Risk assessment of preventable drug-related morbidity in older persons

MISSING IMAGE

Material Information

Title:
Risk assessment of preventable drug-related morbidity in older persons
Physical Description:
Book
Creator:
MacKinnon, Neil John, 1971-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 30256254
oclc - 51624306
System ID:
AA00022319:00001

Table of Contents
    Title Page
        Page i
        Page ii
    Dedication
        Page iii
    Acknowledgement
        Page iv
        Page v
    Table of Contents
        Page vi
        Page vii
        Page viii
        Page ix
    List of Tables
        Page x
        Page xi
    List of Figures
        Page xii
    Abstract
        Page xiii
        Page xiv
    Chapter 1. Introduction
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
    Chapter 2. Conceptual framework
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
    Chapter 3. Review of the literature
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
    Chapter 4. Methods
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
    Chapter 5. Results
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
    Chapter 6. Discussion
        Page 113
        Page 114
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
        Page 122
        Page 123
        Page 124
        Page 125
        Page 126
        Page 127
        Page 128
        Page 129
        Page 130
        Page 131
        Page 132
        Page 133
        Page 134
    Appendix A. Geriatric medicine expert panel members
        Page 135
    Appendix B. Example survey for geriatric medicine expert panel members - round 1 of delphi technique
        Page 136
        Page 137
        Page 138
        Page 139
        Page 140
        Page 141
        Page 142
        Page 143
        Page 144
        Page 145
        Page 146
        Page 147
        Page 148
        Page 149
        Page 150
    Appendix C. Instructions for patient chart abstracter, patient chart abstract form, and sample patients
        Page 151
        Page 152
        Page 153
        Page 154
        Page 155
        Page 156
        Page 157
    Appendix D. Members of the chart abstract reviewer panel
        Page 158
    Appendix E. Instructions for chart abstract reviewer panel
        Page 159
        Page 160
    Appendix F. Personal wellness profile senior assessment
        Page 161
        Page 162
        Page 163
        Page 164
        Page 165
        Page 166
        Page 167
        Page 168
        Page 169
        Page 170
        Page 171
        Page 172
        Page 173
    Appendix G. Round 3 of the geriatric medicine expert panel survey
        Page 174
        Page 175
        Page 176
        Page 177
        Page 178
        Page 179
        Page 180
        Page 181
        Page 182
        Page 183
        Page 184
        Page 185
        Page 186
        Page 187
        Page 188
    List of references
        Page 189
        Page 190
        Page 191
        Page 192
        Page 193
        Page 194
        Page 195
        Page 196
        Page 197
        Page 198
    Biographical sketch
        Page 199
        Page 200
        Page 201
Full Text











RISK ASSESSMENT OF PREVENTABLE DRUG-RELATED MORBIDITY
IN OLDER PERSONS














By


NEIL JOHN MACKINNON


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA


1999




























Copyright 1999

by

NEIL JOHN MACKINNON


















This dissertation is dedicated to my parents, James Elliott and Shirley MacKinnon. Thank you

for your love, support, and examples of faith.
















ACKNOWLEDGMENTS


Professional maturity has much in common with maturity as a person. One attribute
common to both is a world view, an expectation that one thrives best by using one's
power to serve something bigger than oneself. -Hepler and Strand (1990)



Completing a dissertation is one sure way to mature and grow as a person, as most

people who have obtained a Doctor of Philosophy degree would agree. It is a long process, from

the genesis of an initial idea for a research project, to the time the study comes to completion.

The process is filled with times of intense struggle, frustration and difficulty, but also with times

of excitement, learning, and finally, fulfillment. Perhaps the Apostle Paul had this in mind,

when he said in the Book of Romans that "tribulation produces perseverance; and perseverance,

character; and character, hope" (Romans 5:3b-4).

In completing a dissertation, there are many individuals who play key parts throughout

the process. Obviously one group of individuals that deserves considerable acknowledgment is

my dissertation committee. We worked together for countless hours shaping a rough research

idea into a real research project. Charles D. (Doug) Hepler, my committee chair, should take a

lot of credit. He proved to be an excellent mentor and hopefully I can have even a small portion

of the success he has had as a researcher. He also helped me balance the dual roles of graduate

student and research fellow over the past three and a half years.

Earlene Lipowski, a fellow Badger, gave invaluable advice, including encouraging me to

keep a day-by-day project journal, which I have done. Richard Segal helped me initially to

enroll in this graduate program and he always seems to be in a good mood, which is a welcomed









sight on many days. Finally, GeoffVining helped mold my perspective of statistics from a

necessary evil to a useful tool for the researcher.

There are also several other individuals who made considerable contributions to this

research project. They include several people at the Florida Hospital in Orlando, from which my

patient population came: Paul Garrett, Tim Regan, Lisa Hutchison, and Scott Neel. Two

individuals at MEDai, a medical artificial intelligence company, helped to assemble the study

database: Kathleen Costello and David Katz (without pay, I might add).

Several other individuals in the Pharmacy Health Care Administration department also

deserve recognition. The ladies in the office provided not only secretarial support, but also some

great conversation: Delayne Redding, Debbie Kemp and Jennifer Ryder. The other graduate

students also provided valuable input on my dissertation and welcomed diversions away from

my dissertation too, such as racquetball. The other faculty in the department helped to provide a

great learning environment for my three and a half years of graduate school. Finally, the faculty

also helped to directly support my dissertation through the Liberty Funds.

Finally, those individuals who have provided love and support definitely deserve special

recognition. This includes my family (thanks mom and dad!) and my fiance, Leanne Moore, as

well as my Lord and Savior Jesus Christ. Philippians 4:6-7 is one Bible passage from which I

have drawn strength many times: "Be anxious for nothing, but in everything by prayer and

supplication, with thanksgiving, let your requests be made known to God; and the peace of God,

which surpasses all understanding, will guard your hearts and minds through Christ Jesus."














TABLE OF CONTENTS
page

ACKN OW LEDGM ENTS............................................................................................................... iv

LIST OF TABLES ...........................................................................................................................x

LIST OF FIGURES....................................................................................................................... xii

ABSTRA CT ................................................................................................................................. xiii

CHAPTERS

1 IN TRODUCTION ....................................................................................................................... I

The N eed for the Study.............................................................................................................. I
Older Persons and Healthcare............................................................................................. 1
Drug-Related M orbidity and M ortality............................................................................... 2
Preventable Drug-Related M orbidity and M ortality........................................................... 3
Problem Statem ent.....................................................................................................................5
Study Objectives........................................................................................................................5
Rationale and Theoretical Introduction ..................................................................................... 5
Research Questions.................................................................................................................... 8
Research Question 1 ............................................................................................................ 8
Research Question 2............................................................................................................ 8
Research Question 3............................................................................................................ 8
Research Question 4............................................................................................................ 8

2 CON CEPTUAL FRAM EW ORK ................................................................................................ 9

System s Theory and the M education Use System ..................................................................... 9
The System Matrix Applied to the Medication Use System............................................. 12
Feedback and Communication Loops in the Medication Use System ............................. 14
Change and the M education Use System ........................................................................... 19
Biopsychosocial Model of Disease Etiology and Therapy...................................................... 20
Risk Factors for PDRM ........................................................................................................... 24
Summary Of The Conceptual Models To Be Used In This Study .......................................... 27

3 REVIEW OF THE LITERATURE ........................................................................................... 28

Drug-Related M orbidity and M ortality in Older Persons........................................................ 28
Identification of Patients at High Risk of Medical Problems in Older Persons ...................... 31
Prediction M odels for Drug Use in O lder Persons .................................................................. 33
Research Assum options and Hypotheses .................................................................................. 34
Research A ssum ptions...................................................................................................... 35









Research Hypotheses .......................................................................................................36

4 M ETHODS................................................................................................................................ 48

Phase I: Operational Definitions of PDRM ............................................................................. 48
Operationalization of the Study Construct PDRM ........................................................... 49
Review of Literature to Identify Specific Operational Definitions of PDRM in
O lder Persons.............................................................................................................. 52
Survey Developm ent......................................................................................................... 52
Geriatric M medicine Expert Panel....................................................................................... 53
Identification of Consensus-Approved Operational Definitions of PDRM in
Database......................................................................................................................55
Validation of Operational Definitions of Preventable Drug-Related Morbidity ..............55
Phase II: Identification of Risk Factors for PDRM ................................................................. 58
Selection of Possible Risk Factors for PDRM .................................................................. 58
Semantic Hierarchy of Risk Factors for Preventable Drug-Related Morbidity................ 60
Study Population............................................................................................................... 62
Data Collection and Formation of the Study Database..................................................... 62
Statistical Analysis............................................................................................................ 64
Logistic Regression Model with the 18 Hypothesized Risk Factors.......................... 64
Factor Analysis........................................................................................................... 69
Logistic Regression Models with Additional Variables............................................. 70
PDRM and Healthcare Resource Utilization.............................................................. 70

5 RESULTS.................................................................................................................................. 71

Delphi Technique The Geriatric M medicine Expert Panel...................................................... 71
Identification of Consensus-Approved Operational Definitions of PDRM in
Database......................................................................................................................73
Validation of Operational Definitions of Preventable Drug-Related Morbidity ..............77
Phase II: Identification of Risk Factors for PDRM ................................................................. 85
Logistic Regression with all 18 Hypothesized Risk Factors ............................................ 85
Factor Analysis.................................................................................................................. 89
Bivariate Analysis ............................................................................................................. 92
Logistic Regression Models with Additional Demographic Variables ............................ 95
Additional Analyses Related to Risk Factor Identification............................................ 101
Risk Stratification System ............................................................................................... 105
Testing the Hypotheses.......................................................................................................... 105
First Set of Hypotheses ................................................................................................... 105
Hypothesis H 1A ........................................................................................................ 107
Hypothesis H 1B ........................................................................................................ 107
Hypothesis H 1C ........................................................................................................ 107
Hypothesis H ID ........................................................................................................ 107
Hypothesis H IE ........................................................................................................ 107
Hypothesis H 1F ........................................................................................................ 108
Hypothesis H 1G ........................................................................................................ 108
Hypothesis H 1H ........................................................................................................ 108
Hypothesis H II ......................................................................................................... 108
Hypothesis H I J......................................................................................................... 108









Hypothesis H1K........................................................................................................ 109
Hypothesis H1L ........................................................................................................ 109
Hypothesis H 1M ....................................................................................................... 109
Hypothesis H1N........................................................................................................ 109
Hypothesis H10 ........................................................................................................ 109
Hypothesis HIP ........................................................................................................ 110
Hypothesis H1Q........................................................................................................ 110
Hypothesis H1R........................................................................................................ 110
Second Hypothesis .......................................................................................................... 110
Third Hypothesis............................................................................................................. 111

6 DISCUSSION ......................................................................................................................... 113

Use of the Delphi Technique with the Geriatric Medicine Expert Panel.............................. 113
Consensus-Approved Operational Definitions of PDRM Observed in the Study
Population ....................................................................................................................... 116
Validation of Operational Definitions of Preventable Drug-Related Morbidity................... 119
Risk Factors for PDRM ......................................................................................................... 121
Risk Factor 1: Four or M ore Recorded Diagnoses......................................................... 122
Risk Factor 2: Antihypertensive Drug Use ..................................................................... 122
Risk Factor 3: M ale Gender............................................................................................ 123
Risk Factor 4: Four or M ore Prescribers....................................................................... 123
Risk Factor 5: Six or M ore Prescription M edications.................................................... 124
General Discussion on the Final Prediction M odel for PDRM ............................................. 124
General and Specific Risk Factors......................................................................................... 126
PDRM and Healthcare Resource Utilization......................................................................... 127
Potential Limitations.............................................................................................................. 127
Significance ........................................................................................................................... 129
Contribution to the Profession of Pharmacy ......................................................................... 131
Contribution to Healthcare .................................................................................................... 131
Theoretical Contribution........................................................................................................ 132
Conclusions ............................................................................................................................ 133

APPENDICES

A GERIATRIC MEDICINE EXPERT PANEL MEMBERS.................................................... 135

B EXAMPLE SURVEY FOR GERIATRIC MEDICINE EXPERT PANEL MEMBERS
ROUND 1 OF DELPHI TECHNIQUE.............................................................................. 136

C INSTRUCTIONS FOR PATIENT CHART ABSTRACTER, PATIENT CHART
ABSTRACT FORM AND SAM PLE PATIENTS............................................................... 151

D MEMBERS OF THE CHART ABSTRACT REVIEWER PANEL...................................... 158

E INSTRUCTIONS FOR CHART ABSTRACT REVIEWER PANEL................................... 159

F PERSONAL WELLNESS PROFILE SENIOR ASSESSMENT........................................... 161

G ROUND 3 OF THE GERIATRIC MEDICINE EXPERT PANEL SURVEY...................... 174


viii









LIST OF REFEREN CES ............................................................................................................. 189

BIO GRAPHICAL SKETCH ........................................................................................................ 199














LIST OF TABLES


Table page
2.1 Eight Fundamentals (Necessities) for Safe and Effective Drug Therapy............................ 15

2.2 The Fundamentals/Human Agents Cell of the System Matrix Applied to the
M education U se System ....................................................................................................... 16

4.1 Phase I M ethodology............................................................................................................ 50

4.2 Phase II M ethodology .......................................................................................................... 59

4.3 Semantic Hierarchy in Risk Factors for Preventable Drug-Related Morbidity................... 61

4.4 Hypothesized Risk Factors and Their Measurement........................................................... 65

4.5 Additional Demographic Variables and Their Measurement.............................................. 67

5.1 Geriatric Medicine Expert Panel Results............................................................................. 72

5.2 Round 2 of the Delphi Technique........................................................................................ 74

5.3 Clinical Scenarios That Were Rejected As Preventable Drug-Related Morbidities ........... 76

5.4 Patients with Outcomes and PDRM..................................................................................... 78

5.5 Number of PDRMs and Specific Outcomes by Individual Patients.................................... 79

5.6 Chart Abstract Reviewer Panel Results for Hyperglycemia with no................................... 80
Regular HgA I c M onitoring................................................................................................. 80

5.7 Chart Abstract Reviewer Panel Results: Secondary Myocardial Infarction........................ 81
in Patients Without ASA and/or Beta-Blocker Use............................................................. 81

5.8 Sensitivity and Specificity for Both of the Operational Definitions of PDRM................... 82

5.9 Sensitivity and Specificity of the Operational Definition of PDRM (Hyperglycemia
O utcom e) ............................................................................................................................. 83

5.10 Sensitivity and Specificity of the Operational Definition of PDRM (Secondary
M yocardial Infarction Outcome)......................................................................................... 84

5.11 Logistic Regression Model With All 18 Hypothesized Variables...................................... 86









5.12 Correlation Matrix for Significant Variables in Regression Model With All 18
Hypothesized V ariables....................................................................................................... 88

5.13 Classification Table for the Regression Model with the 18 Hypothesized Variables........ 90

5.14 Rotated Factor Matrix Varimax........................................................................................ 91

5.15 Rotated Factor Matrix Harris-Kaiser................................................................................ 93

5.16 Rotated Factor Matrix Varimax Validation Group........................................................... 94

5.17 Bivariate Analysis of Patients With, and Without, PDRM Categorized by
H ypothesized V ariables ...................................................................................................... 96

5.18 Bivariate Analysis of Patients With, and Without, PDRM, Categorized by
Additional Demographic Variables..................................................................................... 97

5.19 Logistic Regression Model Including Additional Demographic Variables........................ 98

5.20 Correlation Matrix for Significant Variables in Regression Model with Additional
V ariables............................................................................................................................ 100

5.21 Rotated (Varimax) Factor Matrix of Five Original Risk Factors and Three New Risk
Factors .............................................................................................................................. 102

5.22 Logistic Regression Model Excluding Antihypertensive Drug Use................................. 104

5.23 Bivariate Analysis of Patients With and Without PDRM, Categorized by Healthcare
R source U tilization ........................................................................................................ 112















LIST OF FIGURES

Figure page
2.1 The M education Use System ................................................................................................ 10

2.2 High Leverage Points W within the M education Use System.................................................. 18

5.1 Risk Stratification System.................................................................................................. 106















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



RISK ASSESSMENT OF PREVENTABLE DRUG-RELATED MORBIDITY
IN OLDER PERSONS

By

Neil John MacKinnon

May, 1999

Chairman: Charles D. Hepler, Ph.D.
Major Department: Pharmacy Health Care Administration


This study had three primary objectives: 1) to create operational definitions of

preventable drug-related morbidity (PDRM), 2) to identify patients who are at particular risk of

PDRM and who may therefore benefit from comprehensive pharmaceutical care, and 3) to create

a risk stratification system for PDRM based on the number of risk factors present in an

individual patient. The study was conducted in two phases. In the first phase, the Delphi

technique was used with a geriatric medicine expert panel to create 52 operational definitions of

PDRM in older persons. Ninety-seven patients who matched these definitions were found in

3365 older persons in a Medicare managed care health plan. This was a health plan offered by

the Florida Hospital Healthcare System, a provider-sponsored network with a Medicare contract,

available to all Medicare beneficiaries in three counties of Central Florida. Chart abstracts from a

sample of these patients were given to a panel of five pharmacists to validate the operational

definitions. Overall, the two operational definitions of PDRM that were validated were found to

have a sensitivity of 87.5 percent and a specificity of 73.5 percent as compared to the panel of

pharmacists.


xiii









In the second phase, a prediction model was created to identify risk factors for PDRM.

The dependent variable was the occurrence of a PDRM as operationally defined, and the

independent variables were the hypothesized risk factors. Forward inclusion logistic regression

models and factor analysis were used to identify risk factors for PDRM. Five risk factors for

PDRM were identified in the final prediction model: four or more recorded diagnoses,four or

more prescribers, six or more prescription medications, antihypertensive drug use, and male

gender. A risk stratification system was developed for PDRM based on the number of risk

factors present in an individual patient. Finally, patients with PDRM were shown to use

significantly more healthcare resources then patients who did not experience a PDRM.














CHAPTER 1
INTRODUCTION


The Need for the Study



Older Persons and Healthcare


The special health care concerns of older persons are important factors in health care

policy and research. This problem is predicted to worsen as the baby boom generation becomes

older and the percentage of the total population that is elderly becomes greater. Currently, the

fastest growing segment of the United States population comprises people age 85 and older. By

the year 2030, 70 million people may be enrolled in Medicare, up from the current 33 million

(Rhinehart, 1996).

Health care resource utilization patterns are different in older persons. This is, in part,

because older persons typically have at least one chronic condition and may have multiple

disease states, experience more morbidity, and have more functional limitations than younger

people. Older persons consume a disproportionate share of health care resources. Forty percent

of all health care expenditures are related to older persons (Fincham, 1996) and older persons

account for a large percentage of all hospital stays. Thirty-eight percent of all medications are

prescribed for older persons, even though older persons only make up twelve percent of the total

population (Dalziel, Byszewski and Ross, 1996). Older persons, then, clearly make a significant

economic impact on the healthcare system.

While the utilization of healthcare services is higher for the geriatric population, a small

subset of older persons is responsible for the majority of the utilization. A study which examined

the utilization patterns of continuously enrolled Medicare beneficiaries over a four year period

concluded that ten percent of the population was responsible for 88 percent of the costs (McCall









and Wai, 1983). High users in the first year tended to be high users in the following years

(McCall and Wai, 1983). Anderson and Knickman (1984) studied the temporal patterns of

Medicare beneficiaries' medical expenditures and concluded that unusually high expenditures

for a specific person in one year allow a good prediction of high expenditures in following years.

Therefore, this subset of older persons consistently uses more health care resources year after

year.


Drug-Related Morbidity and Mortality


While medications are prescribed to millions of older persons in an attempt to improve

their health-related quality of life, often an optimum outcome from these medications will not be

achieved. A drug-related problem (DRP) is any patient and time-specific event or situation

involving the medication regimen that interferes with the achievement of an optimum outcome

(Hepler and Strand, 1990). Eight types of DRPs have been described (Hepler and Strand, 1990;

Strand et al., 1990):

1. Untreated indication: The patient has a medical problem that required drug therapy

(an indication for drug use), but is not receiving a drug for that indication.

2. Improper drug selection: The patient has a drug indication but is taking the wrong

drug.

3. Sub-therapeutic dosage: The patient has a medical problem that is being treated with

too little of the correct drug.

4. Failure to receive drugs: The patient has a medical problem that is the result of his or

her not receiving a drug (e.g.; for pharmaceutical, psychological, sociological, or

economic reasons).

5. Over-dosage: The patient has a medical problem that is being treated with too much

of the correct drug (toxicity).









6. Adverse drug reactions: The patient has a medical problem that is the result of an

adverse drug reaction or adverse effect.

7. Drug interactions: The patient has a medical problem that is the result of a drug-

drug, drug-food, or drug-laboratory interaction.

8. Drug use without indication: The patient is taking a drug for no medically valid

indication.

DRPs may lead to drug-related morbidity, which is the failure of a therapeutic agent to

produce the intended therapeutic outcome (therapeutic malfunction or miscarriage) (Hepler and

Strand, 1990). The drug-related morbidity results from either (a) the production of an adverse or

toxic effect or (b) the failure to produce the desired effect within a reasonable time.

If the DRP is unrecognized or unresolved, then drug-related mortality can occur. Recent

literature reports that the clinical, economic, and humanistic outcomes of drug-related morbidity

and mortality are substantial. Lazarou, Pomeranz and Corey (1998) conducted a meta-analysis of

prospective studies involving drug-related morbidity and mortality and concluded that the

overall incidence of serious drug-related morbidity is 6.7 percent and the incidence of drug-

related mortality is 0.32 percent. This estimate places drug-related morbidity and mortality

between the fourth and sixth top cause of death (Lazarou, Pomeranz and Corey 1998). The real

incidence may actually be higher, as their meta-analysis did not include drug-related morbidity

due to noncompliance, drug administration errors, drug abuse, poisonings or therapeutic failures.

Phillips, Christfeld and Glynn (1998) reported that this problem is growing, as between 1983 and

1993, drug-related mortality increased by 257 percent.


Preventable Drug-Related Morbidity and Mortality


Some drug-related morbidities are not preventable, including those resulting from patient

idiosyncracy, while others are preventable. As described by Hepler and Strand (1990),

preventable drug-related morbidity (PRDM) has four unique elements. Given an adverse clinical









outcome, a pre-existing DRP must have been recognizable and the adverse outcome or treatment

failure must have been foreseeable. In addition, the causes of the DRP and the outcome must

have been both identifiable and controllable. However, there are no published criteria to help

determine what drug-related morbidities are, and are not, preventable. Schumock and Thorton

(1992) have attempted to develop such criteria for adverse drug reactions (one type of DRP).

Although their criteria are primarily drug-oriented, a similar patient-oriented approach for drug-

related morbidity may be possible. Others have used these criteria successfully to determine

whether adverse drug reactions were preventable (Pearson et al., 1994).

The extent of the problem of PDRM was not known until recently. A recent estimate of

the total annual cost of drug-related morbidity and mortality in the ambulatory setting in the

United States by Johnson and Bootman (1995) is $76 billion, with a range from $30.1 to $136.8

billion, in their cost-of-illness model. Although they used the opinions of experts and not actual

utilization data to obtain this estimate, this figure is quite comparable with the costs of other

major diseases, such as asthma and diabetes. Schneider et al. (1995) concluded that the annual

cost to one academic medical center was approximately $1.5 million. Nelson and Talbert (1996)

reported that 16.2 percent of admissions in 452 consecutive patients were due to drug-related

morbidity, and 49.3 percent of these admissions were definitely preventable. Fifty-six percent of

all drug-related hospital admissions in older persons in a study in Denmark were judged to have

been "definitely" or "probably" avoidable (Hallas et al., 1991). Therefore, it appears that PDRM

is a serious problem, especially in older persons.

Often the problem of drug-related morbidity and mortality is under-reported and

underestimated. Nelson and Talbert (1996) concluded that the discharge summary of almost 20

percent of patients with a drug-related admission made no mention of this fact. They and others

have lamented that there have been no rigorous methods for identifying and evaluating drug-

related morbidity and mortality. Underestimation of drug-related morbidity is especially

prevalent among community-living older adults, where (1) they may fail to recognize the





5



symptoms of drug-related morbidity or (2) their clinicians may attribute these symptoms to

aging, rather than to the drugs (French, 1996).


Problem Statement


As indicated by the previous discussion, the problem of PDRM in older persons has been

long recognized. The literature continues to grow with new studies of problem drugs and new

estimates of the extent of the problem. Despite this, little quantitative information is known

about the factors associated with increased risk that an older person would experience PDRM,

especially in the ambulatory setting. A better understanding of this relationship may help to

create more effective intervention strategies and more efficient use of scarce healthcare

resources.


Study Obiectives


There are three primary research objectives of this proposed study. These objectives are

1. To create operational definitions of PDRM,

2. To identify patients who are at particular risk of PDRM (as defined) and who may

therefore benefit from comprehensive pharmaceutical care, and

3. To create a risk stratification system for PDRM based on the number of risk factors

present in an individual patient.


Rationale and Theoretical Introduction


The rationale for this study will be two related models that are based on systems theory.

These two models are the medication use (pharmaceutical care) system and the biopsychosocial

model of disease etiology and therapy. The use of risk factors is the final important aspect of this

study.









The first of these two models the medication use system. Pharmaceutical care has been

defined by Hepler and Strand (1990) as being the cooperative, responsible provision of drug

therapy to achieve definite outcomes intended to improve a patient's quality of life.

Pharmaceutical care is quite different from our current approach to medications use.


Pharmaceutical care differs from traditional drug treatment because ... it is an explicitly
outcome-oriented, cooperative, systematic approach to providing drug therapy, directed
not only at clinical outcomes but also activities of daily life and other dimensions of
health-related quality of life. Preventing, detecting and resolving pharmacotherapeutic
problems before they become adverse outcomes increases the effectiveness of drug
therapy. (Hepler and Grainger-Rousseau, 1995, p.8)

The development of a predictive model for PDRM in older persons may support

prospective prevention of adverse outcomes related to drug therapy.

The usual medication use process is far from the ideal pharmaceutical care system. This

process has been previously described (Hepler and Grainger-Rousseau, 1995). Communication

and cooperation between the patient, physician, and pharmacists may be unsystematic and

ineffective. The patient may develop a drug-related problem, which, if left unresolved, may

develop into drug-related morbidity or even mortality. However, no one may recognize the

problem or attribute it to drug therapy. This flow of care is described as a process, and not a

system, because there is no feedback loop after the patient receives the prescription.

Compared to a drug therapy process, a pharmaceutical care system emphasizes the

prevention, detection, and resolution of drug-related problems before they can become drug-

related morbidities. To do this, a pharmaceutical care system emphasizes monitoring of patients-

in-therapy to detect problems. Monitoring, in turn, increases communication among the patient,

physician, and pharmacist. All three parties need more information (more often) from each other,

not only to detect problems but also to prevent and resolve them.

A second dimension of the conceptual framework of this research is the biopsychosocial

model. According to this model, which is also based on systems theory, many clinical outcomes,

including those of drug therapy, result in part from complex interactions of psychological and









sociological factors. These factors are in addition to the physiological and chemical explanations

provided in the biomedical view. Therefore, it may be possible to identify risk factors for PDRM

from among a wider variety of variables than would be predicted by reasoning from the

biomedical model. Furthermore, this model views illness as being more continuous and less

episodic than the biomedical model and therefore it may be possibly more descriptive of the

medical problems of older persons.

The third dimension of the conceptual framework of this study is the use of risk factors.

Risk factors in this study are those variables that are statistically associated with PDRM (the

outcome event) in older persons. While general risk factors for PDRM will be identified for an

entire geriatric population, the constructed model should be applied on a patient-specific basis.

This approach follows the similar use of probability theories and the rationale behind screening

tests, although this study is not a classical epidemiological study. These tests use population-

based information to identify the risk factors, but then the information is applied on an individual

basis to identify those patients who might be at particular risk for problems. Risk factors may be

used to help allocate scarce healthcare resources and as potential indicators to identify patients

who need interventions. Careful consideration of these risk factors should be incorporated into

the therapeutic and monitoring plans of health care professionals in order to proactively prevent

PDRM in older persons. A goal of identifying older persons at risk for PDRM is to intervene in

their medical care before they experience a PDRM. Risk factors are patient characteristics and

therefore they may or may not be able to be changed (e.g.; it may be hard to change a patient's

gender or drug therapy if it is essential). If a risk factor can not be eliminated in an individual,

then at least prospective management of that risk factor should occur. This study describes an

approach to use population-based information to construct a model for PDRM in older persons

that will be applied on a patient-specific basis.









Research Questions


To achieve the purpose of this study, four research questions will be investigated. These

questions will explore different aspects of risk factor identification of PDRM in older persons.


Research Question 1


What are the issues in developing and using operational definitions of PDRM?


Research Question 2


What are major risk factors for PDRM in older persons?


Research Question 3


Are there general, and disease (or drug), specific risk factors for PDRM in older

persons?


Research Question 4


What is the relationship between PDRM and the utilization ofhealthcare resources?















CHAPTER 2
CONCEPTUAL FRAMEWORK



The conceptual framework for this study consists of two models that incorporate systems

theory. These two models are the medication use (pharmaceutical care) system and the

biopsychosocial model of disease etiology and therapy. The concept of risk factors is a third

fundamental component of this study. Brief descriptions and empirical findings of these models

will be discussed.


Systems Theory and the Medication Use System


Medication use can be described as a process. In general, model patients enter the health

care system when they recognize a health problem and see a physician. The physician then

diagnoses the patient's problem and constructs a therapeutic plan, which is often accompanied

by a prescription. The therapeutic plan is implemented when the patient goes to a pharmacy and

a pharmacist dispenses the prescription and provides advice about the use of the medication. The

typical medication use process is completed when the patient consumes the medication.

Unintended outcomes of this process include drug-related problems, which, left unresolved or

undetected, may lead to drug-related morbidity. As discussed in chapter one, there are many such

unintended adverse outcomes from the current medication use process and several authors have

argued that this current process must be changed (Shane, 1992; Smith and Benderev, 1991).

In contrast, a medication use system emphasizes systematic monitoring and

communication. This model of medication use has been described as being a philosophy of

practice for pharmacists and its knowledge base is formed from systems theory (Hepler, 1996).

The structure of this model is seen in Figure 2.1. Patient "progress" is monitored to prevent,











Prescribing
Evaluation


Prescribing /
Influences / \
(Formulary, /
Education) /

t Recognize Assess Therapeuti,
Patient __ Patient __ Plan,
s Problem Problem Prescriptio





Drug-Related Resolve
Morbidity Patient roduct(s)pense
Products)
Problem and Provid
(if any) /Advice

Monitor o-*
Outcomes
According
to Plan Consume
Administer
Products)


Drug-Related Problem


Figure 2.1 The Medication Use System


(Adopted from Hepler and Grainger-Rousseau, 1995)









detect, and resolve DRPs before they develop into drug-related morbidities. Communication

between the physician, patient, pharmacist, and other health care professionals is critical to

proper functioning of the system. Prevention is important, especially in older persons, since

DRPs occur frequently and affect clinical, psychosocial, and economic outcomes. The

medication use system has a goal of patient-care outcomes, rather than performance of tasks, as

in the medication use process (Smith and Benderev, 1991).

Systems theory is the foundation upon which the medication use system is built. The

phrase "systems theory" may be deceptive, as according to some definitions, it is really a model

and not a theory (Babbie, 1983). Laszlo (1973) refers to systems theory as a philosophy.

Regardless, at the fundamental level, systems theory, or systems thinking, is a body of

knowledge that has gradually developed over the past fifty years into a model, which helps one

to see and influence things from a different, larger perspective (Senge, 1994). Key to systems

theory is the word systems itself. It is frequently used but infrequently understood. A system can

be defined as follows:

Basically, a system is (1) a group of related entities that (2) does something receives
inputs, affects them in some way, and produces outputs to achieve some purpose. Almost
anything in the world can be called a system. A sheep can be considered a system. It
takes in fodder and produces wool and lamb chops. (Nadler and Hibino, 1994, pp. 198-
199)

It is clear from this definition that one of the tenets central to systems theory is seeing the "big

picture" and "whole structures". Senge (1994) states that systems theory is a method to (1) see

wholes, (2) interrelationships rather than things, and (3) patterns of change rather than single

elements. Nadler and Hibino (1994) call this the "systems principle": each problem is just a

piece of a larger system. Many authors argue that these patterns, or structures, underlie complex

situations (Senge, 1994; Nadler and Hibino, 1994; Laszlo, 1973). By learning to recognize and

see them, the way by which one approaches problems will be modified. Senge (1994) elaborates

on these "structures":

one of the most important, and potentially most empowering, insights to come from the
young field of systems thinking is that certain patterns of structure recur again and again.








These 'systems archetypes' or 'generic structures' embody the key to learning to see
structures in our personal and organizational lives. The systems archetypes of which
there are only a relatively small number- suggest that not all... problems are unique.
(Senge, 1994, p.94)

These structures, then, are an important concept in systems theory and allow one to view

problems with a larger perspective.

The medication use system has built upon many of the central principles of systems

theory. These include the interrelationship of the various elements and dimensions of the

medication use system, the importance of communication and feedback loops, identification of

key areas of change, and the individualization of drug therapy goals and monitoring for specific

patients. Systems theory also emphasizes that the medication use system can be viewed as being

a subset of a larger system the health care system or subsets of the medication use system can

be thought of systems in their own right. For example, the act of a physician writing a

prescription for a patient is a complex system that has been extensively studied. The physician

uses multiple inputs when deciding which drug to prescribe, such as the price of the prescription,

the condition of the patient, the information given to him by a pharmaceutical representative,

whether the drug is in his evoked set, and other factors. Similarly, there are many outputs and

feedback mechanisms after he/she has written the prescription, such as whether the patient has

improved, drug use evaluation, and any adverse effects experienced by the patient. One way of

viewing these aspects of the medication use system is through the use of the system matrix.


The System Matrix Applied to the Medication Use System


While it is useful to recognize that structures exist, the size of most systems could

quickly cause one to become overwhelmed by the amount of information in these structures. One

way of organizing this information is to use a method called the system matrix (Nadler and

Hibino, 1994). According to the system matrix, a system consists of elements and dimensions.

The elements of a system are (1) purpose (the mission, aim, need, primary concern, or results

sought from a system), (2) inputs (physical items, information or human beings on which work,








conversion, or processing takes place), (3) outputs (desired and undesired physical items,

information, humans and services that result from processing inputs), (4) sequence (the

conversion, work, process, or transformation by which the inputs become the outputs), (5)

environment (the physical and sociological factors within which the other factors operate), (6)

human agents (those who aid in the steps of the sequence without becoming part of the outputs),

(7) physical catalysts (resources that aid in the steps of the sequence without becoming part of

the outputs), and (8) information aids (include knowledge and data resources that help in the

steps of the sequence without becoming part of the outputs) (Nadler and Hibino, 1994).

Dimensions help clarify the conditions for each element in a specific situation (Nadler and

Hibino, 1994). The dimensions of a system are (1) fundamentals (tangible, overt, observable,

physical, or basic structural characteristics), (2) values and goals (motivating beliefs, human

expectations, global desires, ethics, equity, and moral concerns that can be ascribed in some form

to each element, (3) measures (translate the fundamentals and values dimensions into particular

performance factors and operational objectives), (4) control (comprises methods for ensuring

that the fundamentals, measures, and even value specifications, are maintained as desired during

the operation of the system), (6) interface (the relationships of the fundamentals, values,

measures, and control specifications to other elements and to external systems), and (7) future

(changes in each specification of the other dimensions in the future) (Nadler and Hibino, 1994).

A system matrix clearly delineates the relationships of elements and their interdependences and,

most importantly according to Nadler and Hibino (1994), it prevents the omission of critical

components of the system.

The system matrix can be used to describe the medication use system. As previously

discussed, a system matrix helps to view a system by listing the elements and dimensions of that

system. Chiefly, this allows one to see the structure of a system; in this case, the medication use

system by specifying each element and dimension cell in the system matrix. For example, the








fundamentals (one of the dimensions of a system) have been previously described by Grainger-

Rousseau et al. (1997). As seen in Table 2.1, these eight fundamentals, or necessities, must be

present to ensure that drug therapy will be safe, effective, and humane (Grainger-Rousseau et al.,

1997). The human agents (one of the elements of a system) can be described for these eight

fundamentals for the medication use system. In the medication use system, the main human

agents are the prescriber, the pharmacist, and the patient (and caregiver). Table 2.2 shows how

each of the eight fundamentals of a safe and effective medication use system relates to these

three human agents. The boxes with checkmarks indicate an important relationship between an

element and a dimension. For example, fundamental one (timely recognition of signs and

symptoms) often depends on the action of the prescriber, pharmacist, or patient/caregiver. This

example shows how the system matrix can be applied to one cell for the medication use system.



Feedback and Communication Loops in the Medication Use System


An important concept in systems theory is feedback, also referred as communication

loops. Feedback is defined by Senge (1994) as being a broad concept meaning any reciprocal

flow of influence, and every influence acts as both a cause and an effect. Senge (1994) continues,

The practice of systems thinking starts with understanding a simple concept called
'feedback' that shows how actions can reinforce or counteract (balance) each other. It
builds to learning to recognize types of 'structures' that recur again and again: the arms
race is a generic or archetypal pattern of escalation, at its heart no different from turf
warfare between two street gangs, the demise of a marriage, or the advertising battles of
two consumer goods companies fighting for market share. (Senge, 1994, p.73)

Two types of feedback have been described. The first type is reinforcing (also called amplifying

or positive) feedback which causes growth. The second type is balancing (also called stabilizing

or negative) feedback which counteracts the reinforcing feedback. These feedback cycles or

communication loops often may have "delays", in which the flow of influence is interrupted,

resulting in a slowing down of events (Senge, 1994).








Table 2.1 Eight Fundamentals (Necessities) for Safe and Effective Drug Therapy
(Adopted from Grainger-Rousseau et al., 1997)

1. Timely recognition of drug indications and other signs and symptoms relevant to drug
use with accurate identification of underlying disease. "Correct" therapy for a late or
incorrect diagnosis cannot improve a patient's quality of life.
2. Safe, accessible, and cost-effective medicines. Safe and cost-effective (efficient) drug
products must be legally and financially available.
3. Appropriate prescribing for explicit (clear, measurable, and communicable) objectives.
Explicit therapeutic objectives simplify the assessment of prescribing appropriateness and
are necessary for assessing (monitoring) therapeutic outcomes.
4. Drug product distribution, dispensing, and administration with appropriate patient
advice. Including: (a) ensuring that a patient actually obtained the medicine, (b) negotiating
a regimen that the patient can tolerate and afford, (c) ensuring that a patient (or caregiver)
can correctly use the medicine and administration devices, (d) advising to empower the
patient or caregiver to cooperate in his or her own care as much as possible.
5. Patient participation in care (intelligent adherence). The ambulatory patient or caregiver
should consent to therapeutic objectives and know the signs of therapeutic success, side
effects and toxicities; when to expect them; and what to do if they appear.
6. Monitoring (problem detection and resolution). Many failures can be detected while they
are still problems and before they become adverse outcomes or treatment failures.
7. Documentation and communication of information and decisions. Communication and
documentation are necessary for cooperation in a system.
8. Product and system performance evaluation and improvement. Practice guidelines,
performance indicators, and databases are a useful approach to achieving and maintaining
improved system performance (outcomes).









Table 2.2 The Fundamentals/Human Agents Cell of the System Matrix Applied to the
Medication Use System

Fundamental for safe and effective Human Agents of the Medication Use System

drug therapy Prescriber Pharmacist Patient/ Caregiver

Timely recognition of signs and

symptoms 4 "

Safe, accessible, and cost-effective

medicines

Appropriate prescribing V

Distribution, dispensing,

administration, and patient advice V '

Patient participation V

Monitoring I 1

Documentation and communication V/ / I

System evaluation









The acknowledgement of the importance of feedback in systems theory allows one to

accurately discern the role of individuals in a system. In systems theory, individuals are seen as

simply part of the feedback process, and not as a separate component of a system. Individuals

can act as either reinforcing or balancing feedback. This is different from the common perception

that individuals are somehow special and are different from the other elements of a system.

Senge (1994) notes that considering individuals as a form of feedback implies that an individual

is usually not solely responsible for an action, but that everyone shares in the responsibility for

the event occurring. Individuals may have different levels of influence but most systems

problems are solved by looking at all the types of feedback.

Several feedback loops occur in the medication use system. As seen in Figure 2.2, there

is a feedback loop for aggregated prescribing that includes prescribing data, prescribing (drug

use) evaluation, and prescribing influences such as a formulary and education. A second

feedback loop exists for a single patient's prescription or drug regimen. This is when information

obtained by a pharmacist or another health care professional from the patient can be used to alter

the therapeutic plan. A third feedback loop, which is not formalized in the medication use

process, involves monitoring the patient for drug-related problems, resolving any problems

which exist, and using this information to revise the therapeutic plan as appropriate.

The medication use system contains feedback loops that are reinforcing and it may also

contain delays. Patient monitoring by the pharmacist can be used to reinforce or strengthen the

therapeutic plan. Monitoring includes the determination of which information to collect for the

evaluation of the progress of therapy, the evaluation of achieving the therapeutic objectives, and

responding to evaluations (Grainger-Rousseau et al., 1997). This constant, critical re-evaluation

of the therapeutic plan is an important reinforcing feedback loop that helps facilitate the

provision of pharmaceutical care (Strand et al., 1991). Delays can occur at many places in the

medication use system, such as when pharmacists or patients do not provide timely data to the

physician, such that the therapeutic plan is not revised and poor patient care results.









Prescribing
Evaluation


Prescribing
Influences
(Formulary,
Education)


Drug-Related Resolve
Morbidity Patient
Problem
(if any)

High Monitor
Leverage Outcomes
Point According
to Plan




Dru2-Related Problem


Figure 2.2 High Leverage Points Within the Medication Use System
(Adopted from Hepler and Grainger-Rousseau, 1995)


Prescribing
Data


High
Leverage
Point










Change and the Medication Use System


The next logical step is to ask how one can influence and change the various types of

feedback that exist in a given system. This skill of knowing where to affect a system to produce

the greatest possible intended effect is known as leverage. Senge (1994) explains this concept,

The bottom line of systems thinking is leverage seeing where actions and changes in
structures can lead to significant, enduring improvements.., our nonsystemic ways of
thinking are so damaging specifically because they consistently lead us to focus on low-
leverage changes: we focus on symptoms where the stress is greatest. We repair or
ameliorate the symptoms. But such efforts only make matters better in the short run, at
best, and worse in the long run. (Senge, 1994, p. 14)

In order to change systems the most effectively, then, high-leverage changes should be sought.

By recognizing which feedback mechanisms result in the high-leverage changes, one can target

those feedback points and obtain the intended effect.

In order to most effectively change the medication use system such that PDRM is

minimized, high leverage change points must be identified and successful intervention strategies

must be implemented. Two such high leverage points for the prevention of PDRM within the

medication use system are identified in Figure 2.2: the therapeutic plan and patient monitoring.

The creation and re-evaluation of the therapeutic plan is a key point in the medication use

process. Without a therapeutic plan that includes explicit, realistic objectives, therapy

management is practically impossible (Hepler and Grainger-Rousseau, 1995). Also, the patient

risks for PDRM are generally unknown during the formation of the therapeutic plan (Strand et

al., 1991):

It seems evident that we have accumulated very little hard information that describes
pharmaceutical-care needs, patient risks related to pharmacotherapy, and the drug-
related problems present. Before the concept of... pharmaceutical care can be developed
any further, this information needs to be collected and evaluated. (Strand et al., 1991,
p.550)

Some medications, such as digoxin, have been identified in the literature as being commonly

prescribed inappropriately in the older population (Aronow, 1996). There is also some evidence









that those medications that are prescribed inappropriately lead to PDRM in older adults. French

(1996) argues that up to 25 percent of community-living older persons are at risk for drug-related

morbidity due to inappropriate prescribing, a key part of the therapeutic plan. Ideally, a method

that could prospectively identify these high-risk medications, and incorporate this information

into the therapeutic plans for them, would be of tremendous value.

The monitoring and management of patient outcomes according to the therapeutic plan is

a second key point in the medication use process. As early as 1981, (Campbell, 1981) an

advisory committee on pharmaceutical needs in older persons recommended that identification

of toxic drug effects and drug monitoring be priority functions of the pharmacist. In 1988,

Grymonpre et al. (1988) stated that their study on drug-related adverse patient events confirmed

the need for increased caution and monitoring of all consequences and outcomes of medication

use in older persons. More recently, Johnson and Bootman (1995) recommended that urgent

attention be given to this problem and monitoring should be emphasized to help lessen PDRM.

Patient monitoring is also required to ensure that the desired patient outcomes and goals are

attained (Strand et al., 1991; McDonough, 1996). Identification of medications for which

monitoring is extremely important may greatly assist the pharmacist during patient monitoring

and management to prevent PDRM in older persons.


Biopsychosocial Model of Disease Etiology and Therapy


The previous discussions on (1) the importance of feedback and communication within

the medication use system, and (2) the high leverage points (the therapeutic plan and patient

monitoring), show the importance of accounting for individual patient differences. As well, as

was seen in Table 2.2, many of the fundamentals of safe and effective drug therapy rely on the

cooperative actions of the prescriber, pharmacist, and patient. This holistic, individualized

approach to medication use is complimented by the biopsychosocial model of disease etiology

and therapy.









The biopsychosocial model was first proposed by Engel (1977), in response to the

biomedical model. The biomedical model of disease had molecular biology as its basic scientific

discipline (Engel, 1977) and tried to explain disease as being deviations from 'norms'-

measurable biological variables (DiMatteo, 1991). Engel (1977, 1981) argued that such a model

did not consider the person as a whole, and failed to incorporate psychological, social or

behavioral aspects of illness.

Engel turned to nature, and biology in particular, for an alternative. The knowledge base

from which the biopsychosocial model is built upon is the use of systems theory in biology

(Engel, 1981; Sadler and Hulgus, 1992). Engel (1981) describes this foundation,


systems theory, by providing a conceptual framework within which both organized
wholes and component parts can be studied, overcomes this centuries-old limitation and
broadens the range of the scientific method to the study of life and living systems,
including health and illness (Engel, 1981, p. 103)

and,

systems theory is best approached through the common sense observation that nature is a
hierarchically arranged continuum, with its more complex larger units being
superordinate to the less complex smaller units. (Engel, 1981, pp. 103-104)

As previously discussed, systems theory emphasizes this hierarchy and views each component of

the system, whether it be a person or the biosphere, as not existing in isolation, but being

influenced by its environment. A bee, for example, can not truly be studied without studying its

environment (flowers, other bees, honeycomb, etc.). Systems theory, then, was useful to help

explain how things are ordered in nature and how these different "systems" in nature interacted.

Engel advocated taking this knowledge base of systems theory in biology and applying it

to the medical care of individuals, as well as medical research and teaching. This would allow

for a holistic evaluation of patients considering cultural, social, psychological, and behavioral

dimensions of illness (DiMatteo, 1991), humane, empathic and rigorous medical care (Sadler and

Hulgus, 1992), and patient-centered and patient-specific medical care (Howell, 1992;








Zimmermann and Tansella, 1996). This is especially important in older persons, where medical

information and diagnosis are insufficient in predicting their health care needs and consideration

of these other dimensions are critical (Rock et al., 1996).

This distinction between the biomedical and biopsychosocial models can be seen in the

medication use system. Traditional drug treatment follows a population-based approach to care,

whereas a pharmaceutical care system follows a patient-based approach. In order to improve the

medications use process, the current approach utilizes population-based methods such as drug

formularies and drug use evaluation. These methods follow a biomedical model of disease,

whereby the disease becomes the emphasis and a mechanistic, reductionistic, and dualistic

perspective is employed. Patients who have illness must be deviating from objective somatic

norms and thus there is always a drug of choice that can be chosen for a whole population.

In contrast, a biopsychosocial model approach states that deviation from the somatic

norm is insufficient to explain disease, and biological relationships to illness are complex

interactions with the mind and environment. This model accounts for patient-specific differences

in illness because it holds a monistic perspective that states the mind and body are intimately

related. While a population basis can be employed to change a system, the biopsychosocial

requires that individual patient differences be considered. Variation between patients is expected

and is not necessarily negative. It even allows for patient disagreement with the health

professional's plan of care (Kasahara, Shemon and Holzschuh, 1994). This model incorporates

psychological aspects of illness (such as anxiety or depression), the individual's cultural

expectations about illness, and the present social context of the illness as well as the biological

parameters (DiMatteo, 1991).

The biopsychosocial model fits well with the approach to be used in this proposed

investigation of PDRM in older persons, which will consider these other dimensions of illness.

First, illness presentation in older persons is different from other age groups and this raises some








unique concerns. One of these unique factors is that many older persons have atypical disease

presentations that do not fit a classic disease model. Jarrett et al. (1995) concluded that atypical

disease presentation is associated with adverse hospital outcomes. This makes the issues of

disease diagnosis and treatment far more complex in older persons and further strengthens the

need to account for patient differences. Second, PDRM has been reported and analyzed on a

patient-specific basis. Third, drug-related morbidity often has atypical or paradoxical

manifestations in older persons (French, 1996; Harper, Newton and Walsh, 1989). This may be

because patient-specific differences play a large role in the development of drug-related

morbidity and mortality. Gurwitz and Avomrn (1991) state that patient-specific physiologic and

functional patient characteristics, such as pharmacokinetic and pharmacodynamic changes, are

often very important to predict drug-related morbidity in older persons due to the inter-individual

variability of the aging process. As Strand et al. (1991) explain,

each patient must receive individualized treatment not only because of information
derived from scientific knowledge but also because he or she must be consistently
respected as a unique individual with specific needs. (Strand et al., 1991, p.549)

The biopsychosocial model allows for the inclusion of social, psychological, and behavioral

factors in a prediction model of PDRM in older persons. This includes such things as trouble

paying for medications, difficulty taking medications, patient belief that he/she is on too many

medications, and the patient's own perception of their health. This is important, as, for example,

an older person's self-assessment of their health as being poor has been shown to be associated

with mortality in a study of recipients of community-based long-term care (Fried, Pollack, and

Tinetti, 1998). A prediction model that includes this and other social, psychological, and

behavioral factors may be the best way to identify PDRM in older persons.








Risk Factors for PDRM


The medication use system emphasizes an individualized approach to care. However,

population-based information can be useful in planning for a specific patient. One such type of

population-based information is risk factors. Risk factors are variables that are statistically

associated with an outcome event. Prediction models use risk factors identified from a

population and apply this information to the individual patient. The individual patient is then told

the probability, or risk, of having a certain disease or medical condition.

Probability theory and its applications play a central role in the development and

function of screening tests and risk factors. Medical tests results are rarely positive or negative,

instead they lie on a continuum; test results that are beyond a cutoff threshold are called positive.

This cutoff threshold separates the two separate, but usually overlapping Gaussian distributions

of patients with, and without, the disease (condition). This relates to the concepts of sensitivity

and specificity. Sensitivity is the percentage of patients with a disease (condition) who are

labeled "positive" by the test, whereas specificity is the percentage of patients without the

disease who are labeled "negative" by the test. A test or prediction model with a high specificity

and sensitivity will have a low rate of false negatives and false positives.

Risk factor identification is an important issue, but equally important is the actual use of

these risk factors in an assessment or screening program. Risk assessment generally involves a

prospective investigation of a person's health risks to facilitate interventions before the

occurrence of a preventable health crisis (Kerekes and Thornton, 1996). Such a risk assessment

program may utilize the risk factors previously identified to target certain patients proactively.

Nikolaus et al. (1995) were able to develop an instrument to assess nutritional risk in older

persons. They argue that risk assessment is a major component of the medical management of

older persons. Interventions are often designed based on the risk factors identified. For example,








an intervention program which targets older persons who are frail and who have chronic illness

has demonstrated a significant decrease in emergency room visits, admissions, and length of stay

in a risk assessment program in St. Joseph Medical Center (Swindle, Weyant and Mar, 1994).

Cargill (1992) developed criteria to help identify those patients at highest risk for

problems related to medication noncompliance. These criteria included: multiple medications

(more than three medications per day prescribed), medication regimen changes (a change in the

past six months), multiple prescribers, and memory, sensory, and cognitive deficits (unable to

verbalize name or purpose and frequency of medication, unable to read the label, unable to

calculate how many 10mg pills in a 20mg dose, and unable to judge appropriately administration

times for twice a day dosing regimensXCargill, 1992). Cargill (1992) reported that those older

persons with more risk factors were older and had more medications prescribed, but they did not

have different compliance patterns than those with fewer risk factors.

Identifying possible risk factors for drug-related morbidity based on theory has been a

difficult problem. Strand et al. (1991) used the following approach to identify possible risk

factors,

we identify three categories of risk factors that can affect the type and level of
pharmacotherapeutic risk: (1) risk factors associated with the patient's clinical
characteristics, (2) risk factors associated with the patient's disease, and (3) risk
factors associated with the patient's pharmacotherapy. The interaction of these three
types of risk factors ultimately determines the level of risk associated with a patient's
pharmacotherapy and therefore the level of pharmaceutical care required of the
pharmacist. (Strand et al., 1991, p.549)

The approach used in this study will be different. Possible risk factors for PDRM in older adults

will be identified by careful consideration of the high leverage points within the medication use

system and the biopsychosocial model of disease etiology and therapy. Based on this conceptual

basis and empirical evidence, possible risk factors for PDRM will be entered into logistic

regression models for PDRM. Therefore, risk factors in this study will be those variables that

help to explain some proportion of the variance of PDRM in older persons. Variables with odds








ratios greater than one will be positive risk factors for PDRM, while variables with odds ratios

less than one will be negative risk factors for PDRM. This approach will also allow estimation of

the proportion of variance in PDRM in older persons that can be accounted for by these risk

factors.

Risk factors for PDRM may be used in a variety of ways. First, if health care

professionals know that an individual has one or more risk factors for PDRM, a proactive change

in the medical care of that patient may be made to help prevent the occurrence of PDRM in the

future. This knowledge will hopefully allow health care professionals to anticipate the possibility

of PDRM occurring in individual patients. Boult et al. (1998) argue that high-risk older persons

should be identified so that a comprehensive assessment of their health-related needs may be

performed, and interventions planned to meet these needs. In the medication use system,

physicians, in particular, should consider these risk factors when creating a therapeutic plan for

the patient. Risk factors may be considered as indicators that help direct the health care

professional to patients with potential problems. Koecheler et al. (1989) used six prognostic

indicators (or risk factors without the empirical evidence) for patients who might warrant

pharmacist monitoring. Second, the risk factors identified can be related to the medication use

system for interpretability in order to determine the best possible manner of use. For example, if

a risk factor is known to relate to patient monitoring, then every attempt should be made to

ensure that the patient receives proper monitoring. This is important because often risk factors

can not be eliminated from the individual patient (e.g. gender), so proper management of the

patient becomes paramount. Finally, risk factors may be used to help in the allocation of scarce

healthcare resources.








Summary Of The Conceptual Models To Be Used In This Study


In summary, the conceptual framework that will be used for the study risk assessment of

preventable drug-related morbidity in older persons is based on the application of systems

theory in two models. The first model that incorporates systems theory is the medication use

system. This model applies many of the central themes of systems theory, including the

interdependency of related things, identification of recurring structures, feedback and

communication loops, system goals, and identification of main change (or high leverage) points

(Senge, 1994; Laszlo, 1973; Nadler and Hibino, 1994). In the medication use system itself, a

structure has been proposed and essential elements identified, which includes key

communication and feedback loops between the physician, patient, and pharmacist, a system

goal of improving patient outcomes, and high leverage points such as patient monitoring (Hepler

and Grainger-Rousseau, 1995; Hepler and Strand, 1990). The biopsychosocial model, also based

on systems theory, permits a holistic evaluation of patients that includes cultural, social,

psychological, and behavioral dimensions of illness. Risk factors for PDRM in older persons will

be identified based on these first two models.














CHAPTER 3
REVIEW OF THE LITERATURE





The following review of the literature will focus on drug-related morbidity and mortality

in older persons, identification of older persons at high risk of medical problems, and finally, a

review of prediction models for drug use in older persons.


Drug-Related Morbidity and Mortality in Older Persons

Drug-related morbidity and mortality is a problem of special consideration in older

persons. The incidence of drug-related morbidity seems to be higher in older persons, although it

declines in the last decades of life (80 plus years) and age does not appear to be an independent

risk factor for drug-related morbidity (Carbonin et al., 1991; Gurwitz and Avomrn, 1991). Still,

Campbell (1981) argues the increasing number of older persons and their prominent utilization

of healthcare services, particularly medications, naturally extend itself to concern about the high

risk drug-related morbidity and mortality in this population.

Fincham (1996) gives a succinct statement as to the extent of this problem in older

persons:

providing for consistent and appropriate use of drugs is exceedingly important for the
ambulatory elderly. Studies have shown that when it does not occur, hospitalizations
occur due to noncompliance and to the occurrence of avoidable adverse drug reactions.
Significant predictors of preventable hospital readmission for the elderly include the
occurrence of preventable adverse drug reactions, noncompliance, overdose, lack of a
necessary drug therapy, and underdose. Others have noted that 50% of drug-induced
illnesses that require hospitalization could have been avoided. Elsewhere, researchers
have estimated that 75% of medications are misprescribed for the elderly, with overuse
and underuse rampant. There must be increasing efforts to ensure continuity of care for
the ambulatory elderly to avoid these and other drug-related problems. Because drug use








in the elderly is dynamic and increases with proximity to death, pharmacists are key
players to help the elderly avoid these drug-related problems. (Fincham, 1996, p.525)

The literature is rich with examples to substantiate these concerns. Grymonpre et al. (1988)

determined that 19 percent of all hospital admissions (23 percent of all admissions that involved

prescription drugs) of patients aged 50 and older exhibited at least one type of drug-related

morbidity and mortality in a tertiary care hospital in Manitoba, Canada. The major types of drug-

related morbidity and mortality identified were adverse drug reactions (48 percent), intentional

noncompliance (27 percent), treatment failures (19 percent), alcohol-related problems (14

percent) and medication errors (10 percent) (Grymonpre et al., 1988).

Ostrom et al. (1985) studied medication use in 183 independently living seniors in

Seattle and reported the prevalence of many medication problems. Seventy-five percent of the

older persons had at least one potential medication problem, with a label discrepancy (37

percent), potential drug interaction (27 percent), underuse of medication (24 percent), inability to

read label (14 percent), and failure to open prescription vial (12 percent) being the most common

problems (Ostrom et al., 1985).

Some researchers have identified drugs that have a particularly high risk for drug-related

morbidity and mortality in older persons. Dalziel, Byszewski and Ross (1996) constructed a list

of the top ten problem drugs in the older persons (they failed to provide any empirical evidence

as to why these certain medications made the list, however): Non-steroidal anti-inflammatory

drugs (NSAIDS), benzodiazepines, amitriptyline, fluoxetine, anticholingerics/antihistamines,

over-the-counter drugs/alcohol, cimetidine, centrally-acting antihypertensives/beta-blockers,

digoxin, and irritant laxatives/colacel. The most commonly implicated drugs in 162 cases of

drug-related morbidity and mortality identified in another study were: systemic steroids, digoxin,

nonsteroidal anti-inflammatory agents, methyldopa, calcium channel blockers, beta-blockers,

theophylline, furosemide, sympathomimetics, thiazides, and benzodiazepines (Grymonpre et al.,








1988). In a study of drug-related hospital admissions, hypoglycemic and diuretic agents were the

two most implicated drugs (Nelson and Talbert, 1996).

Beers et al. (1991) used the Delphi technique to develop 30 factors defining

inappropriate medication use in the nursing home setting. Using modified versions of these

criteria, other authors determined that 14.0 to 23.5 percent of older adults living in the

community were using at least one inappropriate drug (Stuck et al., 1994; Wilcox, Himmelstein

and Woolhandler, 1994). Beers' criteria was limited by its failure to include specific reasons why

these "inappropriate" drugs should be avoided, which newer lists have attempted to include

(Buerger, 1998).

Several authors have studied specific types of PDRM in older persons, such as falls and

hip fractures. Ray, Griffin and Downey (1989) studied a population of older persons in

Saskatchewan, Canada, and determined that the risk relative risk of hip fracture was higher for

users of long half-life benzodiazepines (1.7) as compared to those who used short half-life drugs

(1.1). Prudham and Grimley-Evans (1981) concluded that older persons who reported falls in a

one-year period were taking statistically significantly more tranquilizers and diuretics than older

persons who did not report having at least one fall. A case-control study assessed the risk of hip

fractures associated with four classes of psychotropic drugs and determined that there was an

increased risk with concomitant use of long-half-life hypnotic-anxiolytics, tricyclic

antidepressants, and antipsychotics (Ray et al., 1987).

Drug-related morbidity and mortality in older persons has been documented to occur in

many different locations, such as the ambulatory, nursing home, emergency room, and inpatient

settings. In a study conducted in the early 1960's in 178 older persons ambulatory patients with

chronic illness, 59 percent were found to have at least one type of drug-related morbidity and

mortality, with 26 percent of the cases viewed as being serious (Schwartz et al., 1962). Aronow

(1996) looked at the use ofdigoxin in 500 consecutive nursing home admissions and concluded








that 47 percent of the patients had an inappropriate indication for use. In the nursing home

setting, it has been estimated that the total cost of drug-related morbidity and mortality without

the services of consultant pharmacists is $7.6 billion annually (Bootman, Harrison and Cox,

1997). Adams et al. (1987) found that a large percentage of older persons who had an

emergency room visit at a hospital in England had a DRP in the categories of drug interaction

and improper drug selection. Six percent of older persons had a serious drug-diagnosis or drug-

laboratory contraindication and 19.7 percent of patients had a drug-drug interaction, although not

all of these were deemed to be clinically significant (Adams et al., 1987). Ray, Federspeil and

Schaffner's (1980) study of antipsychotic drug use in Tennessee nursing homes suggested that

many older persons were using drugs without an indication. A study of the use of sedative-

hypnotics in hospitalized older persons revealed that 20 percent of the prescriptions exceeded

recognized dosing guidelines and this was associated with a greater severity of illness (Zisselman

et al., 1996).



Identification of Patients at High Risk of Medical Problems in Older Persons


A risk factor is a variable statistically associated with an outcome event. Many

researchers have attempted to identify specific risk factors for health care resource utilization or

morbidity in older persons. These researchers reason that because health care resources are

limited it may be best to find those patients who are at particular risk and concentrate on those

individuals. For example, Nikolaus et al. (1995) identified risk factors associated with

malnutrition in older persons. The main risk factors were a high number of prescription drugs,

social isolation, chronic and painful diseases, and high consumption of alcohol or cigarettes

(Nikolaus et al., 1995). Their study was limited to hospitalized older persons so the authors

admit that further research must be done in other settings. Fowles et al. (1996) compared self-








reported health status (ShortForm-36) and diagnosis (Ambulatory Care Groups) to demographic

information and found that the former two were much better predictors of health care

expenditures in older adults. Therefore, it appears that demographic information is not sufficient

to predict high utilizers of health care and other factors must be considered.

Kramer, Fox, and Morgenstern (1992) describe the approaches taken by seven health

maintenance organizations (HMOs) with Medicare-risk contracts to identify high risk patients.

One of these HMOs, Kaiser Permanente Southern California, identified approximately 35

percent of all inpatient admissions as being a high risk group through the use of the following

criteria: age 80 or above, cerebral vascular accident, new fracture, admitted from a nursing

home, a hospital readmission within ninety days that was unplanned, immobility, activities of

daily living impairment, malnutrition, incontinence, confusion, prolonged bed rest, history of

falls, depression, or existence of social problems (Kramer, Fox and Morgenstern, 1992). Kramer,

Fox and Morgenstern (1992) did not state whether these initiatives to identify high risk patients

were successful.

Stuck et al. (1994) studied patient factors associated with a risk of using an

"inappropriate medication" in older persons by performing a multivariate logistic regression

analysis. It was determined that a depression score was a risk factor, however, age, gender,

income, number of chronic diseases, and activities of daily living score were not predictors

(Stuck et al., 1994). Wilcox, Himmelstein, and Woolhandler (1994) also examined the risk of

using an inappropriate medication in older persons and determined that patient risk factors were

a high number of prescription medications, female gender, people living in the Southern United

States, poor self-rated health status, and Medicaid beneficiaries.

Risk factors associated with drug-related morbidity have also been identified for older

persons. Grymonpre et al. (1988) determined that the risk of a drug-related morbidity in patients

aged 50 and older was related to the number of diseases and number of drugs used, but not to








age, health score, or gender. Hurwitz (1969) reported that predisposing risk factors associated

with drug-related morbidity included an age of 60 or greater, female gender, previous adverse

drug reaction, and history of allergic disease. In a review of the English-language literature,

Gurwitz and Avomrn (1991) found that age is not an independent risk factor for drug-related

morbidity, but rather patient-specific and functional characteristics are more important.

Therefore, it appears that the literature is rich in examples of possible risk factors for PDRM in

older persons.



Prediction Models for Drug Use in Older Persons


There is some evidence in the literature that prediction models can be created to identify

individuals who are at risk of DRPs and these models can facilitate the development of

interventions. Beers et al. (1992) developed an operational definition of inappropriate medication

use in older persons in the nursing home setting and subsequent studies were able to use this

definition to determine the degree of inappropriate use and develop intervention strategies to

help correct this problem. Koecheler et al. (1989) developed six prognostic indicators for

patients who might warrant pharmacist monitoring: (1) five or more medications in present drug

regimen, (2) 12 or more medication doses per day, (3) medication regimen changed four or more

times during the past 12 months, (4) more than three concurrent disease states present, (5)

history of noncompliance, and (6) presence of drugs that require therapeutic drug monitoring.

Evidence of adverse outcomes related to drug therapy was identified in 33.1 percent of charts,

based on the use of these indicators (Koecheler et al., 1989). McGhan, Wertheimer, and Rowland

(1982) used Medicaid data to develop multivariate predictive equations to identify patients with

drug therapy problems.

In a recent study, McElnay et al. (1997) developed a risk model for predicting drug-

related morbidity and mortality in older persons, similar to the approach used in this study. Their

model was able to predict drug-related mortality and morbidity in older persons with a









specificity of 69 percent, a sensitivity of 41 percent, and an overall accuracy of 63 percent

(McElnay et al., 1997). McElnay et al. (1997) identified seven variables which influenced the

risk of drug-related morbidity and mortality: digoxin, antidepressants, chronic obstructive

airways disease, angina, abnormal potassium level, and patient belief that their medication was

in some way responsible for their hospital admission (McElnay et al., 1997).

Wilcox, Himmelstein, and Woolhandler (1994) state that measuring preventable drug-

related morbidity and mortality in the community setting is difficult but it is extremely important

since only a small proportion of drug-related morbidity results in hospital admissions and many

problems may be unreported or unrecognized. Therefore, there appears to be a continued need

for better prediction models of drug-related morbidity and mortality in older persons. Also, the

McElnay study did not consider preventability. Furthermore, there are no operational definitions

of PDRM in the peer-reviewed medical literature. The development of such definitions would

contribute to the conceptual framework of a systems approach to the medication use system.


Research Assumptions and Hypotheses


The research assumptions and hypotheses that will serve as the basis for this study are

proposed within the context of systems theory applied to the medication use system, the

biopsychosocial model of disease etiology and therapy, and the ability to use conditional

probabilities to identify at-risk individuals. As described earlier, while PDRM has been

previously determined to be an important problem within the medication use process, operational

definitions have not been adequately developed and the method to best identify those individuals

at-risk is unknown. Four research questions that address these unresolved problems will be

investigated. Based on the literature reviewed earlier in this chapter, and the conceptual

framework discussed in chapter two, propositions were made for all four of these research

questions.








Research Assumptions


The first research question is directed at the creation of operational definitions of PDRM.

A research assumption related to this research question was proposed. This research assumption

had to be met before the other research questions could be investigated. This research

assumption is as follows:

A1A: Valid operational definitions of PDRM can be developed by a panel of geriatric

medicine experts.

This assumption proposes that operational definitions of PDRM, consisting of criteria for

specific types of PDRMs in older persons, can be developed and tested for validity. Previous

authors have succeeded in creating algorithms for the assessment of adverse drug reactions (one

type of DRP that can lead to PDRM). These algorithms have been tested for validity (Karch and

Lasagna, 1977; Hutchinson et al., 1979; Kramer et al., 1979; Leventhal et al., 1979; Naranjo et

al., 1981). Schumock and Thornton (1992) have created criteria to determine the preventability

of adverse drug reactions, which has been subsequently used by others (Pearson et al., 1994). As

explained in the next chapter, an attempt will be made to demonstrate the validity of the

operational definitions of PDRM, although the lack of an accepted "gold standard" for PDRM is

a limitation.

The consensus method to be used is the Delphi technique. Therefore, research question

one was refined to focus on the usefulness of developing operational definitions of PDRM with

the Delphi technique. The utilization of expert panels via the Delphi technique to generate

consensus on healthcare issues has been quite extensive (Roberts, Sek Khee and Philp, 1994;

Butterworth and Bishop, 1995; Megel, Barna Elrod, and Rausch, 1996). This includes its use to

determine criteria for medication use in older persons. Fouts et al. (1997) used the Delphi

technique to identify risk factors for DRPs in older persons. Beers et al. (1991) used the Delphi

technique with 13 experts to reach consensus on explicit criteria for determining inappropriate

medication use in nursing home residents. Therefore, it seems there is sufficient evidence that










the use of an expert panel in the creation of operational definitions for PDRM in older persons is

reasonable and not without precedent.


Research Hypotheses


The second research question deals with the identification of major risk factors for

PDRM in older persons. The specific hypotheses proposed to address this research question are

the following:

HIA: Digoxin use will be a risk factor for PDRM in older persons.

H1B: Antidepressant drug use will be a risk factor for PDRM in older persons.

H1C: Long-acting benzodiazepine use will be a risk factor for PDRM in older persons.

HID: Antihypertensive drug use will be a risk factor for PDRM in older persons.

HIE: Gastrointestinal disorders will be a risk factor for PDRM in older persons.

H1F: Lung conditions (lung disease, emphysema, bronchitis and asthma) will be a risk

factor for PDRM in older persons.

HI1G: Kidney disease will be a risk factor for PDRM in older persons.

HIHI: A history offalling will be a risk factor for PDRM in older persons.

H1I: Four or more prescribers will be a risk factor for PDRM in older persons.

H1J: Six or more prescription medications will be a risk factor for PDRM in older

persons.

H1K: Four or more recorded diagnoses will be a risk factor for PDRM in older persons.

HIL: A previous adverse drug reaction will be a risk factor for PDRM in older persons.

HIM: High alcohol consumption will be a risk factor for PDRM in older persons.

HiN: Self-assessment ofpoor health status will be a risk factor for PDRM in older

persons.

H10: Trouble paying for medications will be a risk factor for PDRM in older persons.

HIP: Difficulty taking medications will be a risk factor for PDRM in older persons.









H1Q: Patient belief that they are taking too many medications will be a risk factor for

PDRM in older persons.

H1R: Female gender will be a risk factor for PDRM in older persons.



This proposition is based on (1) reports of the risk factors associated with different kinds

of drug-related problems and drug-related morbidity that are found in the peer-reviewed medical

literature, and (2) identification of possible risk factors as they might relate to the medication use

system and the biopsychosocial model of disease etiology and therapy. The rationale for each

individual risk factor is as follows.

Several variables relate monitoring of drug therapy:

HIA: Digoxin use

There is considerable evidence in the medical literature that digoxin use is a risk factor

for adverse drug reactions and drug-related morbidity and mortality. Williamson and Chopin

(1980) determined that digoxin is one of the highest risk drugs for drug-related hospital

readmissions in older persons. Digoxin was the most implicated drug in an inpatient study

involving 193 adverse drug reactions, accounting for 21 percent of all adverse drug reactions

(Ogilvie and Ruedy, 1967b). Digoxin has been labeled one of the top ten problem drugs in older

persons (Dalziel, Byszewski and Ross, 1996), and it was also identified as a risk factor for drug-

related morbidity in older persons (McElnay et al., 1997). A multidisciplinary panel of health

professionals in long-term care listed it as a risk factor for drug-related problems in elderly

nursing home residents (Fouts et al., 1997).

There are several reasons why digoxin use may be a risk factor for PDRM in older

persons. It is a water-soluble drug and has a smaller volume of distribution in older persons,

therefore, it requires a lower dose (Harper, Newton and Walsh, 1989). Digoxin is also implicated

in many drug-interactions and elimination of the drug may be a problem in older persons since

there is an age-related loss of renal function. Digoxin can also cause failure to thrive in older

persons through a diminished appetite (Harper, Newton and Walsh, 1989). Finally, digoxin









toxicity manifestations are often subtle or atypical in older persons (Daiziel, Byszewski and

Ross, 1996).

Many of these problems with digoxin use in older persons may also relate to

inappropriate prescribing, lack of patient advice or poor patient monitoring. As discussed in

chapter two, all these factors can impact PDRM. This is especially true in older persons, as

inadequate patient education on prescribed drugs was a factor that increased the risk of drug-

related morbidity in older persons (French, 1996).

H1B: Antidepressant drug use

There is also considerable evidence in the medical literature for including antidepressant

drug use as a possible risk factor for PDRM in older persons. Certain antidepressants

amitriptylinee and fluoxetine) were identified as being two of the top ten problem drugs in older

persons (Dalziel, Byszewski and Ross, 1996), and McElnay et al. (1997) also determined that

antidepressant use was a risk factor for drug-related morbidity in older persons.

Like digoxin, patient monitoring is extremely important for antidepressants in older

persons. In particular, tricyclic antidepressants are highly bound and there are lower albumin

levels in older persons so the free fraction is greater, increasing the likelihood of drug toxicity

(Harper, Newton and Walsh, 1989). Tricyclic antidepressants have anticholingeric/

antihistaminic side effects that are more evident in older persons.

H1 C: Long-acting benzodiazepine use

There is considerable evidence in the medical literature that long-acting benzodiazepine

use is a risk factor for adverse drug reactions and drug-related morbidity and mortality.

Williamson and Chopin (1980) determined that long-acting benzodiazepines are one of the

highest risk drug classes for drug-related hospital readmissions in older persons, and they were

identified as one of the top ten problem drugs in older persons (Dalziel, Byszewski and Ross,

1996). They were also a risk factor for adverse drug reactions associated with global cognitive

impairment in older persons (Larson et al., 1987). A multidisciplinary panel of health









professionals in long-term care listed it as a risk factor for drug-related problems in elderly

nursing home residents (Fouts et al., 1997).

There are many physiological reasons why long-acting benzodiazepine use may be a risk

factor for PDRM in older persons. Long-acting benzodiazepines are fat-soluble and have a larger

volume of distribution in older persons, leading to increased storage and prolonged half-life

(Harper, Newton and Walsh, 1989). Also, older persons are more sensitive to the effects of

benzodiazepines (Harper, Newton and Walsh, 1989). The oxidative metabolism of long-half life

benzodiazepines is often impaired in older persons (Harper, Newton and Walsh, 1989). As a

result, they can cause depression, falls, confusion and withdrawal symptoms. French (1996)

argues that age-related physiological changes that alter drug kinetics and pharmacological

responses to the prescribed medication are factors that increase the risk of drug-related morbidity

in older persons. Therefore, for long-acting benzodiazepine use, it appears that dosing and drug

monitoring are two critical elements needed to diminish PDRM in older persons.

HID: Antihypertensive drug use

At least two studies have identified antihypertensive drug use as a possible risk factor for

drug-related morbidity. This drug class was a risk factor for adverse drug reactions associated

with global cognitive impairment in older persons in the Larson study (Larson et al., 1987), and

it was one of the highest risk drugs for drug-related hospital readmissions in older persons in

another study (Williamson and Chopin, 1980). Also, centrally-acting antihypertensives/ beta-

blockers was identified as being one of the top ten problem drugs in older persons (Dalziel,

Byszewski and Ross, 1996).

Antihypertensive drugs have been well documented to cause drug-related morbidity in

older persons. Many antihypertensives have central nervous system side effects and may cause

acute confusion, hallucinations, impairment of memory, and reduced ability to perform complex

psychomotor tasks (Harper, Newton and Walsh, 1989). Older persons are also particularly

susceptible to depression and postural hypotension from certain antihypertensives. Therefore, it









appears that monitoring is also an important element of care for older persons who are taking this

class of drugs.

HI E: Gastrointestinal disorders

Gastrointestinal disorders was previously identified in one study has being a risk factor

for drug-related morbidity in older persons (McElnay et al., 1997). This could be because

medications used for gastrointestinal disorders are often used improperly (Tamblyn et al., 1997;

Moride et al., 1997) and many of these medications have been associated with drug-related

morbidity in older persons (Harper, Newton and Walsh, 1989; Dalziel, Byszewski and Ross,

1996). Therefore, it appears that elements three (appropriate prescribing for explicit objectives)

and six (monitoring) of the eight necessities for safe and effective drug therapy may also be

potential problem areas for patients with gastrointestinal disorders.

H1F: Lung conditions (emphysema, bronchitis or asthma)

Chronic obstructive airways disease was previously identified in one study has being a

risk factor for drug-related morbidity in older persons (McElnay et al., 1997). Patients with lung

diseases such as asthma are often on medications that require special monitoring and it has been

argued that these diseases can not be adequately explained by the biomedical model. For

example, severity of asthma symptoms may depend on such things as cleanliness of living

conditions and activities of daily living, which are better explained by the biopsychosocial

model. Therefore including lung conditions as a risk factor appears to be compatible with the

biopsychosocial model of disease and with the importance of monitoring drug therapy.

H 1G: Kidney disease

The presence of kidney disease will be included as a possible risk factor for PDRM

based on both the medical literature and theoretical considerations. A multidisciplinary panel of

health professionals in long-term care listed decreased kidney (renal) function as a risk factor for

drug-related problems in elderly nursing home residents (Fouts et al., 1997). Renal failure was

found to be an associated factor to drug-related morbidity in a retrospective study in Chile

(Zilleruelo, Espinoza and Ruiz, 1987). Older persons may be at a special risk of this. Renal blood









flow and glomerular filtration rate decrease with age (Harper, Newton and Walsh, 1989). This

leads to an elevated drug level and prolonged half-life for drugs excreted by the kidney. This can

cause drugs to accumulate and toxicity to develop. Therefore, prescribing proper doses of many

medications and drug level monitoring for PDRM are very important. As well, patient-specific

differences in renal function are great in older persons, which the biopsychosocial model

considers.

H1H: A history offalling

A history offalling was found to be a risk factor for drug-related morbidity associated

with global cognitive impairment in older persons in one study (Larson et al., 1987). In contrast,

a history offalling was not found to be a risk factor for drug-related morbidity in a later study

(Carbonin et al., 1991).

A history offalling may be a risk factor for PDRM in older persons because many

different drug classes, such as long-acting benzodiazepines, antihypertensives, and others have

been documented to cause falls. Therefore, it seems that appropriate prescribing, patient advice,

patient participation in care, monitoring, and appropriate documentation of previous falls may be

key elements to help prevent PDRM in older persons.

Several variables relate to the importance of communication for optimal drug therapy:

HI 1: Four or more prescribers

Four or more prescribers will be included as a possible risk factor, mainly based on

theoretical considerations. French (1996) did document that several providers prescribing

therapy independently was a factor that increased the risk of drug-related morbidity in older

persons. Theoretically, if a patient has multiple prescribers, PDRM could develop from

competing prescribing objectives, and poor documentation and communication of information

and therapy decisions.

H 1J: Six or more prescription medications

There is considerable evidence in the medical literature that the risk of drug-related

morbidity increases with an increase in the number of medications in the drug regimen. As early









as 1969, Hurwitz observed that patients with drug reactions had significantly more drugs during

their hospital stay then those who did not develop drug reactions. Five or more medications was

found to be the primary risk factor in a study of potential drug-drug interactions (Braverman et

al., 1996), and taking more than four drugs was found to be a risk factor of drug-related

morbidity (Carbonin et al., 1991). A multidisciplinary panel of health professionals in long-term

care said elderly nursing home patients who take nine or more medications are at risk for drug-

related problems (Fouts et al., 1997). Larson et al. (1987) looked at the relationship between the

number of medications and cognitive impairment and determined that the relative odds for

adverse drug reactions related to cognitive impairment was 9.3 for patients taking four or five

drugs, and 13.7 for patients taking six or more drugs. Finally, five or more drugs in a regimen

was a prognostic indicator chosen to identify ambulatory patients who warranted special

pharmacist monitoring (Koecheler et al., 1989).

It appears that as the number of medications in a regimen increases, the opportunity for

inappropriate prescribing, dispensing/administration errors, inadequate patient advice, lack of

patient participation in care, inadequate monitoring, and poor documentation and communication

all increase.

H 1K: Four or more recorded diagnoses

Patients with several diseases appear to be at greater risk for drug-related mortality.

Carbonin et al. (1991) found that more than four active medical problems was a risk factor for

drug-related morbidity. "Patients having more than three diseases" was used as a prognostic

indicator chosen to identify ambulatory patients who warranted special pharmacist monitoring

(Koecheler et al., 1989). A multidisciplinary panel of health professionals in long-term care

listed "more than six active chronic medical diagnoses" as a risk factor for drug-related problems

in elderly nursing home residents (Fouts et al., 1997).

The biopsychosocial model may be important in understanding why patients with several

diseases may be at an increased risk of PDRM. Gurwitz and Avorn (1991), upon reviewing the

medical literature on drug-related morbidity, stated that patient-specific physiologic and









functional characteristics are important in predicting drug-related morbidity. The

biopsychosocial model allows for consideration of the patient-specific differences. As well,

timely recognition of signs and symptoms and appropriate documentation may be even more

important for patients with several concurrent diseases. Patients with several diseases are often in

the care of a general practitioner and one or more specialists, who may not always communicate

their therapeutic plans, and which may be in conflict.

H1 L: A previous adverse drug reaction (ADR)

It appears that patients who had a previous adverse drug reaction are at increased risk for

experiencing another event in the future. A previous adverse drug reaction was previously found

to be a predisposing factor in adverse reactions to drugs (Hurwitz, 1969). A previous adverse

drug reaction was found to be an associated factor with adverse drug reactions in a retrospective

study in Chile (Zilleruelo, Espinoza and Ruiz, 1987). A multidisciplinary panel of health

professionals in long-term care listed it as a risk factor for drug-related problems in elderly

nursing home residents (Fouts et al., 1997). Also, in one study involving 177 patients who had

suffered adverse reactions during hospitalization, 32 percent had a second reaction (Ogilvie and

Ruedy, 1967a).

There are several reasons why these individuals may be at particular risk of PDRM.

They may have physicians, prescribers or patient-specific factors that contribute to poor

prescribing, dispensing, administration, patient advice, patient participation in care, monitoring

or documentation. As well, the biopsychosocial model may be useful to help explain why

specific patients are at risk of experiencing a second drug-related morbidity. The most important

factor, however, may be poor documentation and communication of their previous adverse drug

reaction.

Several variables relate to the biopsychosocial model and patient-specific differences:

HIM: High Alcohol Consumption

There is some evidence in the medical literature that high alcohol consumption is

associated with drug-related morbidity. Alcohol consumption was found to be a risk factor of









drug-related morbidity (Carbonin et al., 1991) and alcohol was listed as one of the top ten

problem drugs in older persons (Dalziel, Byszewski and Ross, 1996). The side effects of alcohol

intake in older persons are potentiated because both metabolism and excretion of alcohol are

altered with aging. In older persons, recognition of alcoholism is often difficult and delayed

because the manifestations may be subtle or erroneously attributable to normal aging (Dalziel,

Byszewski and Ross, 1996). Alcohol, when combined with many medications, can be dangerous

and lead to such events as falls. When a patient consumes alcohol and is taking medications,

patient advice and monitoring are both critical elements. As well, high alcohol consumption is

associated with sociological, behavioral, and psychological factors that can be best explained by

the psychosocial model of disease.

H1N: Self-assessment ofpoor health status

Self-assessment ofpoor health status will be included as a possible risk factor, also

primarily based on theoretical considerations. There is some evidence in the literature that self-

assessment of health as poor is associated with mortality (Fried, Pollack and Tinetti, 1998). In

addition, patient-specific physiologic and functional characteristics are important in predicting

drug-related morbidity (Gurwitz and Avorn, 1991). The biopsychosocial model is important in

helping to explain this as a possible risk factor for PDRM. This model allows the inclusion of

psychological and behavioral elements in illness.

H10O: Trouble paying for medications

Trouble paying for medications will be included as a possible risk factor for PDRM

based on theoretical considerations. A patient who reports that they are having trouble paying for

medications may exhibit poor medication compliance with their medication regimen due to the

cost of the medications. The expense of the drug is a factor that contributes to poor compliance.

Schneider et al. (1991) showed that noncompliance is related to a belief that taking medications

will not result in a successful medical outcome. Thus, trouble paying for medications may also

relate to a reluctance to take medications because of a belief that they will not help to improve

the health of the patient. O'Neil and Poirer (1998) showed that patients with poor perceptions of









their drug regimen had more adverse drug therapy outcomes. One study did determine that older

persons who do not comply with prescribed medicines are at an increased risk of drug-related

morbidity (French, 1996). Again, the biopsychosocial model is important in order to consider

sociological variables in this model. As well, patient participation in care is important to prevent

PDRM, and this includes compliance with medications and attributing medications with an

improved health status.

HIP: Difficulty taking medications

Difficulty taking medications will be included as a possible risk factor for PDRM based

on theoretical considerations. A patient's self-report that they are having difficulty taking

medications may also relate to their attitude of taking medication. Grembowski et al. (1993)

discovered that older persons with a low self-efficacy in health behaviors had poorer health.

Therefore, it is conceivable that many older persons have low self-efficacy in taking medications

and thus develop PDRM. The biopsychosocial model is important in order to consider

behavioral variables in this model. As with the previous variable, patient participation in care is

important to prevent PDRM, and this includes self-efficacy with taking medications. In addition,

French (1996) claims that many older persons experience motor-sensory declines that contribute

to an inability to properly take medications.

H 1Q: Patient belief that they are taking too many medications

Patient belief that they are taking too many medicines will be included as a possible risk

factor for PDRM based on theoretical considerations. This variable may also relate to the

patient's attitude towards medications. Disagreement with the prescribed medication regimen is a

factor that contributes to poor compliance. Highly complex medication regimens are associated

with noncompliance (Haynes, Taylor, and Sackett, 1979). Also, patients who believe they are on

too many medications may have low self-efficacy for taking medications, which may also lead to

poorer health and PDRM. Again, the biopsychosocial model is important in order to consider

behavioral variables in this model. As well, patient participation in care is important to prevent

PDRM, and this includes compliance with medications.









One variable has been included based on empiric evidence, but it does not seem to

directly relate to the conceptual models used in this study:

H1 R: Female gender

Female gender has been identified as being a risk factor for drug-related morbidity in

several studies. Hurwitz (1969) noted that females had significantly more drug-related

morbidities than males. Female gender was found to be an associated factor with adverse drug

reactions in a retrospective study in Chile (Zilleruelo, Espinoza and Ruiz, 1987). The odds ratio

associated with having an ADR in older females as compared to males was found to be 1.9,

although it was not statistically significant (Hallas et al., 1991). Finally, gender was found to be

a determinant of both the frequency and characteristics of ADRs in a prospective drug

surveillance study involving 1920 patients in Chile (Domecq et al., 1980). In contrast,female

gender was not found to be a risk factor for ADRs in one study, although the odds ratio for an

ADR was 1.22 (0.987-1.54, 95 percent confidence interval) (Carbonin et al., 1991). Zadoroznyj

and Svarstad (1990) argue that by excluding female-specific drugs and conditions (e.g.;

pregnancy) there is basically no difference in drug use between males and females.

The reason why female gender may be a risk factor for PDRM does not seem to be easily

explained by any of the models used in this study but will be included for empirical reasons.



The third research question addresses whether there are general risk factors for PDRM

and risk factors which might be drug or disease specific for PDRM. This research question is

obviously closely related to the previous research question. The hypothesis proposed for this

question is as follows:

H2: There will be both general risk factors for PDRM and risk factors that are drug or

disease specific for PDRM



This proposition is based on the reports of the wide variation of risk factors associated

with different kinds of PDRMs that are found in the peer-reviewed medical literature. As the









previous review of the literature showed, there seems to be some general risk factors for PDRM

(poor health status, multiple prescribers, multiple disease states), and some drug-specific

(digoxin, antihypertensives, etc.) and disease-specific (lung disease, kidney disease, etc.) risk

factors for PDRM.



The fourth research question is whether the utilization of health care resources differs

between patients with, and without, PDRM. The specific proposition hypothesized for this

research question is the following:



H3: Older persons that have PDRM will consume more health care resources than those

who do not have PDRM



This hypothesis suggests that there will be a significant difference in the health care

resource utilization patterns between those older persons who do, and do not, have PDRM. Based

on the literature previously reviewed, there is empirical support that those older persons who

experience PDRM do consume more health care resources. Zisselman et al. (1996) concluded

that those older persons who received sedative-hypnotics with doses exceeding the Health Care

Financing Administration (HCFA) guidelines had increased hospital costs and longer lengths of

stay as compared to those who did not receive these drugs or whose dosages did not exceed the

guidelines (although the direction of causality is unknown). The very nature of PDRM often

involves the consumption of valuable resources. For example, Bates et al. (1997) determined that

those patients with a preventable adverse drug event (ADE) had an average increase of 4.6 days

in length of stay and $5857 in total cost, and the annual costs attributable to all preventable

ADEs was estimated to be $2.8 million for a 700-bed teaching hospital. Thus, based on this

evidence, the proposition will be made that those who experience PDRM will have a higher

utilization of healthcare resources.















CHAPTER 4
METHODS



The intent of this study was to (i) determine the issues in developing and using

operational definitions of PDRM, (ii) identify major risk factors for PDRM in older persons, (iii)

examine whether there are general risk factors for PDRM in older persons and drug, or disease,

specific risk factors, and (iv) determine what is the relationship between PDRM and the

utilization of healthcare resources.

To meet these objectives, this study was conducted in two phases. The first phase of the

study concentrated on the dependent variable in this study: preventable drug-related morbidity

(PDRM). In this phase, operational definitions of PDRM were created through the use of a

geriatric medicine expert panel and validated using a chart abstract reviewer panel. The second

phase of the study focused on the independent variables: hypothesized risk factors for PDRM. In

this phase, the risk factors related to PDRM in older persons were identified using the study

database: claims and quality of life data from the Florida Hospital Healthcare System Premier

Care Plan Medicare population. This second phase of the study also involved the creation of a

risk stratification system for PDRM and exploration into the relationship of PDRM and

healthcare resource utilization.


Phase I: Operational Definitions of PDRM


The first research question described above was investigated in Phase I of the study.

Phase I of the study involved the creation of operational definitions of PDRM in older persons.

This was accomplished by a review of the medical literature, and the administration of a survey

to a consensus panel of geriatric medicine experts. The validity of these definitions of PDRM









was explored through the use of a second panel: a chart abstract reviewer panel. These steps of

the Phase I methodology are displayed in Table 4.1.


Operationalization of the Study Construct PDRM


In order to have the geriatric medicine expert panel agree on what is a PDRM, and for

the purpose of the conceptual framework of this study, it was necessary to operationalize the

term PDRM. A PDRM can be defined as an unwanted consequence of the medication use

process that, with appropriate systems, adequately trained personnel and patients or caregivers,

could have detected, predicted, controlled and avoided (Hepler and Strand, 1990). PDRM results

from (a) unacceptable quality of care (e.g., failure to meet consensus guidelines) or (b) occurs

after a drug-relatedproblem (DRP). A review of the literature of PDRM in older persons was

presented in chapter 3 and its relationship to the medication use system and biopsychosocial

model was explored in chapter 2.

An extension of this conceptual definition is that PDRM has four defining

characteristics. The DRP that lead to the PDRM must be recognizable and the likelihood of a

drug-related morbidity must be foreseeable. In addition, the causes) of the DRP (and subsequent

drug-related morbidity) must be identifiable, and those causes must be controllable. Preventable

drug-related morbidity, therefore, results from unrecognized or otherwise unresolved DRPs.

An operational definition of each of these four defining characteristics is as follows:

1. Recognizable. In order for a drug-related morbidity to be recognizable, the DRP

that produced the drug-related morbidity must be observable (Hepler and Strand, 1990). This

was determined by listing specific outcomes (morbidities) and patterns of care. A geriatric

medicine expert panel was then asked to judge whether for most older persons, if health









Table 4.1 Phase I Methodology

Steps of the Phase I Methodology
Review of literature on PDRM, relationship to conceptual framework explored, and
defining characteristics of PDRM studied
Review of literature to identify specific operational definitions of PDRM in older persons
Construction of survey for the Geriatric Medicine Expert Panel, in order obtain consensus
on specific operational definitions of PDRM
Content review of survey by 8 individuals
Pilot test of survey mailed to 40 pharmacists
Revision of survey based on comments/responses from 28 respondents
Selection of Geriatric Medicine Expert Panel members
Administration of survey to Geriatric Medicine Expert Panel -Round 1 of Delphi
Administration of survey to Geriatric Medicine Expert Panel -Round 2 of Delphi
Consensus-approved operational definitions of PDRM obtained from the Geriatric
Medicine Expert Panel
Identification of consensus-approved operational definitions of PDRM in study
population
Abstracted chart reviews of a sample of patients with and without PDRM in study
population
Administration of chart reviews to Chart Abstract Reviewer Panel to test for validity








professionals (physicians, pharmacists, etc.) should be able to recognize significant problems in

this pattern of care.

2. Foreseeable. In order for a drug-related morbidity to be foreseeable, a reasonably

prudent clinician would have recognized that the drug-related morbidity might follow if the

recognized DRP were not resolved. This was determined by listing specific outcomes

(morbidities) and patterns of care. A geriatric medicine expert panel was then asked to judge

whether for most older persons, if health professionals should be able to foresee the possibility

of the outcome, given those problems were not resolved.

3. Causality must be identifiable. Formal attribution algorithms, such as the

Naranjo algorithm, DRAPE algorithm, and the Kramer algorithm, have been used in the past to

identify causality for adverse drug events. Identification of causality for the drug-related

morbidity was determined by listing specific outcomes (morbidities) and patterns of care.

Causality typically involves seeing what to change. A geriatric medicine expert panel was then

asked to judge whether most health professionals should see how to change the pattern of care to

prevent the outcome.

4. Controllable. In order for the cause of a drug-related morbidity to be controllable,

the clinician, patient, or caregiver must have been able to exercise restraint or direction over the

presumed cause of the drug-related morbidity. In order to determine whether the cause of the

drug-related morbidity was controllable, specific outcomes (morbidities) and patterns of care

were listed. A geriatric medicine expert panel was then asked to judge whether most health

professionals should actually change the pattern of care.

A drug-related morbidity was judged to be preventable only if the criteria for all four

defining characteristics were met.

In this study, the standard of care used by the health care professional to assess these

four definitions in specific clinical scenarios was not explicitly stated. It is therefore assumed








that the health care professionals used an implicit standard of care, such as their typical daily

practice or clinical practice guidelines. This approach has the advantage of letting the health care

professionals use their own professional judgement and years of clinical experience to determine

whether a specific clinical scenario is a case of PDRM.



Review of Literature to Identify Specific Operational Definitions of PDRM in Older Persons


As previously discussed in Chapter 3, the literature contains a comprehensive account of

the most commonly occurring drug-related morbidities in older persons. Many of these

morbidities have been hypothesized to be preventable. The literature on drug-related morbidity

since 1967 was reviewed for possible types of preventable drug-related morbidity. The search

was limited to those morbidities that occur in older persons. Peer-reviewed medical articles and

referenced texts were included in the literature review.

Based on the literature review, a list of 50 clinical scenarios (representing possible

PDRMs occurring in older persons) was compiled. These clinical scenarios had to meet the

following inclusion criteria: (1) well-referenced, (2) occur fairly commonly in the geriatric

population, (3) result in serious adverse outcomes, and (4) searchable in the study database.

Clinical scenarios involving specific laboratory values or drug dosages were excluded as this

information was not available in the study database.


Survey Development


A survey instrument was constructed in order to evaluate whether these 50 clinical

scenarios met the four defining characteristics of a PDRM. All clinical scenarios were listed in a

similar format to facilitate reading. In this format, the outcome (morbidity) was listed first, with

the pattern of care which lead to the outcome, listed second.

Expert review and pilot testing were used to assess the content validity of the survey.

The survey, instructions for use, and cover letter were reviewed by eight experts in health









services research from the Department of Pharmacy Health Care Administration at the

University of Florida for ease of use and to determine the relevance of the questions to existing

conceptual frameworks. Based on their feedback, slight modifications were made.

A convenience sample of community, hospital, managed care, consultant, and academic

clinical pharmacists from geographically diverse parts of the United States and Canada was

selected for the pilot test. Before receiving the survey, all individuals were contacted by e-mail,

fax or telephone to inform them to expect the survey. The list of 50 clinical scenarios was split in

half and each half was given to 20 content reviewers, along with a cover letter and instructions

for use. A total of 28 content reviewers completed the survey (15 completed the first half, 13

completed the second half), made comments, and returned the survey in time for analysis, for a

70 percent response rate.

Based on their feedback, slight modifications were made to the survey instrument. As

well, based on the pilot test results, some clinical scenarios were dropped from the survey, and

several new clinical scenarios were added to the survey instrument, leaving a total of 48 clinical

scenarios.


Geriatric Medicine Expert Panel


In order to reach consensus on which clinical scenarios listed in the survey instrument

were actual PDRMs, the Delphi technique was used. As Goodman (1987) explains:

The Delphi technique is a survey method of research which aims to structure group
opinion and discussion. It was first developed in the 1950s by the Rand Corporation in
California as an attempt to eliminate interpersonal interactions as the controlling
variables in decision making, as usually happens when groups of experts interact in
meetings. Its purpose is to generate discussion and enable ajudgement on a specified
topic to be made so that policy decisions can be taken which can claim to represent a
given group's wants and views (Goodman, 1987, pp.729).

Beers et al. (1991) argue that consensus methods, such as the Delphi technique, are useful

because: (1) differences in published opinion may be overcome, (2) they can help create criteria








that address narrow clinical scenarios, and (3) they can incorporate supplemental information,

such as physiological changes in older persons.

Duffield (1993) argues that the choice of panel members is critical in order for the

Delphi technique to work correctly. Participants should be chosen based on their willingness to

participate and their expert knowledge base (Goodman, 1987). With this in mind, a panel of

seven members was chosen by the Chief Medical Officer of the Florida Hospital Healthcare

System, with input from the principal investigator and the Director of Ambulatory Pharmacy at

the Florida Hospital. This panel consisted of physicians with recognized credentials in geriatric

medicine, physician administrators, and a geriatric specialty clinical pharmacist at the Florida

Hospital. These individuals were all thought to be opinion leaders within the Florida Hospital

Healthcare System and have extensive expertise in geriatric medicine. The Geriatric Medicine

Expert Panel members are listed in Appendix A.

The principal investigator explained the survey in depth to each panel member before

they completed the survey. This is because commitment and understanding of the Delphi

technique at the start of the technique influences the time and consideration given to the

technique by the participants (Goodman, 1987). Appendix B contains the cover letter,

instructions for use and survey instrument for the panel members. As Appendix B shows, the

panel members were asked to judge whether each of the 48 clinical scenarios met the four

defining characteristics of a PDRM. One clinical scenario was listed twice in the survey to serve

as a validity check, so there were actually 47 unique clinical scenarios. In addition, there was an

open-ended question at the end of the survey, where panel members could suggest any additional

operational definitions of PDRMs. Prior to the commencement of the Delphi rounds, the

inclusion of operational definitions of PDRM was set as those that were chosen by a majority of

panel members (at least four out of seven members). All seven panel members completed and

returned the survey.








The round two survey was sent to the same seven panel members the following month.

This survey contained the clinical scenarios that had survived round one and several new clinical

scenarios that were suggested by the panel members. For each clinical scenario, the panel

members were given their response (yes/no) from the previous round, the total group response,

and all the comments made by the panel members. By providing comments from the previous

round, consensus is reached quicker, usually in two rounds (Duffield, 1993). Round three

consisted of the results from round two and a letter thanking each expert panel member.



Identification of Consensus-Approved Operational Definitions of PDRM in Database


The consensus-approved operational definitions of PDRMs were identified by first

examining the study database for the outcomes related to the specific operational definitions of

PDRM. This was performed by searching for the diagnosis codes related to these outcomes.

Then, once the outcomes (morbidities) were identified, each patient case was individually

searched to determine whether the associated pattern of care that led to the outcome was

provided or not. If both the outcome and pattern of care matched the specific operational

definition of the PDRM, then it was judged to have been a case of PDRM.



Validation of Operational Definitions of Preventable Drug-Related Morbidity


A chart abstract reviewer panel of five clinical pharmacists was used to further validate

the Geriatric Medicine Expert Panel consensus-approved operational definitions of PDRM.

The first step was to choose the specific operational definitions of PDRM which

occurred often enough in the study database to allow adequate determination of sensitivity and

specificity (confidence intervals that did not include zero). The size was chosen to allow

determination of a sensitivity and specificity of 67 percent. This would be better than the








sensitivity (41 percent) and near the specificity (69 percent) of an existing model to detect total

adverse drug events in older adults (McElnay et al., 1997). It would also be comparable to

models developed to predict adverse drug reactions related to digoxin (sensitivity 92.9 percent,

specificity 61.8 percent) and theophylline (sensitivity 95.8 percent, specificity 84.0 percent)

(Tschepik et al., 1990). Based on this target sensitivity and specificity, only two specific

operational definitions of PDRM occurred often enough to be tested for validity: (1) patients

with secondary myocardial infarction who did not receive ASA and/or a beta-blocker, and (2)

patients with an emergency room visit and/or hospitalization due to hyperglycemia who were on

an oral hypogylcemic and did not have regular hemoglobin Alc monitoring.

Second, abstracted chart reviews of these patients were performed. These chart abstracts

were performed by a primary care pharmacy resident at the Florida Hospital. The instructions for

the chart abstracts, the patient chart abstract review form, and samples are shown in Appendix C.

The chart abstracts were all performed in the same format to allow for ease of reading by the

Chart Abstract Reviewer Panel members. The chart abstracter was given the patient medical

record numbers for all the patients who had the outcome regardless of whether they had the

pattern of care related to that defined preventability. The chart abstracter was blinded as to

whether the patient he was reviewing did, or did not, have a case of PDRM, as defined by the

specific operational definition. All chart abstracts were done from the inpatient charts at the

Florida Hospital and the outpatient laboratory computer system at the Florida Hospital. Chart

abstracts could not be completed for 4 patients with the secondary myocardial infarction

outcome and 1 patient with the hypergylcemia outcome because the medical charts could not be

located. In all, 35 chart abstracts were performed for patients with secondary myocardial

infarction and 31 patients with hypergylcemia.

All of the chart abstracts (n=66) were then given to the Chart Abstract Reviewer Panel,

consisting of five pharmacists at the Florida Hospital. The pharmacists were chosen by the








Clinical Coordinator of Pharmacy at the Florida Hospital based on their availability, willingness

to participate and experience working with the study population. Appendix D shows the

background of the panel members. The principal investigator and chart abstracter met with the

panel members to explain the instructions for use (see Appendix E), and to answer any questions

the panel members might have. These pharmacists were given the two relevant consensus-

approved operational definitions of PDRM and were asked to use them in determining whether

the patients in the chart abstracts actually experienced a PDRM. If four or more of the five panel

members judged a chart abstract to be a case of PDRM, then it was categorized as a PDRM. If

three or fewer panel members judged a chart abstract not to be a case of PDRM, then it was

categorized as not being a case of PDRM. The Fleiss measure of overall agreement was used to

calculate the degree of agreement among the five raters (pharmacists) for classifying each patient

(Fleiss, 1971). Others have advocated using this statistic in situations such as this, when there are

more than two raters (Conger, 1980; Abedi, 1996).

From the results of the Chart Abstract Reviewer Panel, the sensitivity and specificity of

these two specific operational definitions of PDRM was calculated. Sensitivity was calculated as

the percentage of true PDRMs (defined as a PDRM by four or more of the panel members) that

the operational definitions of PDRM labeled as such. Specificity was calculated as the

percentage of abstracted chart reviews that were judged by the Chart Abstract Reviewer Panel

not to be a PDRM, and the operational definition of PDRM labeled as not being a PDRM. A two-

by-two table of true PDRMs and predicted PDRMs was constructed, based on the sensitivity and

specificity.

It should be noted that for most screening tests or instruments, the sensitivity and

specificity of that test is calculated by comparing the outcome of the test to a "gold standard". In

this case there is no generally accepted "gold standard" for PDRM. However, if chart review by

experts is considered to be such a standard, then the panel of five clinical pharmacists is being








used as the "gold standard", acknowledging certain limitations with this approach (see

limitations section in chapter 6). The use of chart reviews by experts is commonly used as a

"gold standard" in many other areas of healthcare, such as peer review organizations (PROs). In

such cases, as in this study, the professional judgement of medical experts reviewing patient

charts is used as the "gold standard".



Phase II: Identification of Risk Factors for PDRM


The methodology for the independent variables will now be discussed. The final three

research questions described at the beginning of this chapter were investigated in Phase II of the

study. Phase II of the study involved the identification of risk factors for PDRM. This was

accomplished by reviewing the medical literature, identifying possible risk factors, and relating

them to the conceptual framework and models used in this study. A database, consisting of

enrollees from the Florida Hospital Healthcare System Premier Care Plan, was constructed to

allow for the measurement of these risk factors. Next, a series of statistical analyses were

performed to identify the risk factors. Finally, the relationship between PDRM and healthcare

resource utilization was explored. These steps of the Phase II methodology are outlined in Table

4.2.



Selection of Possible Risk Factors for PDRM


As previously discussed in Chapters 1 and 3, the literature contains a rich account of risk

factors for the most commonly occurring drug-related morbidities in older persons. The literature








Table 4.2 Phase II Methodology

Steps of Phase II Methodology
Review of literature on risk factors for PDRM and the relationship to conceptual framework
and models explored
Hypotheses generated related to specific risk factors for PDRM in older persons, and
semantic hierarchy developed
Study population identified
Construction of study database in order to identify and measure these hypothesized risk
factors for PDRM
Logistic regression model of all 18 hypothesized risk factors for PDRM
Factor analysis of all 18 hypothesized risk factors for PDRM
Additional logistic regression models to identify other risk factors for PDRM
Risk stratification system developed
Relationship of PDRM and healthcare resource utilization explored








on drug-related morbidity since 1967 was reviewed for possible risk factors for PDRM. Peer-

reviewed medical articles and referenced texts were included in the literature review.



Semantic Hierarchy of Risk Factors for Preventable Drug-Related Morbidity


Next, the relationship of these possible risk factors to the models (pharmaceutical care

and biopsychosocial) discussed in chapter 2 was explored. Other possible variables were also

identified, based on these models. Out of this process, 18 possible risk factors for PDRM were

selected.

Based on this, a semantic hierarchy of risk factors for PDRM was developed and

hypotheses related to the 18 risk factors were stated (chapter 3). A semantic hierarchy, as seen in

Table 4.3, relates constructs to variables, and variables to measurements. Several possible risk

factors relate to monitoring, as described by the medication use system: digoxin use,

antidepressant drug use, long-acting benzodiazepine use, antihypertensive drug use,

gastrointestinal disorders, lung conditions, kidney disease, and a history offalling. Several

possible risk factors relate to patient-provider communication, as described by the medication

use system: four or more prescribers, a previous adverse drug reaction, six or more prescription

medications, and four or more recorded diagnoses.

Several possible risk factors relate to patient-specific aspects of drug use, as described

by the biopsychosocial model: difficulty taking medications, high alcohol consumption, self-

assessment ofpoor health status, trouble paying for medications, and patient belief that they are

taking too many medications. Female gender did not seem to correspond to the conceptual

framework but was included for empirical reasons. Several risk factors also seem to fit more than

one construct: for example, six or more prescription medications may also be related to poor








Table 4.3 Semantic Hierarchy in Risk Factors for Preventable Drug-Related Morbidity


Constructs/ Preventable drug-related morbidity can be predicted by certain
Concepts risk factors.
These risk factors relate to the key concepts of the medication use
system and biopsychosocial model
Monitoring Communication Patient
Specific aspects
Variables Digoxin use Four or more Difficulty taking
Antidepressant drug prescribers medications
use A previous adverse High alcohol
Long-acting drug reaction consumption
benzodiazepine use Six or more Self-assessment of
Antihypertensive drug prescription poor health status
use medications Trouble paying for
Gastrointestinal Four or more medications
disorders recorded diagnoses Patient belief that
Lung conditions they are taking too
Kidney disease many medications
A history offalling

Measurements Digoxin use: Four or more Difficulty taking
(observables) FHHS and PCS claims prescribers: medications:
Examples data PCS claims data Personal Wellness
Dichotomous: l=yes, Dichotomous:l 1=4 or Profile, "How
no=0 more prescribers, 0=3 difficult is taking
or less medications for
you?"
Dichotomous: 1 =
difficult ...very
difficult/can't do it,
O=_________________________0not difficult_








monitoring as well as communication. A more comprehensive discussion of each risk factor is

contained in chapter 3.



Study Population


In order to measure these risk factors, a study population was selected. The study

population was drawn out of the larger pool of enrollees in the Florida Hospital Healthcare

System Premier Care Health Plan. This was a health plan offered by the Florida Hospital

Healthcare System, a provider-sponsored network with a Medicare contract. It was available to

all Medicare beneficiaries who live in Orange, Osceola, and Seminole Counties of Florida and

who were also enrolled in Medicare Part B. By U.S. federal law, however, those individuals who

elected to receive the Medicare hospice benefit and those who had end-stage renal disease were

not eligible for enrollment. Only those enrollees who completed the Personal Wellness Profile

(PWP) Senior Assessment were included in the study. Enrollment into the Premier Care Health

Plan began in January 1997, with approximately 7,000 enrollees by December 1997 and

approximately 50 percent of these individuals completing the PWP. This comparable with the

completion of health risk assessment tools in other Medicare programs (Kerekes and Thornton,

1996). Individuals who were enrolled in the plan anytime during 1997 were included in the study

population.



Data Collection and Formation of the Study Database


The data used in this study consisted of(l) claims which were already collected as a

natural part of the administration of the Premier Care Health Plan and (2) the Personal Wellness

Profile Senior Assessment instrument which was completed by the enrollees. All claims








processed and surveys completed between January 1 and December 31, 1997 were included in

the study.

More specifically, the study database consisted of three parts:

(1). Florida Hospital Healthcare System claims data. This data included all claims made

in the outpatient and inpatient settings for this population, except for prescriptions filled in the

ambulatory setting.

(2). PCS Outpatient Prescription Claims. This data includes all prescriptions filled for

the plan enrollees in the ambulatory setting.

(3). Personal Wellness Profile (PWP) Senior Assessment. The PWP is an instrument

that was given to all Premier Care plan members to complete (see Appendix F for the entire

questionnaire). It is an instrument used to identify enrollees at high risk for health-related

problems. It has been previously used by other health plans and its predictive validity has been

verified and studied by Boult et al. (1994), Pacala, Boult, and Boult (1995), and Pacala et al.

(1997). The PWP contains valuable information related to physical and functional status, which

previous authors have shown to be related to utilization of health care services by older persons

(Branch et al., 1981).

A central database containing these three data sets was constructed by a medical

artificial intelligence company in Orlando, FL called MEDai. This database was completed on

April 15, 1998 with approximately 95 percent of claims from 1997 processed at this time. A

unique patient identifier was used to link these three data sets. All patient names were masked to

protect patient confidentiality. The data set was provided to the principal investigator who

discarded irrelevant data fields. Many variables were dichotomized for the purpose of the study.

Considerable data manipulation was required for three of the study variables. While the study

database as prepared by MEDai did contain a listing of all prescription drugs the patients had

received, the drugs had to be organized into therapeutic classes for three of the study variables








(long-acting benzodiazepine use, antidepressant drug use, and antihypertensive drug use) for

ease of use. For these three variables, an on-line database called Lexicon was used to group all

these drugs by national drug code (NDC) number to the appropriate therapeutic class (Multum

Information Services, 1998). The accuracy of the Multum classification system was assessed for

one drug in each of these three therapeutic classes (triazolam, luvoxamine, and diltiazem). All

the patients in the study database who received these drugs were identified and it was determined

whether the drug was placed into a therapeutic class, and if so, whether it was the correct class.

Table 4.4 contains all 18 hypothesized risk factors and shows how they will be measured in this

database. Table 4.5 contains the additional demographic variables to be considered for inclusion

into the regression models, based on the bivariate analysis.



Statistical Analysis



Logistic Regression Model with the 18 Hypothesized Risk Factors

Initial data analysis was performed using a forward inclusion logistic regression

procedure to determine which of the 18 hypothesized risk factors were significantly associated

with PDRM. Larson et al. (1987) previously used this technique to determine which variables

were associated with global cognitive impairment adverse drug reactions in older persons.

McElnay et al. (1997) also used this technique to determine which variables were risk factors

associated with drug-related morbidity in older persons in the inpatient setting. SAS (SAS

Institute Inc., 1993) and JMP IN (Sail and Lehman, 1996) were used to create the regression

models.








Table 4.4 Hypothesized Risk Factors and Their Measurement


Hypothesized Risk Factor Measurement
Variables related to monitoring
Digoxin use Dichotomous: l=yes, 0=no
Measured from PWP drug question 1.
Antidepressant drug use Dichotomous: l=yes, O=no
~~~____~__________Measured from PCS claims data.
Long-acting benzodiazepine use Dichotomous: 1=yes, 0=no
Measured from PCS claims data.
Antihypertensive drug use Dichotomous: I =yes, 0=no
____________________Measured from PCS claims data.
Gastrointestinal disorders (ulcers or Dichotomous: l=yes, O=no
gastrointestinal bleeding) Measured from PWP question 14.
Lung conditions (emphysema, Dichotomous: I =yes, 0=no
bronchitis or asthma) Measured from PWP question 14.
Kidney disease Dichotomous: 1 =yes, O=no
Measured by PWP question 14.
A history offalling Dichotomous: l=yes, 0=no
_____________________ Measured by PWP question 24.
Variables related to communication
A previous adverse drug reaction Dichotomous: 1 =yes, O=no
Measured by PWP question 3, "Have you had a side
effect due to a medication that caused you to stop that
medication in the last 6 months?"
Four or more prescribers Dichotomous: 1= 4 or more prescribers, 0-=three or fewer
prescribers
Measured through the PCS outpatient prescription data.
Six or more prescription medications Dichotomous: I =six or more prescription medications,
0=five or fewer prescription medications
The number of medications taken by a patient was
~~~____~~______measured by PWP question 24.
Four or more recorded diagnoses Dichotomous: 1 =four or more disease states, 0=three or
fewer disease states. Measured by PWP question 14.
Variables related to biopsychosocial aspects of drug therapy
Self-Assessment ofpoor health status Dichotomous: 1 =poor, 0=better than poor (other)
____________________________ Measured by PWP question one.
Trouble paying for medicines Dichotomous: I =yes, O=no
Measured by PWP question 23, "Do you have trouble
___________________ paying for your medicines?"
Difficulty taking medications Categories: 1= Difficult or Very difficult/can't do it,
0=Not difficult
~~________~~___Measured by PWP question 31.









Table 4.4 Hypothesized Risk Factors and Their Measurement (Continued)


I Hypothesized Risk Factor Measurement
High Alcohol Consumption Dichotomous: 1=yes, 0=no
Measured by PWP question 11, "Do you often have more
_____________________ than 1 to 2 alcoholic drinks in a day?"
Patient belief that they are taking Dichotomous: 1= Patient thinks they are on too many
too many medications medications, 0= patient does not think they are on too
many medications.
Measured by PWP drug question 4, "How do you feel
about the number of medications you are taking?"
Other variables of interest
Female gender Dichotomous: l =female, 0=male
IMeasured by the FHHS data.








Table 4.5 Additional Demographic Variables and Their Measurement


Additional Demographic Variable Measurement
Arthritis Dichotomous: l=yes, 0=no
_______________________ Measured by PWP question 14.
Bladder/Bowel Control Problems Dichotomous: l=yes, 0=no
_______________________ Measured by PWP question 14.
Blind/Trouble Seeing, Even With Glasses Dichotomous: l=yes, O=no
Measured by PWP question 14.
Cancer (Non-skin) Dichotomous: l=yes, 0=no
_______________________ Measured by PWP question 14.
Congestive Heart Failure Dichotomous: l=yes, 0=no
_______________________ Measured by PWP question 14.
Coronary Disease Dichotomous: 1=yes, 0=no
_______________________ Measured by PWP question 14.
Angina Dichotomous: l=yes, 0=no
___________________ Measured by PWP question 14.
Myocardial Infarction Dichotomous: l=yes, 0=no
_______________________ Measured by PWP question 14.
Sciatica Dichotomous: 1=yes, 0=no
Measured by PWP question 14.
Deafness or Trouble Hearing Dichotomous: l=yes, O=no
Measured by PWP question 14.
Diabetes (High Blood Sugar) Dichotomous: l=yes, 0=no
Measured by PWP question 14.
High Blood Pressure Dichotomous: 1 =yes, 0-=no
~~~____~~_________Measured by PWP question 14.
Memory Problems (More Than Typical) Dichotomous: l=yes, 0=no
Measured by PWP question 14.
Stroke Dichotomous: l=yes, O=no
~~~~____~~~______Measured by PWP question 14.
Self-Assessment of Much Worse Health Dichotomous: l=much worse health status, 0=better
Status than much worse
Measured by PWP question 2.
Smoker Dichotomous: l=yes, 0=no
______________ Measured by PWP question 11.
Use of Six or More Over-The-Counter Dichotomous: l=yes, 0-five or fewer OTCs
Medications (OTCs) Measured by PWP question 22.
Warfarin use Dichotomous: 1 =yes, 0=-no
~________~~__Measured by PWP drug question 1.
Theophylline use Dichotomous: l=yes, 0=no
_______________________ Measured by PWP drug question 1.
Cimetidine use Dichotomous: 1 =yes, 0=no
_______________________ Measured by PWP drug question 1.
Phenytoin use Dichotomous: l=yes, 0=no
Measured by PWP drug question 2.
Lives Alone Dichotomous: =yes, 0=no
Measured by PWP question 16.









Table 4.5 Additional Demographic Variables and Their Measurement (Continued)


Hypothesized Risk Factor Measurement
Three or more hospitalizations in Dichotomous: 1 =yes, 0--==two or fewer
previous year hospitalizations
_______________________ Measured by PWP question 17.
Three or more ER visits in previous year Dichotomous: 1 =yes, 0=--two or fewer ER visits
Measured by PWP question 18.
Five or more MD clinic visits in previous Dichotomous: l=yes, 0=four or fewer MD clinic
year visits
Measured by PWP question 19.
Nursing home residence Dichotomous: 1 =yes, 0=no
Measured by PWP question 20.
Use of durable medical equipment Dichotomous: l=yes, 0=no
(oxygen, hospital bed, wheelchair, Measured by PWP question 25.
walker)
Use of home health services (visiting Dichotomous: 1 =yes, 0-no
nurse, physical therapy, homemaker/aide, Measured by PWP question 26.
adult day care)









Factor Analysis

The second step of the statistical analysis was a factor analysis with a varimax

orthogonall) rotation of principal components. This was done on a random selection of 2500

patients from the study population, with the remaining 835 patients serving as a validation group.

All 18 hypothesized risk factors were included in the factor analysis. After the factor analysis

was completed, the number of factors identified and factor scores were studied. While factor

analysis does make assumptions about the normality of the data, dichotomous data can be used

with factor analysis with confidence, provided that sample sizes are large enough (greater than

200 observationsXParry and McArdle, 1991).

The objective of the factor analysis was to (1) reduce the rather large number of

hypothesized variables to a relatively small number of factors, or common traits, and (2)

determine whether these factors matched the structure proposed in the semantic hierarchy. Factor

analysis accomplishes the first objective by focusing on the part of the total variance that is

shared by the variables, assuming that variables consist of common parts. The initial

communality estimates were set as one. The results were kept in perspective as factor analysis is

intended only to be used as a tool to help guide the researcher, not to be used without

consideration of the conceptual framework being used and other factors (Maraun, 1996).

One of two approaches could be used, based on the results of the factor analysis. Had the

factors matched the semantic hierarchy and shown to represent constructs with confidence, then

principal component scores based on these factors would have been entered into a logistic

regression model. However, because the factors did not match the semantic hierarchy and could

not be shown to represent constructs with confidence, the risk factors from the first regression

model were entered into another regression model with the additional variables, taking the factor

analysis results into consideration.








Logistic Regression Models with Additional Variables

Additional logistic regression models were then run, with the risk factors from the first

model entered a priori to adjust for their effects on PDRM. Other demographic variables were

also allowed to enter the model to see if they added significantly to the prediction, based on

statistical significance in a bivariate analysis between patients who did, and who did not, have

PDRM. Because there was a theoretical basis for including the risk factors from the first model,

it was felt that the final model for PDRM must include all of these risk factors, even if it

explained less of the variance of PDRM. Thus, this process incorporated both statistical and

theoretical criterion for deciding which terms to include in the model and this helped to focus

attention on those variables that fit into the conceptual framework and that had the greatest

independent effect on PDRM.

Following the creation of the final prediction model, a risk stratification system was

developed. Patients were categorized according to the number of risk factors they had and a

comparison was made between the patients with, and without, PDRM.


PDRM and Healthcare Resource Utilization


The final component of the methodology was to do another bivariate analysis to

determine the relationship of PDRM to healthcare resource utilization. Each enrollee was

classified as either having PDRM or not. Then, the utilization of health care resources was

compared for the two groups.














CHAPTER 5
RESULTS




The results of this study will be presented in two parts. First, the results of the Delphi

technique and the creation of the operational definitions of PDRM will be shared. Second, the

results of the prediction models for PDRM and risk factors identified will be presented.



Delphi Technique The Geriatric Medicine Expert Panel


As was discussed in the previous chapter, the Delphi technique was used in an attempt to

generate consensus on PDRM. Two rounds were used until consensus was obtained. Appendix G

contains the final list of consensus-approved operational definitions of PDRM. This appendix

also includes all the comments made by the Geriatric Medicine Expert Panel members in either

round I or round 2.

The expert panel agreed that 52 of the clinical scenarios presented to them were actual

PDRMs. Table 5.1 shows the opinions of the panel members after round 1 and round 2. After

round 1, the panel members' agreement with the clinical scenarios ranged from 60.4 percent to

97.9 percent. After round 2, the agreement ranged from 82.8 percent to 100 percent.

Initially, the panel members were presented with 47 unique clinical scenarios. One of

these was listed twice, as a validity check, so the panel members were actually presented with 48

outcomes and patterns of care to evaluate. The validity check received the same score from all

seven panel members for both rounds. After the first round, two clinical scenarios were rejected

(received fewer than four "yes" votes). The panel members were given the opportunity to suggest








Table 5.1 Geriatric Medicine Expert Panel Results


Expert Number Percentage (%) of clinical Percentage (%) of clinical
scenarios the expert felt were scenarios the expert felt were
PDRMs after Round 1 PDRMs after Round 2
1 95.8 100
2 97.9 82.8
3 79.2 89.7
4 60.4 91.4
5 79.2 87.9
6 77.1 91.4

7 97.9 94.8








other operational definitions of PDRMs and 12 were generated. One of these 12 new operational

definitions was, in fact, a duplicate of a previous definition. These 12 new operational definitions

were added to the remaining 46 operational definitions of PDRM and given to the panel

members in round 2. After the second round, an additional four clinical scenarios were rejected,

the two duplicates were removed, leaving 52 operational definitions of PDRM that were

approved.

After round 2 there appeared to be overwhelming consensus on which clinical scenarios

were actual PDRMs. Of the 52 clinical scenarios deemed to be PDRMs by the expert panel, 35

clinical scenarios had the agreement of all seven panel members, 15 clinical scenarios had the

agreement of six out of the seven members, and two clinical scenarios had the agreement of five

out of the seven members. There were no clinical scenarios that only had the agreement of four

panel members. Table 5.2 lists shows how each panel member voted in round 2. Throughout the

two round process, there were six clinical scenarios which did not have at least the support of

four panel members. Table 5.3 lists these clinical scenarios that were rejected as being PDRMs.



Identification of Consensus-Approved Operational Definitions of PDRM in Database


The 52 PDRMs approved by the geriatric medicine expert panel were identified by first

examining the study database for the outcomes related to the specific operational definitions of

PDRM. Overall, 1005 patients with outcomes related to one of the 52 consensus-approved

PDRMs were identified. Next, each patient with one of these outcomes was individually studied

to determine whether the pattern of care associated with a PDRM was provided or not. The

outcome and pattern of care matched a consensus-approved operational definition of PDRM in

158 cases. This represented 97 patients, as several patients had more than one specific









Table 5.2 Round 2 of the Delphi Technique


Number Expert Expert Expert Expert Expert Expert Expert Total
1 2 3 4 5 6 7 "Yes"
I Y Y Y Y Y Y Y 7
2 Y Y Y Y Y Y Y 7
3 Y Y Y Y Y Y Y 7
4 Y Y Y Y Y Y Y 7
5 Y Y Y Y Y Y Y 7
6 Y Y Y Y Y Y Y 7
7 Y Y Y Y Y Y Y 7
8 Y Y Y Y Y Y Y 7
9 Y Y Y Y Y Y Y 7
10 Y Y Y Y Y Y Y 7
11 Y Y Y Y Y Y Y 7
12 Y Y Y Y Y Y Y 7
13 Y Y Y Y Y Y Y 7
14 Y Y Y Y Y Y Y 7
15 Y Y Y Y Y Y Y 7
16 Y Y Y Y Y Y Y 7
17 Y Y Y Y Y Y Y 7
18 Y Y Y Y Y Y Y 7
19 Y Y Y Y Y Y Y 7
20 Y Y Y Y Y Y Y 7
21 Y Y Y Y Y Y Y 7
22 Y Y Y Y Y Y Y 7
23 Y Y Y Y Y Y Y 7
24 Y Y Y Y Y Y Y 7
25 Y Y Y Y Y Y Y 7
26 Y Y Y Y Y Y Y 7
27 Y Y Y Y Y Y Y 7
28 Y Y Y Y Y Y Y 7
29 Y Y Y Y Y Y Y 7
30 Y Y Y Y Y Y Y 7
31 Y Y Y Y Y Y Y 7
32 Y Y Y Y Y Y Y 7
33 Y Y Y Y Y Y Y 7
34 Y Y Y Y Y Y Y 7
35 Y Y Y Y Y Y Y 7
36 Y Y Y Y Y N Y 6
37 YV N Y Y y y Y 6
38 Y N Y Y y y y 6
39 Y N Y Y y y y 6
40 Y Y N Y Y Y Y 6
41 Y Y Y Y N Y Y 6
42 Y N Y Y y y 6








Table 5.2 Round 2 of the Delphi Technique (Continued)


Number Expert Expert Expert Expert Expert Expert Expert Total
1 2 3 4 5 6 7 "Yes"
43 Y Y N Y Y Y Y 6
44 Y Y N Y Y Y Y 6
45 Y Y Y Y N Y Y 6
46 Y Y Y N Y Y Y 6
47 Y Y Y Y Y N Y 6
48 Y N Y Y Y Y Y 6
49 Y Y Y Y N Y Y 6
50 Y N Y Y Y Y Y 6
51 Y Y N Y N Y Y 5
52 Y N Y Y Y Y N 5
53b Y N Y N N N Y 3
54b Y Y N N Y N N 3
55b Y N N N N Y N 2
56b Y N Y N N N Y 3
57c y Y y y y y y 7
58d Y___ Y Y Y Y Y Y 7


a See Appendix G for a description of the PDRM
b Rejected as an operational definition of PDRM
c Duplicate same as PDRM #2
d Duplicate same as PDRM #17








Table 5.3 Clinical Scenarios That Were Rejected As Preventable Drug-Related Morbidities

Clinical Scenario (Outcome and Pattern of Care)
Outcome: Fall and/or hip fracture and/or other bone fracture and/or bone break
Pattern of care: 1. Use of an anti-parkinsonian agent (e.g.; levodopa, bromocriptine,
Benztropine, etc.)
Outcome: Major and/or minor hemorrhagic event
Pattern of care: 1. Use of SQ heparin
2. PTT not done at least every month
Outcome: Digoxin toxicity
Pattern of care: 1. Use ofdigoxin
2. BUN/serum creatinine not done at least every 6 months
3. Digoxin level not done at least every 6 months
Outcome: Fall and/or hip fracture and/or other bone fracture and/or bone break
Pattern of care: 1. Use of a nitrate (e.g.; isosorbide)
Outcome: Acute renal failure and/or renal insufficiency
Pattern of care: 1. Use of allopurinol
2. BUN/serum creatinine not done at least every 6 months
Outcome: Asthma exacerbation and/or status asthmaticus and/or ER visit/hospitalization
Due to asthma
Pattern of care: 1. Diagnosis of asthma
2. Use of theophylline
3. Drug level not done at least every 6 months








operational definition of PDRM. Table 5.4 shows the individual breakdown of each of the 52

consensus-approved operational definitions of PDRM.

As previously mentioned, many patients experienced more than one specific operational

definition of PDRM. Table 5.5 shows the number of PDRMs each patient had, along with the

number of specific outcomes these PDRM represented. This distinction is important to make

because some patients met the criteria for more than one PDRM, but these PDRMs shared the

same outcome (e.g. there are several PDRMs related to major/minor hemorrhagic events with

different patterns of care).



Validation of Operational Definitions of Preventable Drug-Related Morbidity


Pharmacist agreement with the PDRM classification assigned by the operational

definitions of PDRM was acceptable, although it varied for the two specific operational

definitions of PDRM included in the analysis. Tables 5.6 and 5.7 show the individual panel

members' classification for each patient. One patient chart for the hyperglycemia outcome and

four patient charts for the secondary myocardial infarction outcome could not be located and

they were not included in the final analysis.

Overall, the sensitivity of the two specific operational definitions of PDRM was 87.5

percent and the specificity was 73.5 percent (Table 5.8). For the hyperglycemia outcome, the

sensitivity was 93.3 percent and the specificity was 81.3 percent (Table 5.9). For this outcome,

the chart abstracts had to be administered a second time to the Chart Abstract Reviewer Panel

because not all panel members followed the initial instructions, as will be discussed in the next

chapter. For the secondary myocardial infarction outcome, the sensitivity was 82.4 percent and

the specificity was 66.7 percent (Table 5.10).










Table 5.4 Patients with Outcomes and PDRM


PDRM Number Number Percentage PDRM Number Number Percentage
Num- of of (%) of Num- of of (%) of
ber a Patients Patients Patients ber a Patients Patients Patients
with with with the with with with the
Outcome PDRM outcome Outcome PDRM outcome
(%) (%) who have (%) (%) who have
(n=1005) (n=97) b PDRM (n=1005) (n=97) b PDRM
1 28(2.8) 8 (8.2) 28.6 27 14(1.4) 0 (0.0) 0
2 24(2.4) 6 (6.2) 25.0 28 45(4.5) 2 (2.1) 4.4
3 0(0.0) 0 (0.0) 0 29 28(2.8) 3 (3.1) 10.7
4 7(0.7) 1 (1.0) 14.3 30 13(1.3) 0 (0.0) 0
5 24(2.4) 0 (0.0) 0 31 2 (0.2) 0 (0.0) 0
6 3(0.3) 1 (1.0) 33.3 32 7(0.7) 0 (0.0) 0
7 17(1.7) 0 (0.0) 0 33 39(3.9) 2 (2.1) 5.1
8 24(2.4) 0 (0.0) 0 34 16(1.6) 7 (7.2) 43.8
9 0(0.0) 0 (0.0) 0 35 1(0.1) 0 (0.0) 0
10 0(0.0) 0 (0.0) 0 36 32(3.2) 8 (8.2) 25.0
11 45(4.5) 4 (4.1) 8.9 37 32(3.2) 6 (6.2) 18.8
12 28(2.8) 2 (2.1) 7.1 38 2(0.2) 1 (1.0) 50.0
13 27 (2.7) 0 (0.0) 0 39 3 (0.3) 0 (0.0) 0
14 2 (0.2) 0 (0.0) 0 40 0 (0.0) 0 (0.0) 0
15 2(0.2) 0 (0.0) 0 41 14(1.4) 12(12.4) 85.7
16 46(4.6) 4 (4.1) 8.7 42 31(3.1) 3 (3.1) 9.7
17 7(0.7) 1 (1.0) 14.3 43 31(3.1) 3 (3.1) 9.7
18 24(2.4) 10(10.3) 41.7 44 7(0.7) 0 (0.0) 0
19 0(0.0) 0 (0.0) 0 45 16(1.6) 0 (0.0) 0
20 0 (0.0) 0 (0.0) 0 46 24(2.4) 0 (0.0) 0
21 32(3.2) 18(18.6) 56.3 47 31(3.1) 6 (6.2) 19.4
22 45(4.5) 5 (5.2) 11.1 48 24(2.4) 6 (6.2) 25.0
23 0(0.0) 0 (0.0) 0 49 31(3.1) 1 (1.0) 3.2
24 39(3.9) 24(24.7) 61.5 50 42(4.2) 2 (2.1) 4.8
25 12(1.2) 0 (0.0) 0 51 32(3.2) 10(10.3) 31.3
26 45(4.5) 1 (1.0) 2.2 52 7 (0.7) 1 (1.0) 14.3

a See Appendix G for the description of the PDRM.
b Adds up to over 100 percent because some patients had more than one operational definition
of PDRM.








Table 5.5 Number of PDRMs and Specific Outcomes by Individual Patients

Patient Category Number of Patients
with PDRM (%) n=97
1 case of PDRM with 1 specific outcome 61(62.9)
2 cases of PDRM with 1 specific outcome 12 (12.4)
2 cases of PDRM with 2 specific outcomes 6 (6.2)
3 cases of PDRM with 1 specific outcome 4 (8.2)
3 cases of PDRM with 2 specific outcomes 8 (8.2)
3 cases of PDRM with 3 specific outcomes 0 (0.0)
4 cases of PDRM with 1 specific outcome 0 (0.0)
4 cases of PDRM with 2 specific outcomes 3 (3.1)
4 cases of PDRM with 3 specific outcomes 0 (0.0)
4 cases of PDRM with 4 specific outcomes 2 (2.1)
5 cases of PDRM with 1 specific outcome 0 (0.0)
5 cases of PDRM with 2 specific outcomes 0 (0.0)
5 cases of PDRM with 3 specific outcomes 1 (1.0)
5 cases of PDRM with 4 specific outcomes 0 (0.0)
5 cases of PDRM with 5 specific outcomes 0 (0.0)









Table 5.6 Chart Abstract Reviewer Panel Results for Hyperglycemia with no
Regular HgAlc Monitoring

Pt # RPh RPh RPh RPh RPh Total PDRM according to
____#1 #2 #3 #4 #5 "yes" definition
I Y Y Y Y Y 5 Y
2 Y Y Y Y Y 5 Y
3 N N N N Y 1 N
4 Y Y Y Y Y 5 Y
5 Y N N Y N 2 N
6 Y N N N Y 2 N
7 Y Y Y Y Y 5 Y
8 Y Y N Y Y 4 Y
9 Y Y N Y Y 4 Y
10 Y Y Y Y Y 5 Y
11 N N N N Y 1 N
12 N N N N N 0 N
13 N N N N N 0 N
14 Y Y Y Y Y 5 Y
15 Y Y Y Y Y 5 Y
16 N N N N N 0 N
17 N Y N Y N 2 N
18 Y Y Y Y N 4 Y
19 N Y IN Y N 2 N
20 N Y N Y N 2 Y
21 Y Y Y Y Y 5 N
22 N Y N N Y 2 Y
23 N N N N Y I N
24* Y
25 N N Y N N 1 N
26 Y Y Y Y Y 5 Y
27 N Y N N N I N
28 Y Y Y Y Y 5 Y
29 Y Y Y Y Y 5 Y
30 Y N N Y N 2 Y
31 Y Y Y Y Y 5 Y
32 N N N N Y 1 N

* Chart abstract could not be done because patient chart could not be located.









Table 5.7 Chart Abstract Reviewer Panel Results: Secondary Myocardial Infarction
in Patients Without ASA and/or Beta-Blocker Use

Pt # RPh RPh RPh RPh RPh Total "Yes" PDRM according
____#1 #2 #3 #4 #5 _____to definition
1 N N N N N 0 N
2 Y Y Y Y N 4 N
3 N N N N Y I N
4 Y Y N Y N 3 Y
5 Y Y Y Y Y 5 Y
6 N Y Y N N 2 N
7 Y N N Y N 2 N
8 N N Y Y Y 3 N
9 Y Y Y Y Y 5 Y
10* y____________Y
11 Y Y Y N Y 4 Y
12 Y Y Y Y Y 5 Y
13 Y N N Y Y 3 Y
14 Y Y Y Y Y 5 Y
15 Y Y Y Y Y 5 N
16 Y N N Y Y 3 Y
17 Y Y Y Y Y 5 Y
18 Y Y Y Y Y 5 Y
19 Y N Y N Y 3 Y
20 N N N Y Y 2 N
21 N N N Y N I N
22*_________ y
23 Y Y Y Y Y 5 Y
24 Y Y Y Y Y 5 Y
25 Y Y Y Y Y 5 Y
26*0___ ------- Y
27* ______ y
28 Y Y Y N N 3 Y
29 N Y N Y Y 2 Y
30 Y Y Y Y N 4 Y
31 N N N Y N I N
32 Y N Y Y Y 4 Y
33 Y N Y Y N 3 N
34 Y Y Y Y Y 5 Y
35 Y N Y Y Y 4 N
36 N N N Y Y 2 N
37 Y Y Y Y Y 5 Y
38 N Y N Y N 2 N
39 N N N Y N I N

* Chart abstract could not be done because patient chart could not be located.








Table 5.8 Sensitivity and Specificity for Both of the Operational Definitions of PDRM


Operational Definition of PDRM
Yes No
True Yes 28 4
PDRM No1 9 25

Sensitivity was calculated as [28/(28+4)] x 100 = 87.5 percent.
Specificity was calculated as [25/(25+9)] x 100 = 73.5 percent.
True PDRM (Yes) = Four or more of the panel members classified as PDRM,
True PDRM (No) = Four or more of the panel members classified as non-PDRM,
Operational definition of PDRM (Yes) = Classified as a PDRM by the operational
definition,
Operational definition of PDRM (No) Classified as a non-PDRM by the operational
definition.








Table 5.9 Sensitivity and Specificity of the Operational Definition of PDRM (Hyperglycemia
Outcome)


Operational Definition of PDRM
Yes No
True Yes 14 1
PDRM No 3 13

Sensitivity was calculated as [14/(14+1)] x 100 = 93.3 percent.
Specificity was calculated as [13/(13+3)] x 100 = 81.3 percent.
True PDRM (Yes) = Four or more of the panel members classified as PDRM,
True PDRM (No) = Four or more of the panel members classified as non-PDRM,
Operational definition of PDRM (Yes) = Classified as a PDRM by the operational
definition,
Operational definition of PDRM (No) Classified as a non-PDRM by the operational
definition.








Table 5.10 Sensitivity and Specificity of the Operational Definition of PDRM (Secondary
Myocardial Infarction Outcome)


Operational Definition of PDRM
Yes No
True Yes 14 3
PDRM No 6 12

Sensitivity was calculated as [14/(14+3)] x 100 = 82.4 percent.
Specificity was calculated as [12/(12+6)] x 100 = 66.7 percent.
True PDRM (Yes) = Four or more of the panel members classified as PDRM,
True PDRM (No) = Four or more of the panel members classified as non-PDRM,
Operational definition of PDRM (Yes) = Classified as a PDRM by the operational
definition,
Operational definition of PDRM (No) Classified as a non-PDRM by the operational
definition.








The agreement among the five pharmacists in classifying the patients into those with PDRM, and

those without PDRM, was acceptable. Fleiss's measure of overall agreement for the

hyperglycemia patients was 0.652, for the secondary myocardial infarction patients it was 0.674

and overall it was 0.664. Therefore, if a patient was selected at random and classified as either

having, or not having, PDRM by a randomly selected panel member, a second randomly selected

panel member would agree with the first panel member 66.4 percent of the time.

Therefore, because of the consensus reached by the Geriatric Medicine Expert Panel and

the high sensitivity and specificity of the two validated operational definitions, research

assumption A1 A could be met: valid operational definitions of PDRM can be developed by a

panel of geriatric medicine experts.



Phase II: Identification of Risk Factors for PDRM


Phase II involved the identification of risk factors for PDRM. First, the results of the

statistical analyses used to identify the risk factors will be presented. Second, the results related

to the hypothesis for each risk factor will be shown. Third, the results related to the hypothesis of

general risk factors and drug (and disease) specific risk factors will be presented. Finally, the

relationship between PDRM and healthcare resource utilization will be discussed.


Logistic Regression with all 18 Hypothesized Risk Factors


In order to test the first set of hypotheses regarding possible risk factors for PDRM, a

logistic regression analysis was performed. A forward inclusion procedure was used with the

entry level set at p=0.05.

A five-variable risk model was produced (Table 5.11). This model indicates that patients

withfour or more recorded diagnoses were 2.93 times more likely to have PDRM than those

with three or fewer diseases. Patients with antihypertensive drug use are at a much greater risk









Table 5.11 Logistic Regression Model With All 18 Hypothesized Variables


Variable Parameter Standard Chi-Square Odds 95 Percent
Estimate Error (SE) Probability Ratio Confidence
(b)____ ____ Interval
Four or more 0.2683 0.0602 0.0001 1.308 1.162-1.472
prescribers
Four or more 1.0758 0.2475 0.0001 2.932 1.805-4.763
recorded
diagnoses
Female gender -0.6633 0.2393 0.0056 0.515 0.823-0.322
Antihypertensive 0.7023 0.2787 0.0118 2.018 1.156-3.524
drug use
Six or more 0.6525 0.2942 0.0266 1.920 1.079-3.418
prescription
medications
Equation Constant -4.6958 0.2758 0.0001 -