Benzodiazepines, the most prescribed drugs among the sedative, hypnotic, and anxiolytic agents, are associated with potential drug abuse and a significant rebound effect related to withdrawal (
1,
2). Although treatment guidelines recommend the use of other treatment modalities for anxiety disorders in the elderly population, benzodiazepines are still highly prescribed in this population and in many cases for prolonged periods (
3–
6). Studies have reported that up to 32% of community-dwelling older persons in Canada use such drugs (
1,
7–
9). The high rate of benzodiazepine use in Canada may be attributable to the fact that for nearly the entire senior population, such prescriptions are fully covered or partially covered (except for copayments) by provincial drug insurance plans (whereas in the United States, benzodiazepines have not been covered under Medicare Part D). In 2007, close to $100 million was spent on benzodiazepines in Canada (
10).
The use of benzodiazepines in the older adult population has been associated with an increased risk of falls, fractures, and accidents (
11–
13). This is mainly a result of the alteration in physiologic functions observed among elderly persons, such as decreased volume of distribution, hepatic function, and rate of drug clearance, which translate into higher benzodiazepine half-life and increased plasma concentrations (
1). Given the increased prevalence of polypharmacy in the elderly population, the possibility of drug interactions also must be considered. Benzodiazepines have been shown to be susceptible to pharmacological interactions with other sedative-hypnotic agents and with several other drug classes (
14–
17).
Hanlon and colleagues (
18) defined a potentially inappropriate prescription as one that has more potential risk than benefit or one that does not conform to accepted medical standards. Benzodiazepines are reported to be among the principal causes of potentially inappropriate prescribing (
19–
21). Depending on the criteria used (
22–
25), the prevalence of potentially inappropriate benzodiazepine prescriptions has been reported to be as high as 20% among elderly persons living in the community, which corresponds to approximately 50% of benzodiazepine users (
19,
20,
26). A number of studies have shown that inappropriate prescriptions of benzodiazepines or other sedative, hypnotic, and anxiolytic drugs are associated with increased risks of falls (
27,
28) and with higher health care costs (
28,
29); however, other studies have not shown any association between these potentially inappropriate prescriptions and the risk of fracture or other adverse drug reactions (
30,
31). Differences in study designs and criteria used may explain the divergent results.
Because of the prevalence of potentially inappropriate benzodiazepine prescription—as high as 14% in the community-dwelling elderly population of Quebec, a Canadian province with a publicly funded health care system and a pharmaceutical insurance plan covering the elderly population—studying the impact of this issue is of great relevance. This study represents a unique opportunity to document the prevalence of potentially inappropriate benzodiazepine prescription and potential for drug-drug interactions and related excess health care costs in a large representative community-dwelling older adult population by using administrative and survey data, rather than self-reported data alone, and thus increasing the validity of results.
Methods
In this retrospective study, administrative data were used to assess participant characteristics and health service use over a 12-month period after the first prescription of a benzodiazepine. Established criteria (
15,
23) were used to define benzodiazepine prescriptions as inappropriate or appropriate, and the two groups of patients were compared in terms of health service use and total health care costs.
Study sample
The data were from the ESA survey (Enquête sur la Santé des Aînés, or Survey on the Health of the Elderly), a longitudinal study conducted between 2005 and 2009 that used a probabilistic sample (N=2,811) of the older (≥65 years) community-dwelling adult population in Quebec. Self-reported data from the survey were linked to person-level information from the administrative databases of Quebec’s health insurance (Régie de l'assurance maladie du Québec) (RAMQ). Linking had a success rate of 89.9% (N=2,494). A total of 174 participants were not covered by the drug insurance plan and were therefore excluded from the analysis, resulting in an analytic sample of 2,320 patients. The research procedure was previously reviewed and authorized by the Ethics Committee of the University of Sherbrooke Institute of Geriatrics. Additional information about study sampling and data collection has been published elsewhere (
9,
32).
Measures
Independent variables.
On the basis of pharmaceutical claims data from the RAMQ databases, the respondents’ benzodiazepine use was measured for the 12-month period preceding the interview. Benzodiazepine prescriptions were classified as appropriate or potentially inappropriate; a prescription was deemed potentially inappropriate on the basis of Beers criteria updated by Fick and colleagues (
23) or if it entailed a possible drug-drug interaction (
15).
According to updated Beers criteria (
23), long-acting benzodiazepines (half-life >24 hours) and high doses of short- or intermediate-acting benzodiazepines (half-life ≤24 hours) should be avoided. These criteria (see box on the next page) are strictly based on the type of drugs and the dose prescribed rather than on the absence of a valid indication. Beers criteria have generally been considered the most appropriate screening tool to assess the inappropriate use of medication among elderly persons. In addition to their being the most frequently cited in the literature (
33), they are currently used by the Quebec Drug Council and the Canadian Institute for Health Information in their study of potentially inappropriate prescribing.
The criteria for potential benzodiazepine-related drug interactions, developed by Ben Amar (
15), are based on each drug’s pharmacokinetic and pharmacodynamic properties and include drug interactions from the Micromedex index and
Stockley’s Drug Interactions (
16). As shown in the box, most of these drug interactions are related to the inhibition of cytochrome P450 isoforms CYP2C19 and CYP3A4, and the potential consequences include an increased sedative effect (
15,
17).
The respondents’ general medical status was measured by the number of chronic
ICD-9 conditions (
34) reported in the ESA survey. For our analyses, the number was categorized as a dichotomous variable (zero to two conditions versus three or more conditions); the median of three conditions was retained as the cutoff point. The respondent’s mental health status was measured as the presence of a mental disorder (yes or no) as assessed via the diagnostic component of the ESA survey questionnaire, which is based on
DSM-IV criteria (
35). Additional information on the instruments and definitions of the psychiatric disorders studied has been reported elsewhere (
32). Because health status variables are important predictors of health service use, they were used as adjustment variables in our analysis.
Outcomes measurement.
Health service use was assessed for the 12-month period after the first benzodiazepine prescription and was identified from the RAMQ and the MED-ECHO (Quebec’s medical procedure) databases. Health service categories include inpatient stays, outpatient visits, emergency department visits, physician fees, and outpatient medications. From a health care system perspective, the cost valuation was carried out by a direct allocation method (
36). The calculation of unit costs for an inpatient visit (cost per day) or an outpatient consultation (cost per visit) was based on annual budgets and activity reports of the Quebec Ministry of Health. These reports, which include financial reports submitted by each institution in the province, show costs (for example, salaries and cost of medical and office supplies) and activity levels by department, including direct medical and nonmedical services (for example, inpatient stay in a general ward or a psychiatric ward, number of outpatient surgeries, emergency department visits, ambulatory services per visit, nutrition, and housekeeping). The average provincial costs and activity levels for the 2009–2010 fiscal year were retained. Finally, other overhead costs not included in these reports, such as opportunity costs of building and land, were added to unit costs (
37). Physician fees and outpatient medication costs were separately measured within RAMQ databases.
Statistical analysis
Chi square tests were performed to compare respondents’ characteristics at baseline as a function of the presence of potentially inappropriate benzodiazepine prescriptions. Logistic regressions were also employed to describe the association between potentially inappropriate benzodiazepine prescriptions and health service use; the analyses controlled for the respondent’s age, gender, education, marital status, income, region of residence, availability of medical services in the region (number of physicians per 100,000 inhabitants [>116 or ≤116, where 116 represents the first quartile and therefore high availability]), number of days of benzodiazepine use, number of chronic general medical conditions, presence of psychiatric disorders, and self-reported general medical and mental health status. Health service categories used for logistic regressions were emergency department visits (yes or no), inpatient visits (yes or no), and outpatient visits (zero to seven visits versus eight or more visits). Because elderly persons generally have several outpatient consultations per year, the median of this variable (eight visits) was retained as the cutoff point to test the association. The adjusted odds ratio (OR) and its 95% confidence interval (CI) were used as a measure of association.
Accounting for the nonnormal distribution of the cost data and the heteroskedasticity of their log-residuals, a generalized linear model (GLM) with a gamma distribution (log link) was retained to assess the excess costs associated with inappropriate benzodiazepine prescription. This was based on the procedure proposed by Manning and Mullahy (
38), in which a Box-Cox test and a modified Park test were performed to determine the proper link and distribution employed in the GLM. Adjusted incremental costs [Δ$=Exp(LSMeans_2) – Exp(LSMeans_1)] were estimated after control for the above-mentioned potential confounding factors. The potentially inappropriate prescriptions variable was categorized into three mutually exclusive groups: Beers criteria only, drug interactions only, and Beers criteria plus drug interactions.
All adjusted estimates presented were weighted to ensure that the true proportion of older adults within each region and geographical area were reflected in the analysis (
39,
40). Statistical analyses were performed with SAS, version 9.2, and Stata, version 12, with a 5% significance level (two-sided).
Results
On the basis of the weighted sample (N=2,320), the respondents’ mean±SD age was 73.8±6.1; 58.4% were women, and 34.9% had any postsecondary education. Most respondents were from an urban region (59.9%) and reported an annual household income higher than $15,000 (74.8%). Among participants, 59.9% reported at least three chronic general medical conditions. Use of benzodiazepines in the 12 months preceding the survey was reported by 32% (N=744).
Among benzodiazepine users, 9% (N=70) had one or more emergency department visits, 20% (N=149) had one or more inpatient visits, and 68% (N=506) had eight or more outpatient visits. The mean annual health care cost was $6,840±$7,662. In addition, 44% of the benzodiazepine users (N=331) received at least one potentially inappropriate prescription during the study period: 22% of prescriptions (N=163) were inappropriate according to Beers, 15% (N=113) were susceptible to benzodiazepine-related drug interactions, and 7% (N=55) met both criteria. The most frequent potentially inappropriate prescriptions based on the Beers criteria were clonazepam (N=96), a long-acting benzodiazepine, and high doses of short-acting temazepam (N=45). The principal drugs involved in potential benzodiazepine-related drug interactions were diltiazem (N=43), macrolid antibiotics (N=40), trazodone (N=30), and digoxin (N=30).
Table 1 presents data on respondents’ characteristics stratified by the presence of an inappropriate benzodiazepine prescription. Participants with potentially inappropriate prescriptions were more likely than those with appropriate prescriptions to report lower income, worse general medical and mental health status, and the presence of three or more chronic conditions. Those with inappropriate prescriptions were also more likely to use benzodiazepines for more than 90 days during the year.
Table 2 presents results of the logistic regressions examining the association between potentially inappropriate prescriptions and health service use. Compared with participants who had appropriate prescriptions, those with prescriptions meeting Beers criteria were more likely to visit the emergency department (OR=2.35, p=.018); those with prescriptions entailing a possible drug interaction were more likely to be hospitalized (OR=1.95, p=.017), to visit the emergency department (OR=5.80, p<.001), and to have eight or more outpatient visits (OR=2.64, p<.001); and those whose prescriptions met both conditions were more likely to visit the emergency department (OR=3.50, p=.007).
To determine whether potentially inappropriate benzodiazepine prescriptions were related to higher health care costs for Quebec’s health care system, a GLM was used, and the results are presented in
Table 3. Controlling for individual and clinical factors, the multivariate analysis showed that participants whose prescriptions made them susceptible to benzodiazepine-related drug interactions had significantly higher total annual health care costs than those whose prescriptions were appropriate ($3,076 higher, p<.001). Those with prescriptions that met both conditions for inappropriateness also had higher annual costs than those with appropriate prescriptions ($3,102 higher, p<.001). For those with inappropriate prescriptions based on Beers criteria, no significantly higher costs were observed. The cost estimates for participants whose prescriptions entailed possible drug interactions did not differ from the estimates for those whose prescriptions met both conditions for inappropriateness, which suggests that there were no multiplicative effects between these two criteria. Additional factors associated with higher health care costs were older age (>75 years), male gender, and worse general medical health (three or more chronic general medical conditions and mediocre or bad self-perceived general medical health).
Discussion
This study aimed to assess the association between potentially inappropriate benzodiazepine prescriptions and health care service use and the associated excess health care costs. The study used data collected from the ESA survey, a representative sample of community-dwelling French-speaking elderly persons in the province of Quebec.
Studying the inappropriate use of benzodiazepines and drug interactions related to benzodiazepines is of a great relevance because in Canada these drugs are covered by public prescription insurance plans for nearly the entire elderly population and are consequently widely used, whereas in the United States benzodiazepines have been excluded from the Medicare Part D prescription drug plan. (It should be noted that benzodiazepines will be covered in Medicare Part D in the United States beginning in 2013 as a result of the Affordable Care Act.) In the ESA survey, the prevalence of benzodiazepine use was 32%, which is within the range of 9%–41% reported in other studies (
1,
7–
9,
41). Among these users, the prevalence of potentially inappropriate prescriptions according to Beers criteria reached 29%. When possible benzodiazepine-related drug interactions were considered, the prevalence of potentially inappropriate prescriptions was 44%. These results are consistent with those of other studies that used similar study designs and screening tools for inappropriate prescribing (
19,
21,
26,
42).
Potentially inappropriate benzodiazepine prescriptions were associated with increased health service use and with higher health care costs, mostly due to the potential for benzodiazepine-related drug interactions. When the criteria developed by Ben Amar (
15) were used, patients whose benzodiazepine prescriptions made them susceptible to moderate to severe drug interactions were more likely than patients whose prescriptions were deemed appropriate to be hospitalized, to visit the emergency department, and to have a large number of outpatient visits during the year following their first benzodiazepine prescription. Moreover, from the health system perspective, they generated higher annual health care costs. These results are consistent with those of other studies reporting that the concomitant use of benzodiazepines and other drugs deemed to present a risk of major drug interactions increased the odds of injuries and hip fractures, which may translate into emergency department visits and hospitalizations (
14,
43,
44).
In our study, prescriptions that were potentially inappropriate according to Beers criteria were associated only with an increased risk of emergency department visits and were not associated with higher health care costs. However, other studies using similar criteria have reported higher health care costs attributable to potentially inappropriate benzodiazepine prescriptions in the elderly population (
28,
29). Those findings may differ from our findings because the studies did not include high doses of short-acting benzodiazepines and included drugs other than benzodiazepines; in addition, those analyses did not adjust for potential benzodiazepine drug interactions. The relatively small sample size in our study may also account for the absence of an association with higher health care costs. Nevertheless, our results underline the importance of considering drug interactions when assessing inappropriate prescribing.
Beers criteria have been subject to several criticisms, and their clinical relevance is unproven (
45). Recent studies have suggested that dosage, duration of use, and the clinical situation are more important criteria for determining appropriate benzodiazepine use in clinical practices (
31,
46,
47). However, after adjusting for a number of factors, we did not find any significant association between the duration of benzodiazepine use and health service use and related costs.
Although the design of this study did not allow direct evaluation of the incremental cost associated with the management of adverse drug reactions resulting from an inappropriate prescription, the results show a significant association between health care costs and the prescription of benzodiazepines that carry a risk of drug interactions. Seven percent of the community-dwelling elderly population in Quebec were affected by this issue and, on average, the cost of their health care was $3,076 higher per year than elderly persons whose benzodiazepine prescriptions were appropriate. This may represent an economic burden as high as $216 million (in 2009 Canadian dollars) for every million community-dwelling elderly persons in Quebec: per capita incremental costs of $3,076 multiplied by the prevalence rate of 7%, further multiplied by 1,000,000. A previous study showed that the number of benzodiazepine prescribers and the number of pharmacies consulted annually were associated with an increased number of benzodiazepine prescriptions and the presence of potentially inappropriate benzodiazepine prescriptions, which may contribute to an increased risk of drug interactions (
48). These results suggest that a better monitoring system, such as sharing electronic medical and pharmaceutical databases, would be an approach to decreasing benzodiazepine-related drug interactions and their clinical and economic consequences. Another key step would be to increase efforts to educate health care practitioners about drug-drug interactions and the available tools to assess them (such as Micromedex and cytochrome P450 drug-drug interaction tables).
The results of this study should be viewed in light of certain limitations. First, we used data from administrative databases to measure benzodiazepine use, which may not reflect actual consumption. In the same vein, the design of the ESA survey did not allow us to examine use of natural products, over-the-counter drugs, or alcohol consumption in the assessment of benzodiazepine-related drug interactions. However, a study by Tamblyn and colleagues (
49) found that prescription claims databases in Quebec represent one of the most accurate means of determining drugs dispensed to individuals. In addition, the study design did not allow us to directly attribute a hospitalization or an emergency department visit to a potentially inappropriate prescription. Finally, self-reported survey data, which were used mainly as adjusting factors, are susceptible to social desirability and memory bias. However, use of administrative and survey data in the ESA survey permitted control of a number of individual and clinical factors, and the survey has provided new data on the impact of inappropriate benzodiazepine prescribing in a large representative sample of community-dwelling older adults.
Conclusions
The findings highlight the important economic burden associated with benzodiazepine-related drug interactions, which were found to affect 7% of community-dwelling elderly persons in Quebec. Such drug interactions may be preventable with better monitoring of older adults’ drug use, and future studies should therefore focus on evaluating the potential of shared medical and pharmacy electronic databases to decrease life-threatening drug interactions.
Acknowledgments and disclosures
This study was supported by operating grant 200683MOP from the Canadian Institutes of Health Research (CIHR). The ESA survey was supported by CIHR operating grant 200403MOP and grant 9854 from the Quebec Health Research Fund. The authors thank members of the Scientific Committee of the ESA survey for their scientific advice and Cassandra Wong for helpful comments.
Mr. Dionne has been employed as an intern by Pfizer Canada. The other authors report no competing interests.