Prescribing Potentially Inappropriate Psychotropic Medications to the Ambulatory Elderly

Background  Psychotropic agents account for 23% to 51% of all inappropriate medications prescribed based on 1991 inappropriate medication criteria for nursing home residents. The criteria were revised to apply to all people older than 65 years. This study used the revised criteria in ambulatory settings to quantify potentially inappropriate prescription of psychotropic agents and identify associated characteristics.

Methods  The 1996 public use data files of the National Ambulatory Medical Care Survey and the National Hospital Ambulatory Medical Care Survey were analyzed for inappropriate prescription of psychotropic medications for the elderly in office-based settings and outpatient departments. Disease-independent and disease-dependent criteria were analyzed.

Results  Elderly patients were prescribed a psychotropic agent in 8.7% of all visits, antidepressant and antianxiety agents being the most frequently prescribed medications. Commonly, elderly patients receiving psychotropic agents were female, white, aged between 65 and 74 years, and received health care in a metropolitan area. Potentially inappropriate psychotropic agents were prescribed in 27.2% of all visits involving a psychotropic agent for the elderly. Disease-independent criteria (eg, antidepressant agents and long-acting benzodiazepines) accounted for most of the potentially inappropriate prescriptions. Factors positively associated with potentially inappropriate prescriptions included older age, "seen before" status, and antidepressant drug class, while enrollment in Medicaid, antipsychotic drug class, living in the Northeast region, and receiving health care in a metropolitan area were negatively associated.

Conclusions  Potentially inappropriate prescription of psychotropic agents is very common for the elderly patient in the ambulatory setting. By focusing on the agents most frequently involved (eg, amitriptyline and long-acting benzodiazepines), provider characteristics (eg, location), and patient characteristics (eg, age), the greatest impact on potentially inappropriate prescribing can be achieved.

PSYCHOTROPIC AGENTS are frequently prescribed to the elderly for a wide variety of mental health conditions.^1-3 These agents significantly improve patients' conditions but also may lead to deleterious outcomes, especially if not used properly.^2,3 Elderly patients are more susceptible to these outcomes owing to pharmacokinetic and pharmacodynamic changes.^2,4 General concerns regarding medication use by elderly patients led to the 1991 publication by Beers et al⁵ of potentially inappropriate medication criteria for patients older than 65 years who resided in long-term care facilities. Researchers applying these criteria have found that psychotropic agents constitute a large portion of the potentially inappropriate prescriptions for the elderly in various settings (23%, 44%, and 51% in long-term care, office-based settings, and outpatient departments, respectively).^6-8 The applicability of these criteria to settings outside of long-term care has been questioned.

Beers,⁹ using a consensus panel of 6 nationally recognized experts, revised the criteria in 1997 to apply to all people older than 65 years, to incorporate new information, and to include a severity rating. This consensus panel redefined medications that should be generally avoided and identified disease-dependent inappropriate medications. Given that psychotropic agents have constituted a high proportion of potentially inappropriate medications, and that the revised criteria have not been examined in the ambulatory population, we undertook this research. The specific objectives of this study were (1) to examine the nature and extent of potentially inappropriate prescription of psychotropic agents for the elderly in physician offices and hospital outpatient departments and (2) to determine the factors predicting potentially inappropriate prescription of psychotropic agents in these settings.

The National Center for Health Statistics (NCHS) is the primary source of ambulatory care data in the United States. The agency annually conducts the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Medical Care Survey (NHAMCS). These 2 national surveys provide the most comprehensive information about the provision and use of ambulatory care services in the United States.¹⁰ In this study, the 1996 public use data files from the NAMCS and NHAMCS were used to examine potentially inappropriate prescription of psychotropic agents in physician offices and hospital outpatient departments.^11,12

The NAMCS provides comprehensive data on ambulatory care services rendered in office-based settings in the United States. Findings are based on a sample representative of all in-person visits to office-based physicians. In 1996, 2142 office-based physicians participated in the multistage probability sample survey for an overall response rate of 70%. Data were collected using patient records forms (PRFs) by participating physicians or their staff for a systematic random sample of office visits during the randomly assigned reporting period. The PRF included patient demographic information, diagnoses (up to 3 diagnoses), drugs prescribed (up to 6 medications), and disposition of the visit. The NCHS also collected data on physician characteristics prior to the survey implementation. A total of 29,805 PRFs were collected by participating physicians in 1996.

The NHAMCS collected data on ambulatory services provided by hospital outpatient departments and emergency departments in the United States. The outpatient department component of NHAMCS involved a multistage probability sample survey to collect data from a representative sample of visits to hospital outpatient departments. In 1996, a national sample of 486 hospitals participated for an overall response rate of 95%. Data describing the clinical nature of the visits were collected using PRFs similar to the one used in the NAMCS. A total of 235 outpatient departments provided 29,806 PRFs. The NCHS also collected data regarding the provider and institution characteristics. The outpatient department component of the NHAMCS was used for this study.

The data collection for the survey was done by the US Bureau of the Census Housing Survey Branch, and data processing and coding were performed by Analytic Services Inc of Durham, NC. Diagnoses were coded according to the International Classification of Diseases, Ninth Revision, Clinical Modification.¹³ A predefined checklist was also included in the PRF in both national surveys to record other existing conditions. Prescribed medications were coded according to a unique classification scheme developed at the NCHS.¹⁴ Respondents were instructed to record all new or continued medications ordered, supplied, or administered at the visit. Additional details regarding the methods, definitions, and technical information of the NAMCS and NHAMCS can be found in other sources.^11,12,15,16

The data extraction and analysis involved use of SAS software (version 6.03, 1992; SAS Institute Inc, Cary, NC) and the Statistical Export and Tabulation System (version 1.22a, 1991; SETS) (National Center for Health Statistics, US Dept of Health and Human Services). For the purpose of this study, psychotropic drugs were categorized into 6 classes: antipsychotic agents, stimulants, sedative/hypnotic agents, antianxiety agents, and antidepressants. A detailed list of medications under each psychotropic class^17-19 is provided in Table 1. The psychotropic agent classification was based on standard references and published literature. The explicit criteria developed by Beers⁹ were used to define potentially inappropriate psychotropic use by the elderly. The psychotropic drugs considered potentially inappropriate for this study entailed medications that should be generally avoided in the elderly (disease-independent inappropriate medications) and medications that should be avoided because of preexisting disease or condition (disease-dependent inappropriate medications). The disease-independent and disease-dependent inappropriate medications^{9,13,17,18,20-27} are listed in Table 2 and Table 3. Psychotropic drugs considered potentially inappropriate based on dose were not examined because of limitations of the data sources.

Data extraction, coding, and analysis involved 3 steps. First, visits involving psychotropic drugs for patients aged 65 years or older were abstracted. Second, single-entity and combination drugs, along with the diagnoses, were examined for the potentially inappropriate psychotropic medications. Third, certain variables including patient, visit, and provider characteristics were recoded for analytical purposes. Extracted data sets from the NAMCS and NHAMCS were combined to analyze potentially inappropriate psychotropic agent prescribing in ambulatory settings. From the 2 national surveys, 1373 records involving psychotropic agents were extracted in the first step and 309 records involving potentially inappropriate psychotropic agents were extracted in the second step.

National visit estimates were derived based on the patient sampling weight provided by the NCHS to examine the nature and extent of potentially inappropriate psychotropic prescribing in ambulatory care settings (N = 309). All the visit estimates, visit rates, and percentage distributions were calculated based on the information provided by the NCHS for the national surveys.^11,12 To examine predictors of potentially inappropriate psychotropic prescription, multiple logistic regression methods were applied to modified weights to account for the complex design of the national surveys (N = 1373).²⁸ The predictor variables in the model were patient characteristics (ie, age, sex, race, insurance status, and payment method), visit (ie, referral, "seen before" status, injury, and surgery), medication (ie, number of medications, number of psychotropic agents, and psychotropic class), and provider characteristics (ie, nonphysician, ambulatory setting, region of the country, and metropolitan location of health care). Statistical significance was defined as P≤.05 for all statistical analyses.

According to the national surveys, psychotropic medications were prescribed in an estimated 16.55 million visits to physician offices and outpatient departments for the elderly in 1996. This represents 8.70% of all clinic visits by the elderly, for an annual visit rate of 51.97 visits for every 100 persons aged 65 years or older. The elderly receiving psychotropic agents were most often female (68.08%), white (87.74%), between ages 65 and 74 years (53.43%), and obtained care in metropolitan areas (72.17%). The psychotropic prescribing patterns for the elderly in ambulatory settings are given in Table 4.

Approximately 4.5 million visits (27.2% of all visits involving psychotropic agents prescribed for the elderly) involved at least 1 potentially inappropriate psychotropic agent. The percentages of visits involving prescription of potentially inappropriate psychotropic agents in office practice and outpatient departments were 27.95% and 19.28%, respectively. Disease-independent potentially inappropriate agents accounted for 94.14%, and disease-dependent potentially inappropriate agents accounted for 12.13% of all visits involving potentially inappropriate psychotropic agents (Table 5). A visit was counted more than once if multiple criteria were met. Antidepressant agents and antianxiety agents constituted most of the potentially inappropriate prescribing in disease-independent situations, whereas antidepressant agents and sedative/hypnotic drugs made up most of the potentially inappropriate prescription in disease-dependent situations. Overall, potentially inappropriate psychotropic prescribing most often involved antidepressant agents and antianxiety agents, and the least likely involved drugs were antipsychotic agents. Beers' criteria⁹ also included a severity rating scale that incorporated the probability of the adverse outcome occurring and the clinical significance of the problem (Table 2 and Table 3). According to this severity rating scale, 95.34% of the visits involving potentially inappropriate psychotropic medications cause a high-severity type of adverse outcome.

The logistic model was used to evaluate the factors predicting potentially inappropriate psychotropic agent prescribing in ambulatory settings. The factors in Table 6 that provide the predictive power for the visits involving potentially inappropriate psychotropic agents include patient characteristics of age and insurance, visit characteristic of "seen before" status, drug characteristic of drug class, and provider characteristics of region and location. Older patient age, "seen before" status, and antidepressant drug class were positively associated, whereas use of Medicaid, antipsychotic drug class, Northeast region, and metropolitan location were negatively associated with potentially inappropriate psychotropic prescribing in ambulatory settings.

The percentage of visits involving psychotropic medications for the elderly reported in this study (8.7%) is lower than the percentage of community-based elderly patients receiving psychotropic medications (ie, 12.2%-30%).^3,29-32 This might be due to differences in method issues such as definition of psychotropic agents, data collection (eg, number of medications limited to 6 in this study), unit of observation, duration of data collection, and regional characteristics. Studies reporting sex and race information among elderly patients in the community receiving psychotropic agents concurred with our study's findings of greater prevalence among women^29,31,33,34 and whites.²⁹ While our study found that antidepressant drugs were the most commonly prescribed psychotropic agents, followed by antianxiety agents and sedative/hypnotic drugs, previous research on community-dwelling elderly patients found that antianxiety medications were most often prescribed, followed by antidepressant and sedative/hypnotic drugs.^3,29,35 The more frequent prescription of antidepressant agents found in our study may be owing to the increased emphasis on recognizing and treating depression among the elderly¹⁹ since the early 1980s, when data were collected for the comparative studies.^3,29,35 Of the visits involving prescription of psychotropic agents, 18.3% included 2 or more such agents, which is similar to the findings of Dealberto et al,²⁹ who found that 14.5% of their community-based elderly patients prescribed psychotropic agents were receiving 2 to 4 of these drugs. Overall, the demographic characteristics of the patients in our study are similar to those in studies involving community-based elderly patients.^29,31,33,34 However, in our study, the most frequently prescribed psychotropic medication class differed,^3,29,35 and the percentage of visits involving psychotropic agents was lower than the previously reported percentages of elderly patients receiving psychotropic agents.^3,29-32

More than 25% of visits involving psychotropic agents were for one of the potentially inappropriate psychotropic agents, which is greater than the percentage reported using the 1991 criteria of Beers et al⁵ in an office-based setting (16.9%).³⁶ This may be due in part to the different definitions of psychotropic agents used or to the inclusion of outpatient departments in our current study. However, a larger effect would have been expected from the expanded 1997 criteria, which included both disease-dependent criteria and more disease-independent criteria.

Most potentially inappropriate psychotropic medication visits are accounted for by the disease-independent criteria. Since the disease-dependent criteria are much more restrictive, this category would be expected to constitute a smaller percentage of visits. The small percentage may also be because of the limited number of diagnoses (ie, up to 3) that were available in the data source. This research suggests that disease-independent criteria are an important aspect to consider when attempting to have the greatest impact on the prescription of potentially inappropriate psychotropic agents.

Antidepressant and antianxiety classes were the most commonly involved potentially inappropriate psychotropic medications prescribed in this study as well as in the community-based study using the 1991 criteria.³⁶ According to Beers' 1997 list,⁹ these 2 classes made up 5 of the 7 disease-independent criteria and 5 of the 8 disease-dependent criteria. Therefore, it would be anticipated that they would account for most of the potentially inappropriate prescribing. More specifically, most potentially inappropriate psychotropic agent visits involved amitriptyline, an antidepressant agent that can cause substantial anticholinergic effects and sedation.⁹ The antianxiety agents considered potentially inappropriate were primarily composed of the long-acting benzodiazepines, which are associated with sedation, falls, and fractures.⁹ Therefore, by focusing on amitriptyline and long-acting benzodiazepines, physicians can significantly change inappropriate prescribing for and quality of life of the elderly.

The importance of these 2 classes of medications in potentially inappropriate prescription was further validated by a large consensus panel in Canada.³⁷ The criteria produced by this panel also included the avoidance of amitriptyline to treat depression and of long-acting benzodiazepines to treat anxiety, insomnia, or agitation in dementia. While the Canadian criteria are more disease specific, the indications associated with the long-acting benzodiazepines are comprehensive and would be expected to include most of the uses.

Disease-dependent and disease-independent psychotropic medications are considered inappropriate because the risk outweighs the benefit.⁹ Most of the inappropriate psychotropic agents cause a high-severity type of adverse outcome according to Beers' criteria.⁹ Previous research has shown that inadequate or inappropriate drug therapy is an important risk factor for medication-related illnesses among the elderly.³⁸ Therefore, avoiding these medications will significantly reduce the risk of adverse outcomes.

Characteristics associated with a greater likelihood of a visit involving a potentially inappropriate psychotropic agent were patient's older age, seen before status, and antidepressant class. The odds of prescribing a potentially inappropriate psychotropic medication increased by 2% for each additional year of patient age or 10% for every 5-year increment of an elderly patient's age. Previous studies examining all criteria for inappropriate prescription have reported higher rates in patients older than 80 years.^7,39 The positive association of seen before patients suggests that previously seen patients are more than twice as likely to receive potentially inappropriate psychotropic medications than new patients. Consistent with a previous study, our study found the highest odds of potentially inappropriate psychotropic prescription were for antidepressant agents.³⁶ Amitriptyline constituted most of these potentially inappropriate prescription cases, and its use is rarely justified for the elderly because safer antidepressant agents are available.⁹ Because of the consistent findings, efforts focusing on these patient and drug characteristics could provide the greatest impact on inappropriate psychotropic medication prescribing.

Characteristics associated with a decreased likelihood of a visit involving prescription of a potentially inappropriate psychotropic agent included antipsychotic drug class, Medicaid status, metropolitan area, and Northeast region location. The antipsychotic class was least likely to be associated with potentially inappropriate prescription. This may be owing to increased awareness among prescribers or limited use of antipsychotic agents in patients with seizure disorders. The relationship of Medicaid status to decreased usage of potentially inappropriate medications may be partially explained by prescription benefit plans. Medicaid programs involve some systemwide control mechanisms such as prospective and retrospective drug utilization reviews to improve physician prescribing practices.⁴⁰ Consistent with a previous study,⁸ our study found that physicians practicing in metropolitan areas were less likely to prescribe inappropriately, yet no clear explanation exists for this difference. Another geographic factor negatively associated with potentially inappropriate psychotropic prescription was the Northeast region. However, prescribing patterns in other regions might raise some concerns. Therefore, the region and location characteristics associated with lower odds of potentially inappropriate psychotropic use require further examination to understand the relationship before the most effective strategy for modifying prescribing patterns can be deployed.

Although our study provided greater insight into potentially inappropriate psychotropic prescribing patterns for the elderly in outpatient settings, the results were limited by the definitions and data sources used. Because of the limitations of the data sources, our study did not examine psychotropic agents considered potentially inappropriate based on excessive dosage. Therefore, this probably underestimated the extent of inappropriately prescribed psychotropic medications in the elderly. In addition, data collection limitations such as a maximum of 6 medications and 3 diagnoses may have yielded an underreporting of the total number of potentially inappropriate medications prescribed. The independent variables included in the logistic analysis were limited to those available from the data sources. Furthermore, the analyses explored associations between the independent variables and potentially inappropriate prescription, and did not address cause-and-effect relationships between them. In this study, the overall annual visit estimates are considered stable by the NCHS. However, extrapolation of national visits from the subsamples could create potential problems regarding the reliability and confidence level of smaller estimates.^15,16 The inherent disadvantages of using secondary data, such as difficulty in evaluating accuracy caused by errors in data collection, analysis, and reporting, were also limitations of this study.

In summary, potentially inappropriate prescription is a major issue in the effort to optimize care for the elderly while avoiding excessive costs associated with adverse outcomes. Future research should quantify the health care impact of prescribing these potentially inappropriate agents. Currently, inappropriate prescription of psychotropic agents is very common for the elderly in the ambulatory setting, with 27.2% of all psychotropic agent–related visits involving a potentially inappropriate psychotropic agent. A small number of psychotropic medications (ie, amitriptyline and long-acting benzodiazepines) were involved in most of the visits involving potentially inappropriate psychotropic agents. By focusing on these agents and some of the provider characteristics (eg, location) and patient features (eg, patient age), the greatest impact on potentially inappropriate prescribing can be achieved.

Accepted for publication April 6, 2000.

Corresponding author: Jane R. Mort, PharmD, College of Pharmacy–South Dakota State University, 1011 11th St, Rapid City, SD 57701 (e-mail: Jane_Mort@sdstate.edu).