Unnecessary and/or inappropriate use of multiple medications is a national health concern. 1 Higher numbers of concurrent medications are strongly and consistently associated with an increased risk of adverse outcomes, particularly in the elderly. 1–15 Most research on multiple medication use and associated risks has focused on the elderly because they have more chronic illnesses, receive more medications, and are particularly vulnerable to adverse drug effects. 2 Nevertheless, adverse drug-drug interactions (DDIs) occur in patients of all ages and in patients taking as few as 2 drugs.
Recent research on the causes of adverse drug events has emphasized the need for systems-based strategies to prevent adverse drug outcomes, such as developing clinical decision support systems that provide real-time information relevant to medication selection and dosing. 16–18 Several useful strategies have been adopted. These include:
- Implementation of online prescription screening 19 or computerized drug alert systems 20 to notify prescribers about potential hazards of co-prescribing specific pairs of drugs (e.g., the system employed by the U.S. Veterans Affairs [VA] healthcare system)
- Development and application of expert consensus guidelines regarding specific drugs deemed to be inappropriate in patients who have severely diminished physical tolerability 21–23
- Pharmacist assessment of the safety and appropriateness of drug regimens in high-risk patients (e.g., those prescribed large numbers of concurrent medications or geriatric patients) 24–27
- Some combination of computerized decision support for potential medication problems and a pharmacist follow-up evaluation of the patient’s condition (i.e., DoseChecker 28).
Strategies such as these are capable of assessing the safety of multi-drug regimens; however, most are also labor-intensive and costly, so that they tend to be used only for the most vulnerable, highest risk patients. Thus, additional strategies are needed to address the safety of multi-drug regimens in the general patient population.
A possible strategy to improve the safety of multi-drug regimens (i.e., involving 3 or more drugs) would be to identify the most prevalent drug regimens and evaluate their potential to interact in complex ways based on all of their pharmacokinetic and pharmaco-dynamic properties. Since research to date has emphasized prescribing practices in the elderly, it would enhance our understanding to examine multiple medication use and the complexity of treatment regimens in adult outpatients of all ages, particularly those receiving medications associated with more serious adverse outcomes.
A previous study by the same authors assessed levels of multiple medication use and identified the most prevalent drug regimens in a general adult outpatient sample from U.S. Veterans Integrated Service Network 15 (VISN 15). 29 We found high levels of multiple medication use and highly complex and unique drug regimens (i.e., total specific drug entities the patient was receiving regardless of dose, formulation, or administration schedule). The most prevalent drug regimen was monotherapy with aspirin, which occurred in only 48 out of 5,003 patients (i.e., less than 1% of the population). These findings raised the question of whether such heterogeneity in prescribing is pervasive or was due to the broad age range (18–100 years of age) and general nature (all outpatients) of the sample or to other factors. If such heterogeneity in prescribing is pervasive, strategies to identify commonly used drug combinations would be unsuccessful. However, if the heterogeneity is primarily due to the nonspecific nature of the sample, then more homogeneity in drug combinations might be found by restricting the focus to a narrower age range and to patients selected for receiving a specific type of drug treatment.
A literature search for other U.S. studies concerning the frequency of unique drug combinations revealed no studies on this topic. In fact, only a few recent studies have examined overall medication use in adult outpatients of all ages. 30–32 The Slone Epidemiology Unit examined general patterns of medication use in ambulatory U.S. adults in 1998 and 1999. 30 They reported that, during the previous week, 81% were taking at least 1 medication, 50% were taking at least 1 prescription drug, and 7% were taking 5 or more prescription drugs. They also examined the use of over-the-counter drugs and herbal supplements. However, they did not analyze drug combinations or report on the number of persons taking 2, 3, or more drugs.
To the authors’ knowledge, only 2 published studies have examined the complexity and uniqueness of entire drug regimens patients were taking. 31,33 Both studies employed a population-based sample of prescription reimbursement records for patients aged 16 years and older in Fünen County, Denmark. They found highly unique and complex drug regimens in patients aged 70 years or older and in patients receiving 5 or more reimbursed drugs. However, neither study evaluated the nature or the frequency of unique drug regimens in a U.S. patient population, in younger patients, in patients of any age receiving fewer than 5 drugs, or in patients prescribed a specific class of drugs. The present study addresses these questions.
Antidepressants (ADs) were selected as the drug treatment of interest for the present study for several reasons. First, studies have found that ADs, as a class of drugs, are associated with an increased likelihood of adverse outcomes, even after adjustment for potential confounders such as comorbidity. 13, 34–41 If commonly used drug combinations could be identified in outpatients receiving ADs, the safety of those common drug combinations could be evaluated, thereby reducing the potential for adverse outcomes. Second, ADs affect the central nervous system (CNS), and CNS toxicity is responsible for some of the most serious adverse drug reactions. 37,41,42 Third, ADs have been associated with an increased risk of inappropriate prescribing. 6,22,43,44 Fourth, ADs are commonly prescribed and their use has increased markedly in recent years. 45–49 Fifth, ADs are prescribed chronically (i.e., for months to years), during which time other drugs can be added or stopped. A comparison of drug prescribing patterns in patients receiving and not receiving ADs could provide valuable information about potential risks and strategies for evaluating the safety of multi-drug combinations in such patients.
The study presented here expands on the results of our previous study. Its primary objective was to evaluate the extent, uniqueness, and complexity of multiple medication use in younger and older subsets of adult outpatients who were dispensed or not dispensed ADs at VISN 15 facilities. Because the ultimate goal of this research is to develop strategies to identify potential multi-drug interactions, analyses were restricted to drugs that act systemically or gastrointestinally (SG drugs) and thus have the potential to interact with one another.
The study examined the following questions:
- Are adult outpatients who are prescribed ADs similar in age, gender, and ethnicity to those not prescribed ADs?
- How do younger and older adult outpatients (< 60 years or ≥ 60 years of age), who are receiving or not receiving ADs, differ in terms of the extent and complexity of multiple medication use?
Source of the Data
The VA healthcare system was selected for this study because its computerized prescription database system permits examination of prescribing patterns and multiple medication use. This study examined data from VISN 15. In 1999, VISN 15 included 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 study population consisted of all potentially active outpatients listed in the 7 prescription databases in VISN 15, from which 2 stratified random samples were drawn. A computer program was developed and validated for selection of the study sample and extraction of prescription dispensing information.
The VA prescription database is described in more detail elsewhere. 29 Briefly, the record for each prescription contained information on age, gender, race/ethnicity, drug name and formulation, dosing, number of days supply, date of most recent refill, and current status of the prescription.
The first stratified random sample consisted of 3,500 outpatients, 500 from each of the 7 prescription databases, selected as representative of patients receiving ADs, that is, patients with at least 1 active prescription for any AD (VA Drug Classes CN601, CN602 or CN609, see Table 1) within 365 days prior to data extraction (June 1999). The second stratified random sample is described in detail in our previous study. 29 That sample of 7,000 patients (1,000 from each of the 7 prescription databases) was drawn on the same date as the AD sample but with the goal of representing general outpatients with at least 1 active prescription for any medication within 365 days prior to data extraction.
The analyses presented examined data from two subsets of these patient samples. The first subset consisted of the 1,991 patients who were deemed to have a current supply of at least one AD from the sample of 3,500 patients with active AD prescriptions sometime during the previous 365 days (referred to here as the “AD patients”). The second subset consists of the 3,732 patients from the sample of 7,000 patients who had no active AD prescriptions within the previous 365 days, but who were deemed to have a current supply of at least 1 other SG prescription item (referred to here as “NoAD patients”). Patients who were excluded from the analysis may have been taking a prescribed drug(s); however, their supply should have been exhausted if they had taken their dispensed prescription exactly as prescribed. Data from each subset of patients was further broken down for analysis by age (i.e., whether the patients were younger than 60 years of age or 60 years of age or older, based on the age-related bimodal distribution of the VA population, as shown in Figure 1).
Point Prevalence of Number of SG drugs and Levels of Multiple Medication Use
The count of concurrent SG drugs per patient is based on having a current supply of the drug on the day before data extraction. 29 Each drug in a combination product was counted separately (e.g., the combination formulation of hydrochlorothiazide and triamterene was counted as 2 drugs). The levels of multiple medication use in the four patient subsets were compared in two steps:
- Total number of all SG medications
- Number of SG drugs, excluding ADs (Table 1).
SG drug counts were grouped into five categories: 0 drugs, 1 drug, 2–4 drugs, 5–7 drugs, or ≥ 8 drugs. The use of five levels allowed more detailed examination of trends and of different cutpoints (more than 5 or more than 8 SG drugs) for high levels of multiple medication use. 29 Frequencies of specific drug regimens (i.e., the total specific drugs each patient was receiving, irrespective of dose or schedule) were also determined.
The procedures used for data extraction and cleaning are described elsewhere. 29 SAS® (Cary, NC) programming was used to calculate means and standard deviations, median number of drugs, frequencies, and to identify and determine frequencies of specific drug regimens.
SAS® was used for all statistical calculations. Means and standard deviations were calculated for interval data (age). Medians were used to examine distribution of the number of drugs. Mean ages of independent, unequal samples were compared using Student’s t-test with pooled variance estimate. 50 The Chi Square statistic (χ2) was used to compare cross-sectional groups by ordinal variables (i.e., number of drugs, level of multiple medication use). 51 If significant overall results were obtained and expected cell sizes were at least five, partitioned Chi Square tests were performed to determine which categories were contributing to the observed differences. 51 Bonferroni corrections were made to p-values to accommodate multiple comparisons. 50
Demographic characteristics of the samples are presented in Table 2. AD patients were significantly younger than NoAD patients. There was a higher proportion of females in the AD sample, although both samples were predominantly male. Data on race/ethnicity were missing from prescription records for significantly different proportions of each subset, preventing valid comparisons based on race/ethnicity.
Figure 1 shows that AD patients were more likely to be younger than 60 years of age, while NoAD patients were more likely to be 60 years or older. The nadir in both age distributions was observed at age 60 years, suggesting this age as a natural point at which to split the two samples into two groups to adjust for differences in age distributions. Stratification of each sample by age (< 60 years versus ≥ 60 years) yielded 1,100 younger and 891 older AD patients, and 1,195 younger and 2,535 older NoAD patients (age information was missing for 2 patients who were excluded from the analyses).
Younger Patients (< 60 Years of Age)
Younger AD- and NoAD patients were similar in age and gender, lending support to the selection of age 60 as the cutpoint for the two age groups. The small number of females in the younger subsets prevented comparisons of multiple medication use by gender. The average age of younger female AD patients (44.0 ± 7.3 years) was slightly less than younger male AD patients (48.7 ± 6.4 years) (t = –6.26, p < 0.0001). Younger female AD patients were slightly older than younger female NoAD patients (41.8 ± 9.6 years; t = –8.30, p < 0.0001). Data on race/ethnicity were missing for 17.7% of younger AD patients and 31.6% of younger NoAD patients, thereby preventing valid comparisons by this variable.
A total of 947 different drug regimens (i.e., total number of specific drug entities the patient was receiving regardless of dose, formulation, or administration schedule) were dispensed to the 1,100 younger AD patients. The simplest regimens were comprised of one drug and the most complex exceeded 20 different drugs. Twenty-six monodrug regimens were shared by 2 to 10 patients each. Only 4 drug regimens, all of which consisted of monotherapy, were shared by 1% or more of patients: sertraline in 17 patients, fluoxetine in 16 patients, amitriptyline in 14 patients, and paroxetine in 13 patients.
The 837 different drug regimens dispensed to the 1,195 younger NoAD patients included 95 drug regimens shared by at least 2 patients. Of these, only 6 were shared by at least 1% of patients and all consisted of monopharmacy (naproxen in 57 patients, ibuprofen in 28, ranitidine in 28, lisinopril in 18, aspirin in 15 and levothyroxin in 14 patients).
In summary, 83% (917/1,100) of the younger AD patients were receiving a unique drug regimen (total specific drugs irrespective of dose or administration schedule) compared with 62% (742/1,195) of NoAD patients.
Extent and nature of multiple medication use.
Distribution of the levels of multiple SG drug use in the younger patient subsets are shown in the top half of Table 3, with successive SG drug counts based on first including and then excluding ADs.
Drug count including ADs.
Figure 2 compares the median number of SG drugs and levels of multiple medication use in younger AD and NoAD patients. Younger AD patients received a median of 3 more drugs than younger NoAD patients. Because selection criteria required that all AD patients have a current supply of at least 1 antidepressant, no patient in the AD subset could be receiving 0 SG drugs. In the younger patient subset, 91.4% of AD patients, compared with 66.3% of NoAD patients, were receiving at least 2 SG drugs. The proportion of younger AD patients receiving 5 or more SG drugs was more than twice that of younger NoAD patients (51.8% versus 22.3%, χ2 = 214.9, p < 0.0001). Younger AD patients were 4 times more likely than younger NoAD patients to be receiving 8 or more SG drugs (23.9% versus 5.9%, χ2 = 145.9, p < 0.0001).
Drug count excluding ADs.
Figure 2 shows the result of excluding ADs from the SG drug counts. The median number of drugs decreased from 5 to 3 in younger AD patients, while the counts in younger NoAD patients, of course, remained the same (median of 2) because those patients were selected for not having received ADs. Levels of multiple medication use in younger AD patients remained significantly higher than in younger NoAD patients (χ2 = 169.2, p < 0.0001). The percentage of younger AD patients on 5 or more drugs was nearly twice that of younger NoAD patients (39.4% versus 22.3%, χ2 = 84.2, p < 0.0001). Also, the proportion of younger AD patients on 8 or more SG drugs was nearly three times that of younger NoAD patients (16.3% versus 5.9%, χ2 = 64.2, p < 0.0001).
Older Patients (≥ 60 Years of Age)
Although the difference in the mean ages of the older AD and NoAD patients was statistically significant (Table 2), the difference in mean ages was less than 1 year (71.0 versus 71.8 years), suggesting that the split at age 60 years is reasonable. The small number of females prevented detailed analyses by gender. Data on race/ethnicity were missing for 17.1% of older AD patients and 26.2% of older NoAD patients, preventing analyses by race/ethnicity.
A total of 867 different drug regimens occurred in the 891 older AD patients. Of those regimens, 13 occurred in 2 to 6 patients. The 854 remaining regimens were unique, each occurring in only 1 patient.
Of the 2,030 different total drug regimens (monodrug and combination regimens) found in the 2,535 older NoAD patients, 139 regimens were shared by 2 to 71 patients. Only one mono-drug regimen occurred in 1% or more of patients and that was aspirin.
In summary, 96% (854/891) of the older AD patients were receiving a unique drug regimen compared with 75% (1,891/2,535) of NoAD patients.
Extent and nature of multiple medication use.
Distribution of the levels of multiple SG drug use in the older patient subsets are shown in the lower half of Table 3, with successive SG drug counts first including and then excluding antidepressant drugs.
Drug count including ADs.
Figure 2 compares the median number of SG drugs and levels of multiple medication use in older AD and NoAD patients. Older AD patients received a median of 2 more drugs than older NoAD patients. Older AD patients also had higher levels of multiple SG drug use than older NoAD patients (χ2 = 339.2, 4df, p < 0.0001). As noted in the discussion of the younger subsets, because the selection criteria required AD patients to have a current supply of at least 1 antidepressant, no one in that group could be on 0 SG drugs. The percentage of older AD patients receiving 5 or more SG drugs was nearly twice that of older NoAD patients (70.6% vs. 37.8%, respectively, χ2 = 286.0, p < 0.0001). Older AD patients were nearly 3 times more likely to receive 8 or more SG drugs than older NoAD patients (37.6% vs. 12.8%, respectively, χ2 = 850.4, p < 0.0001).
Drug count excluding ADs.
Figure 2 shows the result of excluding ADs from the SG drug counts. The median number of drugs decreased from 6 to 5 in older AD patients, which was 1 more drug than the median of 4 in older NoAD patients. The proportion of older AD patients receiving 5 or more drugs dropped from 70.6% to 55.4% when ADs were excluded from the count, but was still higher than the rate in older NoAD patients (37.8%) (χ2 =83.5, p < 0.0001). The proportion of older AD patients receiving 8 or more drugs was reduced by one-third, but it was still more than twice that of older NoAD patients (29.0% vs. 12.8%, χ2 =122.3, p < 0.0001).
Younger versus Older Patient Subsets
Drug count including ADs.
Based on median number of drugs, the highest levels of multiple medication use occurred in older AD patients (median = 6 drugs), followed by younger AD patients (median = 5 drugs), followed by older NoAD patients (median = 4 drugs), and finally, followed by younger NoAD patients (median = 2 drugs) (Figure 2).
The proportion receiving 5 or more drugs was higher in older than in younger patients in both subsets: (70.6% vs. 51.8%, older vs. younger AD patients; χ2 = 125.5, p < 0.0001; 37.8% vs. 22.3% in older vs. younger NoAD patients; χ2 = 57.2, p < 0.0001). The same was true for patients receiving 8 or more drugs (37.6% vs. 23.6%, older vs. younger AD patients, χ2 = 46.1, p < 0.0001 and 12.8% vs. 5.9%, older vs. younger NoAD patients; χ2 = 40.8, p < 0.0001). In addition, younger AD patients were more likely to be on 8 or more SG drugs than older NoAD patients (23.6% versus 12.8%, respectively; χ2 = 55.0, p < 0.0001).
Drug count excluding ADs.
When ADs were excluded from the count (Figure 2 and Table 3), older AD and NoAD patients received 2 more drugs than younger AD and NoAD patients (medians of 5 vs. 3 drugs in AD patients and medians of 4 vs. 2 drugs in NoAD patients). Compared to older AD patients, younger AD patients were 3 times more likely to be receiving ADs as their only drug (10.4% vs. 3.8%, χ2 = 34.0, p<0.0001). Older AD and NoAD patients were about twice as likely as their younger counterparts to receive 8 or more drugs (29.0% vs. 16.3% in older vs younger AD patients, χ2 = 46.2, p < 0.0001; 12.8% vs. 5.9% in older vs. younger NoAD patients; χ2 = 41.21, p < 0.0001). The proportion of younger AD patients receiving 8 or more drugs was closer to the rate in older NoAD patients (16.3% vs. 12.8%; respectively, χ2 = 7.84, p > 0.05) than to the rate in older AD patients (16.3% vs. 29.0%; respectively, χ2 = 46.2, p < 0.0001).
Summary of Key Findings
In this study, the 1,100 younger and 891 older AD patients, respectively, received 947 and 867 different drug regimens (i.e., total specific drug entities the patient was receiving regardless of dose, formulation or drug administration schedule). The 1,195 younger and 2,535 older NoAD patients received 837 and 2,030 different drug regimens, respectively. Each drug regimen containing 2 or more SG drugs occurred in fewer than 1% of patients, regardless of age or antidepressant use. Younger AD patients were 4 times more likely to receive 8 or more SG drugs than younger NoAD patients. ADs accounted for one-fourth of this difference, so that, when ADs were excluded from the SG drug counts, younger AD patients were 3 times more likely to be receiving 8 or more non-AD SG drugs. The likelihood of being on 8 or more SG drugs was greatest in older AD patients (37.6%), followed by younger AD patients (23.6%), older NoAD patients (12.7%), and, finally, younger NoAD patients (5.9%).
Interpretation of Findings
The large number of unique drug regimens found in both age groups, regardless of whether or not the patients were receiving ADs, suggests that such complexity and uniqueness are pervasive. However, patients receiving ADs showed more diversity in their prescribed regimens than did patients receiving treatment only with non-AD drugs (e.g., blood pressure medications). This finding is consistent with reports of increased psychiatric and medical comorbidity among patients with major depression compared with non-depressed patients. 52–65
In this patient setting, age played less of a role than did treatment with ADs. For example, the rates of multiple medication use in younger patients on ADs most closely resembled rates in older patients not on ADs. These findings regarding younger veterans are consistent with the results of a cross-sectional study by Kazis, et al., which found that physical limitations in the youngest veterans (ages 22–49 years) were almost as severe as in the oldest veterans and that a greater proportion of younger than older veterans had clinical depression and required more mental health services. 66 Most research on multiple medication use has focused on the elderly because they are particularly vulnerable to adverse DDIs. 2 The findings of this study stress the uniqueness of drug regimens and higher levels of multiple medication use among younger patients on ADs in comparison to older patients not on ADs. These findings raise concerns regarding the potential for adverse DDIs in younger patients on ADs. Furthermore, accurate assessment of adverse DDIs involving ADs is especially important, since such DDIs can mimic symptoms of the depression the ADs are intended to treat. 42 That can lead to a misinterpretation of the DDI as a worsening of primary symptoms and, ironically, the prescribing of additional drugs or higher doses of ADs. 42
Limitations of the Study and Generalizability of the Findings
There are several limitations to this study, including the generalizability of findings from a predominantly male sample of VA outpatients in a midwest setting to other adult U.S. outpatients. In addition, comparisons to other studies are somewhat limited by the focus on SG drugs. The results of this study pertain most closely to outpatients at other VA sites and, to some extent, older males in the general U.S. population. Veterans who use the VA health care system are often older, poorer, less educated, and sicker than other patient populations. 67,68 Adjustments for these factors would improve generalizability but would require linking prescription and medical records databases, which was beyond the scope of the present study. Future studies will benefit from recent changes in the VA healthcare system, including the introduction of electronic medical records that facilitate pharmacoepidemiologic and pharmacoeconomic research. 69 VA prescription dispensing data would be expected to underestimate use of non-VA-provided medications and to overestimate use in cases of noncompliance or when a patient is switched to another drug without discontinuing the prescription order for the previous drug. In addition, given the available data, it cannot be said to what extent ADs were prescribed for treatment of depression as opposed to other indications.
The demographic compositions of the two samples are consistent with Veterans Health Administration reports on patient characteristics. 70 The small proportion of females in each sample prevented adequate analyses by gender and, hence, further studies would be needed to address possible effects of gender on complexity and uniqueness of prescribing patterns. Although the substantial proportions of missing data on race/ethnicity are consistent with other studies using VA computerized databases, 71 this fact suggests the need to obtain such data by other means, such as patient surveys.
A main focus of this project was multiple medication use and prescribing patterns in relation to potential DDIs. For this reason, it was decided to count each drug entity separately, even when these drugs occurred as part of a combination formulation. This approach affects comparisons with findings from other studies that counted combination formulations differently.
Other confounding factors are related to unique characteristics of the VA population of patients. For example, they are initially screened to rule out serious medical and psychiatric conditions; they have better access to subsidized healthcare; and older veterans are likely to have seen combat with potentially deleterious effects on health. For a more detailed discussion of these issues, see Part I of this report describing our previous study in this population. 29
The results of this study suggest the need for further research to:
- Determine effective alternative strategies for assessing potential risks associated with multi-drug regimens, such as developing drug classifications or groupings based on pharmacodynamic and pharmacokinetic characteristics relevant to DDIs. For example, a drug would be defined “at risk” for becoming a victim of a DDI based on the fact that its clearance is preferentially mediated by a single, drug-metabolizing enzyme. Other drugs would be classified as “potential perpetrators” based on their ability to induce or inhibit the same drug-metabolizing enzyme.
- Analyze multiple medication regimens, as was done in this study, but replacing individual drugs with functional drug groupings, as proposed in number 1 above, to find common combinations warranting further evaluation.
- Predict the risk and the nature of such DDIs using the system described above and then examine medical records to validate predictions in actual clinical practice.
- Select targeted patient samples so that effects of other potentially relevant variables, including gender, ethnicity, and age, can be assessed.
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. 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.
3. 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.
4. Cooper JW. Adverse drug-reaction-related hospitalizations of nursing home facility patients: A 4-year study. South Med J 1999;92:485–90.
5. Koechler 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.
6. Hanlon JT, Fillenbaum GG, Schmader KE, et al. Inappropriate drug use among community-dwelling elderly. Pharmacotherapy 2000;20:575–82.
7. Colley CA, Lucas LM. Polypharmacy: The cure becomes the disease. J Gen Intern Med 1993;8:278–83.
8. Bates DW, Miller EB, Cullen DJ, et al. Patient risk factors for adverse drug events in hospitalized patients. Arch Int Med 1999;159:2553–60.
9. 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.
10. 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.
11. Beyth RJ, Shorr RI. Epidemiology of adverse drug reactions in the elderly by drug class. Drugs and Aging 1999;14:231–9.
12. 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.
13. Hanlon JT, Schmader KE, Koronkowski MJ, et al. Adverse drug events in high risk older outpatients. J Am Geriatr Soc 1997;45:945–8.
14. 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.
15. Barat I, Andreasen F, Damsgaard EMS. The consumption of drugs by 75-year-old individuals living in their homes. Eur J Clin Pharmacol. 2000;56:501–509.
16. Bates DW, Spell N, Cullen DJ, et al. The costs of adverse drug events in hospitalized patients. JAMA 1997;277:307–11.
17. Leape LL, Bates DW, Cullen DJ, et al. Systems analysis of adverse drug events. JAMA 1995;274:35–43.
18. Routledge PA, O’Mahony MS, Woodhouse KW. Adverse drug reactions in elderly patients. Br J Clin Pharmacol 2004;57:121–6.
19. Halkin H, Katzir I, Kurman I, et al. Preventing drug interactions by online prescription screening in community pharmacies and medical practices. Clin Pharmacol Ther 2001;69:260–5.
20. Raschke RA, Gollihare B, Wunderlich TA, et al. A computer alert system to prevent injury from adverse drug events: Development and evaluation in a community teaching hospital. JAMA 1996;280:1317–20.
21. Samsa GP, Hanlon JT, Schmader KE, et al. A summated score for the medication appropriateness index: Development and assessment of clinimetric properties including content validity. J Clin Epidemiol 1994;47:891–6.
22. Schmader K, Hanlon JT, Weinberger M, et al. Appropriateness of medication prescribing in ambulatory elderly patients. J Am Geriatr Soc 1994;42:1241–7.
23. Fick DM, Cooper JW, Wade WE, et al. Updating the Beers criteria for potentially inappropriate medication use in older adults: Results of a US consensus panel of experts. Arch Intern Med 2003;163:2716–24.
24. Carter BL, Malone DC, Valuck RJ, et al. The IMPROVE study: Background and study design. Am J Health-Syst Pharm 1998;55:62–7.
25. Hamdy RC, Moore SW, Whalen K, et al. Reducing polypharmacy in extended care. South Med J 1995;88:534–8.
26. Jameson JP, VanNoord GR. Pharmacotherapy consultation on polypharmacy patients in ambulatory care. Ann Pharmacother 2001;35:835–40.
27. Hanlon JT, Weinberger M, Samsa GP, et al. A randomized, controlled trial of a clinical pharmacist intervention to improve inappropriate prescribing in elderly outpatients with polyphar-macy. Am J Med 1996;100:428–37.
28. McMullin ST, Reichley RM, Kahn MG, et al. Automated system for identifying potential dosage problems in a large university hospital. Am J Health Syst Pharm 1997;54:545–9.
29. Preskorn SH, Silkey B, Shah R, et al. Complexity of outpatient medication use in the Veterans Affairs healthcare system: Part I. Outpatient use in relation to age and number of prescribers. J Psychiatr Pract 2005;11:5–15.
30. 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.
31. Bjerrum L, Sogaard 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.
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. Rosholm JU, 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.
34. Atkin PA, Veitch PC, Veitch EM, et al. The epidemiology of serious adverse drug reactions among the elderly. Drugs and Aging 1999;14:141–52.
35. Ebly EM, Hogan DB, Fung TS. Potential adverse outcomes of psychotropic and narcotic use in Canadian seniors. J Clin Epidemiol 1997;50:857–63.
36. Lawlor DA, Patel R, Ebrahim S. Association between falls in elderly women and chronic diseases and drug use: Cross sectional study. BMJ 2003;327:712–7.
37. Moore AR, O’Keefe ST. Drug-induced cognitive impairment in the elderly. Drugs and Aging 1999;15:15–28.
38. Weiner DK, Hanlon JT, Studenski SA. Effects of central nervous system polypharmacy on falls liability in community-dwelling elderly. Gerontology 1998;44:217–21.
39. Thapa PB, Gideon P, Cost TW, et al. Antidepressants and the risk of falls among nursing home residents. N Engl J Med 1998;339:875–82.
40. Leipzig RM, Cumming RG, Tinetti ME. Drugs and falls in older people: A systematic review and meta-analysis. I. Psychotropic drugs. J Am Geriatr Soc 1999;47:40–50.
41. Lawson DH. Life threatening drug reactions amongst medical in-patients. Scot Med J 1979;24:127–30.
42. Preskorn SH, Jerkovich GS. Central nervous system toxicity of tricyclic antidepressants: Phenomenology, course, risk factors, and role of therapeutic drug monitoring. J Clin Psychopharmacol. 1990;10:88–95.
43. Muijen M, Silverstone T. A comparative hospital survey of psychotropic drug prescribing. Br J Psychiatry 1987;150:501–4.
44. Willcox SM, Himmelstein DU, Woolhandler S. Inappropriate drug prescribing for the community-dwelling elderly. JAMA 1994;272:292–6.
45. Pincus HA, Tanielian TL, Marcus SC, et al. Prescribing trends in psychotropic medications: Primary care, psychiatry, and other medical specialties. JAMA 1998;279:526–31.
46. Olfson M, Marcus SC, Pincus HA, et al. Antidepressant prescribing practices
of outpatient psychiatrists. Arch Gen Psychiatry 1998;55:310–6.
47. Olfson M, Marcus SC, Druss B, et al. National trends in the out-patient treatment of depression. JAMA 2002;287:203–9.
48. McManus P, Mant A, Mitchell PB, et al. Recent trends in the use of antidepressant drugs in Australia, 1990–1998. Med J Aust 2000;173:458–61.
49. Roberts E, Norris P. Growth and change in the prescribing of anti-depressants in New Zealand, 1993–1997. NZ Med J 2001;114:25–7.
50. Snedecor GW, Cochran WG. Statistical methods. Ames, IA: Iowa State University Press; 1980.
51. Fleiss JL. Statistical methods for rates and proportions. New York: John Wiley and Sons; 1973.
52. Coulehan JL, Schulberg HC, Block MR, et al. Depressive symptomatology and medical comorbidity in a primary care setting. Int J Psychiatry Med 1990;20:335–47.
53. Spollen JJ III, Gutman D. The interaction of depression and medical illness: Depression and comorbid cardiovascular disease. Medscape. www.medscape.com/viewarticle/457165
54. Cohen-Cole SA, Kaufman KG. Major depression in physical illness: Diagnosis, prevalence, and antidepressant treatment (a ten year review: 1982–1992). Depression 1993;1:181–204.
55. Gagnon LM, Patten SB Major depression and its association with long-term medical conditions. Can J Psychiatry 2002;47:149–52.
56. Haug TT, Mykletun A, Dahl AA. Are anxiety and depression related to gastrointestinal symptoms in the general population? Scan J Gastroenterol 2002;3:294–8.
57. Anderson RJ, Freedland KE, Clouse RE, et al. The prevalence of comorbid depression in adults with diabetes: A meta-analysis. Diab Care 2001;24:1069–78.
58. Lacro JP, Jeste DV. Physical comorbidity and polypharmacy in older psychiatric patients. Biol Psychiatry 1994;36:146–52.
59. Wells KB, Rogers W, Burnam A, et al. How the medical comorbidity of depressed patients differs across health care settings: Results from the Medical Outcomes Study. Am J Psychiatr 1991;148:1688–96.
60. Blazer DG, Kessler RC, McGonagle KA, et al. The prevalence and distribution of major depression in the national community sample; The National Comorbidity Survey. Am J Psychiatry 1994;151:979–86.
61. Carney RM, Rich MW, teVelde A, et al. Major depressive disorder in coronary artery disease. Am J Cardiol 1987;60:1273–5.
62. Forester AW, Lipsey JR, Teitelbaum ML, et al. Depression following myocardial infarction. Int J Psychiatry Med 1992;22:33–46.
63. Gonzalez MB, Snyderman TB, Colket JT, et al. Depression in patients with coronary artery disease. Depression 1996;4:57–62.
64. Weissman MM, Bland RC, Canino GJ, et al. Cross-national epidemiology of major depression and bipolar disorder. JAMA 1996;276:293–9.
65. Wells KB, Stewart A, Hays RD, et al. The functioning and well-being of depressed patients: Results from the Medical Outcomes Study. JAMA 1989;262:914–9.
66. 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.
67. Wolinsky FD, Coo RM, Mosely RR, et al. Veterans and nonveterans use of health services: A comparative analysis. Med Care 1985;23:1358–71.
68. Sloan KL, Fishman PA, Malkin JD. Comparing VA to the Group Health Population on the chronic disease score categories. Presentation, 2000 HSR&D Annual Meetings.
69. Ogden JE, Muniz A, Patterson AA, et al. Pharmaceutical services in the Department of Veterans Affairs. Am J Health Syst Pharm 1997;54:761–75.
70. Miller DR, Skinner KM, Kazis LE. Study design and sampling in the Veterans Health Study. J Ambul Care Manage 2004;27:166–79.
71. Kressin N. Challenges in conducting research on racial variations in health care. Washington, DC: Workshop held at the VA Health Services Research and Development Meetings, 2000.