kenney GM, Haley JM, Pan C, Lynch V , Buettgens M. Children’s coverage climb
continues: uninsurance and Medicaid/CHIP eligibility and participation under the ACA.
Robert Wood Johnson Foundation and the Urban Institute. 2016.
78. Ku L, Jones K, Shin P, Bruen B, Hayes K. The states' next challenge—securing primary
care for expanded Medicaid populations. New England Journal of Medicine.
2011;364(6):493-495.
79. Garthwaite CL. The doctor might see you now: The supply side effects of public health
insurance expansions. American Economic Journal: Economic Policy.
2012;4(3):190-215.
80. Wang L, Mandell DS, Lawer L, Cidav Z, Leslie DL. Healthcare Service Use and Costs
for Autism Spectrum Disorder: A Comparison Between Medicaid and Private Insurance.

132
Journal of Autism and Developmental Disorders. 2013;43(5):1057-1064.
81. Chang J, Freed GL, Prosser LA, et al. Comparisons of health care utilization outcomes in
children with asthma enrolled in private insurance plans versus medicaid. Journal of
Pediatric Health Care. 2014;28(1):71-79.
82. Health Care Homes: certification & recertification.
http://www.health.state.mn.us/healthreform/homes/certification/certification/cer
tassess.html. Accessed April 12, 2016.
83. Burton RA DK, Berenson RA. . Patient-Centered Medical Home Recognition Tools: A
Comparison of Ten Surveys’ Content and Operational Details. Washington, DC: The
Urban Institute, Health Policy Center; May 2011.
84. Health Care Homes Performance Measurement and Evaluation
http://www.health.state.mn.us/healthreform/homes/outcomes/ccmeasureimplem
entation.html. Accessed May 13, 2015.
85. Wholey D, Finch M, Shippee N, et al. Evaluation of the State of Minnesota's Health Care
Homes Initiative: Evaluation Report for Years 2010-2014. Minnesota Department of
Health; 2015.
86. Lin C, Romley J, Carlin C. Patient-Centered Medical Home And Annual Health Care
Expenditures Among Children With Special Health Care Needs: New Evidence From
Medical Expenditure Panel Survey. Value in Health. 2016;19(3):A7.
87. Health Care Homes Certified Clinicians Report. 2015;
https://apps.health.state.mn.us/hchcertification/publicReports/publicCertifyClinici
anList.seam. Accessed October 27, 2015.
88. Health Care Homes: Annual Report on Implementation. St. Paul, MN: Minnesota
Department of Health; January 2015.
89. Gunasekara FI, Richardson K, Carter K, Blakely T. Fixed effects analysis of repeated
measures data. International journal of epidemiology. 2013;43(1):264-269.
90. Bethell CD, Blumberg SJ, Stein RE, Strickland B, Robertson J, Newacheck PW. Taking
stock of the CSHCN Screener: a review of common questions and current reflections.
Academic pediatrics. 2015;15(2):165-176.
91. Paustian ML, Alexander JA, El Reda DK, Wise CG, Green LA, Fetters MD. Partial and

133
incremental PCMH practice transformation: implications for quality and costs. Health
services research. 2014;49(1):52-74.
92. Carlin CS, Flottemesch TJ, Solberg LI, Werner AM. System Transformation in
Patient-Centered Medical Home (PCMH): Variable Impact on Chronically Ill Patients'
Utilization. The Journal of the American Board of Family Medicine. 2016;29(4):482-495.
93. Higgins S, Chawla R, Colombo C, Snyder R, Nigam S. Medical homes and cost and
utilization among high-risk patients. The American journal of managed care.
2014;20(3):e61-71.
94. Mosquera RA, Avritscher EB, Samuels CL, et al. Effect of an enhanced medical home on
serious illness and cost of care among high-risk children with chronic illness: a
randomized clinical trial. JAMA. 2014;312(24):2640-2648.
95. van Hasselt M, McCall N, Keyes V , Wensky SG, Smith KW. Total Cost of Care Lower
among Medicare Fee-for-Service Beneficiaries Receiving Care from Patient-Centered
Medical Homes. Health services research. 2015;50(1):253-272.
96. Schwenk TL. The patient-centered medical home: one size does not fit all. JAMA.
2014;311(8):802-803.
97. Rhodes KV , Basseyn S, Gallop R, Noll E, Rothbard A, Crits-Christoph P. Pennsylvania's
Medical Home Initiative: Reductions in Healthcare Utilization and Cost Among Medicaid
Patients with Medicaland Psychiatric Comorbidities. Journal of General Internal
Medicine. 2016;31(11):1373-1381.
98. Health Insurance Marketplace Quality Rating System Measure Technical Specifications.
Centers of Medicare & Medicaid Services; Spetember 2015.
99. Core Set of Children's Health Care Quality Meeasures for Medicaid and CHIP (Child
Core Set). Centers for Medicare & Medicaid Services; July 2014.
100. Raven MC, Lowe RA, Maselli J, Hsia RY . Comparison of presenting complaint vs
discharge diagnosis for identifying “nonemergency” emergency department visits. JAMA.
2013;309(11):1145-1153.
101. Douglas R. Wholey MF, Katie M. White, Jon Christianson, Rob Kreiger, Jessica Zeglin,,
Suhna Lee LG. Evaluation of the State of Minnesota’ s Health Care Home Initiative
2010-2012. Minnesota: Minnesota Department of Health, Minnesota Department of

134
Human Services, University of Minnesota; January 2014.

Abstract (if available)

Abstract The objectives of this dissertation are to investigate the impacts of the patient-centered medical home model (PCMH) for children with special health care needs (CSHCN) and to improve our understanding in the differences between private insurance and Medicaid in health care expenditures, utilization, and quality of care. Chapter 2 updated the literature and evaluated whether the effects of the PCMH evolved with time using parent-reported 2008-2012 Medical Expenditure Panel Survey. Findings in this chapter suggested that PCMH was associated with an improvement in the parent-reported quality of care and access to preventive services, such as well-child care, with unchanged expenditures for CSHCN. However, these children were less likely to access mental health services when receiving care from a PCMH. Chapter 3 described the differences in health care expenditures and patterns of care between Medicaid and privately insured CSHCN. This cross-sectional analysis of 2013 insurance claims showed that, after accounting for baseline and demographic differences, Medicaid CSHCN were more likely than privately insured CSHCN to rely on emergency department, urgency care, and other non-hospital and non-office-based setting as the source of care. They also had significantly higher average expenditures when accessing this care, suggesting a less optimal care patterns. Chapter 4 quantified the PCMH effects for CSHCN in Minnesota using longitudinal insurance claims. The difference-in-difference analysis using 2007-2014 insurance claims suggested that the PCMH in Minnesota did not appear to either effectively reduce cost or improve quality of care for CSHCN. We found a significant decrease in urgent care expenditure immediately and an increase in evaluation and management expenditure, mainly driven by the lagged PCMH effects. Current PCMH may not benefit CSHCN, whom the model was originally designed to care for, the most. Further research with a longer time horizon and a focus disease area or risk stratification could help to better target efforts to optimize primary care delivery for CSHCN.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The impact of Patient-Centered Medical Home on a managed Medicaid plan

PDF

Three essays on estimating the effects of government programs and policies on health care among disadvantaged population

PDF

Economic, clinical, and behavioral outcomes from medical and pharmaceutical treatments

PDF

Value in health in the era of vertical integration

PDF

Discriminating changes in health using patient-reported outcomes

PDF

Developing an agent-based simulation model to evaluate competition in private health care markets with an assessment of accountable care organizations

PDF

Long-term impacts of childhood adversity on health and human capital

PDF

The impact of treatment decisions and adherence on outcomes in small hereditary disease populations

PDF

Effects of a formulary expansion on the use of atypical antipsychotics and health care services by patients with schizophrenia in the California Medicaid Program

PDF

Investigation of health system performance: effects of integrated triple element method of high reliability, patient safety, and care coordination

PDF

An evaluation of risk predictors for disease progression in Veteran Affairs patients with chronic hepatitis C

PDF

Essays on health insurance programs and policies

PDF

Advances and applications for economic evaluation methods in health technology assessment (HTA)

PDF

A series of longitudinal analyses of patient reported outcomes to further the understanding of care-management of comorbid diabetes and depression in a safety-net healthcare system

PDF

Three essays in health economics

PDF

Investigating racial and ethnic disparities in patient experiences with care and health services use following colorectal cancer diagnosis among older adults with comorbid chronic conditions

PDF

Essays on macroeconomics of health and labor

PDF

Impact of surgery on health-related quality of life in patients with acoustic neuroma

PDF

Racial/ethnic differences in colorectal cancer patient experiences, health care utilization and their association with mortality: findings from the SEER-CAHPS data

PDF

Essays on improving human interactions with humans, algorithms, and technologies for better healthcare outcomes

Asset Metadata

Creator Lin, Chia-Wei (author)

Core Title Delivering better care for children with special health care needs: analyses of patient-centered medical home and types of insurance

School School of Pharmacy

Degree Doctor of Philosophy

Degree Program Pharmaceutical Economics and Policy

Publication Date 12/12/2018

Defense Date 11/01/2017

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag children with special health care needs,health care expenditure,health care home,health care utilization,Medicaid,OAI-PMH Harvest,patient-centered medical home,quality of care

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Romley, John A. (committee chair), Carlin, Caroline (committee member), Joyce, Geoffrey (committee member)

Creator Email cwlin114@gmail.com,linchiaw@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c40-461410

Unique identifier UC11267169

Identifier etd-LinChiaWei-5951.pdf (filename),usctheses-c40-461410 (legacy record id)

Legacy Identifier etd-LinChiaWei-5951.pdf

Dmrecord 461410

Document Type Dissertation

Rights Lin, Chia-Wei

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA