Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Cost analysis of three pharmacy counseling programs for diabetics in a health maintenance organization
(USC Thesis Other)
Cost analysis of three pharmacy counseling programs for diabetics in a health maintenance organization
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may
be from any type of computer printer.
Hie quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and im proper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand corner and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in
reduced form at the back of the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6" x 9" black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly
to order.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
A Bell & Howell Information Company
300 North Zeeb Road. Ann Arbor. M l 48106-1346 USA
313/761-4700 800/521-0600
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
COST ANALYSIS OF THREE PHARMACY COUNSELING PROGRAMS FOR
DIABETICS IN A HEALTH MAINTENANCE ORGANIZATION
by
Robert Arthur Gerber
A Thesis Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF ARTS
(Economics)
August 1996
Copyright 1996 Robert Arthur Gerber
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UMI Number: 1381584
UMI Microform 1381584
Copyright 1996, by UMI Company. Ail rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
UNIVERSITY O F SOU TH ERN CALIFORNIA
T H E GRADUATE SCHOOL.
UNIVERSITY PARK
LOS ANGELES. CALIFORNIA 80007
This thesis, w ritten by
RoberV ^r-V K yjr & e r b e r_ _
under the direction of Thesis Com m ittee,
and approved by a ll its members, has been pre
sented to and accepted by the D ean of The
Graduate School, in partial fulfillm ent of the
requirements fo r the degree of
/Ac x &Wt - o f __________
D ttx
D ate ___
THESIS COMMITTEE
hatrmax
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
Dedication
Page
iii
Acknowledgment iv
List of Tables V
List of Figures vi
Chapter One: Introduction 1
Chapter Two: Problem
Rationale 3
Kaiser Permanente Project 5
Potential Impact 6
Hypothesis 7
Chapter Three: Literature Review 8
Chapter Four: Methods
Research Population 13
Patient Identification 17
General Methods 19
Cost Information 21
Quality of Life Variables 22
Statistical Model 22
Chapter Five: Descriptive Results 29
Chapter Six: Cost-of-Dlness for Diabetes 38
Chapter Seven: Cost Analysis of Pharmaceutical Care 45
Chapter Eight: Discussion 58
Bibliography 63
Appendix 67
ii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Dedication
This thesis is dedicated to my wife, Rosalie. She has supported me in every
of the word throughout all of my endeavors.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Acknowledgment
I would not have not been able to complete this thesis without the guidance of
my advisor Professor Gordon Liu. His patience, guidance and encouragement during the
writing o f my thesis is greatly appreciated. I also want to thank the other two committee
members, Professor Jeff McCombs for his enlightenment and constant support and
Professor Jeff Nugent for his elucidating questions about my work.
iv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Tables
Page
Table 1: Percentage of New Prescriptions Filled at One Site 24
Table 2: Independent and Dependent Variables 28
Table 3: Baseline Characteristics for NIDDM Patients (Area Wide Portion) 30
Table 4: Baseline Characteristics for IDDM Patients (Area Wide Portion) 31
Table 5: Baseline characteristics for EDDM and NIDDM Patients (Randomized) 37
Table 6: OLS of Baseline Log Total Costs (Area Wide) 39
Table 7: Total Costs and Utilization for NIDDM Patients (Area Wide) 40
Table 8: Total Costs and Utilization for IDDM Patients (Area Wide) 41
Table 9: Total Costs and Utilization for IDDM Patients (Randomized) 42
Table 10: Total Costs and Utilization for NIDDM Patients (Randomized) 44
Table 11: OLS Results for IDDM & NIDDM Patients (Area Wide) 48
Table 12: OLS Results for IDDM & NIDDM Patients (Area Wide) 49
Table 13: 2SLS Results for EDDM & NIDDM Patients (Randomized Portion) 53
Table 14: OLS Results for IDDM & NEDDM Patients (Randomized Portion) 55
Table 15: 2SLS Results for EDDM & NIDDM Patients (Randomized Portion) 57
V
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
List of Figures
Figure I : Medication Profiles for Diabetic Patients (Area Wide)
Figure 2: Total Costs at Baseline for IDDM Patients (Randomized)
Figure 3: Total Costs at Baseline for NIDDM Patients (Randomized)
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter One: Introduction
Diabetes mellitus is a disease of metabolic dysregulation, most notably abnormal
glucose metabolism, accompanied by characteristic long-term complications.
Complications that are specific to diabetes include lesions of the eye (retinopathy),
kidney failure (nephropathy) and atrophy of the peripheral nerves (neuropathy).1
Patients with all forms of diabetes of sufficient duration, including Type I or insulin-
dependent diabetes mellitus (IDDM) and Type II or non-insulin-dependent diabetes
mellitus (NIDDM) are vulnerable to these complications, which cause serious morbidity.
Diabetes is also accompanied by a substantial increase in atherosclerotic disease of large
vessels, including cardiac, cerebral, and peripheral vascular disease.2 This macrovascular
atherosclerotic disease causes serious morbidity and the largest fraction of mortality
among diabetics.2
Diabetes mellitus and its complications are now the third leading cause of death
in the United States.1 Prevalence of diabetes for Americans has been reported to range
from 3.1% to 4.5%.1 - 3 Seven percent of hospital admissions are due to diabetes4 and the
mean length of hospital stay for diabetics is 6.2 days.5 Diabetes is the leading cause of
new cases of blindness; diabetics are 25 times more prone to blindness than non
diabetics.4 Diabetics also have a 17-fold increased risk of end stage renal disease, a 2-5
fold increase in myocardial infarction and 2-3 fold increase of stroke than non-diabetics.4
The total annual economic costs in 1992 for diabetes totaled $57 billion.5 Direct costs
l
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
were estimated at $37.3 billion (hospital care) and indirect costs $19.7 billion (treatment
of complication of diabetes such as kidney disease and cardiovascular disease).5
Pharmacists are in an important position to provide valuable services to diabetic
patients. Diabetic management requires a thorough education program about the
disease, medication, blood and urine testing and hygiene.1 This requires input from all
health professionals involved in the care. Poor control of diabetes is often the result of
medication error, misinterpretation of test results and ignorance of the disease.
Pharmacists provide a unique role in the management of diabetic patients because they
probably see the patient more regularly than any other health care professional and can
have a significant impact on patients’ well-being. The pharmacist’s easy access to
patients offer an opportunity to help the patients maintain proper therapeutic regimen
and provide referral, if necessary, to the physician. In addition, the pharmacist is able to
educate and reinforce crucial information to the patient.
Due to the prevalence and costs of diabetes, programs and policies that decrease
spending or increase patients' well-being would provide a great service to patients and
payers. A study examining the costs associated with pharmaceutical care for diabetic
patients would enable policy and decision makers to evaluate the pharmacist role in
treating the disease.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter Two: Problem
Rationale
Diabetes is a costly and chronic condition affecting many Americans. Research
into alternate methods for treating diabetes and preventing costly life altering
complications is needed. Many health care professionals are involved in the care of
diabetic patients. The pharmacist, one of these caregivers, can participate in providing
cost-effective care.
As discussed by Hepler and Strand6 , pharmacists’ roles have changed over the
years. In the early years of pharmacy, traditional activities such as preparing and selling
medications, dominated. Today, the roles of pharmacists have changed to provide more
direct patients care services. Pharmacists, Hepler argues, must accept their social
mandate to ensure safe and effective drug therapy.
Effective monitoring of medications is now more critical than ever because of the
vast number of medications that can cause harm. Manasse7 reviewed the literature, and
found that in 1987, approximately 12,000 deaths and 15,000 hospitalizations caused by
adverse drug reactions (ADRs) were reported to the Food and Drug Administration
(FDA). However, Manasse believes this may only reflect 10% of the actual number due
to under reporting.
3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The potential to increase morbidity and mortality can occur for different
medication related reasons; ADRs are just one “route” for medication failure. Other
causes according to Hepler and Strand are: 1) inappropriate prescribing, 2) inappropriate
delivery of the medication caused by the pharmacy being out of stock or the patient is
dispensed the wrong medication, 3) inappropriate behavior by the patient by being non
compliant with the drug regimen, 4) patient idiosyncrasy, and 5) inappropriate
monitoring. Manasse summarizes the literature by suggesting that drug therapy involves
risk and that the cost of morbidity may be substantially greater than the drug treatment
costs.
Hepler and Strand6 define pharmaceutical care as:
...the responsible provision of drug therapy for the purpose of achieving definite
outcomes that improve a patient’s quality of life. These outcomes are (1) cure of
a disease, (2) elimination or reduction of a patient symptomatology, (3) arresting
or slowing of a disease process or (4) preventing a disease or symptomatology.
To carry out the mandate in pharmacy and to ensure safe and effective care,
pharmaceutical care must be provided.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Kaiser Permanente Project
The Kaiser Permanente (KP) Pharmaceutical Care model extends the professional
services of pharmacists beyond that required by the California State Board of Pharmacy.
In addition to mandatory pharmacist counseling as governed by the State Board,
pharmacists in the KP model intervene during the course of therapy. Pharmacists review
all medications, monitor, consult with physicians, provide patient education and follow-
up. The KP model provides more than just point-of-service consultations. Pharmacists
follow a detailed outline to systematically manage the diabetic patient.
A pharmaceutical care algorithm (see Appendix A) was developed by the
Kaiser/USC team as a tool to provide optimal diabetic patient care. Pharmacists review
the medication profile for any drug interactions and ascertain if this is a new or refill
prescription from the computer system. If it is a new prescription, the pharmacist will
determine if the patient has any disease state contraindications (i.e., pregnancy, liver
disease, renal disease) or duplicate therapies. The pharmacist will also assess the
appropriateness of dosage and type of medication. If there are any problems, the
prescribing physician will be contacted.
Before dispensing a refill or new medication, the pharmacist counsels the patient
on the use of medication, signs and symptoms of problems due to toxicity, side effects,
hypoglycemia, hyperglycemia; storage requirements; and the importance of compliance.
Additionally, the pharmacist determines how the patients will monitor their glucose
levels. If there are any problems the patient will either be referred back to their primary
5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
care physician or to a diabetic educator. Patients requiring insulin will be asked if they
know the proper technique for drawing, mixing (if necessary), and administering insulin.
If the patient has not previously received proper instruction, then the pharmacist will
explain the proper technique and refer the patient to a diabetic educator or diabetic
education class, if necessary.
With a refill medication, the pharmacist determines patient compliance, whether
there were any adverse effects or if there were any other problems with the medication.
If the patient was non-compliant or experienced some type of reaction, then the
physician is notified. Depending on the circumstances, the pharmacist would recommend
the patient to go to a diabetic education class.
Potential Impact of Pharmaceutical Care on Diabetic Patients
Hepler and Strand6 developed the pharmaceutical care model as an attempt to
provide optimal medication therapy to patients. Since diabetes is a chronic condition,
there will be many encounters with pharmacists and they can guide medication therapy in
a manner to prevent a disease or symptom, to cure a disease, to arrest or slow a disease
process, and to reduce symptoms of a disease.
Pharmaceutical care for diabetes occurs on different levels depending on
individual preferences of the pharmacist and patient. The pharmacist is in a position to
influence different areas of patients medical therapy. Optimal patient care can occur by
following a protocol, monitoring patients and increasing compliance. This in turn should
6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
decrease the risk for morbidity by producing more consistent blood glucose levels, which
results in less direct costs. The potential value of examining different practices of care
for diabetics would enable pharmacists to focus their practice to best meet the needs of
the patient and society. This creates an opportunity to study and compare costs
associated with different models of pharmacy counseling with a model having greater
emphasis on pharmaceutical care.
Hypothesis
The primary objective is to examine the effects of three different pharmacy
consultation models (Control, State, Kaiser-Permanente) for patients with IDDM and
NIDDM in a managed care environment. The hypothesis is that diabetic patients who
receive consultation from the Kaiser Permanente (KP) model will have decreased total
health care costs and total drug costs compared to those patients in the State or Control
model. The second objective is to measure the cost-of-illness associated with NIDDM
and IDDM.
7
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter Three: Literature Review
Clinical research to diabetes is extensive, but research on economic issues is
limited. To date, there are no pharmacoeconomic evaluations on pharmaceutical care
interventions in the diabetes population. The economic literature on diabetes research
can be categorized by two methods: descriptive and evaluative. Descriptive studies
analyze mainly cost-of-illness and this is where the majority of research has occurred;
there has been relatively few cost-effectiveness studies. Most studies on diabetes
mellitus have not distinguished IDDM from NIDDM and many fail to include diabetes as
a secondary diagnosis.8 Internationally, there have been no reported studies on the costs
of NIDDM. In the United States, although, published data on cost-of-illness exists for
NEDDM, no literature exists for IDDM.
Studies on the cost-of-illness are the most common form of economic analysis in
the diabetic literature. Some studies only measure direct costs, while some include
indirect and intangible costs. Huse et al? estimated the total economic burden of Type
II diabetes to be $19.8 billion (1986 dollars). This amount includes $11.6 billion in
direct health care expenditures, $2.6 billion in forgone productivity related to disability
and $5.6 billion in forgone productivity related to premature mortality. They examined
all age groups and found that individuals greater than 65 years consumed more of the
health care expenditures. Health care expenditures were estimated using ICD-9 codes
for diabetes and its complications and multiplying the number of occurrences with
8
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
hospital, drug and nursing home costs. Secondary costs related to complications were
estimated by multiplying the proportion of complications related to diabetes by the
overall costs for these conditions. Direct health care expenditures were $2,206 for men
greater than 65 years and $3,073 for women greater than 65 years.
Rubin, Altman, and Mendelson1 0 estimated prevalence and annual health care
costs for diabetics and non-diabetics from the National Medical Expenditures survey.1 1
Diabetics were identified by self reports of physician diagnosis of diabetes, a history of
taking diabetic medications, or an encounter with the health care system specifically
related to diabetes. Results showed that annual health care expenditures (in 1992
dollars) were more than three times greater for diabetics ($9,943) than non-diabetics
($2,604). Total health care expenditures for diabetes were $105.2 billion while total US
healthcare expenditures in 1992 were $720.5 billion. They did not differentiate between
EDDM and NIDDM. Therefore, one of every seven healthcare dollars was spent for
diabetic patients.
Internationally, there have been a few studies examining the costs of diabetes. In
France, Triomphe, Flori, and Lanoe1 2 measured direct cost for Type I and Type II
diabetes. Type U costs were similar to non-diabetics, but Type I was higher, attributed
to higher costs of drug therapy. Annual direct costs were FFr 7,711 (1993 FFr). Similar
studies were done in Denmark and the United Kingdom.1 3 ' 1 4
In 1994, Stem and Levy1 5 , estimated the direct costs of Type I diabetes.
Estimates were done by taking an imaginary patient, in Israel, who contracted the disease
at a certain age and then followed over 35 years. Treatment costs were estimated by
9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
applying and multiplying cost information according to current guidelines by the
American Diabetic Association. Results showed that the average direct costs for treating
Type I diabetes was £ 7,100 (£ 1993).
In Argentina, Gagliardino et al.1 6 estimated the annual costs of outpatient control
and treatment of Type I and Type II diabetes by surveying patient files of sixty
diabetologists. One thousand records were included in order to reflect the overall
prescribing habits to the doctors in the community. Based on the data, they prepared an
“average” common prescribing habit for a standardized patient. Results show that Type
I and Type II patients would annually spend $1,221 and $330 (US dollars, 1992),
respectively. Complications for diabetic patients were one ketoacidosis episode, one
acute myocardial infarction, and amputation of two toes. The annual costs of these
complications result in $632, $3,415 and $1,707, respectively.
Evaluative studies examine costs of interventions and outcomes. The use of
proper cost-benefit, cost-effectiveness and cost-utility analysis on diabetes related issues
have been quite limited.1 7 There have been only a few studies evaluating procedures or
treatments in diabetes. Most of the evaluations examine strategies for identifying and/or
treating diabetic retinopathy.1 7
Overall, screening diabetic patients has been shown to be cost-effective in
reducing blindness.1 8 - 2 2 Kaplan, Hartwell, and Wilson2 3 evaluated the cost-effectiveness
of diet and exercise intervention in Type II diabetics. Costs examined were health
professional resource charges and laboratory charges. Outcomes measured were quality
of well being and well years gained. Programs to reduce obesity in Type II diabetes
1 0
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
result in a cost per well year gained of $10,870, which is comparable to other advocated
health care programs.
Scheffler, Feuchtbaum, and Phibbs2 4 analyzed cost-effectiveness o f a diabetes
pregnancy intervention program. They examined program treatment costs, savings in
hospital charges and average hospital length of stay. Only the most severe cases of
diabetes showed significant results compared to less severe cases.
There have been no cost-utility studies published for diabetes. Researchers have
examined only quality of life in diabetic patients without simultaneously examining costs.
Jacobson, De Groot, and Samson2 5 examined the effects of Type I and Type II diabetes
on patients’ perceptions of their quality of life and compared this to the properties of a
generic measurement (Rand Short Form 36) versus a diabetes specific measurement
(Diabetes Quality of Life Measure). According to the authors, quality of life is affected
by marital status for both Type I and Type II, with separated or divorced individuals
experiencing lower levels of quality of life. The severity and frequency of complications
are also associated with a lower quality of life. For Type II diabetic patients, insulin
treatment was associated with lower levels of satisfaction.
Weinberger et al.2 6 examined the relationship between glycemic control and
health related quality of life in 275 Type II patients who were monitored for one year.
The Rand SF-36 questionnaire was administered as baseline and at one year. No linear
or curvilinear relationship between glycosylation and SF-36 scores (p = 0.15) were
found, even after controlling for five covariates identified a priori.
1 1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In 1992, Nerenz2 7 assessed quality of life for two cohorts of diabetic patients
(116 and 117 patients, respectively) using the Rand SF-36. Tight glycemic control was
associated with lower rating on various SF-36 dimensions for all patients in the first
cohort and for Type I patients in the second cohort.
Mayou, Bryant, and Turner2 8 compared quality of life of Type I with Type II
patients. They examined 121 Type I and 57 Type H patients using a social difficulty
questionnaire and a mental state (Profile of Mood States) mood scale. Twenty-seven
percent of Type H patients reported considerable loss of enjoyment in social life.
Significantly fewer subjects mentioned moderate or great effects on at least one leisure
activity (33% compared with 51% of Type II) but more rated some aspect of work as
affected (46% compared with 23% of Type II group).
There are some deficiencies in the current diabetic pharmacoeconomic research.
Broadly, there are relatively few studies done in this field. Most studies estimate the cost
of diabetes without differentiating between IDDM and NIDDM. More specifically, there
are few cost analyses examining interventions in diabetes and none looking at pharmacy
counseling interventions. The greatest concentration thus far is on the cost-effectiveness
of screening and preventing diabetic retinopathy. No studies examine pharmacists’
interventions for diabetic patients. The proposed study will contribute to the diabetic
literature in two ways. First, we will provide estimates of cost-of-illness for IDDM and
NEDDM. Second, we will estimate the relative effectiveness on costs associated with
different models of pharmacy counseling.
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter Four: Methods
Research Population
The Kaiser Permanente/USC Consultation Study (for full study description see
McCombs et al.2 9 ) investigated three alternate models of patient consultation: Kaiser
Permanente (KP), State, and Control models in pharmacies throughout Southern
California. Each pharmacy model provides different levels of patient care, ranging from
the Control model, which counsels patients only when deemed necessary to the KP
model, which identifies target medications and follows specific algorithms. The target
medications have a high potential to cause an adverse event and include: digoxin, anti
diabetics, multi-dose inhalers, anti-epileptics, procainamide, quinidine, theophylline, non
steroidal anti-inflammatory (NSAIDs), warfarin and beta blockers.
The Kaiser/USC study consists of two separate studies: randomized and area
wide. Nine pharmacies participated in the randomized portion of the study and were
assigned to provide services according to one of three models: Control, State and
Kaiser-Permanente (KP). Only nine pharmacies participated in the randomized
component; this was due to the proximity of three pharmacies (known as a “triplet”)
within the same medical center campus. In total there were three “triplets." Over six
thousand patients were randomized to one of three pharmacy models in three service
areas. Patients were stratified according to age greater than eighteen years, using one
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
target medication (as defined above), having five or more medications with at least one
target medication, and having five or more medications without a target drug
The area wide portion combined six distinct geographic regions in Southern
California. Pharmacies were assigned to provide the same three pharmaceutical care
models as the randomized portion. More pharmacies were converted to participate in
the state model (n=64), opposed to the KP model (n=17), or the control model (n=17).
The study period was from March 1993 to March 1995. Data was collected
from November I, 1991 to March 30, 1995. Baseline health use data was collected for
one year (calendar year 1992) prior to study implementation. Afterwards, data was
collected during the first year of study (April 1993 to March 1994) and the second year
(April 1994 to March 1994).
Computerized data was available for patient-level information and services.
These include patient demographic data and health care utilization information (hospital
admissions, diagnosis, DRG category, clinic visits, length of stay, and prescription data).
Patient questionnaire’s include: the Rand Short Form-36, the global assessment scale, the
Moriski scale (a tool to assess compliance), patient satisfaction with the pharmacy and
knowledge about their medications.
Computerized prescription data was used to construct a chronic disease score
(CDS). CDS, developed by a multi-disciplinary group, measures the chronic disease
status of a patient. The scale scores classes of medications according to certain criteria
based on complexity of the regimen, progression of disease, and contribution to target
diseases. It has been shown to be highly correlated with itself from one year to the next,
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
reliable as a valid measure of health status, correlated with physician related disease
severity, and found to predict hospitalization and mortality.3 0 The CDS provides an
alternative to abstracting medical records, which requires time and can be difficult,
especially with a large number of patients.
Control Model
The control model emphasizes pharmacy practice prior to implementation of
mandatory pharmacist counseling. Prior to 1990, this method of counseling was
predominant. Pharmacists provide consultation only when deemed necessary or if
requested by a patient. However, in 1990, the Omnibus Reconciliation Act 1990
(OBRA 1990) mandated that pharmacists provide consultation to Medicaid patients. In
addition several states, including California, required pharmacists to counsel all patients
on every new or changed prescription. The State Board of Pharmacy provided a waiver
to continue practicing pharmacy without providing mandatory counseling.
State Model
The pharmacists in the State model follow their current normal professional
responsibilities as authorized by the California State Board of Pharmacy. The law
requires, at a minimum, pharmacists to counsel patients (either written or verbal) on all
new or changed medications. Prior to dispensing a new medication, pharmacists must
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
explain directions of use, potential hazardous side effects, storage requirements,
importance of compliance and other precautions.
KP Model
The KP model extends the professional services of pharmacists beyond the State
model. In addition to the mandatory pharmacist counseling as governed by the State
Board, pharmacists intervene during the course of therapy. Pharmacists review all
medications, monitor, consult with physicians, provide patient education and follow-up.
The KP model differs from the other two models in three significant ways. First, only
high risk patients are selected to receive KP pharmaceutical care. Second, high risk
patients receive more than just point-of-service consultations. Third, pharmacists follow
a detailed outline to properly manage the diabetic patient.
A pharmaceutical care algorithm is shown in Appendix A. When patients present
a prescription to the pharmacist for an anti-diabetic medication, the pharmacist will first
review the medication profile for any drug interactions and ascertain if this is a new or
refill prescription from the computer system. If it is a new prescription, the pharmacist
will determine if the patient has any disease state contraindications (i.e., pregnancy, liver
disease, renal disease) or a duplication in therapy. The pharmacist will also assess the
appropriateness of dosage and type of medication. If there are any problems, the
prescribing physician will be notified.
1 6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Before dispensing any medication (refill or new), the pharmacist counsels the
patient on the proper use of medication, signs and symptoms of problems due to toxicity,
side effects, hypoglycemia or hyperglycemia, the importance of compliance, and storage
requirements. Also, the pharmacist determines how the patients will monitor their
glucose levels. If there are any problems, patients will either be referred back to their
primary care physician or to a diabetic educator. Patients who require insulin will be
asked if they know the proper technique for drawing, mixing (if necessary), and
administering insulin. If the patient does not know the technique, then the pharmacist
will instruct the patients on proper technique and refer the patient to a diabetic educator
or diabetic education class.
With a refill medication, the pharmacist will determine if the patient was
compliant with the medication, whether the patient experienced any adverse effects or
had any other problems with the medication. If the patient was non-compliant or
experienced some type of reaction, then the physician was notified. Depending on the
circumstances, the pharmacist recommends that the patient go to a diabetic education
class.
Patient Identification
The study consists of four cohorts, two IDDM (area-wide and random) and two
NIDDM (area-wide and random). Patients were identified as being either IDDM or
NIDDM according to their prescription records, which identified anti-diabetic
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
medications by AHFS classification number (AHFS number 68:20). Diabetic
medications are comprised of two main therapeutic classes: insulin and oral
hypoglycemic agents.3 1 The oral hypoglycemic agents (acetohexamide, chlorpropamide,
tolazamide, tolbutamide, glipizide and glyburide) are exclusively used by NIDDM
patients; some NIDDM patients may also require insulin therapy in conjunction with oral
hypoglycemic agents.1 EDDM patients administer only insulin. There are some NEDDM
patients who have difficulty controlling their blood glucose levels and may be required to
use insulin alone. Since these patients only account for a small percentage of NIDDM
patients and the medical records were not available for complete diagnosis information,
IDDM and NIDDM patients will be separated according to medication therapy:
Type Medication
IDDM Insulin only
NIDDM Oral hypoglycemic agent +/- Insulin
To avoid overestimation of a cohort, an assessment of the entire patient
medication profile was examined. If a patient received insulin during the baseline period
and an oral agent in a subsequent period, then he would be classified as NIDDM. Thus
if any patient received any oral hypoglycemic agent during the study period, he was
classified as NIDDM. This assumes that an IDDM patient would not require any oral
hypoglycemic agent (a valid assumption). Hence, IDDM patients would have had to
receive only insulin during the study period.
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
General Methods
The study has two objectives. The first is to estimate the cost-of-illness for
NIDDM and IDDM. The second is to measure the relative impact of costs associated
with different models of pharmacy counseling. A payer’s (i.e., managed care
organization) perspective will be taken in this analysis. Only direct health care costs
associated with patient outcomes will be measured. These include costs for
hospitalization, office visits and medications. Excluded from the model are costs for the
provision of pharmaceutical services and indirect costs (i.e., loss productivity). The
different pharmacy models were approximately equal in staff and no extra personnel
were required to operate the KP model. Therefore, any cost differences in pharmacy
models small.
Faced with many interventions, a managed care organization, like Kaiser
Permanente, must allocate resources. When there is more than one intervention for a
given condition, a health planner should determine the cost-effectiveness of each
intervention. Thus, the goal is to maximize the health benefits of the population to be
treated.
There are two reasons for choosing not to do a cost-effectiveness analysis. First,
analysis by multivariate regression showed no significant effect of pharmacy consultation
on any quality of life domains. Therefore, it would be nonsensical to report a cost-
1 9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
effectiveness ratio. A planner faced with this situation should immediately suspend any
cost-effectiveness ratios and only examine costs associated with the intervention.
Second, the perspective is from the payers. Payers (it could be argued) are less
concerned with quality of life improvements than decreasing costs. Managed care
organizations seek to maximize profits. Therefore, this conjecture, coupled with no
significant findings from the quality o f life analysis, leads to focus on a cost analysis.
A cost-of-illness model estimates all the costs associated with a particular
disease. All direct health care costs associated with treatment of a diabetic patient are
included; there is no differentiation between diabetic and non-diabetic related
hospitalizations or procedures. Once average cost estimates are obtained, these will be
applied to a larger cohort (i.e., the diabetic population in the United States) to estimate
the overall costs o f diabetes in the US.
There are three basic steps in a cost analysis. These are to define the
intervention, identify relevant costs, and measure costs. By examining costs, these
evaluation methods identify the efficiency and effectiveness of programs. In addition to
their economic value, the results could also aid policy decisions. Alternative treatments
can be evaluated for their appropriateness to use valuable resources.3 2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Cost Information
Health care utilization costs include those for hospital services, physician
reimbursement for hospital services, clinic visits, self-pay (out-of-pocket) and
medications. McCombs2 9 developed the methodology to estimate costs, because Kaiser
Permanente does not assign prices to their services. Hospital costs were based on
patients admitting diagnosis related group (DRG) and estimated according to Medicare
payment rates and out-of-pocket costs set by Health Care Financing Administration
(HCFA) [in $1993 dollars].3 3
Detailed prescription files containing medication, quantity, and average wholesale
price (AWP) from individuals estimate costs. Medication costs' estimates were derived
by multiplying the AWP by the quantity of the medication. No other costs were used to
estimate drug costs, even though it is likely that a managed care pharmacy receives
rebates or price reductions from wholesalers or manufacturers. In addition, since the
study site was a managed care facility, no dispensing fee was assigned.
Office visits costs were estimated strictly from the number of visits multiplied by
an estimate of $70 per visit. Other information regarding type of visits or any other
ancillary services was not available and therefore was not included for cost estimates.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Quality of Life Variables
Quality of life was assessed using the Rand Short-Form 36 (SF-36) Health
Survey. Surveys were administered during baseline, the first year, and the second year.
The SF-36 is a standardized, 36 item questionnaire that can be completed by the patient
or can be administered by an interviewer. The questionnaire can be completed in 10 to
15 minutes and measures health across eight dimensions: physical functioning, social
functioning, physical role functioning, emotional role functioning, mental health, energy,
bodily pain and general health perceptions (for further description see appendix B). It
has been tested for reliability and validity. It has been found to have internal consistency,
consistency over time, content validity, and construct validity.3 4 Only baseline variables
will be reported. Future studies will examine any changes in the health domains and
respective cause and effect relationship.
Statistical Model
Area wide and randomized patients were analyzed separately by diabetic type
(EDDM and NEDDM). Baseline comparisons for proportions were analyzed and tested
for significance by the two-tail Fisher exact test; means were analyzed by analysis of
variance (ANOVA). Regression analysis estimated the relationship between costs and
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the independent variables. All statistical analyses were conducted using the Statistical
Analysis System (SAS).3 5
In the randomized portion, estimation of pharmacy model effects may be biased
due to patient self-selection into a pharmacy model. Pharmacies in a “triplet” were
located within the same medical campus. The models were in close enough proximity
that individuals were able to go to different pharmacies than ones originally assigned.
Table 1 shows that few patients in the randomized portion had filled all their
prescriptions at one model. Only 0.9%, 17.1%, and 6.9% of patients had all their new
prescriptions filled at the KP, State, and Control model, respectively. The higher
percentage of State filled prescriptions may be due to more pharmacies assigned to the
State model.
The Area wide sample was located in different areas throughout Los Angeles and
so individuals may not travel to different pharmacies. Results confirm this,
approximately 90% of patients had all their new prescriptions filled at one pharmacy site.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 1 - Percentage of New Prescriptions Filled at One Site
IDDM NIDDM
% of patients
that filled afi
new diabetic
RX’s at that
site
% that filled at
least 50% of
new diabetic
RX’s at that site
% that filled
afi new
diabetic
RX’s at that
site
% that filled
at least 50%
of new
diabetic
RX’s at that
site
AREA WIDE
KP
State
Control
96.6%
97.5%
87.9%
100%
100%
100%
89.3%
97.3%
78.4%
100%
100%
94.0%
RANDOMIZED
KP
State
Control
0.9%
17.1%
6.9%
11.1%
54.8%
32.3%
1.4%
24.0%
3.0%
18.3%
56.8%
23.6%
Two-Staeed Least Square
If the pharmacy consultation models are shown to be endogenous, a two-stage
least square (2SLS) model is required to provide a consistent and unbiased estimate.3 6
Formally, a general model from the regression equation is as follows:
Y, = a + p i X i + f c X z + e , ( 1 . 1)
where Xf is hypothesized to be an endogenous variable and X2 is an explanatory
exogenous variable. The first stage requires simplification of the equation and using an
2 4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
ordinary least squares (OLS), regress Xi on the exogenous variables in the model and a
dummy variable that is always unity to allow for the intercept term.3 7
The Wu-test3 8 tests the model for endogeneity and rejects the null hypothesis of
endogeneity. If the error term is significant (as measured by the t-test, assuming there is
only one error term), then there is endogeneity and the second stage is performed using
the predicted value from the endogenous variable. In this model, there was greater than
one variable hypothesized to be endogenous, therefore testing for significance requires
the F-test to be used as follows:
„ (e'rHr-e'e^lq
e'e/(N-k) ’ (L2)
where e'^ris the restricted sum of squares (residual), e'e is the unrestricted sum of
squares (residual), q is number of restrictions, and N -k are the degrees of freedom.
Independent Variables
Variables are listed in Table 2. The percentage of medications filled at each
pharmacy consultation model was the primary variable used to test the hypothesis.
Patient demographic characteristics include age, gender, ethnicity (White, Black, Asian,
American Indian, Latino), work status, marital status, smoking and drinking history, and
education. Due to low percentages of different ethnic breakdowns, a dummy variable
2 5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
was created to classify individuals as being either white or non-white. Another dummy
variable was employment status. Patients were defined as being employed if they
worked either full- or part- time; “not employed” consist of individuals who were retired,
disabled, students or housewives. Marital status was defined as either being currently
married or not. Previous marriages and widows were defined as “not married.” To be
identified as a smoker, individuals had to be currently smoking at least one cigarette per
day. Alcohol consumption was based on the number of drinks. Those who drank at
least once a month to daily were defined as drinkers; non-drinkers were those who drank
a few times in the past twelve months to not at all. Educational status of individuals was
divided among three categories: not a high school graduate, high school graduate, and
college graduate. Income was not used due to the high percentage of missing
information.
Health-related independent variables were distinguished among co-morbidities
and utilization. To identify and control for certain co-morbidities, patients who had
medications for cardiovascular disease, hypertension, hyperlipidemia, pain, and
depression were used as a proxy for co-morbidity. Additionally, the chronic disease
score was used to assess and control for other disease states. To control for previous
health care utilization, two continuous variables, number of clinic visits and number of
hospital days were used. Baseline cost variables and total number of new prescriptions
were included in the regression analysis.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Dependent Variables
Total direct costs and drug costs will be the dependent variables. Comparisons
of the three pharmaceutical consultation models will be determined by their relative
effectiveness within the model. Each dependent variable will be analyzed separately by
multivariate regression analysis to identify the quantifiable relationship between the
dependent variables and explanatory variables.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 2: Independent and Dependent Variables
Variable Description
Independent
Age Age at baseline
Married Currently married = 1, Otherwise = 0
White White = 1, Asian, Black, American Indian, Hispanic = 0
Female Female = 1, Male = 0
Highgrad High school graduate, but not a college graduate
Collgrad College graduate
Employed Current work status: 1 = full or part time, 0 = otherwise
Smoke I = currently smoke, 0 = currently does not smoke
Drink I = drinks more than once/month, 0 = drinks less than
once/month
Visit Number of Outpatient Clinic visits during one year baseline
Hospital Costs Hospital reimbursement during baseline
Drug Costs Total drug costs during baseline
STATE Assigned to State model (for area wide)
KP Assigned to KP model (for area wide)
TSTATE Percentage of new prescriptions filled at the State Model
TKP Percentage of new prescriptions filled at the KP Model
TKP (error) Residual Term for TKP
TSTATE (error) Residual Term for TSTATE
TKP (pred) Predicted value for TKP
TSTATE (pred) Predicted value for TSTATE
Tnew Drug Totai number of new prescriptions during study period
CDS Chronic disease score
CVD Patient had medications for cardiac disease
Lipid Medications for hyperlipidemia (cholestyramine, clofibrate,
colestipol, gemfibrizol, lovastatin, niacin, probucol, simvastatin)
HTN Patient had medications for hypertension
PVD Patient had medications for coronary and peripheral vascular
disease (dicumarol, heparin, pentoxifylline, warfarin)
Pain Patient had medications for pain (codeine, meperidine, fentanyl,
hydromorphone, methadone, morphine, oxycodone, darvocet)
Depression Patient had medications for depression
Health Baseline global assessment scale (range 0-100)
Dependent
LTotal Costs Natural log of total costs during study period
LTotal Drug Natural log of total drug costs during study period
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter Five: Descriptive Results
Area Wide
Overall, there were 449 diabetic patients divided among the three pharmacy
consultation models (State, n=133; KP, n=152; and Control, n=164). Seventy-three
percent of diabetics were NIDDM (n=330) and 26.5% were IDDM (n=l 19). Baseline
characteristics are shown in tables 3 and 4. The following data presented indicates some
similarities and significant differences among the groups.
Age and gender distributions were similar across cohorts. The average age was
approximately 60 years and there was a slightly higher frequency of females. The
majority of patients attended at least some college. Working status varied;
approximately one-third of NIDDM patients were retired, while the frequency for IDDM
patients ranged from 25.6% (control) to 44.7% (State). There was a significant
difference (p = 0.021) in the proportion of IDDM disabled between the State and
Control group. The majority of patients were married. There was a significant
difference (p = 0.029) in the number of married persons between the State and KP model
for NIDDM patients. Patients were predominantly White; the next largest groups were
Blacks and Latinos. No American Indians were noted in the IDDM cohort. In NIDDM
patients, there was a significant difference between Blacks in the State and KP group (p
= 0.013); differences existed for Asians in the Control and KP model (p = 0.044).
29
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 3: Baseline Characteristics for NIDDM Patients (Area Wide Portion)
State vs. State KP
vs.Control KP vs.Control
State
(n=86)
KP (n=119) Control (n=125) Fish (p) Fish (p) Fish (p)
Variable Freq % Freq % Freq %
Female 49 57 64 53.8 65 52
Grade 4 4.65 9 7.56 4 3.2
High school (some) 9 10.5 13 10.9 9 7.2
High school grad 18 20.9 24 20.2 25 20
College 35 40.7 41 34.5 54 43.2
College grad 15 17.4 15 12.6 24 19.2
College (post grad) 4 4.65 10 8.4 9 7.2
Married 54 62.8 92 77.3 88 70.4 0.029
Smoke 10 11.6 15 12.6 9 7.2
Part time 6 6.98 11 9.24 8 6.4
Full time 26 30.2 35 29.4 50 40 0.107
Retired 36 41.9 39 32.8 42 33.6
Disabled 9 10.5 16 13.5 11 8.8
Housewife 7 8.14 7 5.88 7 5.6
Student 0 0 1 0.84 2 1.6
Black 18 20.9 10 8.4 21 16.8 0.013 0.056
Indian 1 1.16 3 2.52 0 0
Asian 10 11.6 6 5.04 16 12.8 0.113 0.044
Latino 10 11.6 25 21 26 20.8 0.092 0.095
CVD 50 58.1 48 40.3 56 44.8 0.018 0.069
Lipid 9 10.5 23 19.3 15 12 0.118
HTN 54 62.8 64 53.8 76 60.8
PVD 6 6.98 6 5.04 5 4
Anxiety 13 15.1 23 19.3 20 16
COPD 18 20.9 27 22.7 17 13.6 0.069
Crohn 0 0 1 0.84 0 0
Depression 6 6.98 19 16 8 6.4 0.055 0.024
Epilepsy I 1.16 6 5.04 2 1.6
Glaucoma 0 0 3 2.52 5 4 0.081
Gout 2 2.33 4 3.36 5 4
Manic 2 2.33 0 0 1 0.8
Neoplastic 5 5.81 7 5.88 11 8.8
NSAID 22 25.6 51 42.9 45 36 0.012
Pain 21 24.4 25 21 22 17.6
Parkinsons 0 0 0 0 2 1.6
Psychiatric 1 1.16 2 1.68 1 0.8
Rheumatoid 5 5.81 1 1 9.24 10 8
Thyroid 4 4.65 7 5.88 4 3.2
Ulcer 10 11.6 16 13.5 12 9.6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 4: Baseline Characteristics for IDDM Patients (Area Wide Portion)
State vs. State KP
vs.Cont KP vs.Control
State (n=47) KP (n=33) Control (n=39) Fish (p) Fish (p) Fish (p)
Variable Freq % Freq % Freq %
Female 26 55.32 20 60.61 17 43.6
Grade I 2.13 3 9.09 3 7.7
High school 5 10.64 3 9.09 5 12.8
High school (grad) 12 25.53 9 27.27 8 20.5
College (some) 18 38.3 8 24.24 10 25.6
College (grad) 4 8.51 5 15.15 7 17.9
College (post grad) 6 12.77 3 9.09 4 10.3
Married 30 63.83 21 63.64 29 74.4
Smoke 7 14.89 6 18.18 7 17.9
Part time 6 12.77 I 3.03 4 10.3
Full time 11 23.4 14 42.42 15 38.5 0.089
Retired 21 44.68 11 33.33 10 25.6 0.076
Disabled 1 2.13 4 12.12 7 17.9 0.021
Housewife 5 10.64 0 0 1 2.6 0.074
Student 0 0 1 3.03 0 0
Black 10 21.28 5 15.15 6 15.4
Indian 0 0 0 0 0 0
Asian 3 6.38 1 3.03 3 7.7
Latino 5 10.64 8 24.24 3 7.7 0.097
CVD 21 44.68 15 45.45 19 48.7
Lipid 4 8.51 0 0 4 10.3
HTN 27 57.45 14 42.42 23 59
PVD 2.13 I 3.03 4 10.3
Anxiety 11 23.4 4 12.12 9 23.1
COPD 4 8.51 11 33.33 5 12.8 0.008 0.049
Crohn 0 0 0 0 0 0
Depression 10 21.28 3 9.09 3 7.7 0.129
Epilepsy 0 0 2 6.06 1 2.6
Glaucoma 4 8.51 3 9.09 3 7.7
Gout 0 0 0 0 3 7.7 0.089
Manic 0 0 0 0 0 0
Neoplastic 3 6.38 2 6.06 4 10.3
NSAID 16 34.04 11 33.33 16 41
Pain 12 25.53 3 9.09 5 12.8 0.084
Parkinsons 0 0 1 3.03 0 0
Psychiatric 2 4.26 0 0 0 0
Rheumatoid 5 10.64 5 15.15 4 10.3
Thyroid 5 10.64 2 6.06 1 2.6
Ulcer 4 8.51 9 27.27 3 7.7 0.033 0.054
3 1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Medication profiles at baseline are shown in figure 1. Both types of diabetic
patients have similar medication profiles. There are no differences between NIDDM and
IDDM patients. The most frequent classes observed are medications for hypertension,
CVD, inflammation, pain, anxiety, COPD, depression, and rheumatoid arthritis. Some
slightly notable differences with the medication classes o f hypertension and lipids occur
with less frequency by IDDM patients. Medications for glaucoma occur slightly less
frequently with NIDDM.
In general, both IDDM and NIDDM patients had some form o f cardiovascular
disorder (one co-morbidity associated with diabetes). Approximately 43-51% were
receiving cardiovascular drugs during baseline. Of these medications, anti-hypertensives
were seen with the greatest frequency. Significant differences existed for NIDDM
patients between the State and KP model for cardiovascular medications (p=0.018).
Additionally, there was a significant difference (p=0.024) in the amount of anti
depressant medication dispensed between the KP model (16%) and the Control model
(6.4%).
In EDDM patients, there were significant differences in the proportion of
medications for COPD between the State and KP model (p=0.008) and the KP and
Control (p=0.049). At least one-third of EDDM patients were on a non-steroidal anti
inflammatory (NSAID). The use of NSAIDs among NIDDM patients ranged from
25.6% to 42.9%, with a significant difference between the State and KP model
(p=0.012). There was a significant difference (p=0.03) among IDDM patients in the use
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of anti-ulcer medications between the State and KP model. Other noteworthy
medication patterns seen in diabetic patients are medications for anxiety and pain.
Approximately 94% of patients completed the RAND SF-36 questionnaires at
baseline. Results for quality of life domain information are shown in Appendix F. There
were no significant differences in these respects among the three pharmacy consultation
models for both IDDM and NIDDM patients. Overall general health status scores
averaged 60. There were relatively higher scores for mental health, social functioning
and role-emotional.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 1 - Medication Profiles for Diabetic Patients
(Area Wide)
60.0%
50.0%
40.0% • ■
30.0% • ■
20.0% - ■
10.0% ■ ■
0.0%
NIDDM
□ IDDM
«
M N N
3 4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Descriptive Results for Randomized Portion
The percentages of prescription filled at the three models for IDDM and NIDDM
patients are shown in table 1 . The adherence rate to any single site was not high.
Pharmacies in the triplet site were located in the same medical center; therefore a patient
could easily switch between pharmacy consultation models. IDDM patients who had all
their prescriptions filled at a single model were low: 0.9%, 17.1%, and 6.9% for the KP,
State and Control model. Results for NIDDM patients are similar, 1.4%, 24.0% and
3.0% for the KP, State, and Control model. Because patients self-select into a model of
their choice, original assignments may not be used (if they are proven to be endogenous).
Therefore, unlike the area wide sample, the randomized portion will not be separated by
original pharmacy assignment, rather it will separated by diabetic classification.
Baseline results for IDDM patients are shown in table 5. Two-hundred fifty-
three patients (31.2%) were identified as having IDDM. There were slightly more
females and the average age was 54.5 years. They were more likely to be white and have
some college education. The second largest ethnic group was Black (35.2%).
Patients averaged 8 medications and the average chronic disease score was 3.7
during the baseline period. Profiles indicate that a large percentage take medications for
several cardiovascular disorders. Anti-hypertensive agents were used by over 50% of
IDDM patients. NSAEDs and pain medications were used by over 25%, and other
medications for COPD, anxiety, rheumatoid arthritis and peptic ulcers were taken by
approximately 15% of patients.
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Five hundred fifty-nine NIDDM patients (68.8%) were randomized to one of the
pharmacy models. The average age of NIDDM patients was 58.7 years and there were
slightly more males than females. The majority had been to college and almost half were
currently employed. One-third were retired. The majority of patients were White and
30% were Black.
The average number of medications was 6.8 with an average chronic disease
score of 3.5. Like IDDM patients, cardiovascular drugs made up the largest proportion
o f medications. Specifically, 42%, 10%, and 57% of the patients received medications
for cardiovascular disease, hyperlipidemia, and hypertension, respectively. Other
medications used frequently were: anxiolytics (11.6%), respiratory (16.5%),
antidepressants (7.3%), NSAIDs (36.1%), analgesics (17.9%), rheumatoid arthritis
(9.3%), and peptic ulcer (12.5%).
3 6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 5: Baseline characteristics for EDDM and NIDDM Patients (Randomized)
Variable IDDM NIDDM
Female 54.2% 47.8%
Grade 2.8% 3.9%
High school (some) 9.9% 12.7%
High school (grad) 19.8% 20.4%
College (some) 41.5% 39.0%
College (grad) 17.8% 16.1%
College (post grad) 7.5% 7.0%
Married 60.9% 65.5%
Smoke 15.4% 15.9%
Part time 8.7% 6.4%
Full time 36.8% 41.9%
Retired 30.0% 33.6%
Disabled 10.3% 6.4%
Housewife 7.5% 7.2%
Student 1.6% 0.4%
Black 35.2% 30.6%
Indian 0.4% 2.0%
Asian 5.5% 7.2%
Latino 12.3% 16.3%
CVD 45.5% 42.4%
Lipid 13.8% 9.5%
HTN 54.2% 56.7%
PVD 5.1% 3.9%
Anxiety 15.0% 11.6%
COPD 13.4% 16.5%
Crohn 0.4% 0.2%
Depression 9.9% 7.3%
Epilepsy 4.0% 1.4%
Glaucoma 4.3% 3.4%
Gout 4.7% 3.6%
Manic 0.8% 0.2%
Neoplastic 7.1% 5.0%
NSAID 34.8% 36.1%
Pain 24.9% 17.9%
Parkinsons 1.2% 0.7%
Psychiatric 2.4% 0.9%
Rheumatoid 13.0% 9.3%
Thyroid 9.1% 6.1%
Ulcer 15.4% 12.5%
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter Six: Cost-of-Dlness Results
Area Wide
Total health care costs consisting of hospitalizations, office visits, and
medications and health care utilization data (office visits, hospital admissions and number
of hospital days) for NIDDM and IDDM patients are shown in tables 7 and 8,
respectively. The data shows non-interventional (baseline) and interventional (year 1 and
year 2) periods. Baseline cost information provides useful data for the medical costs;
treatment data provides comparisons. However, one should realize that in this section
only averaged data is included without any multivariate analysis. Further cost analysis is
presented in the following chapter.
Combining all the area wide diabetics, average total annual costs at baseline were
$3,140, $3,463 and $4,594 for Control, State and KP model, respectively. Drug costs
represented 29.5%, 32.4% and 23.8% of total costs for the Control, State and KP
model, respectively. Therefore, pharmaceuticals in the diabetic population represent
approximately one-third of total annual costs.
Differences between the pharmacy consultation models can be explained by
differences in the demographic variables. Area wide subjects were not randomized and
therefore may have different characteristics. A separate multivariate analysis (see table
6) indicates that the assigned pharmacy model at baseline did not have a significant
impact on baseline total costs.
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 6: OLS of Baseline Log Total Costs (Area Wide)
Dependent Variable: Baseline Log Total Costs
IDDM NIDDM
Variable Beta T-Value Prob> |T| Beta T-Value Prob> |T|
Intercept 6.1228 10.567 0.0001 5.9758 12.849 0.0001
Age 0.0012 0.201 0.8415 0.0073 1.358 0.1756
STATE -0.1074 -0.660 0.5115 -0.0041 -0.034 0.9732
KP 0.2705 1.574 0.1200 0.2080 1.825 0.0691
Female -0.1365 -1.041 0.3013 -0.0919 -0.886 0.3767
Employed -0.2784 -1.723 0.0893 -0.1368 -1.175 0.2412
Married -0.0075 -0.052 0.9586 0.0793 0.712 0.4770
White 0.2472 1.647 0.1039 0.1032 1.019 0.3093
High Grad -0.1660 -0.802 0.4252 -0.0234 -0.167 0.8674
College Grad 0.0355 0.157 0.8757 0.0361 0.220 0.8257
CVD -0.2251 -1.135 0.2602 0.1511 1.077 0.2823
CDS 0.2142 3.769 0.0003 0.0859 2.011 0.0454
Health 0.0015 0.467 0.6418 -.00006 -0.022 0.9824
TNewDrugs 0.0487 4.865 0.0001 0.0730 7.908 0.0001
HTN 0.2195 0.965 0.3377 0.1156 0.835 0.4046
Lipid 0.6227 2.400 0.0190 0.0228 0.151 0.8803
Pain 0.3487 1.775 0.0801 0.4678 3.515 0.0005
Depression 0.3178 1.598 0.1145 -0.3732 -2.104 0.0364
Smoke -0.2172 -1.044 0.2999 -0.0263 -0.144 0.8859
Drink 0.2493 1.540 0.1281 -0.3018 -2.022 0.0442
R2 0.7318 0.5014
Mean 7.528 7.536
F Value 10.195 13.497
N 90 274
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Baseline total costs for all NIDDM patients are shown in table 7. Costs were not
normally distributed and were highly skewed towards the right (range $112 to $38,264).
Median costs for all NIDDM patients were $1,668. Average baseline total costs were
$2,876, $3,329 and $4,835 for the Control, State and KP model, respectively. The top
10% of patients had costs greater than $9,000. There was a significant difference
(alpha=0.05) between the KP and Control groups. Drug costs comprised 31.6%, 29.6%
and 23.2%, for the Control, State and KP model respectively. The number of initial
visits varied among the three groups. There was a significant difference (alpha=0.05)
between the KP and Control group and between the KP versus State group
(alpha=0.01).
No other differences were noted across baseline utilization variables. Average
number of hospital admissions was 0.192, 0.256, and 0.378 for the Control, State, and
KP model, respectively. Average length of stay was 0.696, 1.453, 1.798 days for the
Control, State, and KP model, respectively. The number of new medications was
slightly higher in the KP group (average=9.2), but this was not significant.
Table 7: Total Costs and Utilization for NEDDM Patients (Area Wide)
Variable Median Mean Standard Dev Minimum Max
Total costs (Baseline) 1,668 3,700 5,751 112 38,264
Total drug costs 778 1,007 906 22 5,791
(Baseline)
Total costs (yrl+2) 3,493 6,975 8,393 10 61,677
Total drug costs (yrl+2) 1,453 1,908 1,734 10 11,213
Days (yrl+2) 0 2.24 5.98 0 60
Admissions (yrl+2) 0 0.51 1.0 0 8
Visits (yrl+2) 17 23.61 20.1 0 106
4 0
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Baseline total costs and utilization for IDDM patients are shown in table 8.
Costs range from a low of $391 to a high of $32,225. The top 10% of patients had total
costs greater than $10,000. Median costs were $1,872. Average annual total costs for
baseline period were $3,987, $3,713 and $3,729, for the Control, State and KP model,
respectively. Drug costs represented 24.7%, 36.8%, and 26.3%, respectively for the
three groups.
Hospital admissions were 0.33 per patient on average. Lengths of hospital stay
were similar (approximately 1.3 days) for the three groups. A significant difference
(p=.05) was noted for the number of visits. The average number in the control was
highest with 18.4 visits.
Table 8: Total Costs and Utilization for IDDM Patients (Area Wide)
Variable Median Mean Standard Dev Minimum Maximum
Total costs (Baseline) 1,872 3,808 5,165 391 32,225
Total drug costs 731 1,135 1,807 61 18,204
(Baseline)
Total costs (yrl+2) 3,744 9,071 14,542 10 106,868
Total drug costs 1,414 1,898 1,719 10 8,678
(yrl+2)
Days (yrl+2) 0 3.1 8.5 0 63
Admissions (yrl+2) 0 0.69 1.46 0 10
Visits (yrl+2) 20 27.8 28.2 0 184
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Random
Total costs at baseline for IDDM patients in the randomized portion are shown in
table 9. Results are not normally distributed; histogram results show that they are
skewed to the right (see figure 2); the range is $174 to $38,011. Five percent of patients
had costs greater than $16,000. Median costs were $1,908. Mean total costs were
approximately $4,500, of this amount 24% ($1,064) was for medications. Patients had
an outpatient visit on average, it appears, on a regular basis and had an average hospital
length of stay in the hospital of 1.32 days.
Table 9: Total Costs and Utilization for IDDM Patients (Randomized)
Variable Median Mean
Total costs (Baseline) 1,908 4,497
Total drug costs 747 1,065
(Baseline)
Total costs (yrl+2) 4,079 8,389
Total drug costs (yrl+2) 1450 1,942
Days (yrl+2) 0 2.63
Admissions (yrl+2) 0 0.83
Visit (yrl+2) 17 23.1
Standard Minimum Maximum
Dev
6,026 175 38,011
1,078 36 9,502
11,510 10 79,097
1,941 10 16,227
6.55 0 40
1.69 0 16
23.6 0 230
4 2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 2: Total costs at baseline for IDDM patients (Randomized)
160
140
120
100
80
60
40
2 0
0
$175 $5,219 $10,264 $15,309 $20,354 $25,399 $30,444 $35,489
Results for NIDDM patients are shown in table 10. Total costs are skewed to
the right with 90% of patients having costs less than $6,000 (see figure 3). The top 5%
in costs are greater than $10,000 with a maximum of $58,680. Median costs were
$1,342. Mean total costs were $2,731, which is approximately 60% less than for total
costs for EDDM patients. Total drug costs were about one-third ($947) of total costs.
Drug costs were similar for both types of diabetics. The mean number of admissions was
0.18 and the average length of stay was 0.66 days. Average number of out-patient visits
(9.8) was less than that for IDDM patients.
4 3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 3: Total costs at Baseline for NTDDM Patients (Randomized)
450
400
350
300
250
200
150
100
5 0
0 > t i >
$108 $7,747
m I — I " T '
$15,387 $23,027 $30,667 $38,307
1 l " -i — r ~ i - - - - - - - - 1 —
$45,947 $53,587
Table 10: Total Costs and Utilization for NTDDM Patients (Randomized)
Variable Median Mean
Total costs (Baseline) 1,342 2,731
Total drug costs 712 947
(Baseline)
Total costs (yrl+2) 2,984 6,576
Total drug costs (yrl+2) 1,461 1,844
Days (yrl+2) 0 1.84
Admissions (yrl+2) 0 0.47
Visit (yrl+2) 15 19.0
Standard Dev Minimum Maximum
4,716 106 58,680
929 27 10,025
14,920 10 240,888
1,614 10 10,158
8.65 0 159
1.62 0 30
17.3 0 189
4 4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter Seven: Cost Analysis Results
Area Wide
Total Costs
I next employed a statistical approach to identify the relationship between the
independent variables, total costs and drug costs. Self selection was not evident in the
area wide population by the Wu-test. The majority of patients had their prescriptions
filled entirely at one pharmacy consultation model; endogeneity test produced non
significant results for the percentage of medications filled at a pharmacy consultation
model and total medications filled.
Multivariate regression analyses were performed for overall costs and drug costs
as the dependent variables. The statistical distribution of total costs did not appear to
follow a normal distribution. However, after a logarithmic transformation, log of total
costs exhibits a well shaped normal distribution. Additionally, White’s4 0 test for
heteroscedasticity showed that the variance of the error term was constant
(homoscedastic). Baseline hospital costs, drug costs and out-patient visits were included
as independent variables to control for costs occurring during the study period. Beta
coefficients for each variable indicate not only the percentage difference in the outcome
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
related to that variable but also the direction of the difference compared to the program
serving as the control group in the model.
Ordinary least squares (OLS) results for both types of patients are shown in table
11. In the analysis of EDDM patients, there were no significant impact on total costs for
the two pharmacy consultation models compared to the control group. Total number of
all medications filled at all the models was positively correlated (P-value is about +0.028)
and highly significant at the 0.01% level.
Analysis of NIDDM patients also produced non-significant results for pharmacy
consultation model. However, other variables were found to be significant. Age is
positively correlated with total costs, as expected. The elderly have more health related
illnesses and require more care, in general. The variable for total medications had a P-
value of 0.017 and is significant at the 2% level. Another significant variable is the anti
depressants. The P-value was -0.4571 and was significant at the 1% level. Thus it
impacts costs significantly. Diabetes, in conjunction with depression, is a particularly
critical combination because of the consequences of depression. Depression can lead to
decreased health care utilization because the patient may be mentally too sick to seek
help and this causes an increased threat for long term morbidity.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
OLS results are shown in table 12. Pharmacy consultation models did not make
a significant difference in the model for both IDDM and NIDDM patients. For IDDM
patients, the KP model showed a negative trend, but it was not highly significant (p =
0.07). As before, total medications are highly significant. 3-values were 0.007 and
0.015 for IDDM and NIDDM patients, respectively; both were significant.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 11: OLS Results for IDDM & NTDDM Patients (Area Wide)
Dependent Variable: Log Total Costs (year I + year 2) for Area Wide Sample
IDDM NIDDM
Variable Beta T-Value Prob> |T| Beta T-Value Prob> |T|
Intercept 7.1454 9.640 0.0001 6.2909 13.607 0.0001
Age 0.0173 2.115 0.0381 0.0120 2.243 0.0258
STATE -0.1243 -0.576 0.5663 0.0037 0.031 0.9754
KP 0.2088 0.954 0.3434 -0.0254 -0.224 0.8229
Visits 0.0220 1.477 0.1444 0.0094 1.602 0.1104
Drug costs 0.0001 0.600 0.5505 0.0002 2.673 0.0080
Hospital costs -0.00003 -0.144 0.8859 -0.00006 -0.634 0.5267
TNewDrug 0.0288 5.499 0.0001 0.0297 8.529 0.0001
Female -0.3250 -1.911 0.0603 -0.0226 -0.220 0.8264
Employed 0.1541 0.730 0.4681 -0.1484 -1.271 0.2049
Married -0.1893 -1.024 0.3096 0.1907 1.737 0.0836
White -0.0150 -0.075 0.9405 0.0546 0.546 0.5857
High Grad -0.1236 -0.447 0.6563 0.1512 1.093 0.2754
College Grad -0.3409 -1.186 0.2398 0.1623 1.000 0.3185
CVD -0.1026 -0.394 0.6951 0.0786 0.568 0.5705
CDS -0.0742 -0.964 0.3384 -0.0474 -1.104 0.2706
Health -0.0042 -1.031 0.3060 0.0024 0.862 0.3896
HTN 0.4720 1.635 0.1067 0.2105 1.520 0.1299
Lipid 0.3303 0.845 0.4012 -0.0350 -0.239 0.8110
Pain -0.2555 -0.986 0.3275 -0.0217 -0.165 0.8692
Depression -0.0621 -0.241 0.8106 -0.4571 -2.604 0.0098
Smoke -0.4404 -1.688 0.0960 0.0289 0.159 0.8736
Drink 0.1090 0.532 0.5964 -0.0662 -0.444 0.6575
R2 0.6808 0.4916
Mean 8.502 8.3413
F Value 6.593 11.077
N 90 274
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 12: OLS Results for IDDM & NIDDM Patients (Area Wide)
Dependent Variable: Log Total Drug Costs (year 1 + year 2) for Area Wide Sample
IDDM NIDDM
Variable Beta T-Value Prob> |T| Beta T-Value Prob> |T|
Intercept 6.2308 13.890 0.0001 6.4120 17.517 0.0001
Age 0.0059 1.194 0.2365 -0.0002 -0.006 0.9955
STATE -0.2165 -1.658 0.1020 0.0566 0.585 0.5594
KP -0.2411 -1.821 0.0731 -0.1023 -1.139 0.2557
Visits -0.0003 -0.003 0.9973 0.0016 0.356 0.7220
Drug costs 0.0005 6.709 0.0001 0.0004 7.605 0.0001
Hospital costs -0.00006 -0.474 0.6373 -0.00008 -0.932 0.3522
TNewDrug 0.0071 2.250 0.0277 0.0157 5.708 0.0001
Female 0.0937 0.910 0.3658 -0.0344 -0.423 0.6727
Employed -0.1173 -0.917 0.3623 0.0297 0.321 0.7483
Married 0.0551 0.492 0.6241 0.0649 0.747 0.4556
White 0.0207 0.171 0.8649 0.1229 1.551 0.1220
High Grad -0.0322 -0.192 0.8479 -0.1148 -1.048 0.2955
College Grad 0.0195 0.112 0.9110 -0.0457 -0.356 0.7223
CVD 0.0935 0.593 0.5552 -0.0041 -0.038 0.9696
CDS 0.0091 0.197 0.8442 0.0170 0.500 0.6173
Health 0.0014 0.587 0.5594 -0.0013 -0.617 0.5381
HTN -0.0316 -0.181 0.8568 0.1418 1.293 0.1971
Lipid 0.0267 0.113 0.9102 0.0753 0.649 0.5166
Pain 0.1928 1.230 0.2230 -0.0355 -0.341 0.7336
Depression -0.1527 -0.977 0.3318 -0.1903 -1.369 0.1722
Smoke -0.2539 -1.608 0.1124 -0.0382 -0.266 0.7907
Drink 0.2259 1.823 0.0728 -0.0501 -0.425 0.6713
R2 0.7700 0.5700
Mean 7.368 7.306
F Value 10.347 15.183
N 90 274
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Randomized
Total Costs
Total costs in year I and year 2 were log linearly transformed and analyzed as the
dependent variable. After logarithmic transformation, the log of total costs exhibits a
normal distribution. Results of the Wu-Hausman test3 8 ,3 9 (shown in Appendix C)
indicate that the percentage of medications filled at the KP and State models were
endogenous. The null hypothesis of no endogeneity was rejected at the 5% significance
level. Error or residual terms for the percentages of prescriptions filled at the KP and
State model was significant and tested by the F-test. Additionally, each error term was
analyzed separately for endogeneity and found to be significant by the t-test. Total
number of medications was tested and rejected for endogeneity (p=0.0001). Because the
two variables were endogenous, the second stage of the 2SLS was performed. Results
for 2SLS are shown in table 13. Additionally, because of endogeneity, the 3 coefficients
for the KP and State model are defined as:
Percent Change In ( Total Costs)
Change in the Percentage o f Prescriptions Filled
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Predicted values for the percentages of prescriptions filled at both the KP and
State models were determined from the following explanatory variables: age, gender,
marital status, gender, education, employment, compliance, chronic disease score,
number of existing medications at baseline, and the use of medications for depression,
thyroid dysfunction, pain, hypertension, and anti-inflammation. Following the Wu-test,
the predicted values were substituted in the second stage.
IDDM patients counseled from the State model showed decreasing overall costs.
The results show that the percentage of prescriptions filled at the State model has a
significantly negative effect on total costs. State pharmacies had a much greater negative
effect than the KP or Control model. Patients who have their prescription filled at a
State model reduce costs by 7.8% (p=0.008) for a one unit increase in prescriptions filled
at the State model compared to those in the Control model.
Among NIDDM patients, however, benefits from consultation from the KP
model was revealed. Predicted values indicate that the percentage of prescriptions filled
correlates with a decrease in total costs by 21.9%. This was highly significant at the
0.01% level. The State model also was negatively correlated with a P value of -0.99 and
highly significant at the 0.01% level.
As expected, the total number of prescriptions filled across both cohorts was
positively correlated (P-value=0.05) and highly significant (p=0.0001). The total number
of medications and drug costs are key predictors for total costs. The only medication
class that was significant in influencing costs were cardiovascular medications. NIDDM
5 1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
patients taking anti-hypertensive medications show a 31% increase in total costs
(p=0.00l3).
Another notable finding is that baseline health status significantly decreases total
costs in NIDDM patients. A decrease in costs of approximately 0.8% with a
corresponding significance levels of 0.01%. Health status was not significant for IDDM
patients. Age was positive and significant across both cohorts, which is to be expected.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 13: 2SLS Results for IDDM & NIDDM Patients (Randomized Portion)
Dependent Variable: Log Total Costs (year 1 + year 2) for Randomized Sample
IDDM NIDDM
Variable Beta T-Value Prob> |T| Beta T-Value Prob> |T|
Intercept 6.9020 15.784 0.0001 6.9385 20.913 0.0001
Age 0.0165 2.225 0.0273 0.0141 3.117 0.0019
TKP (pred) -0.0944 -1.238 0.2172 -0.2191 -4.780 0.0001
TSTATE (pred) -0.0782 -2.691 0.0078 -0.0997 • -7.366 0.0001
Visits -0.0074 -1.046 0.2971 0.0180 3.484 0.0005
Drug costs 0.0001 2.276 0.0240 0.0003 8.159 0.0001
Hospital costs 0.00003 1.703 0.0902 0.00004 4.603 0.0001
TNewDrug 0.0484 7.505 0.0001 0.0549 14.323 0.0001
Female 0.0585 0.459 0.6471 0.3218 4.311 0.0001
Employed 0.1409 0.896 0.3714 0.0025 0.032 0.9743
Married 0.1586 1.159 0.2481 0.3355 3.496 0.0005
White 0.1835 1.448 0.1494 0.1015 1.417 0.1573
High Grad 0.1212 0.506 0.6135 0.0874 0.894 0.3718
College Grad -0.0930 -0.422 0.6733 -0.1585 -1.406 0.1604
CVD -0.0390 -0.173 0.8625 0.0005 0.006 0.9953
CDS -0.0008 -0.018 0.9860 -0.0388 -1.309 0.1912
Health -0.0046 -1.189 0.2360 -0.0080 -4.160 0.0001
HTN 0.2528 1.195 0.2336 0.3118 3.245 0.0013
Lipid 0.2814 1.393 0.1652 -0.0066 -0.050 0.9597
Pain 0.0965 0.480 0.6319 0.1651 1.800 0.0725
Depression -0.1695 -0.807 0.4207 -0.0940 -0.705 0.4813
Smoke 0.1739 1.026 0.3061 0.0352 0.357 0.7214
Drink -0.2648 -1.733 0.0848 -0.0062 -0.076 0.9393
R2 0.4668 0.6084
Mean 8.472 8.1533
F Value 7.560 32.202
N 212 478
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Comparing the OLS (shown in table 14) and 2SLS estimates show that reliance
on the OLS would dramatically underestimate the effect of pharmacy consultation
services. OLS estimates show an increase in total costs for NIDDM patients using the
State and KP model. Results are insignificant for the KP model, but show a positive
trend. Results from OLS in the IDDM cohort are insignificant.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 14: OLS Results for IDDM & NIDDM Patients (Randomized Portion)
Dependent Variable: Log Total Costs (year 1 + year 2) for Randomized Sample
IDDM NIDDM
Variable Beta T-Value Prob> |T| Beta T-Value Prob> |T|
Intercept 6.7275 14.908 0.0001 6.4719 20.855 0.0001
Age 0.0152 2.749 0.0066 0.0138 3.538 0.0004
TSTATE 0.0059 0.571 0.5689 0.0263 3.996 0.0001
TKP 0.0014 0.095 0.9244 0.0155 1.838 0.0668
Visits -0.0060 -0.872 0.3841 0.0018 0.393 0.6943
Drug costs 0.0001 2.225 0.0273 0.0002 5.630 0.0001
Hospital costs 0.00004 2.305 0.0223 0.00003 3.066 0.0023
TNewDrug 0.0281 4.052 0.0001 0.0178 3.521 0.0005
Female 0.0416 0.315 0.7533 0.0913 1.253 0.2108
Employed 0.0100 0.073 0.9420 -0.0687 -0.835 0.4044
Married 0.2214 1.740 0.0834 0.0393 0.526 0.5995
White 0.1135 0.890 0.3747 0.0764 1.034 0.3018
High Grad 0.1695 0.869 0.3860 0.2681 2.767 0.0059
College Grad -0.0432 -0.196 0.8446 0.1376 1.214 0.2252
CVD -0.1184 -0.662 0.5086 0.1414 1.488 0.1373
CDS -0.0195 -0.410 0.6821 -0.0243 -0.787 0.4319
Health -0.0034 -0.917 0.3604 -0.0060 -3.039 0.0025
HTN 0.1652 0.934 0.3517 0.1714 1.775 0.0765
Lipid 0.0019 0.010 0.9922 0.1257 1.043 0.2973
Pain -0.2010 -1.206 0.2294 0.1182 1.213 0.2256
Depression -0.2707 -1.260 0.2091 -0.3629 -2.674 0.0078
Smoke 0.1170 0.673 0.5018 -0.2056 -2.093 0.0369
Drink -0.1921 -1.231 0.2197 -0.0880 -1.017 0.3098
R2 0.4264 0.5620
Mean 8.4727 8.1533
F Value 6.421 26.600
N 212 478
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Drug Costs
Drug costs were log transformed and analyzed as the dependent variable.
Endogeneity was also present and found to be significant by the F-test for the percentage
of prescriptions filled at the KP and State model. Wu-test results are shown in Appendix
D. 2SLS results are shown in table 15.
2SLS results for NIDDM patients show that receiving consultation at the KP
model results in a 12.4% decrease in drug costs. These results were significant at the
2% level. State model was significant at the 0.01% level and showed a 6% decrease in
costs. 2SLS results for IDDM patients show no significant results for either the State or
KP model.
Total number of medications was highly significant across both cohorts. Each
number of new medications increased drug costs by 3%. Analysis by drug classification
indicates that anti-hypertensives and antihyperlipidemia therapy correlate significantly to
drug costs for IDDM and NIDDM patients. There is an approximate 47% increase in
drug costs for both cohorts. Antihyperlipidemic agents show a 29% and 33% increase in
drug costs for IDDM and NIDDM patients, respectively, compared to those who do not
take those medications.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 15 - 2SLS Results for IDDM & NIDDM Patients (Randomized Portion)
Dependent Variable: Log Total Drug Costs (year 1 + year 2) for Randomized Sample
IDDM NIDDM
Variable Beta T-Value Prob> |T| Beta T-Value Prob> |T|
Intercept 5.9529 20.237 0.0001 6.2192 20.985 0.0001
Age 0.0123 2.466 0.0145 0.0090 2.247 0.0251
TKP (pred) -0.0097 -0.190 0.8498 -0.1240 -3.028 0.0026
TSTATE (pred) -0.0287 -1.471 0.1429 -0.0598 -4.940 0.0001
Visits -0.0093 -1.956 0.0519 0.0018 0.401 0.6883
Drug costs 0.0003 6.796 0.0001 0.0004 10.367 0.0001
Hospital costs -.00001 -0.622 0.5347 .00002 2.340 0.0197
TNewDrug 0.0295 6.799 0.0001 0.0320 9.346 0.0001
Female -0.0339 -0.395 0.6933 0.2431 3.646 0.0003
Employed 0.0958 0.906 0.3663 0.0406 0.580 0.5625
Married 0.0361 0.393 0.6951 0.1586 1.851 0.0649
White 0.2805 3.289 0.0012 0.1122 1.754 0.0802
High Grad 0.0362 0.225 0.8221 0.0995 1.139 0.2551
College Grad -0.1219 -0.822 0.4121 -0.0585 -0.581 0.5617
CVD -0.0900 -0.594 0.5533 -0.0582 -0.711 0.4773
CDS -0.0605 -1.867 0.0635 -0.0241 -0.912 0.3625
Health -0.0020 -0.784 0.4337 -0.0060 -3.529 0.0005
HTN 0.4652 3.270 0.0013 0.4701 5.478 0.0001
Lipid 0.2904 2.137 0.0338 0.3352 2.835 0.0048
Pain -0.0916 -0.677 0.4992 0.1076 1.315 0.1893
Depression -0.1541 -1.091 0.2767 0.1180 0.991 0.3224
Smoke -0.0736 -0.646 0.5192 -0.0629 -0.714 0.4757
Drink -0.1818 -1.769 0.0786 0.0400 0.544 0.5867
R2 0.6701 0.5771
Mean 7.3262 7.2651
F Value 17.544 28.25
N 212 478
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter Eight: Discussion
Diabetes mellitus is a disease of metabolic dysregulation, most notably abnormal
glucose metabolism, accompanied by characteristic long-term complications.
Complications that are specific to diabetes include lesions of the eye (retinopathy),
kidney failure (nephropathy) and atrophy of the peripheral nerves (neuropathy).1
Patients with all forms of diabetes of sufficient duration are vulnerable to these
complications, which cause serious morbidity. Diabetes is also accompanied by a
substantial increase in atherosclerotic disease of large vessels, including cardiac, cerebral,
and peripheral vascular disease.2 This macrovascular atherosclerotic disease causes
serious morbidity and the largest fraction of mortality among diabetics.2
Diabetes mellitus and its complications are now the third leading cause of death
in the United States.1 Almost 6% of Americans have diabetes1 , 7%-8% of hospital
admissions are due to diabetes4 and the mean length of hospital stay for diabetics is 6.2
days.5 Diabetes is the leading cause of new cases of blindness; diabetics are 25 times
more prone to blindness than non-diabetics.4 Diabetics also have a 17-fold increased risk
of end stage renal disease, a 2-5 fold increase in myocardial infarction and 2-3 fold
increase of stroke than non-diabetics.4 The total annual economic costs in 1992 for
diabetes totaled $57 billion.5 Direct costs were estimated at $37.3 billion (hospital care)
and indirect costs $19.7 billion (treatment of complication of diabetes such as kidney
disease and cardiovascular disease).5
58
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Pharmacists provide care on a daily basis to diabetic patients. However, little is
known about the impact of pharmacist counseling on diabetics. Using data from Kaiser
Permanente patients, we find that patients assigned to the randomized portion self
selected into different pharmacy models. The percentage of medications filled at a
pharmacy consultation site was found to be an endogenous variable by the Wu-test.
Total health care costs and drug costs were estimated by 2SLS procedures across two
cohorts: IDDM and NIDDM patients. Total costs consist of direct medical costs
(hospitalizations, office visits and medications); excluded from the model were indirect
costs or the time costs of the pharmacists to provide services (this was assumed to be
similar for all three pharmacy models).
This study provides the first cost analysis estimates of pharmacist counseling
utilizing pharmaceutical care principles. The results show that patients who receive
counseling from the KP and State models have lower total health care costs. In the
randomized sample, pharmacist counseling in the KP model resulted in a 21.9% decrease
in total costs for each new prescription filled for NIDDM patients. The State model also
showed a decrease in costs (10%), which is lower than the KP model. This indicates that
pharmacy counseling from the KP or State model applied to the diabetic population can
reduce costs when compared to the Control model. However, a greater reduction in
costs occur with fully implemented pharmaceutical care found in the KP model, opposed
to current practices (State model). OLS results in the non-randomized portion indicate
no significant impact of pharmacy counseling. Lack of significant results could be due to
the inherent problems associated with non randomization. For example, there were
59
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
significant differences in baseline values for patients in the three groups in the area wide
sample.
In the randomized portion, drug costs were reduced by NIDDM patients using
the KP and State model. There was a 12% and 6% reduction in costs for the KP and
State model, respectively. There were no significant results in the area wide sample.
The study also provides valuable information for cost of illness modeling.
Median costs for IDDM patients during the baseline year were $1,872 (area) and $1,908
(random). For NIDDM patients, median costs were $1,668 (area) and $1,342 (random).
Cost distribution was heavily skewed to the right and there were a few high cost outliers
up to $40,000. Mean costs for IDDM patients were $3,808 (area) and $4,497 (random);
mean costs for NIDDM patients were $3,700 (area) and $2,731 (random).
Drug costs contribute in most cases about one-fourth of total costs and in some
cohorts one-third. Mean drug costs for IDDM patients were $1,135 (area) and $1,065
(random); mean drug costs for NIDDM patients were $1,007 (area) and $947 (random).
IDDM patients have greater total costs and drug costs than NIDDM patients.
Differences could be due to the increased costs of insulin. In the area wide sample, there
is a $128 difference in drug costs and a corresponding difference in total costs of $108.
In the randomized portion, drug costs differ by $188 and total costs differ by $1,766.
Therefore, in the area wide portion, drug costs could explain the differences in costs.
However, clearly in the randomized portion, other factors are contributing to overall
costs.
60
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Diabetes has been estimated to effect 3 to 4.5 % of Americans (approximately
7.8 to 11.7 million individuals).1 ’ 4 Multiplying cohort percentages by mean costs ($4,497
for IDDM and $2,731 for NIDDM), estimated total yearly direct costs would be $38.36
billion (IDDM = $16.31 billion and NIDDM = $22.05 billion).
These results are similar to those obtained by Huse9 for NIDDM projections.
Base case cost estimates were $2,206 for men greater than 65 years and $3,073 for
women greater than 65 years. Total costs for NIDDM were $11.6 billion (1986 dollars).
When costs are inflated to 1996 dollars using the medical care component of the
consumer price index, results are $21.85 billion. However, Huse included costs for
nursing home expenditures, which ours did not. Other comparisons have to be made
cautiously because o f the inclusion of different cost estimate or using indirect costs.
In addition to cost information, medication profiles provide useful insight to
other disease states accompanying diabetes. Cardiovascular medications, especially
those used for hypertension were most frequently used across cohorts. This is not
surprising since cardiovascular conditions are a common co-morbidity associated with
diabetes. Pharmacists should also be aware of medications such as those for anxiety,
COPD, depression, pain, inflammation and ulcers. If pharmacists are not able to access
patients medical record, a review of the prescription data base would be beneficial.
Knowledge of key medications could alert the pharmacist to provide additional
counseling on concurrent diseases.
Preliminary results of baseline quality of life data give useful information about
diabetics. Randomized patients mean quality of life domain scores for EDDM patients
61
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
are: 57.4 (health status), 61.6 (pain), 52.3 (general health), 71.9 (mental), 70.4
(physical), 70.0 (role-emotional), 61.8 (role-physical), 74.1 (social), and 52.7 (vitality).
Mean scores for NIDDM patients are: 60.8 (health status), 62.8 (pain), 57.9 (general
health), 72.8 (mental), 69.1 (physical), 73.0 (role-emotional), 61.3 (role-physical), 74.8
(social), and 54.6 (vitality). It would be difficult to make any comparisons to previous
quality of life studies because of different methodologies and treatment variables. For
example, Nerenz2 7 studied quality of life associated with three different levels of glucose
control.
As a policy implication, all diabetic should, at a minimum, receive counseling
from pharmacists when receiving medications. This method has been proven to reduce
total costs (i.e., the use of the State model versus Control). Even further, if program
costs are similar, than the KP model is more effective in reducing costs. Applying more
concentrated efforts on known high health care utilization patients with standard
algorithm can produce even greater results.
Pharmacists have a professional and social obligation to patients. Because
pharmacists see diabetic patients frequently, probably more than any other health care
professional, they are in an excellent position to provide care required to keep them
healthy by improving compliance, monitoring medications, and interacting with the
patients primary care physician. Pharmacists involvement and interaction with physicians
and other health care professionals would also provide other opportunities regarding
patient care.
62
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Bibliography
1 . White JR, and Campbell RK. Diabetes. In: Clinical Pharmacy and Clinical
Therapeutics, Fifth Edition; eds Herfindal ET, Gourley DR, and Hart LL, 307-332.
Williams and Wilkins; 1992.
2. Nathan, DM. Long-term Complications of Diabetes Mellitus. N Engl J Med
328:1676-1685; 1993.
3. Rubin RJ, Altman WM, and Mendelson DN. 1994. Health Care Expenditures for
People with Diabetes Mellitus. J Clin Endocrinol Metab 78(4): 809A - 809F; 1992.
4. Podolsky S. Clinical Diabetes: Modem Management. New York: Appleton-
Century-Crofts. 1980
5. American Diabetes Association. Direct and Indirect Costs o f Diabetes in the United
States in 1992. Washington, DC: The American Diabetes Association; 1992.
6. Hepler CD, and Strand LM. Opportunities and Responsibilities in Pharmaceutical
Care. Am JH osp Pharm A1:533-543; 1990.
7. Manasse HR. Medication Use in an Imperfect World: Drug Misadventuring as an
Issue of Public Policy, Part 1. Am J Hosp Pharm 46: 929-944; 1989.
8. Leese B. The Costs of Diabetes and Its Complications. Soc SciM ed 10: 1303-1310;
1992.
9. Huse DM, Oster G, Killen AR, Lacey MJ, and Colditz GA. The Economic Costs of
Non-Insulin-Dependent Diabetes Mellitus. JAMA 262(19): 2708-2713; 1989.
10. Rubin RJ, Altman WM, and Mendelson DN. 1994. Health Care Expenditures for
People with Diabetes Mellitus. J Clin Endocrinol Metab 78(4): 809A - 809F; 1992
11. Agency for Health Care Policy and Research. National Medical Expenditures
Survey. Washington, DC: US Department of Health and Human Services. 1987.
12. Triomphe A, Flori YA, and Lanoe JL. The Costs of Diabetes to Society and the
Individual. In: Diabetes in Europe, Williams R, Papoz L, Fuller J, editors. John
Libbey and CO. Ltd, London. 153-160; 1993.
13. Gerard K, Donaldson C, and Maynard AK. The Costs of Diabetes. Diabet Med 6:
164-170; 1989.
6 3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
14. Laing W, and Williams DR. Diabetes, a Model fo r Health Care Management.
Office of Health Economics, Paper no. 92, London, October; 1989.
15. Stem Z and Levy R. The Direct Cost of Type I Diabetes Mellitus in Israel. Diabetic
M ed \ 1: 528-533; 1994.
16. Gagliardino JJ, Olivera EM, Barragan H, and Puppo RA. A Simple Economic
Evaluation Model for Selecting Diabetes Health Care Strategies. Diabetic M ed 10:
351-354; 1993
17. Songer TJ. The Economic Costs of NIDDM. Diab Metab Rev 8(4): 389-404; 1992.
18. Javitt JC, Canner JK, and Sommer A. Cost-Effectiveness of Current Approaches to
the Control of Retinopathy in Type I Diabetics. Ophthalmology 96: 255-264; 1989.
19. Javitt JC, Canner JK, Frank RG, Steinwachs DM, and Sommer A. Detecting and
Treating Retinopathy in Patients with Type I Diabetes Mellitus: A Health Policy
Model. Ophthalmology 97: 483-495; 1990.
20. Dasbach EJ, Fryback DG, Newcomb PA, Klein R, and Klein BEK. Cost-
Effectiveness of Strategies for Detecting Diabetic Retinopathy. M ed Care 29: 20-39;
1991.
21. Sculpher MJ, Buxton MJ, Ferguson FA, Humphreys JE, et al. A Relative Cost-
Effectiveness Analysis of Different Methods of Screening for Diabetic Retinopathy.
Diabetic Med 8:644-650; 1991.
22. Lairson DR, Pugh JA, Kapadia AS, et al. Cost-Effectiveness of Alternative Methods
for Diabetic Retinopathy Screening. Diabetes Care 15: 1369-1377; 1992.
23. Kaplan RM, Hartwell SL, and Wilson DK. Effects of Diet and Exercise
Interventions on Control and Quality of Life in Non-Insulin-Dependent Diabetes
Mellitus. J Gen Intern M ed 2: 220-228; 1989.
24. Scheffler RM, Feuchtbaum LB and Phibbs CS. Prevention: The Cost-Effectiveness
of the California Diabetes and Pregnancy Program. Am J Public Health 82: 168-
175; 1992.
25. Jacobson AM, De Groot, M, and Samson JA. The Evaluation of Two Measures of
Quality of Life with Type I and Type II Diabetes. Diabetes Care 17(4): 267-274;
1994.
6 4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
26. Weinberger M, Kirkman MS, Samsa GP, Cowper PA, Shortliffe EA, Simel DL, and
Feussner JR. The Relationship Between Glycemic Control and Health-Related
Quality of Life in Patients with Non-Insulin-Dependent Diabetes Mellitus. Med Care
32(12): 1173-1181; 1994.
27. Nerenz DR, Repasky DP, Whitehouse FW, and Kahkonen DM. Ongoing
Assessment of Health Status in Patients with Diabetes Mellitus. M ed Care 30(5):
MSI 12-MS 124; 1992.
28. Mayou R, Bryant B, and Turner R. Quality of Life in Non-Insulin Dependent
Diabetes and a Comparison with Insulin-Dependent Diabetes. Jour Psychosom Res
34(1): 1-11; 1990.
29. McCombs JM, Cody M, Johnson K, et al. Forthcoming. Measuring the Impact of
Patient Counseling in the Outpatient Pharmacy Setting: The Research Design of the
Kaiser Permanente/USC Consultation Study. Clin Therapeutics Forthcoming.
30. Von Korff M, Wagner EH, and Saunders K. A Chronic Disease Score from
Automated Pharmacy Data. J Clin Epidemiol 45(2): 197-203; 1992.
31. Facts and Comparisons. Eds, Olin BR, Hebei SK and Dewein AC. Facts and
Comparisons, Inc., JB Lippincott Co: 129F-131; 1995.
32. Drummond MF, and Stoddart GL. Principles of Economic Evaluation of Health
Programs. World Health Stat O 38: 355-376; 1985.
33. Health Care Financing and Administration Office of Research and Demonstration.
Washington, D.C. U.S. G.P.O., 1993.
34. Brazier J, Jones N, and Kind P. Testing the Validity of the Euroqol and Comparing
It with the SF-36 Health Survey Questionnaire. Quality o f Life Research 2(3): 169-
80; 1993
35. SAS Release Version 6.08. 1992. SAS Institute INC., Cary, North Carolina
36. Greene WH. Econometric Analysis 2n d edition. Prentice Hall International.
Englewood Cliffs, New Jersey; 1990.
37. Johnston, JJ 1990. Econometric Methods 3rd edition. McGraw-Hill Inc. New York,
New York; 1984.
38. Wu DM. Alternative Tests of Independence Between Stochastic Regressors and
Disturbances. Econometrica. 41:733-750; 1973.
6 5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
39. Hausman JA. Specification Tests in Econometrics. Econometrica, 46:1251-1271;
1978.
40. White, H. A Heteroskedasticity-Consistent Covariance Matrix Estimator and a
Direct Test for Heteroskedasticity. Econometrica, 48, 817-838; 1980.
41. Medical Outcomes Trust. How to Score the SF-36 Health Survey. Boston: Medical
Outcome Survey; 1993.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix A: Diabetic Pharmaceutical Care Model
Assessment Monitoring
Considerations
Interventions Patient
Education
Review profile for drag
interactions: cimetidine.
NSAIDs, rifampin,
salicylates, thiazdes,
corticosteroids, beta-
blockers, warfarin,
gemfibrizol
Corticosteriods,
thiazides may decrease
hypoglycemic effect of
oral antidiabetics:
Alcohol, beta-blockers.
salicylates and some
NSAIDs may decrease
this effect.
If significant
interaction, call MD
with suggestion to
adjust dosage or change
agents or order follow-
up monitoring. If
interaction appears to be
minor, send
communication note to
MD and counsel patient.
In some cases, it may
appropriate to
recommend chemstrips.
diastix. etc.. for home
monitoring.
Counsel patient
regarding
potential drug
interaction and
therefore the
need for urine or
blood glucose
monitoring.
Is this a new prescription or
has patients been on this
agent before?
Was the agent well
tolerated by the
patient?
If patient complains of
significant side effects,
call MD. Perhaps a trial
on a different agent is
warranted. If side
effects appear to be
minor, inform MD via
communication note.
Instruct patient
on possible side
effects: Insulin:
hypoglycemia.
lipodystrophy.
Sulfonylureas:
heartburn.
diarrhea
(glipizide); taste
alteration
(tolbutamide):
allergic skin
reactions,
eczema,
disulfiram-Iike
rxns.
If previously on this agent,
was patient compliant?
Interview patient.
Compare patient report
to pharmacy
prescription.
Inform MD if patient is
non-compliant, either
via telephone call or
communication note
depending on assessed
severity. Dosage
adjustment and therapy
changes may be made
on the assumption that
the medication is being
taken correctly.
Instruct patient
on the
importance of
compliance,
consequences of
non-compliance,
etc. Consider
aids for
compliance if
necessary (medi-
planner,
magniguide.
etc.)
67
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix A (Continued): Diabetic Pharmaceutical Care Model
Assessment Monitoring
Considerations
Interventions Patient
Education
If new prescription, does
patient have any disease
state
precautions/contraindicatio
ns (hypersensitily,
pregnancy, liver disease,
renal disease) to its use?
Insulin - N/A
Sulfonylureas should
be avoided in
pregnancy due to
potential
teratogenicity. Patients
with hepatic
dysfunction may have
an exaggerated
response to
sulfonylureas
Call MD. Suggest
alternative treatment.
Is patient using home blood
or urine glucose monitoring
and dosage titration?
Interview patient. If patient unsure, call
MD to verify and to
assure that patient has
proper equipment
(accucheck, etc.) and
knowledge to perform
home monitoring.
Instruct patient
regarding signs
and symptoms of
hypo-and
hyper- glycemia.
importance of
compliance,
techniques for
injection of
insulin, storage,
potential adverse
drug reactions,
technique,
interpretation
and importance
of home blood
sugar
monitoring.
Is dosage appropriate? See dosage guidelines. If dosage appears to be
abnormal, call MD and
verify and perhaps
suggest alternative
regimen.
If prescription is for insulin,
has patient been instructed
on how to draw insulin into
syringe and perform
injection?
Interview patient. Instruct patient
and demonstrate
process.
Is patient going to be
mixing insulins?
Interview patient. Instruct patient
and demonstrate
process.
6 8
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix B: Summary of SF-36 Adapted (adapted from Medical Outcomes
Trust)4 1
Dimension Summary
Physical Functioning Extent to which health limits physical activities such as self-
care, walking, climbing stairs, bending, lifting an moderate and
vigorous exercise
Extent to which physical health interferes with work or other
daily activities, including accomplishing less than wanted,
limitations in the kind of activities, or difficulty in performing
activities.
Intensity of pain and effect of pain on normal work, both
inside and outside the home.
Personal evaluation of health, including current health, health
outlook and resistance to illness.
Feeling energetic and full of life versus feeling tired and worn
out.
Extent to which physical health or emotional problems
interfere with normal social activities.
Extent to which emotional problems interfere with work or
other daily activities, including decreased time spent on
activities, accomplishing less, and not working as carefully as
usual.
General mental health, including depression, anxiety,
behavioral-emotional control and general positive effect.
6 9
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Role Functioning -
physical
Bodily pain
General health
Vitality
Social functioning
Role-functioning -
emotional
Mental health
Appendix C - WU test results Randomized (WDM & NIDDM)
Dependent Variable: Log Total Costs (year 1 + year 2)
IDDM NIDDM
Variable Beta T-Value Prob> |T| Beta T-Value Prob> |T|
Intercept 6.8965 15.676 0.0001 6.9127 20.573 0.0001
Age 0.0166 2.227 0.0272 0.0143 3.151 0.0017
TSTATE -0.0821 -2.535 0.0121 -0.0934 -4.864 0.0001
TKP -0.0991 -1.263 0.2081 -0.2142 -4.567 0.0001
TNewDrug 0.0501 5.633 0.0001 0.0523 7.597 0.0001
TKP (error) 0.0967 1.238 0.2172 0.2075 4.491 0.0001
STATE (error) 0.0792 2.686 0.0079 0.0993 6.504 0.0001
Visits -0.0077 -1.069 0.2864 0.0180 3.492 0.0005
Drug costs 0.0001 2.230 0.0269 0.0003 8.128 0.0001
Hospital costs 0.00003 1.679 0.0949 0.00003 4.364 0.0001
Female 0.0556 0.433 0.6657 0.3164 4.203 0.0001
Employed 0.1454 0.915 0.3612 0.0003 0.005 0.9963
Married 0.1561 1.133 0.2589 0.3311 3.443 0.0006
White 0.1873 1.462 0.1454 0.0991 1.381 0.1679
High Grad 0.1210 0.502 0.6160 0.0962 0.969 0.3333
College Grad -0.0914 -0.413 0.6804 -0.1458 -1.259 0.2086
CVD -0.0381 -0.168 0.8664 0.0060 0.066 0.9477
CDS -0.0015 -0.031 0.9750 -0.0373 -1.254 0.2105
Health -0.0046 -1.173 0.2421 -0.0080 -4.168 0.0001
HTN 0.2577 1.208 0.2286 0.3099 3.229 0.0013
Lipid 0.2943 1.413 0.1592 -0.0119 -0.090 0.9281
Pain 0.1054 0.515 0.6071 0.1705 1.847 0.0654
Depression -0.1722 -0.815 0.4161 -0.0961 -0.722 0.4708
Smoke 0.1764 1.035 0.3022 0.0292 0.295 0.7684
Drink -0.2690 -1.743 0.0830 -0.0081 -0.099 0.9209
R2 0.4670 0.6115
Mean 8.473 8.153
F Value 6.863 29.772
N 212 478
WU F value 7.14* 28.92*
* The critical F(2, qo) at the 5% level is 3.0
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix D - WU test results Randomized (IDDM & NIDDM)
Dependent Variable: Log Total Drug Costs (year 1 + year 2)
IDDM NIDDM
Variable Beta T-Value Prob> |T| Beta T-Value Prob> |T|
Intercept 5.9456 20.169 0.0001 6.2342 20.710 0.0001
Age 0.0125 2.493 0.0135 0.0089 2.203 0.0281
TSTATE -0.0339 -1.562 0.1199 -0.0634 -3.686 0.0003
TKP -0.015 -0.304 0.7616 -0.1268 -3.018 0.0027
TNewDrug 0.0318 5.332 0.0001 0.0335 5.434 0.0001
TKP (error) 0.0223 0.427 0.6697 0.1294 3.125 0.0019
STATE (error) 0.0288 1.458 0.1465 0.0604 4.415 0.0001
Visits -0.0097 -2.015 0.0453 0.0018 0.399 0.6898
Drug costs 0.0003 6.679 0.0001 0.0004 10.374 0.0001
Hospital costs -.00001 -0.642 0.5214 .00002 2.346 0.0194
Female -0.0379 -0.440 0.6604 0.2462 3.650 0.0003
Employed 0.1015 0.954 0.3414 0.0418 0.596 0.5518
Married 0.0332 0.360 0.7191 0.1611 1.870 0.0621
White 0.2856 3.327 0.0011 0.1137 1.768 0.0778
High Grad 0.0360 0.223 0.8237 0.0944 1.061 0.2893
College Grad -0.1190 -0.802 0.4238 -0.0658 -0.635 0.5260
CVD -0.0885 -0.584 0.5601 -0.0613 -0.743 0.4576
CDS -0.0610 -1.878 0.0619 -0.0250 -0.939 0.3484
Health -0.0020 -0.770 0.4425 -0.0060 -3.525 0.0005
HTN 0.4720 3.302 0.0011 0.4712 5.480 0.0001
Lipid 0.3085 2.211 0.0282 0.3384 2.844 0.0047
Pain -0.0797 -0.582 0.5615 0.1045 1.264 0.2070
Depression -0.1572 -1.110 0.2684 0.1192 0.999 0.3184
Smoke -0.0705 -0.617 0.5379 -0.0595 -0.669 0.5040
Drink -0.1871 -1.809 0.0720 0.0411 0.557 0.5777
R2 0.6729 0.5780
Mean 7.3260 7.26509
F Value 16.112 25.907
N 212 478
Wu F Value 3.35* 13.52*
* The critical F(2, oo) at the 5% level is 3.0
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix F - OLS test results Randomized (IDDM & NIDDM)
Dependent Variable: Log Total DRUG Costs (year 1 + year 2) for Randomized Sample
IDDM NIDDM
Variable Beta T-Value Prob> |T| Beta T-Value Prob> |T|
Intercept 5.8876 20.013 0.0001 5.9731 22.161 0.0001
Age 0.0126 3.519 0.0005 0.0085 2.508 0.0125
TSTATE -0.0026 -0.333 0.7395 0.0094 1.646 0.1004
TKP 0.0078 0.780 0.4363 0.0160 2.184 0.0295
Visits -0.0088 -1.966 0.0507 -0.0082 -2.045 0.0414
Drug costs 0.0003 6.692 0.0001 0.0003 8.937 0.0001
Hospital costs -.00001 -0.801 0.4243 .00001 1.498 0.1349
TNewDrug 0.0245 5.427 0.0001 0.0124 2.845 0.0046
Female -0.0439 -0.510 0.6104 0.1078 1.703 0.0893
Employed 0.0466 0.521 0.6028 -0.0007 -0.011 0.9912
Married 0.0624 0.753 0.4523 -0.0208 -0.320 0.7490
White 0.2648 3.183 0.0017 0.1007 1.569 0.1174
High Grad 0.0327 0.258 0.7969 0.1979 2.352 0.0191
College Grad -0.1106 -0.771 0.4419 0.1064 1.082 0.2800
CVD -0.1336 -1.146 0.2531 0.0209 0.254 0.7998
CDS -0.0651 -2.099 0.0371 -0.0167 -0.623 0.5334
Health -0.0018 -0.761 0.4479 -0.0048 -2.804 0.0053
HTN 0.4254 3.687 0.0003 0.3849 4.588 0.0001
Lipid 0.2095 1.620 0.107 0.4266 4.078 0.0001
Pain -0.1932 -1.777 0.0771 0.0727 0.860 0.3902
Depression -0.1929 -1.378 0.1698 -0.0448 -0.380 0.7039
Smoke -0.0874 -0.771 0.4416 -0.2045 -2.398 0.0169
Drink -0.1648 -1.621 0.1067 -0.0079 -0.106 0.9157
R2 0.6668 0.5528
Mean 7.32620 7.2651
F Value 17.281 25.626
N 212 478
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Appendix F: QUALITY OF LIFE Variables (Area Wide)
IDDM
Control State Model Kaiser Model F
Value
Pr>F
Variable Mean Std
Dev
Mean Std Dev Mean Std Dev
Health status 59.71 22.02 54.80 19.99 55.93 21.84 0.54 0.584
Pain 58.39 26.59 63.02 24.46 61.80 30.70 0.31 0.735
General health 47.32 24.31 47.47 25.30 58.89 24.73 2.40 0.095
Mental Health 70.00 17.91 66.98 21.07 71.74 17.19 0.62 0.540
Physical 64.77 31.32 60.78 35.67 64.67 33.30 0.18 0.833
Role - Emotional 74.29 41.29 75.00 39.46 71.11 41.74 0.09 0.917
Role - Physical 56.62 43.65 52.84 42.87 50.52 42.33 0.17 0.843
Social 71.62 31.68 67.93 33.61 67.97 33.44 0.16 0.856
Vitality 51.44 21.18 50.91 23.27 52.92 21.98 0.08 0.924
NIDDM
Control State Kaiser F
Value
Pr>F
Variable Mean Std Dev Mean Std Dev Mean Std Dev
Health status 61.45 18.58 57.71 18.55 59.32 19.47 0.99 0.371
Pain 63.84 28.35 64.27 27.89 58.63 29.39 1.30 0.274
General Health 59.44 22.86 56.54 22.71 54.56 21.48 1.41 0.247
Mental Health 73.12 18.73 74.48 18.60 69.62 18.57 1.85 0.159
Physical 66.66 27.70 63.19 33.08 61.20 29.65 1.01 0.366
Role - Emotional 73.50 39.96 66.87 43.12 67.54 40.76 0.88 0.418
Role - Physical 62.67 41.97 54.07 45.16 51.32 44.73 2.12 0.122
Social 74.09 30.29 71.22 32.43 71.21 31.11 0.33 0.722
Vitality 55.46 20.12 55.30 20.66 50.63 22.03 1.89 0.153
7 3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Effects of a formulary expansion of the use of SSRIs and health care services by depressed patients in the California Medicaid program
PDF
Assessment of prognostic comorbidity in hospital outcomes research: Is there a role for outpatient pharmacy data?
PDF
Controlling for biases from measurement errors in health outcomes research: A structural equation modeling approach
PDF
Compliance study of second-generation antipsychotics on patients with schizophrenia
PDF
An enquiry into the effectiveness of export assistance organizations serving southern California SMEs
PDF
Business group affiliation in India
PDF
A methodology to identify high -risk patients with diabetes in the California Medicaid populations (Medi -Cal)
PDF
Decentralizing of public finance: Centralizing forces in developing countries
PDF
Imputation methods for missing items in the Vitality scale of the MOS SF-36 Quality of Life (QOL) Questionnaire
PDF
A new paradigm to evaluate quality-adjusted life years (QALY) from secondary database: Transforming health status instrument scores to health preference
PDF
Economic analysis of ground lease-based land use system
PDF
Essays on organizational forms and performance in California hospitals
PDF
How financial statements in Mexican firms reflect changes in the financial/economic environment from 1978 to 1996
PDF
Income inequality and economic growth: A theoretical and empirical analysis
PDF
An analysis of SME export assistance needs
PDF
A study of employee health plan choice and medical cost: Panel data probit regression and sample selection model
PDF
Assessing the cost implications of combined pharmacotherapy in the long term management of asthma: Theory and application of methods to control selection bias
PDF
An inquiry into trade liberalization of the Asia-Pacific Economic Cooperation (APEC)
PDF
Debt reduction by way of inflation: The case of Lebanon
PDF
Alteration of the in vitro and in vivo processing of a polypeptide, BBI, through conjugation with palmitic acid
Asset Metadata
Creator
Gerber, Robert Arthur
(author)
Core Title
Cost analysis of three pharmacy counseling programs for diabetics in a health maintenance organization
School
Graduate School
Degree
Master of Arts
Degree Program
Economics
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
economics, general,health sciences, health care management,Health Sciences, Pharmacy,OAI-PMH Harvest
Language
English
Contributor
Digitized by ProQuest
(provenance)
Advisor
Liu, Gordon (
committee chair
), McCombs, Jeffrey S. (
committee member
), Nugent, Jeffrey B. (
committee member
)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c16-4523
Unique identifier
UC11341167
Identifier
1381584.pdf (filename),usctheses-c16-4523 (legacy record id)
Legacy Identifier
1381584.pdf
Dmrecord
4523
Document Type
Thesis
Rights
Gerber, Robert Arthur
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 au...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA
Tags
economics, general
health sciences, health care management
Health Sciences, Pharmacy