University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
RACIAL/ETHNIC DIFFERENCES IN COLORECTAL CANCER PATIENT EXPERIENCES,
HEALTH CARE UTILIZATION AND THEIR ASSOCIATION WITH MORTALITY:
FINDINGS FROM THE SEER-CAHPS DATA

By
Stephen Chen

A Thesis Presented to the
FACULTY OF THE USC KECK SCHOOL OF MEDICINE
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
APPLIED BIOSTATISTICS AND EPIDEMIOLOGY

May 2020

Copyright 2020 Stephen Chen
ii

Acknowledgments
The interpretation and reporting of these data are the sole responsibility of the authors. The authors
acknowledge the efforts of the National Cancer Institute; the Centers for Medicare & Medicaid Services;
Information Management Services (IMS), Inc.; and the Surveillance, Epidemiology, and End Results
(SEER) Program tumor registries in the creation of the SEER-CAHPS database.

iii

Table of Contents
Acknowledgments .......................................................................................................................... ii
List of Tables ............................................................................................................................... iv
Abstract ............................................................................................................................................v
1 Introduction ................................................................................................................................. 1
2 Methods.............................................................................................................................3
a. Data Source .............................................................................................................3
b. Study Sample ...........................................................................................................4
i. Race/ethnicity ..............................................................................................4
ii. Healthcare Utilization ..................................................................................5
iii. Patient Experiences with Care .....................................................................6
iv. Covariates ....................................................................................................6
v. Mortality ......................................................................................................7
vi. Statistical Analysis .......................................................................................7
3 Results ...............................................................................................................................8
a. Regression Analyses ..............................................................................................8
i. Multivariate logistic regression .....................................................................9
ii. Cox proportional hazards regression ..........................................................10
4 Discussion ...................................................................................................... 11
References ................................................................................................................ 33

iv

List of Tables
Table 1. Distribution of sociodemographic covariates for patients diagnosed with colorectal
cancer between 1997-2011 ........................................................................................................... 20
Table 2a. Means of quality of care composite variables and patient ratings by the levels of
demographic variables ...................................................................................................................23

Table 2b. Means of quality of care composite variables and patient ratings by the levels of
demographic variables ..................................................................................................................26
Table 3. Frequency of utilization counts by demographic covariates (categorically defined as ‘1’
=0 utilization, ‘2’= 1-3 utilizations, ‘3’ = 4-9 utilizations, ‘4’ = >9 utilizations). ........................ 29
Table 4. Multivariate ordinal logistic regression model of patient ratings and covariates including
counts of visits variable. ............................................................................................................... 32
Table 5. Multivariate cox regression model stratified by race for each patient rating including
visit count variable and demographic covariates ..........................................................................38

v
Abstract
Background: The National Cancer Institute (NCI) and Centers for Medicare and Medicaid Services
(CMS) have collaborated to create linked data resources to improve our understanding of patterns of care,
health care costs, and trends in utilization. However, existing data linkages have not included measures of
patient utilization or experiences with patient outcomes. Therefore, our objective is identify the influence
of health care utilization defined as (inpatient, outpatient, and doctor visits) as a covariate of the
relationship between patient experience and colorectal cancer outcomes and whether the counts of
utilization explains/modify the relationship between patient experience and colorectal cancer mortality.
We hypothesize that higher counts of visits (within a six-month period after colorectal cancer diagnosis
and prior to CAHPS survey) should lower patient experiences with care for Asians but result in better
outcomes while higher counts of visits should result in a higher experience rating and result in better
outcomes for African American patients.
Methods: We used the SEER registry data linked with patient surveys from the Consumer Assessment of
Healthcare Providers and Systems (CAHPS) completed after diagnosis date. We examined responses
from 6340 subjects about various aspects of their care such as the ability to get needed care, and to get
care quickly. We calculated frequencies chi squares of association for our demographic covariates by
patient experience ratings and utilizations. Covariates that our models adjusted for as potential
demographic confounders are age at diagnosis, poverty level, geographic region, marital status, or
educational attainment, Medicare type, age at CAHPS survey, survey year, and survey method. The study
hopes to identify particular usage patterns of services that differed by racial/ethnicity from Medicare
claims files that can enlighten researchers how the utilization counts for three kinds of encounters
(inpatient(short stay, long stay, or skilled nursing facility), outpatient (preventative and acute treatment),
and doctor visits as determined by documentation from Medicare claims files. From there, the data was
used to determine if there was significant association between these utilizations and patient experiences
within 6 months prior to the CAHPS survey and the CAHPS survey ratings after diagnosis.
vi

Results:
A total of 6340 subjects were identified. A majority of the cohort were non-Hispanic white patients
(75.4%), followed by non-Hispanic black patients, Hispanic and Asian patients (8.6%, 8.1%, and 7.3%,
respectively). After controlling for potential confounders, Asian patients had significantly lower odds for
giving a rating of “excellent” (a score that is >9 for global ratings and <90 for composite ratings). With
the exception of health care, obtaining patient care quickly and health plan, Asians were significantly
more likely to submit a “not excellent” rating for customer service (OR 1.62; 95% CI: 1.08 – 2.43),
physician communication (OR 1.55; 95% CI: 1.1 – 1.98), getting care quickly (OR 1.96; 95% CI:1.54 –
2.50), getting needed care (OR 1.76 95% CI: 1.38 -2.24), getting needed prescription drugs (OR 1.59;
95% CI: 1.25 -2.03). Among non-Hispanic white patients those who had a one unit increase in rating for a
specialist physician had a 5.3% decreased in risk for mortality (HR 0.95; 95% CI: 0.90, 1.00) compared to
the null. There was also a significant decrease in risk of dying for non-Hispanic white patients for a one-
unit increase in getting prescription drugs HR 0.99; 95% CI :0.99, 1.00). However, the hazard ratio is
much smaller in effect size. Among Hispanic patients, a similar effect that for every one-unit increase in
rating for specialist physicians (HR 0.83; 95% CI: 0.70, 0.98), one sees a resulting decrease in risk of
18.4%. in Hispanic patients for mortality. Hispanics also had a significant decrease in risk for health care,
Hispanic patients are 22% less at risk for mortality if they were diagnosed with colorectal cancer HR
0.78; 95% CI: 0.68, 0.90). Among Asian patients there was significant increase for risk by 66% for a one-
unit increase in rating for specialist physicians HR 1.66; 95% CI 1.03, 2.65).
Conclusion:
Asians reported both poorer patient experiences and increased risk per unit increase of patient experience
rating, but there was not a significant effect modification of utilization on the association of patient
experience and mortality as our study. We did observe a protective effect for patient rating in non-
Hispanic whites and Hispanics which was not observable for non-Hispanic blacks/Asians indicating racial
disparities in healthcare to further investigate.

1

Introduction:
Cancer mortality differs by race/ethnicity.
2

Generally, minorities in the United States have worse
outcomes compared to non-Hispanic Whites. For example, compared with non-Hispanic white women,
African-American, and American Indian/Alaska Native women had a 20.7 and 18.7 fold times the risk of
mortality from colorectal cancer despite no significant differences in cancer stage by race or endoscopy
completion
21
. It was concluded disparities result from other factors and differences in the receipt of
effective cancer treatment.
The process of receiving appropriate cancer treatment is complex and fraught with many administrative
barriers that can lead to racial/ethnic disparities in timely receipt of cancer treatment. Due to these
complexities, the Institute of Medicine (IOM) and the National Cancer Institute (NCI) have released
several reports that emphasize the importance of cancer treatment that focuses on the needs of patients
and the use of patient-reported measures to evaluate the quality of care that they receive. Thus, the
racial/ethnic disparities in late stage at diagnosis and treatment emphasize the importance of high-quality
medical care.
The Consumer Assessment of Healthcare Providers and Systems (CAHPS) is a survey of patient
experience with care—a measure of quality of care. The measures are defined into three domains (1)
patient-centeredness, (2) timeliness, and (3) “realized access” to care. Patient-centeredness is evaluated on
several questions that reflect the patients’ experience of their interactions with the doctor. Timeliness is
related to swiftness of obtaining care e.g. a doctor’s appointment or medication. Realized access is an
aspect of health care effectiveness that addresses specifically whether families perceive problems in
obtaining necessary care or referrals.
9
In a study by Collins et al. (2017)- a joint test of the interactions
between the composite scores and the 5 largest racial/ethnic/language subgroups was statistically
significant (P <0.0001), suggests that the domains varied across subgroups. Doctor communication had
the strongest relationship with care ratings for non-Hispanic whites and English-preferring Hispanics.
2

Getting needed care had the strongest relationship for Spanish-preferring Hispanics and Asian/Pacific
Islanders. Doctor communication and getting care quickly were strongest for African Americans.

In prior studies, Asians who received care in the same healthcare setting rated or reported their experience
much poorer than other groups. Studies have identified that Asians exhibit a greater tendency of avoiding
a top-box response to questions framed as “rating” versus “reporting” and that the magnitude of response
style can be reduced considerably by using questions (reporting vs rating) and answer scales that target
more objectively defined experiences.
1
However, several articles have noticed this discrepancy among
ethnic groups e.g. (Chung et al. 2018) where the CAHPS ratings of Asians are generally lower compared
to non-Hispanic Whites but African Americans broadly show a higher CAHPS rating when compared to
non-Hispanic Whites.
1
In a study by Mayer et al. 2016 Asian patients exhibited lower ‘extreme response
tendency’ (giving extreme left or right ratings on a scale) than Non-Hispanic whites in response to
standardized scenarios. Because CAHPS responses are predominantly near the positive end of the scale
and the most responsive scenario is most typical of the score observed in real-world settings, lower
positive end ratings in Asians may partially explain observations of lower observed mean CAHPS scores
for Asians in real-world settings.

Patient utilization counts may also be an indicator for a patient’s valuation of their experience.
8
Research
shows that there is a connection between patient utilization(in services used) and the effect on patient
experience rating of their care (CAHPS).
18
Despite there being a large amount of literature that have
shown this finding, it is largely unclear the driving determinants and predictors for these ratings.
1
A
factors that has potentially been raised for the differences found between in ratings among minorities are
cultural factors that lead to systemic bias(es) for these patient ratings.
1
On top of this, the Mollica et al.
2016 study demonstrates that there is a link between patient experiences via CAHPS ratings and patient
outcomes in mortality. At this moment we can make a connection between patient utilization and patient
experience and how they affect each other and understand that patient experience(s) tell us something

3

about outcomes. (Mollica et al. 2017). From there, we also can show that there are discrepancies in
patient experiences by ethnicity.

Previous research from the SEER registry-CAHPS Survey on breast cancer patients show significant
correlations between CAHPS responses in rating of their experiences with customer service and stage of
diagnosis for two ethnic groups: African and Asian Americans.
7
However, these studies do not account
for variations in the number of health care visits nor do they examine the influence of CAHPS measures
on cancer mortality. Therefore, this study attempts to bridge the gap between what is observed in
CAHPS-related literature of there being these disparities in ratings of Asian versus African-American
patients as well their correlated outcomes by exploring disparities in utilization by race/ethnicity and
determining if it is a potential predictor for the outcome of mortality in colorectal cancer patients. We
hypothesize that the quantity of health care utilization is a confounder for the relationship between
CAHPS measures and mortality. Should our study discover the driving factors behind what raises or
lowers a given patients’ CAHPS rating, it would then give researchers greater leeway in understanding
the etiology of disparities in outcomes of mortality within healthcare systems.

Methods:
Data Source
We use data from the National Cancer Institute’s Surveillance, Epidemiology, and End Results-Consumer
Assessment of Healthcare Providers and Systems (SEER-CAHPS
®
) that links three secondary data
sources: 1) SEER cancer registry data, 2) Centers for Medicare and Medicaid Services (CMS) CAHPS
surveys and 3) administrative or billing data through Medicare claims and enrollment records. Briefly, the
SEER cancer registry data is available to identify cancer diagnosis, socioeconomic, demographic,
prognostic, and initial treatment factors. The SEER cancer registry program collects and maintains patient
demographic (date of birth, race/ethnicity, sex, marital status), tumor and clinical prognostic information
4

(e.g. date of diagnosis, tumor stage (clinical and pathologic) and size, cell differentiation, axillary lymph
node involvement, receptor-status, histology) and treatment information (surgery, radiation) on all
individuals diagnosed with cancer while residing in participating cancer registry areas. The data also
contain neighborhood-level socioeconomic information. The Consumer Assessment of Healthcare
Providers and Systems (CAHPS) survey of patient experiences are national probability-sample surveys of
Medicare beneficiaries that assess a variety of measures relating to perceived quality of and access to
care. The CAHPS survey was developed to assess patient experiences with health care providers and the
care that patients received as a method of measuring quality of care. CMS has surveyed a nationally
representative random sample of Medicare beneficiaries annually since 1998 on a wide variety of
measures of perceived quality and access to care. The average response rate for Medicare patients is
71%
8
.
Study Sample
We conducted a retrospective cohort study that includes participants diagnosed with colorectal cancer
2002-2011. Colorectal cases were identified using data from the National Cancer Institute’s Surveillance,
Epidemiology, and End Results (SEER) cancer registry and CAHPS questionnaire data. We used a
procedure similar to other published studies using SEER-CAHPS surveillance data (Mollica; Halpern).
Surveys were filtered to include individuals who were 1) 65 years of age or older at diagnosis, 2) were
diagnosed with colon or rectal adenocarcinoma (International Classification of Diseases for Oncology,
Third Edition codes C180, C182-C189, C199, C209) as a single, first primary cancer, and 3) submitted a
CAHPS survey no fewer than six months after cancer diagnosis.
We developed a mutually exclusive variable by defining race/ethnicity identified in SEER with the
following categories: Non-Hispanic white, Non-Hispanic black, Hispanic, Non-Hispanic Asian, and
other. If a patient was categorized as being of Hispanic ethnicity in either SEER, Medicare or CAHPS
datasets then the patient was considered to be Hispanic. We then used the self-reported race variable from
CAHPS to categorize a patient’s race. If race was missing, we used the race variable from the SEER
5

registry and then Medicare. Within “Other” race, we excluded subjects if there were missing gender
information due to small sample size (N=5).
Race/ethnicity
Healthcare Utilization
Patient utilization of medical services was assessed as the number of visits 6 months prior to the date of
taking a CAHPS survey but after diagnosis for three visit types (inpatient, outpatient, doctor visit). Patient
utilization was defined by counts of visits where visits were categorized based on distribution into 4
categories of [0], [1-3], [4-9], and [>9]. Patient utilization is further stratified in our study into three broad
services: physician visits(carrier claims from NCH), inpatient(acute hospital services where patient is
admitted to hospital ), and outpatient services (services that do not require hospital admission but are not
in an emergency setting either). These visit types were classified in Medicare claims files according to
documentation provided by SEER-CAHPS (https://healthcaredelivery.cancer.gov/seer-
cahps/medicare/claims.html) where four different files from the registry were manipulated and merged to
produce tables including utilization(number of visit) counts, demographic covariates, and the four global
rating and five composite scores defined in the methods. The data was taken from MEPPAR (Medicare
Provider Analysis and Review), NCH (National Claims History), outpatient Medicare claims, and PEDSF
(Patient Entitlement and Diagnosis Summary File). The files were filtered for: 1) cancer at a site
designated as colorectal cancer 2) cancer only in primary site 3) survey date is non-missing 4) more than
6 months have passed from diagnosis to survey date and 5) selected for only the first survey taken after
this 6-month time period.

6

Patient Experiences with Care
Patient experience was assessed by five composite measures: 1) customer service, 2) physician
communication, 3) getting care quickly, 4) getting needed care, and 5) getting needed prescription drugs.
We also used four single-items (“global”) that measure patients’ experiences with their overall care,
health plan, personal doctor, and specialist physician using a 0-to-10 rating scales. Variables assessing
patients’ quality of care included patient ratings of care and composite scores of patients’ experiences
with care. Composite scores of patient experiences in M-CAHPS were assessed on a 100-point scale and
include: getting all needed care (2 items), getting care quickly (2 items), quality of physician
communication (4 items), health plan customer service (2 items), and access to prescriptions (1 item).
Both composite scores and global ratings were then binarized into an “excellent” or “not- excellent”
categorization based on whether the composite scores were greater or less than 90 and if global ratings
were greater or less than 9. Ratings of care were assessed on a 10-point scale and include one question per
item: rating of their health plan, of their health care, of their primary physician, of their specialist, and of
their prescription drug plan. Questions can be found at this link:
(https://www.ahrq.gov/sites/default/files/wysiwyg/cahps/surveysguidance/hp/about/measures_hp50_2109
.pdf). Following NIH recommendations, we examined these variables using the linear mean scoring
method.

Covariates
Demographic information was collected from SEER, CMS Medicare claims, and CAHPS data sets. In
certain cases, variables such as race/ethnicity were constructed from multiple data sets in order to ensure
data completeness. State of residence was recoded into geographical regions using the divisions in the
U.S. Census Bureau (CITE). Other variables include race (non-Hispanic white, non-Hispanic black),
gender (male or female), marital status(unmarried, married, unknown), poverty level of the census tract in
which the patient resided (0% to <5% (low poverty), 5% to <10%, 10% to <20, 20% to 100% (high
poverty), and missing), educational attainment (High school or less, college and higher, and missing),

7

mode of survey(mail, phone, unknown), Medicare insurance type (fee-for-service (FFS) and Medicare
Advantage (MA) with and without part D coverage), number of comorbid conditions [0, 1, 2+], ages for
[<65], [65-69], [70-74], 7[5-79,] [80-84], [>85] for both age at diagnosis and age at survey, as well as
time from diagnosis to survey (greater or less than 3.82 years), and survey year ([1997-2000],
[2001-2005], [2007-2011]).

Mortality.
Date of patient death was taken from Medicare records indicating the date the patient died through
December 31, 2013, the date of last follow-up. If a patient did not have a date of death, they were
assumed to be alive through the end of follow-up. Of the 6340 patient who met inclusion and exclusion
criteria, 19.0% (1205 deaths) of cohort participants died during the study period and the mean survival
time was 3013 days and a median of 3029 days. This translates to a grand total of average survival of 8.25
years and a median survival of 8.29 years among our entire cohort. Among the 1205 individual who died
the average mean survival time was 1898.35 days and median survival was 1795 days which is 5.20 years
on average and a median of 4.92 years rounded to the hundredth.

Statistical Analysis
Descriptive analyses included analyzing the distribution of covariate variables by race (Table 1).
Additionally, we examined the means of quality of care composite variables and patient ratings by the
levels of demographic variables (Table 2). Table 3 then described categories of utilization levels
compared to our covariates. Differences in these tables among race were determined by a chi-square test
of association (p <0.05).
We used multilevel logistic regression to assess the association between health care utilization and
excellent ratings of patient experience after diagnosis (defined as being greater than or equal to 9 or less
than 9) (Table 4). The logistic regression is modeled in a descending fashion such that the odds ratio
8

predicts the likelihood of obtaining a “not excellent” composite or global rating. Covariates included race,
gender, age at survey, marital status, poverty indicator, education level, geographic region, Medicare type,
mode of survey assessment, comorbidities, and cancer stage (Table 4).
Multivariable Cox proportional hazards models were used to determine the association between patient
utilization counts, race/ethnicity, and patient mortality after cancer diagnosis which compared the risk for
mortality(death) per one unit increase in a given rating at the level of a covariate(s) (Table 5). The data
was analyzed via SAS (Statistical Analysis Systems) software.

Results
A total of 6340 subjects were identified after applying inclusion and exclusion criteria. A majority of the
cohort were non-Hispanic white patients (75.4%), followed by non-Hispanic black patients, Hispanic and
Asian patients (8.6%, 8.1%, and 7.3%, respectively). Table 1 describes the baseline characteristics of
subjects by race. The mean age of the total cohort was 72.2 (± 8.8) years. A smaller proportion of Non-
Hispanic black patients were married (38.9%) compared to other racial/ethnic groups. The poverty rates
varied greatly by race and were highest for non-Hispanic black patients (53.9%) and Hispanic patients
(28.8%) compared to non-Hispanic white patients (12.5%) and Asian patients (12.7%). Non-Hispanic
black and Hispanic patients were less likely to report completing some college or higher education
(23.1%, 20.0%) than other race groups. Roughly half of the cohort (50.0%) lived in the West region,
however, the regions varied by racial and ethnic groups. A considerably higher percentage of non-
Hispanic black patients lived in the South (51.6%) and (95.7%) of Asian patients lived in the West. Non-
Hispanic white patients had the lowest percentage of people with FFS only (28.4%) compared to Hispanic
patients (47.4%).
The mean composite score and rating of care in patients’ experiences with the health care delivery system
are shown in Table 2 while Table 3 lays out the frequency distribution of patient’s healthcare utilization
by counts of visits. In a chi-square test of association (p<0.05), Table 3 shows that utilization differ
9

significantly by each level of covariate with the exception of marital status and age at diagnosis. The
largest frequency of visits used was the first category of visits (where visits =0) with 47.9% of the
utilizations measured being in that category. Non-Hispanic whites had the largest frequencies of each visit
rate, comprising 70.9% of 0 visits, 73.4% of the visit rate [1,3], 82.7% of visit rate [4,9], and 81.6% of the
visit rate [>9]. Asian had the lowest frequencies for visit rate(s) in that order ranging from 9.5% of no
visits down to 4.5% of visit rate [>9]. Each outcome has a different number of subjects because the valid
response rate for each item varied by survey year, visit experiences, and Medicare programs. The number
of subjects included in the analyses ranged from 1738 (customer service) to 5618 (health plan) and the
mean rating of experiences varied from 79.4 (getting care quickly) to 90.3 (getting needed prescriptions).
There were significant racial/ethnic differences in mean composite scores and global rating. Asian
patients reported the lowest mean score for getting need prescription drugs (72.9 ± 28.5), customer service
(73.5±28.8), and getting needed care (80.3±23.8). Non-Hispanic blacks rated highest for doctor
communication (90.1 ±16.2), customer service (80.5 ±27.1), and primary physician care (8.6 ±2.0).

Regression Analyses
Multivariate logistic regression
The results from the adjusted logistic regression models are presented in Table 4, which show the
association of composite and global ratings of care and demographic covariates that includes adjustment
for categorical counts of visits. For Asians – there was largely a consistent pattern of significant
association between being Asian and having a composite or global rating that was “not excellent” (<9 or
90) in comparison to non-Hispanic whites. With the exception of health care rating, obtaining patient
care quickly and health plan rating, Asians were significantly more likely to submit a “not excellent”
rating for customer service (OR 1.62; 95% CI: 1.08 – 2.43), physician communication (OR 1.55; 95% CI:
1.21 – 1.98), getting care quickly (OR 1.96; 95% CI: 1.54 – 2.50), getting needed care (OR 1.76; 95% CI:
1.38 -2.24), getting needed prescription drugs (OR 1.59; 95% CI:1.25 -2.03), and specialist rating (OR
10

1.71; 95% CI: 1.28 -2.29) compared to non-Hispanic whites. Asian patients compared to non-Hispanic
white patients had a 62% increase in likelihood of rating not excellent for customer service, 55% increase
in likelihood of rating not excellent for physician communication, 96% increase in likelihood of rating not
excellent for getting needed prescription drugs, and a 71% increase in likelihood of rating not excellent
for a global rating of their specialist(s). We did not observe a significant association between any of the
race groups and patient experiences with care among non-Hispanic white, non-Hispanic black, and
Hispanic colorectal cancer patients after their diagnosis.

Cox proportional hazards regression
The results from adjusted cox regression models are presented in Table 5 which shows the association
between composite and global ratings and mortality among each racial/ethnic group. For non-Hispanic
white patients, a one unit increase in rating for a specialist physician was associated with a 5% decreased
in risk for mortality (HR 0.95; 95% CI:0.90-0.99). There was also a significant decrease in risk for non-
Hispanic white patients for a one-unit increase in getting prescription drugs (HR 0.99; 95% CI: 0.99,
1.00) (appears inclusive of null due to rounding). However, the hazard ratio is much smaller in effect size.
Among Hispanic patients, we observe a similar effect where a one-unit increase in rating for specialist
physicians (HR 0.83; 95% CI: 0.70, 0.98), was associated with a decrease in risk of death by 17%. For
every one-unit increase in health care rating, Hispanic patients also had 22% lower risk for mortality HR
(0.78; 95% CI: 0.68, 0.90). We did observe an indication for Asian patients having increased risk for
mortality for a one-unit increase in rating. Among Asian patients there was significant increase in risk for
death by 66% for a one-unit increase in rating for specialist physicians (HR 1.66; 95% CI: 1.03, 2.65)
compared to non-Hispanic whites. There was also a significant increase for risk found for one-unit
increase(s) in customer service for Asian patients (HR:1.27, n/a)– however small sample size for that cox
regression was unable to converge for a confidence interval making the result not necessarily as
meaningful. Our study did not find significant risk association with patient rating for non-Hispanic black
patients.
11

Discussion:
In this large retrospective cohort study, we used data from the SEER-CAHPS linked dataset to investigate
whether there are differences in the association of patient experiences among colorectal cancer patients
after their cancer diagnosis and mortality taking into account patient utilization ( in visitation counts) with
aspects of their medical care. For every composite rating and global rating with the exception of primary
physician and health plan global ratings, the odds for Asian patients to rate “not excellent” experience
(<9) was significantly higher compared to non-Hispanic white patients, which is consistent with domain
literature on CAHPS rating tendencies on Asian patient rating tendencies.
1,7,13
However, what was not
seen in our study was a significant association of healthcare utilization as defined by counts of visits with
patient rating(s). After controlling for health care utilization, we did not observe differences in patient
experiences with care between Black and Hispanic patients compared to non-Hispanic white patients.
Although there was not a specific interaction found between ratings and utilization equating to a
significant modification in association with mortality, what we did observe are a few measures of patient
experiences among non-Hispanic whites, Hispanics, and Asian that are significantly associated with
mortality. For non-Hispanic whites, what is observed are significant protective effects for better reports of
patient experiences with receiving prescription drugs and ratings of their specialist. This protective effect
is also observed for Hispanics where when increases in their scores for the global health care rating and
specialist ratings indicate an improvement in mortality outcomes. Interestingly, what is observed for
Asian patients is as their ratings increased for their specialist, they had a much higher risk for mortality.
This was an effect that was observed in the Farias et al., (2020) paper where increases for rating equated
to worse outcomes and an assumable reverse effect that was similarly found in this study. There were
several other measures of patient ratings among Asians, though not significant, also producing high
hazard ratios given a rather small population in our study (465 Asian patients). The results show
12

something that was not expected in that we see that there is a protective effect and significant association
for non-Hispanic whites that score increases are significantly associated with improvements in their
outcomes with regards to colorectal cancer. In other words, the more satisfied these patients are with their
care, the more successful their chances of surviving are. On the other hand, for Asian patients we see that
(in particular for rating specialists) as they feel more positive about their care, their outcomes are
significantly worse with a rather large effect size. Similarly, we did not see the kind of positive
association in African-Americans as we did with non-Hispanic whites. This finding gives additional
support to what is known in domain literature on racial disparities in healthcare which have found
discrepancies for minorities as this pattern of increased mortality in our study was not observed in non-
Hispanic whites.
2,3
Of course, there are likely other factors that contribute to these results and given our
study was not assessing cancer stage for patient diagnosis, there may be a form of confirming bias for
patients who were already in a later stage of colorectal cancer resulting in mortality. However, our results
do show there are significant linkages between patients experience and mortality in non-Hispanic whites
and Hispanics as well as significant linkages between the number of utilizations a patient uses and
mortality outcomes or Asians. Both patient experience and healthcare utilization show significant
differences by racial/ethnic group.

Our study had a few limitations. First, patient experiences with care was assessed at one snapshot in time
after diagnosis for colorectal cancer and may presumably vary prior to diagnosis due to the use of medical
services and health care needs. However, in a sensitivity analysis where we stratified the analysis based
on the time from survey completion to the date of diagnosis, we found that patient experiences with care
were similarly associated with patent experiences even among those patients that completed the survey
greater or less than the median time difference from diagnosis to survey of about 3.82 years after their
date of diagnosis. Second, questions on patient experiences with care were related to overall medical care
rather than care related specifically to colorectal cancer screening and diagnosis. However since our data
captures inpatient visits(e.g. hospital stays), doctor visits(primary and specialist) as well as outpatient
13

visits(referrals), our study captures a comprehensive experience that includes patients’ feeling about care
specific to their diagnosis of colorectal cancer. Third, the study population was limited to patients age 65
and older enrolled in Medicare fee-for-service and Medicare advantage who were diagnosed with cancer
within the SEER-registry regions; thus, results may not be generalizable to younger cancer survivors.
Finally, to be included in the study, individuals had to be able to complete a CAHPS survey and although
proxies were able to complete a survey on behalf of the individual, we may have fewer individuals in
poorer health after diagnosis. Another limitation to our study was the sample size we were able to use
given inclusion/exclusion criteria. From our table 1, one can observe that the non-Hispanic white
demographic has far more subjects than other racial/ethnic groups being >10 times the size of the smallest
racial/ethnic group (Asian). As a result, had there been more data available to the study for minority
groups, there may have been greater significance in patterns we observed between patient rating and
utilization and/or with outcomes.
In conclusion, our study did not find results to fully support our predicted hypothesis. The study did find
interesting results in that it demonstrated by and large that as Asian patients had more utilization they are
indeed rating significantly worse than other groups and that positive ratings among their population group
are actually associated with worse outcomes which indicates a gap in healthcare that should further
continue to be explored. We did not find a significant link between non-Hispanic black patient rating and
utilization.
We did observe that for Hispanics and non-Hispanic white patients, better ratings may be associated with
a lower risk of mortality. Our study was successful in determining that even a simple quantification of
healthcare utilization and examining patient experiences are significant factors for mortality and that
greater research should be undertaken to fully unpack these pathways that are critical for improvement in
the outcomes for ethnic populations.

14
Table 1: Distribution of sociodemographic covariates for patients diagnosed with colorectal cancer between 1997-2011 (bold indicates chi-
square p<0.05)

Total Non-Hispanic white Non-Hispanic Black Hispanic Asian

6340 4784 (75.4%)

545 (8.6%)

546 (8.6%)

465 (7.3%)

Mean age at survey
(±SD)
76.6(±8.4) 77.2(±8.3) 73.4 (±8.7) 75.6(±8.8) 76.4(±8.2)

Gender (M/F)

Male 2997(47.3%) 2318 (48.5%) 177(32.5%) 280(51.3%) 222(47.7%)
Female 3343(52.7%) 2466(51.5%) 368(67.5%) 266(48.7%) 243(52.3%)

Age at survey

<65 333(5.3%) 208(4.4%) 67(12.3%) 39(7.1%) 19(4.1%)
65-69 941 (14.8%) 658 (13.8%) 110 (20.2%) 88(16.1%) 85(18.3%)
70-74 1073(16.9%) 796(16.6%) 97(17.8%) 99(18.1%) 81(17.4%)
75-89 1361(21.5%) 1022(21.4%) 127(23.3%) 119(21.8%) 93(20.0%)
80-84 1437(22.7%) 1136(23.8%) 85(15.6%) 117(21.4%) 99(21.3%)
85 or older 1195(18.9%) 964(20.2%) 59(10.8%) 84(15.4%) 88(18.9%)

Marital status
Unmarried 2303 (38.2%) 1660 (34.7%) 303 (55.6%) 215(39.4% ) 125(26.9%) (26.9%)
Married 3717 (61.7%) 2875 (60.1%) 212 (38.9% ) 304 (55.7 %) 326 (70.1%)
Unknown
320 (5.1%) 249(5.2%) 30 (5.5%) 27 (4.9%) 14(3.0%)

Census Tract
Poverty

0% to <5% (low
poverty)
1678 (26.5%) 1426 (29.8%) 31(5.7%) 84 (15.4%) 137(29.5%)
5% to <10% 1764 (27.8%) 1454(30.4%) 68 (12.5%) 103 (18.9%) 139(29.9%)
10% to <20% 1758(27.7%) 1283(26.8 %) 150 (27.5%) 197 (36.1%) 128 (27.5%)
20% to 100% (high
poverty)
1106 (17.5%) 596(12.5%) 294(53.9%) 157(28.8%) 59(12.7%)
Missing 34(0.5%) 25 (0.5%) 2(0.4%) 5(0.9%) 2(0.4%)

15

Education level
High school or less
3414(53.9%) 2415 (50.5%) 356 (65.3%) 398(72.9%) 245(52.7%)
College and higher 2429(38.3%) 2023 (42.3%) 126(23.1%) 109(20.0%) 171(36.8%)
Missing 497(7.8%) 346 (7.2%) 63(11.6%) 39(7.1%) 49(10.5%)

Time from survey
to diagnosis

Greater or equal to
median time (1396-
3.82 years)
days ~ 3.82 years)
3171(50%) 2393(50.0%) 261(47.9%) 282(51.6%) 235(50.5%)
Less than or equal to
median time (1396
days ~ 3.82 years)
3169(50%) 2391(50.0%) 284(52.1%) 264(48.4%) 230 (49.4%)
Cancer Stage
0 78.5(±27.6)
79.7(±27.5)

529 (11.1%) 72 (13.2%) 49 (8.8%) 59(12.7%)
1 78.5(±27.6)
79.7(±27.5)

1525(31.9%) 177(32.5%) 179(32.8%) 152(32.7%)
2 3213(50.6%) 2441(51.0%) 245(45.0%) 288(52.8%) 239(51.4%)
3 264(4.2%) 199(4.2%) 35(6.4%) 19(3.5%) 11(2.4%)
4 121(1.9%) 90 (1.9%) 16(2.9%) 11(2.0%) 4 (0.8%)

Region

West 3185(50%) 2206(46.1%) 107(19.6%) 427(78.2%) 445(95.7%)
Midwest 619(9.8%) 546(11.4%) 59(10.8%) 10(1.8%) 4(0.9%)
Northeast 1095(17.2%) 914(19.1%) 98(18.0%) 71(13.0%) 12(2.6%)
South 1441(22.7%) 1118(23.4%) 281(51.6%) 38(7.0%) 4(0.9%)
Medicare type

FFS only 2069(34.2%) 1358(28.4%) 259(45.7%) 258(47.4%) 204(43.9%)
FFS PDP 998(15.7%) 795(16.6%) 59(10.8%) 81(14.8%) 63(13.6%)
MA only 1285(20.3%) 1024(21.4%) 95(17.4%) 89(16.3%) 77(16.6%)
MA PDP 1699(26.8%) 1412(29.5%) 107(19.6%) 98(18.0%) 82(17.6%)
Missing

289(4.6%) 195(4.1%) 35(6.4%) 20(8.6%) 39(8.4%)

16

Mean age at dx
(±SD)
72.2((±8.8) 72.7(±8.6) 69.2(±9.1) 71.1(±9.2) 71.9± (8.5)
Age at dx (N,%)
<65 1218(19.2%) 843(17.6%) 160(29.4%) 122(22.3%) 93(20%)
65-69 1375(21.7%) 1005(21%) 130(23.9%)

130(23.8%) 110(23.7%)
70-74 1035(16.3%) 784(16.4%) 93(17.0%) 81(14.8%) 77(16.6%)
75-79 1285(20.3%) 991(20.7%) 91(16.7%) 110(20.2%) 93(20%)
80-84 987(15.6%) 805(16.8%) 52(9.5%) 71(13.0%) 59(12.7%)
85 or older 440(6.9%) 356(7.4%) 19(3.5%) 32(5.9%) 33(7.1%)

Survey method
Mail
5221(82.4%)

4019(63.3%)

383(70.3%)

412(75.5%)

407(87.5%)

Phone 232(3.7%) 160 (3.3%) 19(3.5%) 39(7.1%) 14(3.0%)
Unknown 887(14%) 605(12.7%) 143(26.2%) 95(17.4%) 44(9.5%)
Count of
comorbidities

0 4074(64.2%) 3126(65.3%) 319(58.5%) 336(61.5%) 291(62.6%)
1 1518(23.9%) 1108(23.1%) 149(27.3%) 139(25.5%) 122(26.2%)
2+ 750(11.8%) 550(11.5%) 77(14.1%) 71(13.0%) 52(11.2%)

Survey year

1997-2000 1069(16.9%)
1074(16.9%)
2232(35.2%)
3039(47.9%)
866 (18.1%) 63 (11.6%) 92(16.9%) 53(11.4%)
2001-2005 2232(35.2%) 1684(35.2%) 204(37.4%) 198(36.3%) 146(31.4%)
2007-2011 3039(47.9%) 2234(46.7%) 278(51.0%) 256(46.9%) 266(57.2%)

17
Table 2a: Means of quality of care composite variables and patient ratings by the levels of demographic variables
Bold sig, round to 10
th
add SD (bold indicates Wald’s chi-square p<0.05)

Customer Service Physician
communication
Getting care quickly Getting needed
care
Getting needed
prescription drugs

Total analyzed: 1738(27.4%) 4728(74.6%) 4786(75.5%) 4666(73.5%) 5355(84.4%)

Mean score: 79.4(±26.8) 89.1(±17.0) 83.9(±23.5) 87.5(±20.2) 90.3(±20.5)

Gender (M/F)

Male 79.5 (±26.3) 89.0(±16.8) 83.1(±24.1) 87.9(±19.6) 89.9(±21.0)
Female 79.32 (±27.2) 89.2(±17.2) 84.6(±23.0) 87.0((±20.8) 90.5(±20

Age at survey

<65 78.9(±25.8) 87.5(±20.0) 82.0 (±24.3) 82.7(±24.0) 85.3(±23.2)
65-69 78.9(±27.2) 89.9(±16.8) 83.7((±24.0) 87.9(±19.7) 89.0(±22.2)
70-74 79.9(±28.2) 89.9(±15.6) 85.4((±22.6) 88.5(±19.7) 90.7(±19.6)
75-89 81.3(±24.4) 89.9(±16.1) 84.4(±23.1) 87.4(±20.4) 91.0(±19.0)
80-84 78.4(±26.6) 88.2((±17.7) 82.4((±24.6) 87.6(±19.6) 90.7(±20.4)
85 or older 78.4(±28.4) 88.4((±17.3) 84.3((±22.7) 87.5((±20.3) 91.1((±20.2)

Marital status
Unmarried 79.5(±26.6)

89.2(±16.8)

90.7(±19.8)

87.6(±19.9)

90.7(±19.8)

Married 79.5(±26.6)

88.9(±17.5) 89.4(±21.9) 87.3(±20.9) 89.4(±21.9)
Unknown 77.3(±28.2) 89.3(±16.4) 84.1(±22.9)

86.4(±18.9) 92(±18.9)
125 (26.9%)
326 (70.1%)

14(3.0%)

Census Tract Poverty

0% to <5% (low poverty) 78.0(±26.4) 88.8(±17.2) 84.8(±22.0) 87.4(±19.7) 90.9(±19.8)
5% to <10% 81.7(±25.6) 89.1(±17.1) 84.1(±23.7) 87.8((±19.5) 91.6((±19.4)
10% to <20% 76.3(±29.9) 89.3(±16.9) 83.6(±24.0) 87.8((±19.9) 89.5((±20.8)
20% to 100% (high
poverty)
79.7(±28.3) 89.2(±16.6) 82.8(±24.4) 86.3((±22.8) 88.3((±22.7)
Missing 70.8(±28.5) 92.0(±17.6) 74.7(±34.1) 88 (±17.0) 92.6(±17.8)

18

Education level
High school or less 79.3(±26.9) 89.1(±17.4) 83.4(±24.7) 88.0(±20.6) 89.9(±21.5)
College and higher 79.6(±26.1) 89.2(±16.4) 84.8(±21.7) 87.0(±19.8) 91.0(±18.6)
Missing 79.1(±29.4) 89.0(±17.2) 82.3(±24.8) 85.8(±20.1) 88.3(±22.6)
Time from survey to
diagnosis

Greater or equal to
median time (1396-3.8
days ~ 3.82 years)
78.0(±26.6) 89.6(±16.6) 90.3(±20.3) 83.1(±24.2) 86.2(±21.0)
Less than or equal to
median time (1396 days ~
3.82 years)

80.5(±26.8)

88.6(±17.4)

90.2(±20.7)

84.6(±22.8)

88.6(±19.5)
Cancer Stage
0 78.5(±27.6) 88.3(±17.9) 83.4(±24.1) 86.8(±19.6)
86.9(±21.0)

90.1(±20.5)
1 79.7(±27.5) 89.2(±17.2) 83.3(±23.8) 86.8(±19.6)
86.9(±21.0)

90.4(±20.0)
2 79.5(±26.1)
79.7(±27.5)

89.1(±16.9) 83.9(±23.7) 87.9(±19.7) 90.2(±20.8)
3 79.5(±26.1) 91.0(±14.5) 89.7(±17.2) 88.4(±20.6) 90.1(±21.8)
4 74.2(±30.8) 89.1(±15.4) 82.4(±24.1) 87.2(±21.4) 90.6(±18.9)

Region

West 79.4(±26.7) 88.4(±17.5) 82.8(±23.8) 86.5(±20.8) 90.4(±20.4)
Midwest 77.9(±29.4) 90.1(±16.6) 86.3(±21.9) 90.4(±17.7) 91.1(±19.9)
Northeast 78.0(±27.2) 89.0(±17.3) 84.1(±23.9) 87.0(±20.5) 89.9(±21.2)
South 81.1(±25.4) 90.4(±15.8) 85.2(±23.2) 88.7(±19.4) 89.9(±20.2)
Medicare type

FFS only 81.4(±24.2) 89.7(±16.4) 81.8(±25.3) 85.3(±21.2) 89.2(±20.6)
FFS PDP 82.9(±25.7)

87.6(±17.7) 85.2(±21.6) 89.8(±20.4) 91.8(±20.7)
MA only 73.1(±27.5) 90.0(±16.9) 83.9(±23.7) 86.2(±20.1) 88.5(±21.3)
MA PDP 72.9(±31.6) 88.5(±17.7) 85.3(±22.5) 88.5(±19.2) 92.1(±19.3)
Missing

79.5(±27.8)

90.9(±14.4) 86.0(±21.9) 89.6(±17.0) 90.7(±19.7)

19

Age at dx (N,%)

Age at Diagnosis
<65 78.5(±26.8)
(cmp_getndcare)26.
89.4(±17.4) 82.8(±24.7) 85.2(±21.5) 87.9(±21.8)
65-69 79.1(±26.80)

90.5(±15.9) 84.6(±23.2) 89.1(±19.4) 91.0(±20.1)
70-74 82.4(±26.0) 88.5(±17.1) 83.9(±23.5) 87.4(±20.3) 90.1(±21.1)
75-79 79.8(±25.1) 88.8(±17.1) 83.6(±23.8) 87.1(±20.5) 91.2(±18.6)
80-84 76.7(±29.6) 87.6(±18.2) 84.2(±23.0) 87.3(±19.7) 90.8(±21.1)
85 or older 80.7(±26.0) 89.8(±15.7) 84.7(±22.3) 89.9(±18,4) 91.3(±19.5)

Survey Mode:

Survey method
Mail 79.8(±26.4) 89.2(±16.8) 84.5(±22.5) 89.2(±16.8) 90.9(±19.4)
Phone 78..4(±28.7) 86.9(±19.8) 82.7(±25.6) 86.8(±19.4) 88.5(±23.6)
Unknown 76.8(±28.3) 89.2(±17.8) 79.5(±28.9) 89.2(±17.8) 86.3(±25.2)

Count of comorbidities
0 79.6(±27.4) 88.7(±17.4) 84.3(±23.6) 88.4(±20.2) 90.6(±20.9)
1 79.2(±25.2) 90.0(±15.8) 83.0(±23.8) 85.9(±20.1) 89.6(±20.0)
2+ 78.8(±26.5) 89.5(±17.2) 83.6(±22.9) 85.6(±20.3) 90.3(±19.0)

Survey year
1997-2000 80.8(±28.8) 86.5(±18.3) 85.5(±21.3) 91.1(±19.9) 91.2(±21.2)
2001-2005 79.2(±26.4) 89.3(±17.1) 83.4(±24.1) 86.4(±20.4) 88.8(±22.1)
2007-2011 78.8(±26.0) 90.0(±16.3) 83.7(±23.9) 86.5(±20.0) 90.9(±19.0)

Race/ethnicity:
Non-Hispanic white: 80.1(±26.2) 89.30(±16.8) 91.6(±18.6) 88.5(±18.9) 85.3(±22.4)
Non-Hispanic black 80.5(±27.1) 90.1(±16.2) 85.1(±25.9) 84.6(±24.2) 82.1(±25.1)
Hispanic 77.2(±28.7)

89.0(±17.8) 86.9(±25.1) 86.0(±23.2) 81.7(±25.5)
72 Asian 73.5(±28.8)

86.1(±18.5) 86.5(±24.2) 80.3(±23.8) 72.9(±28.5)

20

Table 2b: Means of quality of care composite variables and patient ratings by the levels of demographic variables
Bold sig, round to 10
th
add SD (bold indicates Wald’s chi-square p<0.05)

Health Care Primary Physician Specialist Physician Health Plan

Total analyzed: 4782(75.4%) 3276(51.7%) 3589(56.7%) 5618 (88.6%)

Mean score: 9.0(±`1.5) 8.4(±2.0) 9.0(±1.6) 8.6(±1.8)

Gender (M/F)

Male 8.9(±1.5) 8.3(±2.0) 9.0(±1.6) 8.4(±1.8)
Female 9.0(±1.6) 8.5(±2.0) 9.0(±1.7) 8.7(±1.8)

Age at survey

<65 8.8(±1.9)
proc sort data=steph;
by svy_mode;
run;

%macro use(z);
proc univariate data=steph;
var &z;
by svy_mode;
run;

%mend use;
%use(rate_md);
%use(rate_pdp);
%use(rate_spec);
%use(rate_plan);
quit;

7.8(±2.4) 8.6(±2.0) 8.0(±2.2)
65-69 9.0(±1.5) 8.3(±2.0) 9.0(±1.6) 8.3(±1.8)
70-74 9.0(±1.5) 8.5(±1.8) 9.0(±1.6) 8.4(±1.9)
75-89 9.0(±1.5) 8.4(±2.0) 9.0(±1.6) 8.6(±1.8)
80-84 9.0(±1.6) 8.6(±1,9) 9.0(±1.7) 8.7(±1.7)
85 or older 8.9(±1.6) 8.6(±2.0) 8.9(±1.6) 8.8(±1.7)

Marital status
Unmarried 8.98(±1.6)

8.5(±2.0)

8.545(±2.0)

8.3(±2.0)

8.4(±2.0)
8.9(±1.8)

8.6(±1.9)

Married 8.97(±1.5) 8.3(±2.0)

9.0(±1.2)

8.5(±1.7)

Unknown
8.95(±1.5)

8.4(±2.0)

8.545(±2.0)

8.3(±2.0)

8.4(±2.0)

8.8(±1.95)

8.5(±1.7)

Census Tract Poverty

0% to <5% (low poverty) 8.9(±1.6) 8.3(±2.0) 9.0(±1.6) 8.5(±1.8)
5% to <10% 9.0(±1.5) 8.5(±2.0) 8.9(±1.8) 8.6(±1.8)
10% to <20% 9.1(±1.5) 8.7(±1.9) 8.8(±1.8) 8.6(±1.9)

21
20% to 100% (high poverty) N/A 8.3 (±2.5) 9.3(±1.3) 8.8(±2.1)
Missing 8.9(±1.6) 8.3(±2.0) 9.0(±1.6) 8.5(±1.8)

Education level
High school or less 9.0(±1.6)

8.6(±2.0) 8.9(±1.8) 8.7(±1.9)
College and higher 9.0(±1.4) 8.2(±1.9) 9.0(±1.5) 8.4(±1.8)
Missing
9.0(±1.5)

8.2(±2.3)

9.0(±1.6)

8.5(±1.8)
Time from survey to diagnosis
Greater or equal to median time
(1396 days~ 3.82 years)
9.0(±1.5) 8.4(±2.0)

8.9(±1.7) 8.5(±1.8)
Less than or equal to median
time (1396 days ~ 3.82 years)
9.0(±1.6) 8.4(±2.0) 9.0(±1.6) 8.6(±1.8)
Cancer Stage
0 8.9(±1.5)

8.3(±2.1)

8.9(±1.6)

8.4(±1.9)

8.5(±1.9)

1 8.96(±1.6)

8.5(±1.9)

8.9(±1.7)

8.5(±1.9)
2 9.0(±1.5)
79.7(±27.5)

8.4(±2.1)

9.0(±1.6)

8.6(±1.8)
3 9.1(±1.5)

8.6(±1.9)

9.1(1.6)

8.7(±1.7)
4 8.9(±1.5) 8.4(±1.9) 8.9(±1.6)

8.6±2.5)

Medicare type
FFS only 9.0(±1.6) 8.5(±1.9) 8.887(±1.7) 8.6((±1.7)
FFS PDP 8.9(±1.6) N/A 9.001(±1.6)) 8.5(±1.8)
MA only 9.0(±1.6) 8.3(±2.2) 8.9(±1.7) 8.6((±1.7)
MA PDP 9.0(±1.5) N/A 9.026(±1.6) 8.7(±1.8)
Missing

9.1(±1.4) 8.5(±1.9) 9.1(±1.7) 8.6((±1.9)

22

Age at diagnosis (N,%)
<65 8.9 (±1.6) 8.1(±2.2) 8.9(±1.8) 8.2 (2.0)
65-69 8.9(±1.6) 8.4(±1.9) 9.0(±1.5) 8.6(±1.8)

70-74 9.1(±1.4) 8.4(±2.0) 9.0(1.6) 8.5(±1.8)
75-79 9.0(±1.6) 8.5(±2.0) 9.0(±1.5) 8.7(±1.7)
80-84 8.9(±1.6) 8.6(±1.9) 8.9(±1.9) 8.7(±1.7)
85 or older 9.0(±1.5) 8.7(±1.9) 8.9(±1.7) 8.8(±1.6)

Survey Mode:
Survey method )
Mail 9.0(±1.5) 8.4(±2.0) 9.0(±1.6) 8.6(±1.8)
Phone 8.7(±1.9) 8.6(±2.2) 8.7(±2.1) 8.7(±1.9)
Unknown 8.9(±1.7) 8.5(±2.1) 8.9(±1.8) 8.5(±2.0)

Count of comorbidities
0 9.0(±1.6) 8.4(±2.1) 9.0(±1.7) 8.6(±1.8)
1 9.0(±1.4) 8.5(±1.8) 8.9(±1.6) 8.5(±1.8)
2+ 8.96(±1.6) 8.4(±2.0) 8.95(±1.6) 8.5(±1.9)

Survey year
1997-2000 8.9(±1.6) N/A 9.0(±1.7) 8.6(±1.8)
2001-2005 9.0(±1.6) 8.4(±2.1) 8.9(±1.6) 8.6(±1.8)
2007-2011 9.0(±1.5) 8.5(±1.9) 9.0((±1.6) 8.5(±1.8)

Race/ethnicity:
Non-Hispanic white: 9.0(±1.5) 8.4(±2.0) 9.0(±1.6) 8.6(±1.8)
Non-Hispanic black 9.1(±1.6) 8.6(±2.0) 8.8(±2.1) 8.6(±1.9)
Hispanic 9.0(±1.6)

8.5(±2.1) 8.9(±1.8) 8.6(±1.96)
Asian 8.9(±1.5) 8.4(±2.0) 8.7(±1.4) 8.5(±1.8)
Other 9.8 (±0.4) 10.0(±0.0) 10.0(±0.0) 5.5(±1.0)

23
Table 3: Frequency of utilization counts by demographic covariates (categorically defined as ‘1’ =0 utilization, ‘2’= 1-3 utilizations, ‘3’ = 4-9
utilizations, ‘4’ = >9 utilizations) (bold indicates Wald’s chi-square p<0.05)

Visits
(visits =0)
Visits
(visits= 0-3)
Visits
(visits= 4-9)
Visits
(Visits >9)

Total (No)
(No(analyzed:

6340 3035(47.9%) 916(14.4%) 885 (14.0%) 1504 (23.7%)

Gender (M/F)

Male 1455(47.9%) 467(51.0%) 420(47.5%) 655(43.6%)
Female 1580(52.1%) 449 (49.0%) 465(52.5%) 849(56.5%)

Race

Non-Hispanic white 2153(70.9%) 672(73.4%) 732(82.7%) 1227(81.6%)
Non-Hispanic black 285(9.4%) 103(11.2%) 53(6.0%) 104(6.9%)
Hispanic 310(10.2%) 84(9.2%) 46(5.2%) 106(7.1%)
Asian 287(9.5%) 57(6.2%) 54(6.1%) 67(4.5%)

Marital status
Unmarried 1132(37.3%) 341(37.2%) 284(32.1%) 546(23.7%)
Married 1744(57.5%) 525(57.3%) 559(63.2%) 889(59.1%)
Unknown 159(5.2%) 50(5.5%) 42(4.8%) 69(4.6%)

Census Tract Poverty

0% to <5% (low
poverty)
794(26.2%) 242(26.4%) 230(26.0%) 412(24.6%)
5% to <10% 859(28,3%) 255(27.8%) 247(27.9%) 403(26.8%)
10% to <20% 864(28.5%) 227(24.8%) 238(26.9%) 429(28.5%)
20% to 100% (high
poverty)
502(16.5%) 187(20.4%) 163(18.4%) 254(16.9%)
Missing 16(0.5%) 5.0(0.5%) 7.0(0.8%) 6.0(0.4%)

24

Education level
High school or less 1679(55.3%) 501(54.7%) 474(53.6%) 760(50.5%)
College and higher 1126(37.1%) 333(36.4%) 345(38.8%) 625(41.6%)
Missing 230(7.6%) 82(9.0%) 66(7.5%) 119(7.9%)

Time from survey to
diagnosis

Greater or equal to
median time (1396-3.8
days ~ 3.82 years)
1417(46.7%) 498(54.3%) 511(57.7%) 745(49.5%)
Less than or equal to
median time (1396
days ~ 3.82 years)
1618(53.3%) 418(45.6%) 374(42.3%) 759(50.5%)

Cancer Stage
0 358(11.8%) 95 (10.4%) 106(12.0%) 150(10.0%)
1 968(31.9%) 300(32.8%) 288(32.5%) 477(31.7%)
2 1543(50.8%) 476(52.0%) 450(50.9%) 744(49.4%)
3 114(3.8%) 26(2.8%) 24(2.7%) 100(6.7%)
4 52(1.8%) 19(2.1%) 17(1.9%) 33(2.2%)

Region

West 1799(59.2%) 430(46.9%) 366(41.4%) 590(39.2%)
Midwest 230(7.6%) 89(14.4%) 105(11.9%) 195(31.5%)
Northeast 405(13.3%) 172(15.7%) 175(19.8%) 343(22.8%)
South 601(19.8%) 225(24.6%) 239(27.0%) 376(25.0%)
Medicare type
FFS only 1631(53.7%)

383(41.8%) 39(4.4%) 16(1.1%)
FFS PDP 846(27.9%) 105(11.5%) 31(3.5%) 16(1.6%)
MA only 99(3.3%) 164(17.9%) 356(40.2%) 666(44.3%)

25

MA PDP 240(13.1%) 202(22.1%) 455(51.4%) 802(53.3%)
Missing

219(7.2%) 62(6.8%) 4(0.5%) 4(0.3%)

Age at diagnosis
(N,%)

<65 586(19.3%) 187(20.4%) 183(20.7%) 262(17.4%)
65-69 659(45.3%) 207(22.6%) 213(24.1%) 296(21.5%)
70-74 497(16.4%) 147(16.1%) 131(14.8%) 260(17.3%)
75-79 621(20.5%) 182(19.9%) 169(19.1%) 313(20.8%)
80-84 471(15.5%) 129(14.1%) 127(14.4%) 260(26.3%)
85 or older 201(6.6%) 64(7.0%) 62(7.0%) 113(7.5%)

Survey Mode:
Survey method
Mail 2511(82.8%) 740(80.8%) 734(82,9%) 1236(82.2%)
Phone 147(4.8%) 24(2.6%) 24(2.7%) 37(2.5%)
Unknown 377(12.4%) 152(17.0%) 127(14.4%) 231(15.4%)

Count of
comorbidities

0 2072(68.3%) 580(63.3%) 551(62.3%) 869(57.8%)
1 673(22.2%) 229(25.0%) 245(27.7%) 371(24.7%)

2+ 290(9.6%) 107(11.7%) 89(10.1%) 264(17.6%)

Survey year
1997-2000 671(22.1%) 77(8.4%) 113(12.8%) 213(14.2%)
2001-2005 986(32.5%) 384(41.9%) 309(34.9%) 553(36.8%)
2007-2011 1378(45.4%) 455(49.7%) 463(52.3%) 738(49.1%)

26

Table 4: Multivariate ordinal logistic regression model of patient ratings and covariates including counts of visits variable (bold
indicates Wald’s chi-square p<0.05

Customer Service
Physician
communication
Getting care
quickly
Getting needed
care
Getting needed
prescription drugs

Total analyzed: 1738(27.4%) 4728(74.6%) 4786(75.5%) 4666(73.5%) 5355(84.4%)
Visits

Visits = 0
1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Visits = [0,3]
0.99 (0.75-1.31) 1.04 (0.85 – 1.26) 0.96 (0.80 -1.15) 1.04 (0.86 -1.27) 1.19 (0.98 -1.44)
Utilization [4,9] 1.33 (0.84-2.11) 0.85 (0.65 – 1.10) 0.90 (0.70 – 1.15) 1.08 (0.83 -1.40) 1.21 (0.93-1.58)
Utilization > 9
1.16 (0.75-1.80) 0.83 (0.65 -1.06) 0.99(0.78-1.24) 1.15 (0.90 -1.48) 1.33(1.04-1.72)

Gender (M/F)

Male 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Female 1.01 (0.82 -1.25) 0.90 (0.79 -1.03) 0.91 (0.80 -1.03) 1.05 (0.92- 1.19) 0.93 (0.81 – 1.06)
Age at survey
<65 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
65-69 0.85 (0.54 -1.35) 0.83 (0.60 -1.16) 0.80 (0.60 -1.08) 0.74 (0.54– 1.00) 0.80 (0.59 -1.09)
70-74 0.83 (0.49 -1.42) 0.87(0.60 -1.27) 0.73 (0.52 -1.03) 0.70 (0.50 -1.00) 0.85 (0.71 -1.19)
75-89 0.94 (0.50 – 1.75) 0.84 (0.54 -1.29) 0.80 (0.60 -1.24) 0.67 (0.44 – 1.01) 0.76 (0.50 – 1.14)
80-84 0.84 (0.40 – 1.76) 1.03 (0.62 -1.70) 0.97 (0.61 -1.54) 0.55 (0.34 –0.89) 0.78 (0.48 -1.26)
85 or older
0.61 (0.25 -1.49) 1.19 (0.66 -2.15) 0.94 (0.54-1.63) 0.53(0.30 - 0.94) 0.75 (0.42 -1.33)
Race/ethnicity
Non-Hispanic white 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Non-Hispanic black 0.81 (0.56 – 1.16) 1.02 (0.80 – 1.31) 1.24 (0.97 – 1.57) 1.13 (0.88 -1.46) 1.26 (0.99 -1.59)
Hispanic

1.19 (0.85 – 1.66) 0.93 (0.73 – 1.18) 1.13 (0.91 – 1.40) 0.91 (0.73 - 1.16) 1.20 (0.95 -1.60)
Asian 1.62 (1.08 – 2.43) 1.55 (1.21 – 1.98) 1.96 (1.54 – 2.50) 1.76(1.38 -2.24) 1.59 (1.25 -2.03)

Marital status

Unmarried 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)

27
Married
1.0 (0.80-1.33) 1.00 (0.86 -1.15) 1.03 (0.90 -1.18) 1.03 (0.89 – 1.18) 0.90 (0.78 -1.04)
Unknown
1.3 (0.83-2.06) 1.04 (0.76 - 1.41) 1.07 (0.77 - 1.42) 1.36 (1.01 -1.83) 0.71 (0.52 -0.97)

Census Tract Poverty

0% to <5% (low poverty)
1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
5% to <10% 1.00 (0.84-1.32) 0.95 (0.80 -1.13) 0.99 (0.84 – 1.16) 0.98 (0.83 -1.16) 0.85 (0.71 -1.01)
10% to <20% 1.00 (0.81-1.34) 0.96 (0.81 -1.15) 1.01 (0.85 – 1.19) 0.96 (0.81 -1.14) 1.06 (0.89 -1.26)
20% to 100% (high poverty)
1.1 (0.82-1.58) 1.05 (0.85 – 1.30) 1.07 (0.87 – 1.30) 0.91 (0.74 -1.13) 1.03 (0.83 -1.27)
Missing
0.51 (0.14-1.89) 0.77 (0.32 – 1.84) 1.45 (0.67 – 3.16) 1.15 (0.50- 2.65) 0.98 (0.41 -2.34)
Seer Region
West 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Midwest
1.08 (0.84 – 1.38) 0.83 (0.65 – 1.06) 0.84 (0.67 – 1.04) 0.75 (0.59 – 0.94) 0.99 (0.78 -1.25)
Northeast
1.31 (0.98 – 1.75) 1.03 (0.87 -1.25) 0.91 (0.76 -1.09) 1.03(0.85 – 1.23) 1.14 (0.94 – 1.38)
South
0.87 (0.67 – 1.15) 0.92 (0.77 – 1.09) 0.82 (0.69 - 0.96) 0.80 (0.68 -0.96) 1.13 (0.95 – 1.34)
Education level

High school or less 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
College and higher 1.07 (0.86 – 1.31) 1.06 (0.92 -1.21) 1.05 (0.92 -1.19) 1.20 (1.05 -1.38) 1.18 (1.04 -1.35)
Missing
0.93(0.61 -1.43) 1.02 (0.78 -1.33) 1.11(0.88 -1.41) 1.39(1.07 -1.80) 1.23 (0.91 -1.57)
Medicare type

FFS only
1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
FFS PDP 1.1 (0.8 -1.3) 1.01(0.80 – 1.27) 0.91 (0.73 – 1.14) 0.80 (0.64 -1.01) 0.49(0.34 – 0.59)
MA only
1.93 (1.32 -2.81) 1.12 (0.85 – 1.46) 0.99 (0.77 – 1.27) 0.95 (0.73 -1.23) 0.93 (0.73 -1.19)
MA PDP 1.3 (0.9 -2.1) 1.20 (0.93 – 1.54) 0.90 (0.71 – 1.13) 0.81 (0.63 -1.03) 0.42 (0.32 – 0.54)
Missing 0.9 (0.6 -1.6) 1.02 (0.74 –1.40) 0.75 (0.56 –1.02) 0.73 (0.53 – 0.99) 0.91 (0.67 – 1.23)
Survey Mode
Mail 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Phone 1.11(0.70 -1.80) 0.94 (0.66 -1.32) 0.99 (0.69 – 1.42) 1.07 (0.77 -1.49) 1.27(0.89 -1.83)
Unknown
0.97 (0.72-1.32) 0.99 (0.81 – 1.22) 1.09 (0.91 -1.32) 0.99 (0.81 -1.22) 1.17 (0.98 -1.41)

28

Time from survey to
diagnosis

Less than or equal to
median time (1396
days ~ 3.82 years)
1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Greater than or equal
to median time (1396
days ~ 3.82 years)
1.22 (0.94 – 1.59) 0.96 (0.81 -1.14) 1.05 (0.90 -1.24) 1.10 (0.90 -1.20) 0.97 (0.82 -1.14)
Cancer Stage
0 1.0 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
1 0.94 (0.67 – 1.33) 0.94 (0.76 -1.17) 1.06 (0.86 -1.30) 0.90 (0.73 -1.11) 1.04 (0.84 -1.29)
2 1.00 (0.72 -1.38) 0.93(0.75 -1.14) 0.99 (0.81 -1.20) 0.89 (0.73 -1.09) 1.01(0.82 – 1.24)
3 0.90 (0.52– 1.54) 0.72 (0.49 – 1.06) 0.75 (0.53 – 1.06) 0.79 (0.56 – 1.11) 0.98 (0.68 -1.41)
4 0.95 (0.48 – 1.90) 1.17 (0.72 – 1.88) 1.20 (0.75 – 1.88) 1.02 (0.63 -1.64) 1.06 (0.64 – 1.74)
Age at diagnosis
(N,%)

<65 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
65-69
1.17 (0.79 -1.74)
0.90 (0.69 -1.18) 0.94 (0.74 -1.20) 0.80 (0.62 -1.04)

0.72 (0.56 -0.93)
70-74 0.85 (0.49-1.45)

1.10 (0.77 -1.58) 0.93 (0.67 -1.29) 1.04 (0.76 -1.52)

0.84 (0.59 -1.18)
75-79 1.26 (0.66 -2.41) 0.97 (0.64 -1.48) 0.92 (0.62 -1.30) 1.28 (0.85 -1.93)

0.86(0.57 -1.30)
80-84 1.26 (0.58 -2.75) 0.98 (0.59 -1.62) 0.87 (0.55 -1.40) 1.47 (0.90 -2.42)

0.79 (0.48 - 1.31)
85 or older 1.46 (0.56 – 3.75) 0.70 (0.38 -1.28) 0.90 (0.52- 1.59) 1.02 (0.56 -1.84) 0.81 (0.44- 1.48)
Count of
comorbidities

0 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
1 1.08 (0.84 – 1.38) 0.99 (0.84 – 1.88) 1.17 (1.01 – 1.35) 1.20 (1.02 -1.40) 1.23(1.05 -1.43)
2+ 1.13 (0.83 – 1.54) 1.08 (0.88 –1.32) 1.14 (0.94 -1.37) 1.22 (1.01-1.49) 1.14 (0.94 -1.39)
Survey year
1997-2000 1.00 (ref) 1.00 (ref)
1.00 (ref) 1.00 (ref) 1.00 (ref)
2001-2005
1.47 (1.04 – 2.08) 0.71 (0.56 – 0.90)

Validated

0.93 (0.74 – 1.17) 1.70 (1.35 -2.14) 0.74(0.56 – 0.98)
2007-2011 1.78 (1.23 – 2.59) 0.63 (0.49 – 0.80) 0.88 (0.70 -1.11) 1.68 (1.32 -2.13) 0.68 (0.52 -0.89)

29

Health Care Primary Physician Specialist Health Plan

Total analyzed: 4782(75.4%) 3276(51.7%) 3589(56.7%) 5618 (88.6%)
Visits (counts)
Visits = 0 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Visits = [0,3] 1.02 (0.85 -1.23) 1.00 (0.81 – 1.23) 0.84 (0.69 -1.05) 1.10 (0.93 -1.31)
Visits = [4,9] 0.88 (0.69 -1.13) 1.19 (0.86 -1.66) 0.57 (0.43-0.79) 1.09 (0.87 -1.36)
Visits > 9 0.97(0.75 -1.25) 1.38 (1.01-1.89) 0.86 (0.75 -0.99) 0.96 (0.77 -1.18)
Gender M/F

Male 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Female 0.79 (0.69 -0.89) 0.86 (0.74 -1.00) 0.90 (0.70 – 1.00) 0.81 (0.72 – 0.91)
Age at survey
<65 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
65-69 0.91 (0.67 -1.23) 0.73 0.50 -1.05) 0.71 (0.51 -1.01) 0.95 (0.71 – 1.27)
70-74 0.93 (0.66 - 1.32) 0.73 0.50 -1.10) 0.62 (0.42 -0.92) 0.88(0.64 -1.22)
75-89 0.79 (0.53 -1.17) 0.73 0.50 -1.17) 0.60 (0.38 -0.94) 0.69 (0.47 – 1.00)
80-84 0.86 (0.54 -1.36) 0.69 (0.39 -1.21) 0.50 (0.29 -0.84) 0.63 (0.41 -0.97)
85 or older 1.02 (0.59 -1.75) 0.75 (0.38 – 1.46) 0.54 (0.29- 1.02) 0.56 (0.34 -0.93)

Marital status
Unmarried 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Married 0.99 (0.87-1.13) 1.17 (1.00 -1.38) 1.03(0.88- 1.21) 0.99 (0.87 -1.12)
Unknown 1.08 (0.81 -1.43) 1.04 (0.75 – 1.45) 0.99 (0.72 -1.38) 1.08 (0.84 – 1.40)

Census Tract Poverty

0% to <5% (low poverty) 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
5% to <10% 0.98 (0.84 -1.14) 0.89 (0.73 -1.10) 0.98 (0.82-1.18) 0.98 (0.84 -1.14)
10% to <20% 0.91(0.77 -1.06) 0.76 (0.62 -0.93) 0.97(0.80– 1.17) 0.87 (0.74 -1.01)
20% to 100% (high poverty) 0.92 (0.76 -1.12) 0.64 (0.51 – 0.81) 1.1 (0.91 -1.45) 0.81 (0.67 – 0.97)
Missing 0.44 (0.19 -1.01) 0.69 (0.28 -1.68) 0.56(0.22 -1.46) 0.54 (0.26 -1.13)

Education level
High school or less 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
College and higher 1.27 (1.11 – 1.43) 1.79 (1.53 -2.11) 1.27 (1.10 -1.47) 1.58 (1.40 -1.78)
Missing 1.03 (0.81 -1.31) 1.23 (0.89 -1.70) 1.18 (0.88 -1.58) 1.38 (1.10 – 1.74)

30
Time from survey to
diagnosis

Less than or equal to median
time (1396-3.8
days ~ 3.82 years)

1.00 (ref)

1.00 (ref)

1.00 (ref)

1.00 (ref)
Greater than or equal to median
time (1396 days ~ 3.82 years)
0.99 (1.11 -1.43)

0.99 (0.82 – 1.19) 1.18 (0.99 -1.42) 1.13 (0.98 -1.31)
Cancer Stage
0 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
1 0.93 (0.76 -1.13) 0.86 (0.67 -1.11) 0.96(0.76 -1.21) 0.89 (0.73 – 1.07)
2 0.91 (0.75 – 1.10) 0.9 3(0.73 -1.18) 0.78(0.62– 0.98) 0.81 (0.68 – 0.97)
3 0.79(0.56-1.11) 0.84 (0.55 -1.26) 0.75(0.52-1.07) 0.75 (0.55 -1.02)
4 1.02 (0.65 -1.63) 0.76 (0.43 – 1.34) 0.89 (0.53 -1.51) 0.69 (0.45 -1.05)

Seer Region
West 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Midwest 0.89 (0.72 -1.10) 1.07 (0.82 -1.40) 0.94(0.7 -1.20) 0.93 (0.77-1.13)
Northeast 1.01(0.85 – 1.24) 1.33 (1.05 – 1.69) 1.06 (0.87-1.29) 1.00 (0.85- 1.18)
South 0.3 (0.70- 0.97) 1.09 (0.90 -1.33) 0.85(0.71 -1.03) 0.91 (0.79 -1.06)

Medicare type

FFS only 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
FFS PDP 1.15 (0.93 -1.41) N/a 0.93 (0.73-1.19) 0.95 (0.78 – 1.16)
MA only 1.11 (0.87 -1.13) 0.90 (0.70 -1.20) 1.33(0.98 -1.80) 0.80 (0.64 -1.00)
MA PDP 1.27 (1.01 -1.60) N/a 1.21 (0.92 -1.58) 0.91 (0.74 -1.12)
Missing

0.83 (0.62 – 1.11) 1.01 (0.76 -1.33) 0.66 (0.47- 0.90) 1.04 (0.80- 1.12)
Age at diagnosis (N%)
<65 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
65-69 0.86 (0.69 – 1.13) 0.87 (0.65 – 1.17) 1.28 (0.97 – 169) 0.85 (0.68-1.07)
70-74 1.03(0.74 -1.42) 0.85 (0.57 -1.27) 1.33 (0.91 -1.93) 1.00 (0.74 -1.36)
75-79 0.87 (0.67 -1.09) 0.73 (0.45 -1.16) 1.20 (0.77 -1.86) 0.84 (0.59 -1.20)
80-84 0.89 (0.56 – 1.41) 0.65 (0.37 -1.17) 1.51 (0.88 -2.59) 1.06 (0.69 -1.64)
85 or older 0.76 (0.44 – 1.32) 0.54 (0.27 – 1.07) 1.40 (0.73 - 2.68) 1.02 (0.61 -1.71)

31

Count of comorbidities
0 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
1 1.05 (0.90 -1.22) 1.1 (1.00 -1.33) 1.17 (0.98 -1.39) 1.17 (1.02 -1.35)
2+ 1.03 (0.85 - 1.24) 1.2 (1.00 -1.51) 1.07 (0.86 -1.33) 1..29 (1.07 -1.54)

Survey year
1997-2000 1.00 (ref) N/a 1.00 (ref) 1.00 (ref)
2001-2005 0.95 (0.76 -1.17) 1.00 (ref) 0.90 (0.72 – 1.24) 1.16 (0.94 -1.42)
2007-2011 0.95 (0.91 – 1.00) 0.91 (0.78 -1.06) 0.95(0.83 -1.19) 0.95 (0.83 -1.09)

Survey Mode
Mail 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Phone 1.08 (0.79 – 1.47) 0.75 (0.34 – 1.66) 1.21 (0.84 -1.75) 0.69 (0.51 -0.92)
Unknown 1.11 (0.92 – 1.34) 0.69 (0.57 – 0.84) 0.98 (0.77 – 1.24) 0.88 (0.74 -1.04)
Race/ethnicity:
Non-Hispanic white: 1.00 (ref) 1.00 (ref) 1.00 (ref) 1.00 (ref)
Non-Hispanic black 0.75 (0.57 – 0.99) 0.78 (0.59 -1.02) 1.05 (0.77 -1.44) 0. 93(0.75-1.15)
Hispanic 0.89 (0.70 -1.14) 0.98 (0.75 -1.27) 0.85 (0.64 -1.15) 1.00 (0.82 -1.37)
Asian 1.16 (0.90 -1.49) 1.16 (0.88 -1.52) 1.71 (1.28 -2.29) 1.00 (0.81-1.43)

32
Table 5: Multivariate cox regression model stratified by race for each patient rating including visit count variable and demographic
covariates, values with a * indicate that values were significant with (p<0.05) and may include null due to rounding, **indicates
model did not converge due to small sample size

Total(used) Non-Hispanic
white
Non-Hispanic
Black
Hispanic Asian

Subjects 6340 4784 545 546 465
Events (mortality) 1205 960 99 94 52
Customer Service 1738 1.00(0.99-1.00) 1.01(0.98-1.03) 1.00(0.98-1.01) **1.28(n/a)
Doctor communication 4728 1.00(0.99-1.00) 0.99(0.98-1.01) 0.99(0.98-1.01) 1.02(0.99-1.06)
Getting Prescription
Drugs
5355 *0.99(0.99-1.00) 1.00(0.99-1.01) 1.00(0.99-1.01) 1.01(0.99-1.03)
Get Needed Care 4666 1.00(0.99-1.00) 1.00(0.99-1.01) 0.99(0.98-1.00) 1.01(.0.99-1.03)
Getting Care Quickly 4786 1.00(1.00-1.00) 1.01(0.99-1.02) 1.00(0.99-1.01) 1.01(1.00-1.03)
Health Care 4782 0.98(0.94-1.02) 0.87(0.75-1.01) 0.78(0.68-0.90) 1.35(0.95-1.94)
Primary Care Physician 3276 1.02(0.97-1.08) 0.94(0.79-1.12) 0.87(0.73-1.05) 0.71(0.53-0.95)

Specialist Physician 3589 0.95(0.90-0.99) 0.90(0.77-1.05) 0.83(0.70-0.98) 1.66(1.03-2.65)
Health Plan 5618 0.98(0.95-1.02) 0.92(0.81-1.05) 0.94(0.83-1.07) 1.30(1.00-1.69)

33
32
References:
1) Chung, Sukyung, et al. “Racial/Ethnic Differences in Reporting versus Rating of
Healthcare Experiences:” Medicine, vol. 97, no. 50, Dec. 2018, p. e13604. DOI.org
2) .Ooi SL, Martinez ME, Li CI. Disparities in breast cancer characteristics and outcomes by
race/ethnicity. Breast Cancer Res Treat. 2011;127(3):729-738.
3) Silber JH, Rosenbaum PR, Clark AS, et al. Characteristics associated with differences in
survival among black and white women with breast cancer. JAMA. 2013;310(4):389-397.
4) Hewitt M, Greenfield S, Stovall E. From Cancer Patient to Cancer Survivor: Lost in
Transition. Washington, DC2005.
5) Taplin SH, Anhang Price R, Edwards HM, et al. Introduction: Understanding and
influencing multilevel factors across the cancer care continuum. J Natl Cancer Inst
Monogr. 2012;2012(44):2-10.
6) Vaughan Sarrazin, Mary, et al. “Disparities in Inter-Hospital Helicopter Transportation
for Hispanics by Geographic Region: A Threat to Fairness in the Era of Thrombectomy.”
Journal of Stroke and Cerebrovascular Diseases, vol. 28, no. 3, Mar. 2019, pp. 550–56.
DOI.org
7) Farias, Albert J., et al. “Racial/Ethnic Differences in Patient Experiences with Health
Care in Association with Earlier Stage at Breast Cancer Diagnosis: Findings from the
SEER-CAHPS Data.” Cancer Causes & Control, 3 Jan. 2020, doi:10.1007/s10552-019-
01254-3.
8) Chawla et al. 2015. Unveiling SEER-CAHPS®: A New Data Resource for Quality of Care
Research

34

31
9) Collins, Rebecca L., et al. “What Matters Most to Whom: Racial, Ethnic, and Language
Differences in the Health Care Experiences Most Important to Patients.” Medical Care,
vol. 55, no. 11, Nov. 2017, pp. 940–47. DOI.org
10) Hays, Ron D., et al. “Measurement Equivalence of the Consumer Assessment of
Healthcare Providers and Systems (CAHPS®) Medicare Survey Items between Whites
and Asians.” Quality of Life Research, vol. 26, no. 2, Feb. 2017, pp. 311–18. DOI.org
11) Menendez, Mariano E., et al. “Patient Satisfaction in an Outpatient Hand Surgery Office:
A Comparison of English- and Spanish-Speaking Patients.” Quality Management in
Health Care, vol. 24, no. 4, 2015, pp. 183–89. DOI.org
12) Taira, D A et al. “Do patient assessments of primary care differ by patient
ethnicity?.” Health services research vol. 36,6 Pt 1 (2001): 1059-71.
13) Mayer, Lauren A., et al. “Less Use of Extreme Response Options by Asians to
Standardized Care Scenarios May Explain Some Racial/Ethnic Differences in CAHPS
Scores:” Medical Care, vol. 54, no. 1, Jan. 2016, pp. 38–44. DOI.org.
14) Lyratzopoulos G, Elliott M, Barbiere JM, et al. Understanding ethnic and other socio-
demographic differences in patient experience of primary care: evidence from the
English General Practice Patient Survey. BMJ Qual Saf 2012; 21:21–9.

15) Mead N, Roland M. Understanding why some ethnic minority patients evaluate medical
care more negatively than white patients: a cross sectional analysis of a routine patient
survey in English general practices. BMJ 2009; 339:b3450.
16) Hausmann LR, Gao S, Mor MK, et al. Understanding racial and ethnic differences in
patient experiences with outpatient health care in Veterans Affairs Medical Centers. Med
Care 2013; 51:532–9.
17) Rodriguez HP, von Glahn T, Grembowski DE, et al. Physician effects on racial and ethnic
disparities in patients’ experiences of primary care. J Gen Intern Med 2008;23:1666–72.

35

32
18) Mollica, Michelle A., et al. “Examining Colorectal Cancer Survivors’ Surveillance Patterns
and Experiences of Care: A SEER-CAHPS Study.” Cancer Causes & Control, vol. 28, no.
10, Oct. 2017, pp. 1133–41. DOI.org
19) Fenton JJ, Jerant AF, Bertakis KD, Franks P. The Cost of Satisfaction: A National Study
of Patient Satisfaction, Health Care Utilization, Expenditures, and Mortality. Arch Intern
Med. 2012;172(5):405–411. doi:10.1001/archinternmed.2011.1662
20) Hargraves JL, Hays RD, Cleary PD. Psychometric properties of the Consumer
Assessment of Health Plans Study (CAHPS) 2.0 adult core survey. Health Serv
Res. 2003;38(6, pt 1):1509-152714727785Hargraves JL, Hays RD, Cleary PD.
Psychometric properties of the Consumer Assessment of Health Plans Study (CAHPS)
2.0 adult core survey. Health Serv Res. 2003;38(6, pt 1):1509-152714727785
21) Issaka, Rachel B., et al. “Racial Disparities in Colorectal Cancer Mortality: the Role of
Endoscopy Wait-Time and Stage at Diagnosis.” Journal of Racial and Ethnic Health
Disparities, 2020, doi:10.1007/s40615-020-00721-x.

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Racial/ethnic differences in patient experiences with health care in association with the Gleason’s score level at prostate cancer diagnosis: findings from the SEER-CAHPS data

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

The impact of the COVID-19 pandemic on cancer care delivery

PDF

Association between informed decision-making and mental health-related quality of life in long term prostate cancer survivors

PDF

A comparison of three different sources of data in assessing the adolescent and young adults cancer survivors

PDF

Disparities in gallbladder, intra-hepatic bile duct, and other biliary cancers among multi-ethnic populations: a California Cancer Registry study

PDF

Factors associated with agreement between childhood cancer survivors (CCS) and their parents on knowledge and attitudes about health-related quality of life, treatment and follow-up care

PDF

Racial/ethnic variation in care preferences and care outcomes among United States hospice enrollees

PDF

Carcinogenic exposures in racial/ethnic groups

PDF

The association between Echocardiography use in the Intensive Care Unit and 28-day mortality among patients with septic shock: a systematic review

PDF

Ancestral/Ethnic variation in the epidemiology and genetic predisposition of early-onset hematologic cancers

Asset Metadata

Creator Chen, Stephen (author)

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

School Keck School of Medicine

Degree Master of Science

Degree Program Applied Biostatistics and Epidemiology

Publication Date 05/05/2020

Defense Date 05/04/2020

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

Tag cancer outcomes,colorectal cancer,Consumer Assessment of Healthcare and Providers (CAHPS),health care disparities,OAI-PMH Harvest,patient experiences

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Farias, Albert (committee chair), Hwang, Amie (committee member), Sutton, Cecilia (committee member)

Creator Email chen649@usc.edu,sstephen.chen@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-297694

Unique identifier UC11663785

Identifier etd-ChenStephe-8429.pdf (filename),usctheses-c89-297694 (legacy record id)

Legacy Identifier etd-ChenStephe-8429.pdf

Dmrecord 297694

Document Type Thesis

Rights Chen, Stephen

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