Multiple medication use is associated with increased health risks and costs. 1–24 Most research on multiple medication use has focused on high-risk patients, examining adverse drug reactions, drug-drug interactions (DDIs), medication errors, inappropriate prescribing, and noncompliance in hospitalized, elderly, or nursing home patients. 2–30 A better understanding is needed of the relationship between multiple medication use and age, number of prescribers, and complexity of medication regimens (i.e., all drugs a patient is taking irrespective of dose or schedule). Medications that act systemically or gastrointestinally (“SG medications”) are of most concern because they have the greatest potential to interact. The identification of frequently used combinations of SG medications is the first step in identifying common potentially hazardous combinations.
A few recent studies 31–34 have examined overall medication use in adult outpatients of all ages. A series of Danish studies 32–34 provided the most extensive analysis to date of multiple medication use in a population-based sample that included both younger and older persons. The Danish studies analyzed records from a computerized prescription reimbursement database of subsidized prescription medications in Fünen County, Denmark. Bjerrum et al. 32 reported that, on an average day in 1994, 8.7% (SD = 0.2%) of the patients were exposed to 2–4 drugs and 1.2% were exposed to 5 or more drugs. The proportion of patients receiving 5 or more medications increased with age up to 90 years, so that two-thirds of patients aged 70 years or older were receiving 5 or more medications. Rosholm et al., 34 found that 26,337 elderly patients aged 70 years or older were receiving 21,293 different drug combinations and that the 10 most prevalent combinations were found in only 2.7% of the elderly patients. Bjerrum et al. 33 examined drug combinations in 5,443 patients aged 16 years or older who were on ≥5 drugs. They found a total of 3,890 different drug combinations, of which the 10 most prevalent combinations occurred in only 3% of patients. However, the findings of the Danish studies do not necessarily reflect U.S. prescribing patterns and did not include non-subsidized medications, such as anxiolytics. The two analyses of complexity of drug regimen in these studies limited their focus to elderly patients and patients receiving 5 or more medications, and they did not relate complexity of drug regimen to different levels of multiple medication use.
The Slone Survey 35 obtained self-reported information on prescription, over-the-counter (OTC), and herbal drug use from a population-based sample of U.S. adult outpatients aged 18 years or older. This general survey of 2,590 adults found that 7% had taken 5 or more prescription drugs in the preceding week, an indication of the prevalence of multiple medication use in the U.S. adult population.
The potential for adverse drug combinations is a special problem for patients with psychiatric disorders because these patients are frequently receiving multiple medications for a variety of reasons. First, psychiatric illnesses, such as major depressive disorder, occur at an increased frequency in patients with other medical illnesses. 36–39 Second, patients with one psychiatric illness are at increased risk for other psychiatric disorders, leading to frequent use of polypsychopharmacology. 40 Third, patients with depressive and anxiety disorders have been found to be high utilizers of healthcare services and thus may be more likely to receive symptomatic medication treatment. 38,41–45 It is important to keep in mind that drugs do not interact based on their therapeutic indications but rather based on their pharmaco-dynamics and pharmacokinetics. Hence, “psychiatric” and nonpsychiatric medications can interact and produce clinically meaningful changes in patient outcomes. For all of these reasons, clinicians who prescribe psychiatric medications must consider all of the medications that a patient is taking, including OTC medications, illicit substances, herbal products, and even dietary substances. For example, ibuprofen, an OTC analgesic, can cause serious and even life-threatening elevations in lithium levels by affecting the rate of its tubular reab-sorption. 46 Thus, while the study described here focused on general medication use, rather than specifically considering the prescribing of psychotropic agents, the findings are relevant for all prescribers of psychiatric medications. A follow-up study (Part II), described in a second article in this issue of the journal, specifically considered how antidepressants contribute to the complexity of prescribing patterns. 47
The present study examined multiple medication use and complexity of drug regimens in an adult outpatient population of U.S. veterans treated in a Midwest Veterans Affairs (VA) Integrated Service Network (VISN 15). This study builds upon the findings of both the Danish and Slone studies by evaluating 5 levels (0, 1, 2–4, 5–7 and 8 or more SG drugs) of multiple medication use in an adult U.S. outpatient sample in relation to age, number of prescribers, and complexity and uniqueness of drug regimens. The use of 5 levels permitted more sensitive comparisons of proportions of patients receiving 2–4, 5–7, and 8 or more SG drugs.
The use of prescription records from the VA health-care system has several distinct advantages for evaluating multiple medication use in a U.S. setting. First, the VA provides prescription and OTC medications as well as vitamins, minerals, medical supplies, and diagnostic agents. Second, the low cost per refill ($2 each in 1999) encourages patients to rely on VA prescriptions as the source of most of their medications. Third, VA prescription data are based on dispensing rather than prescribing records, and dispensing records have been found to provide more accurate estimates of patient drug exposure than self-reports. 48 In addition, each VISN in the VA includes a variety of outpatient and inpatient services ranging from primary care to tertiary specialty care, so that it is possible to obtain a broad sample of the types of patients who may be seen in diverse healthcare settings.
The objectives of this study were to examine the following in a VA outpatient population:
- the extent and nature of multiple medication use in relation to age and number of prescribers, and
- the frequency and complexity of drug regimens in relation to levels of multiple medication use.
For the purpose of this study, drug regimen was defined as the specific SG drug or combination of specific SG drugs the patient was receiving without regard to dose or drug administration schedule.
Source of the Data
This study examined data from VISN 15, which has 8 medical centers and 24 community-based outpatient clinics that provide primary and subspecialty care. Two of the medical centers used a single prescription database, so that 7 prescription databases were examined. Review and approval from the VA Human Studies Subcommittee and Research Committee was obtained prior to study initiation.
The sampling pool consisted of all “potentially active” outpatients in pharmacy databases at each site. Potentially active is defined here as those who had at least 1 prescription with an expiration date up to 365 days prior to the day of data extraction (June 1999). A stratified, random sample of 1,000 active patients was selected from each of the 7 prescription sites in VISN 15. A computer program was developed and validated for selection of the study sample and extraction of prescription dispensing data. For each prescription, data collected included prescription site identification number, unique patient identifier, patient age, gender, and race/ethnicity; prescriber, clinic, prescription number, prescription name/dosage formulation, number of days supply of prescription, expiration date of prescription, VA drug classification code for prescription (see U.S. Pharmacopeia DI/Micromedex 1999), date of last clinic visit, date of last refill, dosing instructions (“Sig”), number of refills, and prescription status (active, mailed, expired).
Prescriptions were considered current when the number of days of supply equaled or exceeded the number of days since the last refill. Of the 7,000 patients sampled, records for 3 patients were excluded as invalid entries, so that 6,997 potentially active patients comprised the full sample. These patients had 46,365 prescriptions that could have been filled during the past year. Of these prescriptions, 21,488 were determined to be current and were used to identify 5,003 currently active patients (71.5% of the original sample) who, if compliant, should have had a current supply of at least one prescription item. This report focuses on the subset of 5,003 patients and excludes the 1,994 previously active patients who would have been able to refill a prescription during the previous year but whose supply should have been exhausted before the data extraction date. The accuracy of the percentages estimated from the sample of 5,003 currently active patients are as follows: for 1%, ± 0.3%; for 10%, ± 0.8%; and for 50%, ± 1.0%. Breakdown analyses by subgroups of patients would be less accurate.
Point Prevalence of Multiple Medication Use and Drug Prescribing Patterns
The VA prescription database includes supplies, diagnostic agents, irrigants, vitamins, minerals, topical medications, locally acting medications (other than gastrointestinal agents), gastrointestinal medications, and systemically acting medications. However, only drugs that could interact systemically or gastrointestinally (“SG drugs”) and hence, potentially cause a clinically significant adverse DDI were analyzed in this study. Combination products were separated into constituent drugs (e.g., acetaminophen and codeine in a combination product would be counted as two separate drugs). The data presented here represent a 1-day point prevalence of the extent and nature of multiple SG medication use in a VISN outpatient population (N = 5,003). A total of 4,857 currently active patients (97.1%) were deemed to have a current supply of at least one SG drug. The remaining 146 patients (2.9%) had prescriptions only for non-SG items (i.e., supplies, topical or locally active medications, vitamins, or minerals). The following non-SG items were excluded from the analyses: 52 vitamins/minerals, 12 vaccines, and 176 topical/locally acting dermatological, otic, ophthalmic, or nasal drug entities.
Defining Levels of Multiple Medication Use
The term “polypharmacy” literally means 2 or more medications, but operational definitions have varied greatly from one study to another. The term has often been used to suggest the overuse of medications. In the Danish studies described above, the investigators defined “minor” polypharmacy as use of 2–4 medications and “major” polypharmacy as use of 5 or more medications. 32–34 They made this distinction because, by definition, there can be no risk of adverse DDIs with monopharmacy. In contrast, studies suggest some risk of adverse DDIs with the use of 2–4 medications and substantially higher risks of adverse DDIs with use of 5 or more medications. 2,19,22,24 Reports of exponentially increasing risk of adverse DDIs with increasing numbers of concomitant medications 2,19,22,24 suggested that a more detailed breakdown of the category of “5 or more” medications into “5–7” and “8 or more” medications would provide more useful information about potential risks. In addition, no polypharmacy definitions published to date have distinguished between drugs with or without the potential to cause DDIs. However, when evaluating polypharmacy, only SG drugs generally place patients at risk for DDIs. Consequently, this study limited analyses to SG drugs and looked at patients taking 0, 1, 2–4, 5–7, and 8 or more SG drugs.
To avoid possible confusion arising from varying definitions of polypharmacy, this report uses the term “multiple medication use” to refer to multiple SG drug use.
Samples were extracted at each site in EXCEL spreadsheet format. DBMS/COPY™ (Conceptual Software, Inc., Houston, TX) was used to convert the files into SAS® (Cary, NC) format. SAS® was used for data cleaning, which involved identifying and removing exact duplicate prescriptions, identifying incomplete and out-of-range data, correcting miscoded drugs, and identifying duplicate patient identification numbers within and across sites. SAS® was used to identify, separate out, and perform consistency checks for individual drugs in single or combination formulations. Additional SAS® programs were used to identify and determine frequencies of all unique drug combinations.
SAS® was used for all statistical calculations. Means and standard deviations were calculated for interval data (age). Means, standard deviations, and medians were used to examine distribution of the number of drugs. The Chi Square (χ2) statistic was used for comparisons of cross-sectional subgroups (i.e., age group, number of prescribers) by ordinal variables (i.e., levels of multi-SG drug use). 49 Kruskal-Wallis tests were used for comparisons of independent samples (i.e., prescription sites) with ordinal variables, 50 and, if significant overall results were obtained and sample sizes permitted, Wilcoxon two-sample tests were performed to determine which groups were contributing to the observed differences. 50
Comparison with 1999 National Survey
The demographics of the currently active patient sample in this study were compared with the 1999 VA National Survey and the VISN 15 subset of the 1999 National Survey (Table 1). 51 The demographic profiles of the three groups are similar, with mean ages ranging from 59 to 62 years and males making up approximately 95% of all three groups. The small number of females in the currently active patient sample prevented adequate analysis of levels of multiple medication use and drug regimens for females. Prescription data did not include information on race/ethnicity in over a quarter of the currently active sample, so that it was considered invalid to compare race/ethnicity in this study.
Nature and Extent of Multiple Medication Use Involving SG Drugs
Extent of multiple medication use.
Prescribing patterns in the 5,003 currently active patients are shown in Figure 1 and Tables 2–4. The median patient in this sample was receiving 4 SG drugs (Table 2). Of the 5,003 currently active patients, 80% had current prescriptions for at least 2 SG drugs, indicating a potential risk for adverse DDIs. Over one-third of patients (38%) were receiving 5 or more drugs.
A total of 394 SG drugs were used to treat this population of 5,003 patients. Only 88 (22%) of these drugs occurred in at least 1% of patients (Table 2). Each of the other 306 drugs occurred in fewer than 1% of patients. The 20 SG drugs that were most commonly dispensed (Table 3) were used by 5.5%–26.5% of patients. The most frequently dispensed medication was aspirin. Nearly one-half of the other top 20 SG drugs were cardiovascular agents (lisinopril, simvastatin, hydrochlorothiazide, atenolol, furosemide, digoxin, felodipine, nitroglycerin and diltiazem). The remaining 10 drugs included 2 antidiabetic agents (glyburide and insulin), 2 anti-asthma/COPD agents (albuterol and ipratropium), 2 gastrointestinal agents (ranitidine and lansoprazole), 2 analgesic agents (acetaminophen and ibuprofen), 1 agent (terazosin) commonly used to treat benign prostatic hyperplasia and/or hypertension, and 1 laxative (docusate sodium).
No single antidepressant was in the top 30 drugs used in this population. The most frequently used antidepressants were sertraline (rank 31, 4.0% of patients), trazodone (rank 33, 3.6%), amitriptyline (rank 35, 3.5%), fluoxetine (rank 46, 2.2%), and paroxetine (rank 48, 2.2%). However, the SSRIs, if considered as a class, would have fallen in the top 20.
Factors influencing levels of multiple medication use.
Numbers of drugs and levels of multiple medication use were examined in relation to age group, numbers of prescribers, and site. As expected, increasing levels of multiple medication use were associated with increasing patient age (p < 0.0001, Figure 2) up to 80 years of age. The proportion of patients receiving monopharmacy initially decreased with increasing age, ranging from 31.0% (age under 40 years) to 11.5% (ages 70–79 years), but increased slightly after the age of 79 years (15.3%). A corresponding increase in the proportion of patients receiving 2 or more drugs reveals a shift from monopharmacy to multiple medication use with increasing age. Forty percent or more of patients in all age groups received 2–4 drugs. Among the patients 50 years of age or older, approximately 40% received 5 or more drugs.
Of the 5,003 currently active patients, 70% had 1 prescriber, approximately 20% had 2 prescribers, and fewer than 10% had more than 2 prescribers (Figure 3). Patients with multiple prescribers received a larger number of SG drugs, with the number of SG drugs increasing significantly with increasing numbers of prescribers (p < 0.0001). Among patients with 1 prescriber, 28% received ≥5 drugs. Among the 1,084 patients with 2 prescribers, 53% received ≥ 5 SG drugs. Of the 321 patients with 3 prescribers, 82% received ≥5 drugs. A total of 113 patients (2.4% of 5,003) had 4 or more pre-scribers and over one-half of these patients received ≥8 drugs. Of these 113 patients, 90 (79.6%) had 4 prescribers, 15 (13.2%) had 5 prescribers, 7 (6.2%) had 6 prescribers, and 1 patient (0.9%) had 7 prescribers.
There was statistically significant variation across the 7 prescription sites in levels of multiple SG drug use (p < 0.0001). Mean number of drugs ranged from 3.9 ± 2.8 SD to 4.5 ± 3.3 SD across sites. One site had small, but statistically significant higher overall levels of multiple SG drug use than the others (Bonferroni corrected p = 0.042). 52 However, the magnitude of the differences in proportions on each level of multiple drug use between that site and the others ranged only from 0.5% to 2.5%.
Complexity of SG Drug Regimens
The numbers of SG drugs, SG drug regimens, and unique regimens are summarized in Tables 2 and 4. The 5,003 currently active patients received 394 SG drug entities and 3,819 regimens (irrespective of dose and drug administration schedule). Although the differences in subset sizes limit comparisons, the majority of patients taking 2–4 drugs (76%), nearly all patients taking 5–7 drugs (99%), and all patients taking 8 or more drugs (100%) were receiving a unique regimen. Every drug regimen composed of more than 2 drugs occurred in fewer than 1% of patients. Of the 3,819 regimens composed of 2 or more SG drugs, 3,553 were used by only 1 patient, so that 71% of currently active patients were receiving unique regimens.
The 25 most frequently occurring regimens (i.e., specific SG drug or combination of SG drugs the patient was receiving without regard to dose or drug administration schedule) were shared by 10–48 patients and all but one of the regimens consisted of monopharmacy. The only regimen among the top 25 involving more than 1 drug consisted of 2 blood pressure drugs, hydrochlorothiazide and triamterene, which are generally prescribed in a combination formulation. There was no regimen involving 3 or more drugs in the top 25 regimens. The most frequently occurring drug regimen involving 3 or more drugs ranked 35th in frequency, was shared by 7 patients, and consisted of albuterol, ipratropium, and beclomethasone, which are commonly prescribed as a combination formulation to treat asthma/COPD.
Figure 4 suggests how such a large number of unique SG drug regimens might arise. The figure illustrates the rate at which the prevalence of a specified drug combination decreases when the number of drugs in the combination increases. Hence, 26.5% of the 5,003 patients received aspirin and 10.6% received furosemide, but only 4.2% received both aspirin and furosemide. The percentage decreased to fewer than 2% of patients when the analysis looked at the combination of aspirin, furosemide and a third SG drug, digoxin. It decreased further, to fewer than 1% of patients receiving a combination of those 3 SG drugs plus lisinopril—despite the fact that this regimen was the most common 4-drug regimen in this population and that each of the 4 drugs was among the most commonly prescribed drugs in this population (Table 3). Of the 28 patients taking these 4 drugs, 1 patient received only these 4 drugs, 2 others received the same regimen plus a 5th drug (another cardiovascular-related agent, simvastatin), and the remaining 25 received unique drug regimens of 5 or more drugs.
Summary of Key Findings
The 5,003 outpatients in this sample received 394 different SG drugs, 306 of which were used by fewer than 1% of patients. Over one-third (38%) of the patients received 5 or more drugs. Level of multiple medication use increased with age (up to age 80) so that, by age 50 years, approximately 40% received 5 or more drugs. Proportions receiving 8 or more drugs doubled with each additional prescriber (i.e., from 1 to 2 prescribers, from 2 to 3 prescribers and from 3 to 4 or more prescribers). Most patients on at least 2 SG drugs were receiving a unique SG drug regimen.
Interpretation of Findings
The high levels of multiple medication use found in this sample are not surprising, given that VA patients have been reported to be older and sicker than the general population. 53 The increase in level of multiple medication use with increasing age is consistent with other reports. 31,33,34 The nature of the sampled patient populations, data sources, and drugs included in the counts of number of concomitant drugs in this study differ in several key ways from those used in either the Danish studies 32–34 or the Slone study, 35 and hence prevent direct comparisons of levels of multiple medication use. Nevertheless, all of these studies, with their diverse nature and populations, underscore the extent and complexity of multiple medication use in clinical practice.
Although other studies have reported associations between a higher numbers of prescribers and high levels of multiple medication use, the detailed analysis of this study demonstrated a doubling in the proportion of patients receiving 8 or more drugs for each additional prescriber, up to 4 or more prescribers (three-fourths of that group had 4 prescribers).
The findings of this study provide the first evidence of the uniqueness of drug regimens in a U.S. outpatient sample and corroborate findings of two Danish studies. 33,34 This study also found striking increases in the proportions of patients receiving unique drug regimens with increasing levels of multiple medication use, 7% of patients on monopharmacy, 76% of patients on 2–4 drugs, 99% of patients on 5–7 drugs and 100% of patients on ≥8 drugs were receiving unique drug regimens. The fact that no regimens occurred frequently except in patients on monopharmacy is striking.
The uniqueness and complexity of drug regimens implies that no single prescriber will have extensive experience with even a small fraction of the drug regimens that patients receive. This raises concerns regarding the likelihood that any single prescriber in this setting can be familiar with the combined effects of all of the drugs comprising the total regimen a patient is receiving.
The average VA outpatient received a mean of 4.2 drugs, raising concerns about the potential for complex DDIs involving multiple medications. Most drug alert systems are based on the effects of one drug on a single co-prescribed drug, rather than on the reality that multiple drugs are frequently used together in the most medically challenging patients.
Limitations of the Study
VA prescription dispensing records do not capture medications not provided by the VA or patient noncompliance. A study of VA medication profiles found that about 60% of computerized records included drugs patients were no longer taking and 87% of patients were taking 1 or more non-VA-provided medications. 54 However, since the focus of the current study was on potential risks from dispensed drugs, the findings remain valid despite this caveat.
While some studies of polypharmacy have also considered the appropriateness of prescribed drugs and dosages for the patient’s condition, 6,17,18 the diversity and complexity of the SG drug regimens in the current sample prohibited such detailed evaluation of all of the drugs and drug regimens found.
Generalizability of the Findings
Regional differences in patient populations and in specific disease frequencies have been noted across VISNs. 51 The data in Table 1 support the premise that the sample examined here adequately reflects VISN 15 and, to some extent, the overall VA patient population as regards age and gender. 53,54 The present findings have limited generalizability to other U.S. outpatients, except perhaps for adult male general outpatients. However, the uniqueness and complexity of drug regimens—even when only 2–4 drugs were prescribed— suggest that common drug regimens are not likely to be found in a more heterogeneous U.S. population of outpatients.
Nevertheless, a number of unique potentially confounding factors must be considered when extrapolating from these findings to other populations. First, the armed forces do their best to screen out individuals with serious medical and psychiatric conditions, so that a general veteran population starts out healthier than the general population. Second, a veteran population being seen in the VA Medical System by definition has access to health care whereas 25% of Americans have no insurance and another 25% are underinsured. The latter is particularly true for elderly individuals living on fixed incomes. These facts mean that the general U.S. population may be taking fewer medications than an age-matched VA population. Third, the older veterans in this survey are more likely to have seen combat than the younger veterans. Combat is deleterious to one’s physical and mental health (e.g., posttraumatic stress disorder). While assessing the effect of these potential confounds is beyond the scope of this paper, they should be considered when extrapolating from this population to non-Veteran populations.
The findings of this study suggest the need for further initiatives in several areas:
- Education for prescribers, pharmacists, health care organizations, patients and others regarding the complexity of drug prescribing in clinical practice today and how that complexity limits our ability to detect adverse drug regimens and identify the potential risks, outcomes, and healthcare costs of prescribing a specific drug regimen.
- New research strategies that better address risks of DDIs in all patients receiving two or more SG drugs, such as:
- Determining whether focusing on treatment for a specific chronic illness will yield common patterns of drug regimens or improve our understanding of factors contributing to high levels of multiple medication use and complexity. An analysis of multiple medication use and complexity of drug regimens related to antidepressant use or non-use has been completed and is reported in Part II of the study in this issue of the journal. 47
- Extending analysis to regimens involving pharmacodynamically and pharmacokinetically defined drug classes and evaluating whether common drug class combinations occur and can be prioritized for evaluation of risks.
1. Healthy People 2010: National health promotion and disease prevention objectives. Washington, DC: U.S. Department of Health and Human Services; 2001. Also available at http://web.health.gov/healthypeople/document
. Accessed March 3, 2004.
2. Alarcon T, Barcena A, Gonzalez-Montalvo JI, et al. Factors predictive of outcome on admission to an acute geriatric ward. Age and Aging 1999;28:429–32.
3. Bates DW, Miller EB, Cullen DJ, et al. Patient risk factors for adverse drug events in hospitalized patients. Arch Intern Med 1999;159:2553–60.
4. Beyth RJ, Shorr RI. Epidemiology of adverse drug reactions in the elderly by drug class. Drugs and Aging 1999;14:231–9.
5. Classen DC, Pestotnik SL, Evans RS, et al. Adverse drug events in hospitalized patients: Excess length of stay, extra costs, and attributable mortality. JAMA 1997;277:301–6.
6. Colley CA, Lucas LM. Polypharmacy: The cure becomes the disease. J Gen Intern Med 1993;8:278–83.
7. Cooper JW. Adverse drug-reaction-related hospitalizations of nursing facility patients: A 4-year study. Southern Med J 1999;92:485–90.
8. Cluff LE, Thornton G, Seidl L, et al. Epidemiological study of adverse drug reactions. Trans Assoc Am Physicians 1965;78: 255–68.
9. Cullen DJ, Sweitzer BJ, Bates DW, et al. Preventable adverse drug events in hospitalized patients: A comparative study of intensive care units and general care units. Crit Care Med 1997;25:1289–97.
10. Fattinger K, Roos M, Vergeres P, et al. Epidemiology of drug exposure and adverse drug reactions in two Swiss departments of internal medicine. Br J Clin Pharmacol 2000;49: 158–67.
11. Field TS, Gurwitz JH, Avorn J, et al. Risk factors for adverse drug events among nursing home residents. Arch Intern Med 2001;161:1629–34.
12. Flaherty JH, Perry HM III, Lynchard GS, et al. Polypharmacy and hospitalization among older home care patients. J Gerontol 2000;55A:M554–9.
13. Grymonpre RE, Mitenko PA, Sitar DS, et al. Drug-associated hospital admissions in older medical patients. J Am Geriatr Soc 1988;36:1092–8.
14. Gurwitz JH, Soumerai SB, Avorn J. Improving medication prescribing and utilization in the nursing home. J Am Geriatr Soc 1990;38:145–54.
15. Gurwitz JH, Field TS, Harrold LR, et al. Incidence and preventability of adverse drug events among older persons in the ambulatory setting. JAMA 2003;289:1107–15.
16. Hanlon JT, Schmader KE, Koronkowski MJ, et al. Adverse drug events in high risk older outpatients. J Am Geriatr Soc 1997;45:945–8.
17. Hanlon JT, Fillenbaum GG, Schmader KE, et al. Inappropriate drug use among community-dwelling elderly. Pharmacotherapy 2000;20:575–82.
18. Koecheler JA, Abramowitz PW, Swim SE, et al. Indicators for the selection of ambulatory patients who warrant pharmacist monitoring. Am J Hosp Pharm 1989;46:729–32.
19. Kellaway GS, McCrae E. Intensive monitoring for adverse drug effects in patients discharged from acute medical wards. NZ Med J 1973;78:525–8.
20. Lazarou J, Pomeranz BH, Corey PN. Incidence of adverse drug reactions in hospitalized patients: A meta-analysis of prospective studies. JAMA 1998;279:1200–5.
21. Nolan L, O’Malley K. Prescribing for the elderly. Part I: Sensitivity of the elderly to drug reactions. J Am Geriatr Soc 1988; 36:142–9.
22. Smith JW, Seidl LG, Cluff LE. Studies on the epidemiology of adverse drug reactions; V. Clinical factors influencing susceptibility. Ann Intern Med 1966;65:629–40.
23. Veehof LJ, Stewart RE, Meyboomde Jong B, et al. Adverse drug reactions and polypharmacy in the elderly in general practice. Eur J Clin Pharmacol 1999;55:533–56.
24. Williamson J, Chopin JM. Adverse reactions to prescribed drugs in the elderly: A multicentre investigation. Age Ageing 1980;9:73–80.
25. Christensen DB. Drug-taking compliance: A review and synthesis. Health Serv Res 1978;13:171–87.
26. Col N. Fanale JE, Kronholm P. The role of medication noncompliance and adverse drug reactions in hospitalizations of the elderly. Arch Intern Med 1990;150:841–5.
27. Evans L, Spelman M. The problem of non-compliance with drug therapy. Drugs 1983;25:63–76.
28. Goulding MR. Inappropriate medication prescribing for elderly ambulatory care patients. Arch Int Med 2004;164:305–12.
29. Hanlon JT, Artz MB, Pieper CF, et al. Inappropriate medication use among frail elderly inpatients. Ann Pharmacother 2004;30:9–14.
30. Wilcox SM, Himmelstein DU, Woolhandler S. Inappropriate drug prescribing for the community-dwelling elderly. JAMA 1994;272:292–6.
31. Nolan L, O’Malley K. Prescribing for the elderly: Part II. Prescribing patterns: Differences due to age. J Am Geriatr Soc 1988;36:245–54.
32. Bjerrum L, Rosholm JU, Hallas J, et al. Methods for estimating the occurrence of polypharmacy by means of a prescription database. Eur J Clin Pharmacol 1997;53:7–11.
33. Bjerrum L, Søgaard J, Hallas J, et al. Polypharmacy: Correlations with sex, age, and drug regimen: A prescription database study. Eur J Clin Pharmacol 1998;54:197–202.
34. Rosholm J-U, Bjerrum L, Hallas J, et al. Polypharmacy and the risk of drug-drug interactions among Danish elderly: A prescription database study. Dan Med Bull 1998:45:210–3.
35. Kaufman DW, Kelly JP, Rosenberg L, et al. Recent patterns of medication use in the ambulatory adult population of the United States: The Slone Survey. JAMA 2002;287:337–44.
36. Coulehan JL, Schulberg HC, Block MR, et al. Depressive symptomatology and medical comorbidity in a primary care clinic. Int J Psychiatry Med 1990;20:335–47.
37. Coulehan JL, Schulberg HC, Block MR, et al. Medical comorbidity of major depressive disorder in a primary medical practice. Arch Intern Med 1990;150:2363–7.
38. Fulop G, Strain JJ, Stettin G. Congestive heart failure and depression in older adults: Clinical course and health services use 6 months after hospitalization. Psychosomatics 2003;44: 367–73.
39. Robertson MM, Trimble MR. Depressive illness in patients with epilepsy: A review. Epilepsia. 1983;24(suppl 2):109–116.
40. Kupfer DJ, Frank E. Comorbidity in depression. Acta Psychiatr Scand Suppl 2003;418:57–60.
41. Levenson JL, Hamer RM, Rossiter LF. Relation of psychopathology in general medical inpatients to use and cost of services. Am J Psychiatry 1990;147:1498–503.
42. Badamgarav E, Weingarten SR, Henning JM, et al. Effectiveness of disease management programs in depression: A systematic review. Am J Psychiatry 2003;12:2080–90.
43. Katon WJ. Clinical and health services relationships between major depression, depressive symptoms, and general medical illness. Biol Psychiatry 2003;3:216–26.
44. Chisholm D, Diehr P, Knapp M, et al., for the LIDO Group. Depression status, medical comorbidity and resource costs. Evidence from an international study of major depression in primary care (LIDO). Br J Psychiatry 2003;183:121–31.
45. Series HG. Drug treatment of depression in medically ill patients. J Psychosom Res 1992;36:1–16.
46. Ragheb M, Ban TA, Buchanan D, et al. Interaction of indomethacin and ibuprofen with lithium in manic patients under a steady-state lithium level. J Clin Psychiatry 1980;11: 397–8.
47. Silkey B, Preskorn SH, Golbeck A, et al. Complexity of medication use in the Veterans Affairs Healthcare System: Part II. Antidepressant use among younger and older outpatients. J Pract Psychiatry 2005;11:16–26.
48. Beardon PH, McGilchrist MM, McKendrick AD, et al. Primary noncompliance with prescribed medication in primary care. BMJ 1993;307:846–8.
49. Fleiss JL. Statistical methods for rates and proportions. New York: John Wiley & Sons; 1973.
50. Lehmann EL. Nonparametrics: Statistical methods based on ranks. San Francisco: Holden-Day; 1975.
51. Perlin J, Kazis LE, Skinner K, et al. Health status and outcomes of veterans: Physical and Mental Component Summary Scores (SF-36) 1999 large health survey of veteran enrollees executive report. Department of Veterans Affairs, Veterans Health Administration Office of Quality and Performance, Washington, DC and VHA Health Assessment Project Center for Health Quality, Outcomes, and Economic Research, Edith Nourse Rogers Memorial Veterans Hospital, Bedford, MA, May 2000.
52. Feller W. An introduction to probability theory and its applications, Vol. 1, 3rd ed. New York: John Wiley & Sons; 1968.
53. Kazis LE, Ren XS, Lee A, et al. Health status in VA patients: Results from the Veterans Health Study. Am J Med Qual 1999;14:28–38.
54. Kaboli P. Does the VA computerized medication profile agree with a structured medication history? Poster presentation, Veterans Affairs Health Services Research and Development Annual Meetings, Washington, DC, February 12–14, 2003.