- Email: [email protected]
Benzodiazepine Use in Older Adults Enrolled in a Health Maintenance Organization
Benzodiazepine Use in Older Adults Enrolled in a Health Maintenance Organization Shelly L. Gray, Pharm.D., M.S., Anne Elise Eggen, Ph.D. David Blough, Ph.D., Dave Buchner, M.D. Andrea Z. LaCroix, Ph.D.
Objectives: The authors examined patterns of benzodiazepine use in older adults. Specifically, they describe prevalence and incidence of benzodiazepine use during the index year, describe persistence and intensity of benzodiazepine use over a 4-year period; and examine factors associated with benzodiazepine use in the upcoming year. Methods: Authors performed a secondary analysis of data collected as part of a health promotion intervention trial conducted from 1986 to 1992 in older health maintenance organization enrollees (N⳱1,505). Benzodiazepine use was ascertained from computerized pharmacy records. Demographic characteristics, health status, and health behaviors were ascertained from mailed questionnaires. Results: During the index year, the prevalence and incidence of benzodiazepine use was 12.3% and 6.6%, respectively. Of those using during the index year, 16% of new users and 63% of previous users continued to use for the following 3 years. The factors significantly associated with benzodiazepine use in the following year were female gender, high school education, higher chronic disease score, higher levels of self-reported pain and stress, low-to-normal body mass index (BMI), and self-reported nervous disorder. Conclusions: New users had low intensity of use and a low probability of continuing use over the following 3 years. A very small percentage of this sample had evidence of daily use for 4 years. Of concern, benzodiazepines were used by the segment of the sample that were at greatest risk for hip fractures (women with normal/low BMI). Clinicians should assess the need for continued benzodiazepine use at regular intervals. (Am J Geriatr Psychiatry 2003; 11:568–576)
T
he use of benzodiazepines by older adults is common. A major concern with use of benzodiazepines is the potential for long-term use, overuse, and
dependence. Although these agents are effective for short-term management of insomnia and anxiety, long-term use is generally not recommended because
Received October 24, 2002; revised January 29, 2003; accepted February 6, 2003. From the University of Washington, School of Pharmacy, Seattle, WA (SLG,DB), Department of Pharmacoepidemiology and Pharmacy Practice, University of Tromsø, Norway (AEE), Department of Health Services, School of Public Health and Community Medicine, University of Washington, Seattle, WA (D.Buchner), and the Department of Epidemiology, University of Washington, Seattle, and Center for Health Studies, Group Health Cooperative of Puget Sound (AZL). Send correspondence to Dr. Shelly L. Gray, University of Washington, School of Pharmacy, Box 357630, Seattle, WA 98195. e-mail: [email protected] Copyright 䉷 2003 American Association for Geriatric Psychiatry
568
Am J Geriatr Psychiatry 11:5, September-October 2003
Gray et al. of risk of adverse events in elderly patients and limited efficacy for insomnia.1 The negative health effects of benzodiazepine use have been well described; some of these include increased risk for falls, hip fractures, depressive symptoms, and impaired cognition. Once starting on a benzodiazepine, older patients are more likely to use benzodiazepines chronically.2–4 Several studies have examined correlates of benzodiazepine use in population-based samples of older persons in the United States5,6 and Europe.7 Many studies have been cross-sectional and focused on correlates of prevalent use of benzodiazepines5,6 or had limited availability of patient-specific information.7 Use of benzodiazepines in older adults has been associated with several demographic factors, including white race,5,6,8 female gender,5,6 geographic location (Forsyth County, NC, or Washington County, MD),6 and higher level of education.8 Users of benzodiazepines tend to have poorer health than nonusers. For example, users of benzodiazepine were more likely to have coronary artery disease,5,6 have fair or poor self-reported health,5,6 report a nervous or emotional disorder,5,6 have depressive symptoms,5,8 have impaired functioning,5 report sleep problems,8 use over-the-counter sleep aids,6 and use more health services.5 Several investigators in Europe,2,7,9 Australia,10 and Canada4 have examined long-term benzodiazepine use, but little information exists regarding the extent and intensity of long-term benzodiazepine use among elderly individuals in the United States. For example, what is the likelihood that an older person using a benzodiazepine continues to have evidence of use over several years? The objectives of this study were to 1) describe prevalence and incidence of benzodiazepine use during the index year; 2) describe persistence and intensity of benzodiazepine use over a 4-year period; and 3) examine factors associated with benzodiazepine use in the upcoming year.
METHODS Study Population This was a secondary analysis of data from a disability-and-falls prevention trial conducted between 1986 and 1992,11 targeting multiple risk factors among elderly persons residing in the community.
Am J Geriatr Psychiatry 11:5, September-October 2003
The study sample was recruited from a random sample of men and women age 65 and older enrolled in Group Health Cooperative of Puget Sound (GHC). GHC is a large-staff model health maintenance organization (HMO) providing medical care to more than 370,000 people, including more than 40,000 older adults. To summarize, mailed questionnaires (up to two mailings) were sent to participants age 65 years and older, drawn from a random sample of 5,240, who received care from three large Seattle GHC clinics (after exclusion of those deemed too ill to participate by their primary care physicians [8%]). Of the original random sample, 36% returned completed consent forms and questionnaires; 13% refused participation; 2% were ineligible because they were institutionalized, seriously ill, or out of the area; and 41% failed to respond. Of those returning completed questionnaires, an additional 400 individuals were excluded because they reported difficulty with ambulation or with one or more self-care activities of daily living (ADL), thus yielding a sample of 1,559 for the original trial. The details of the eligibility statistics have been previously described.11 Participants were more educated, more affluent, and less likely to smoke than nonrespondents, but both groups had similar chronic disease scores and health status.12 The final sample for this analysis consisted of 1,505 subjects who were alive at the first follow-up interview and were continuously enrolled for the 4-year followup period or until the time they were censored. Data Collection Data for this analysis were taken from automated pharmacy files and from mailed questionnaires obtained at baseline and in Follow-Up Years 1 and 2. Participants were asked to provide information regarding demographics, health status, and health behaviors. A telephone interview was conducted if a participant failed to return a questionnaire, occurring in fewer than 5% of participants. The pharmacy files were available for 1 year before baseline data and for 4 years after baseline. Benzodiazepine exposure. Exposure to benzodiazepines was determined from GHC automated pharmacy files. The pharmacy files included information regarding the therapeutic class, drug form, strength, date dispensed, and quantity dispensed for each prescription filled at GHC’s outpatient pharmacies since
569
Benzodiazepine Use in Elderly Persons March 1977. Previous research indicates that most medications prescribed by GHC physicians are filled in GHC pharmacies.13 A survey of GHC enrollees treated for various types of pain during 1989 through 1990 found that 93% of prescriptions for diazepam were filled at GHC pharmacies. Each participant was classified as using or not using a benzodiazepine the year after the baseline interview (Year 1, index year), and for each of 3 years after the index year (Year 2, Year 3, and Year 4). Previous benzodiazepine use was defined as filling a benzodiazepine prescription in the 6 months before the baseline interview. The intensity of benzodiazepine use was calculated for each year according to the following three steps: First, the total number of milligrams for each benzodiazepine fill was calculated by multiplying the number of pills dispensed by the tablet strength. Second, this sum was standardized by dividing by the minimum effective dose per day according to standardized references:14,15 diazepam (4 mg), alprazolam (0.75 mg), chlorazepate (15 mg), lorazepam (2 mg), flurazepam (15 mg), oxazepam (30 mg), temazepam (15 mg), triazolam (0.125 mg), clonazepam (0.5 mg), and chlordiazepoxide (15 mg). Third, all standardized doses within the 1-year period were summed to create the total standardized dose (TSD). The TSD represents the number of days of use of the benzodiazepine over 1 year if used at the minimum effective dose. The TSD was categorized as: infrequent use (up to and including 20 TSD); occasional (20.1 through 150 TSD); frequent (150.1 through 300 TSD); and daily (⬎300 TSD). Explanatory variables. Several demographic, health status, and health behavior factors were examined that may be associated with benzodiazepine use. Demographic variables included age, sex, education (completed high school; did not complete high school), income (less than $15,000 annual income; at or above $15,000) and living status (alone; with others). Health behavior variables included smoking status at baseline (current, past, never) and alcohol use (non-drinker, up to 1 drink/day, ⬎1 drink/day). Body mass index ([BMI] weight in kilograms divided by height in meters squared) was calculated from self-reported weight and height; BMI was categorized as ⬍25 (low–normal weight) or ⱖ25 (overweight).16 Very few individuals (N⳱17) were considered underweight according to the NIH guidelines (BMI
570
⬍18.5); therefore, these individuals were not examined separately. Presence of chronic conditions (arthritis, cancer, diabetes, stroke, lung disease, heart condition, nervous/emotional problem) was determined by self-report. A measure of psychological well-being, was obtained from a 10- item positive affect scale with items adapted from Veit and Ware.17 This scale correlates moderately with depressive symptoms assessed by the Geriatric Depression Scale (r ⳱ –0.59).18 Subjects were asked to indicate, by means of a Likert scale, about degree of pain they had experienced on a daily basis (0 ⳱ no pain, to 10 ⳱ very severe pain). Subjects were asked about how much stress they had experienced for the past year (0⳱none; 10⳱a great deal). The chronic disease score (CDS) is a measure derived from computerized pharmacy data reflecting the number and severity of several chronic conditions (excluding symptomatic conditions, e.g., insomnia, depression). The CDS has been shown to be highly predictive of death, hospitalization, and ambulatory utilization in a variety of GHC study populations.19 Statistical Analysis Missing data. Multiple imputation was performed to accommodate missing values for 10 covariates by use of Statistical Solutions software.20,21 Table 1 identifies the variables that were imputed and the sample size for each variable. The proportion of data that was imputed for an individual variable did not exceed 16%. A limitation with including only individuals with complete data (complete case analysis) in the analyses is that the remaining sample may not be representative of the target population, and the estimates may be biased. Using the method of propensity scores and Bayesian predictive distributions, missing data were randomly imputed, and five complete datasets were obtained. Each model was estimated with each of the five datasets, and the results were combined. Multiple imputation takes into account the uncertainty in imputing missing values, resulting in conservative estimates of standard errors. Statistical analysis. SAS and STATA statistical software were used to perform all statistical analyses. The analysis proceeded in three phases. In the first phase, descriptive statistics were used to summarize the intensity of benzodiazepine use during the index year (year after baseline interview) according to new or
Am J Geriatr Psychiatry 11:5, September-October 2003
Gray et al. previous use. In the second phase, of those using in the index year, “survival” plots were used to examine continued use over the following 3 years. Participants who died were censored after the last interview. The actuarial method was used to obtain the plots, since the data are naturally grouped by interview year. In the third phase, univariate analyses were performed to examine the association of each variable with benzodiazepine use at baseline. In multivariate models, benzodiazepine use over 3 years was modeled as a binary outcome (use/no use), and correlaTABLE 1.
tions among outcomes measured on the same subject were accounted for by the use of generalized estimating equations (GEE). GEE is similar to a logisticregression analysis; however it allows for repeated measures on each subject, thereby greatly increasing the effective sample size. The following variables were included as time-varying covariates: stress, pain, positive affect score, living status, alcohol use, and previous benzodiazepine use. The multivariate models were constructed in two ways: Model 1 included demographic, health status, and health be-
Description of sample at baseline and univariate associations with benzodiazepine use Univariate Associationsa
Characteristics Demographics Age-group, years 65–69 70–74 75 and older Female Graduated from high schoolc (N⳱1,470) Income ⬍15,000c (N⳱1,262) Live alonec (N⳱1,453) Health status Cancer Diabetes Stroke Emphysema Heart disease Self-reported nervous disorderc (N⳱1,398) Positive affect,c,d mean (SD) (N⳱1,437) Stress,c,f mean (SD) (N⳱1,441) Pain,c,f mean (SD) (N⳱1,339) Fair/poor self-rated healthc (N⳱1,488) Chronic disease score 0 1–3 ⱖ4 Health behaviors Smoking statusd (N⳱1,473) Current Former Never Alcohol used (N⳱1,436) Non-drinker 1 drink/day ⱖ1 drink/day Overweight (BMI ⱖ25) (N⳱1,393)
b
N
%
Odds Ratio
95% Confidence Interval
p
530 493 482 889 1,238 432 507
35.2 32.8 32.0 59.0 84.2 34.0 34.9
1.00 1.06 1.18 1.78 1.45 1.43 1.39
Reference 0.76–1.47 0.85–1.63 1.35–2.35 0.96–2.20 1.06–1.93 1.07–1.80
0.727 0.331 ⬍0.0001 0.077 0.018 0.013
62 98 14 108 174 88 45.1 (8.2) 36.0 (26.5) 20.9 (20.1) 158
4.1 6.5 0.9 7.2 11.6 6.3
10.6
1.48 0.69 0.46 1.62 1.58 3.20 0.93e 1.12e 1.13e 1.86
0.84–2.60 0.40–1.18 0.12–1.83 1.01–2.60 1.14–2.20 2.08–4.93 0.88–0.98 1.08–1.17 1.07–1.18 1.35–2.57
0.171 0.173 0.271 0.047 0.006 ⬍0.0001 0.010 ⬍0.0001 ⬍0.0001 0.0002
604 539 362
40.1 35.8 24.1
1.00 1.82 3.02
Reference 1.30–2.56 2.14–4.27
0.0005 ⬍0.0001
104 546 823
7.1 37.1 55.9
1.00 1.09 0.66
Reference 0.82–1.46 0.35–1.24
0.555 0.195
438 620 378 700
30.5 43.2 26.3 50.3
1.00 0.90 0.97 0.58
Reference 0.64–1.27 0.71–1.31 0.45–0.76
0.563 0.823 ⬍0.0001
Note: SD: standard deviation; BMI: body mass index. a Logistic regression was used for univariate associations for each variable and benzodiazepine use. Imputed data were used for those variables with missing values. The use of multiple imputation methodology does not allow the calculation of an overall measure of fit of the model. b This column provides the distribution of each variable for observed data (not imputed). Total number of observations were specified for variables with missing values in the first column. c Variable was imputed. d Range: (0–60); higher score is better. e Per 10-unit increase. f Range: (0–100).
Am J Geriatr Psychiatry 11:5, September-October 2003
571
Benzodiazepine Use in Elderly Persons havior characteristics. A backward-selection method was used to determine which subset of predictors was significantly related to benzodiazepine use. Model 2 included the covariates from Model 1, but also included a variable indicating previous benzodiazepine use. Two-way interactions with sex were examined for both models.
RESULTS Benzodiazepine Use During the Index Year The distribution of the demographic, health status, and health behavior variables at baseline are presented in Table 1. The average age was 72.5 years (standard deviation [SD]: 5.5), 59% were women, and 10.6% rated their health as fair or poor. During the index year, the prevalence and incidence of benzodiazepine use was 12.3% (N⳱185) and 6.6% (N⳱94), respectively. One hundred and six individuals (7.0%) were using benzodiazepines classified as sedatives (temazepam, triazolam, flurazepam); 69 (4.6%) were using agents classified as anxiolytics (diazepam, alprazolam, lorazepam, chlordiazepoxide, clorazepate). and 10 (0.7%) used both. Benzodiazepines with a long elimination half-life (predominately diazepam and flurazepam) were used by 40.6% of users, but this pattern depended on whether an individual was a new (27.7%) or previous user (51.0%). The intensity of use varied according to whether an individual was a new or previous user during the index year (Figure 1), with infrequent use occurring in 50.0% and 8.8%, respectively. The median (range) TSD for new and previous users was 59.4 (2.5–666.6) and 100.0 (6.7– 900), respectively (Wilcoxon rank sum; z ⳱ –7.076; p⬍0.0001). Persistence of benzodiazepine use among users during the index year. Of those using during the index year, we examined the persistence of use for the next 3 years. Overall, 61.6%, 47.0%, and 38.9% of participants, respectively, used for 1, 2, and 3 consecutive years after the index year. Figure 1 shows the persistence of use according to new or previous use in the index year. Of those using during the index year, 16% of new users and 63% of previous users continued to use for the following 3 years. The probability of continuing use for the 3 years also varied according to the intensity of use during the index year. For new
572
users, the percent continuing use for 3 years after the index year was 6.3% for infrequent users, 18.4% for occasional users, 40.0% for frequent users, and 75.0% for daily users. For previous users, the percent continuing use for 3 years after the index year was 12.5%, 61.0%, 72.7%, and 92.3%, for infrequent, occasional, frequent, and daily users, respectively. Table 2 describes the intensity of use for the 72 individuals who filled at least one prescription for 4 consecutive years. The distribution of intensity of use appears to be stable over the 4 years. The majority of participants (50%–63%) had occasional benzodiazepine use (20–150 TSD) during each of the 4 years. Approximately 17%–21% of users or 0.8%–1.0% of the sample had evidence of daily use (⬎300 TSD). Use of benzodiazepines with a long elimination half-life were used by 50%–53% during each of the 4 years. Factors Associated With Use Over the Following Year Univariate analysis. The unadjusted odds ratios for use of benzodiazepines according to demographic, health status, and health behavior variables are presented in Table 1. Users were more likely to be female, have a high school education, live alone, and have an income of less than $15,000. Furthermore, users of benzodiazepines differed from non-users on FIGURE 1.
Distribution of benzodiazepine use for participants during the index year (Nⴔ185), according to frequency of use
Past
Daily
New
Frequent Occasional Infrequent
0
20
40
60
80
Percent of Users Note: Infrequent: 1–20 TSD; occasional: 20.1–150 TSD; frequent: 150.1–300 TSD; daily: ⬎300 TSD. TSD: total standardized dose.
Am J Geriatr Psychiatry 11:5, September-October 2003
Gray et al. many health-related variables: they were more likely to report fair or poor health, a nervous disorder, emphysema, and heart disease. Users showed higher average levels of pain and stress, lower positive affect scores, and higher chronic disease scores. Multivariate analysis. The factors that remained significantly associated with benzodiazpine use in the following year in the multivariate model included being female, having a high school education, higher chronic disease score, higher levels of self-reported pain and stress, and self-reported nervous disorder at baseline (Table 3, Model 1). Overweight participants were 28% less likely to use a benzodiazepine than low- or normal-weight participants (Table 3, Model 1). When including past use, only chronic disease score, BMI, and self-reported nervous disorder remained significantly associated with use (Table 3, Model 2). Previous benzodiazepine use was the strongest predictor for subsequent benzodiazepine use; those with evidence of benzodiazepine use in the 6 months prior to the interview were 36 times more likely to use in the following year (95% confidence interval [CI]: 25.6–49.5). The interactions between all TABLE 2.
variables and sex were examined for both models and were not significant. The most parsimonious model was derived from Model 2 and included the following variables: BMI, chronic disease score, self-reported nervous disorder, and past benzodiazepine use. We estimated the probability of using a benzodiazepine in the following year, based on the presence of all risk factors identified from this multivariate model (low BMI, chronic disease score [CDS] ⱖ4, self-reported nervous disorder), with and without previous benzodiazepine use. The probability of benzodiazepine use in the following year for an individual with previous use, with all risk factors, was 0.90 (95% CI: 0.84–0.93). However, the probability of use in the following year was only 0.18 (95% CI: 0.12–0.26) for an individual with all risk factors and no previous use.
DISCUSSION The prevalence and incidence of benzodiazepine use was 12.3% and 6.6%, respectively. We found a lower
Distribution (percentages) of benzodiazepine use for participants for 4 consecutive years (Nⴔ72) according to intensity of use
Infrequent Occasional Frequent Daily Total
Index(Nⴔ72)
Year 1(Nⴔ72)
Year 2(Nⴔ72)
Year 3(Nⴔ72)
5.6 59.7 13.9 20.8 100
5.6 50.0 25.0 19.4 100
4.2 62.5 16.7 16.7 100
8.3 62.5 11.1 18.1 100
Note: Infrequent (1–20 TSD); occasional (20.1–150 TSD); frequent (150.1–300 TSD); daily (⬎300 TSD); TSD: total standardized dose.
TABLE 3.
Multivariate model of characteristics associated with benzodiazepine use in Year 1 after the index year, using general estimating equations (GEE) Model 1
Model 2
Characteristics
OR
95% CI
p
OR
95% CI
p
Female Graduated from high school Self-reported nervous disorder (10-unit increase) Stress (10-unit increase) Pain (10-unit increase) Chronic disease score (CDS) 0 1–3 ⱖ4 Overweight (BMI ⱖ25) Previous benzodiazepine use
1.52 1.60 2.83 1.06 1.06
1.14–2.04 1.04–2.45 1.79–4.48 1.02–1.11 1.01–1.12
0.007 0.041 ⬍0.0001 0.005 0.022
1.15 1.37 1.82 1.0037 1.0024
0.87–1.50 0.87–2.16 1.19–2.77 1.00–1.01 1.00–1.01
0.326 0.170 0.006 0.128 0.358
1.00 1.83 3.11 0.72
Reference 1.30–2.58 2.20–4.39 0.54–0.97
0.003 ⬍0.0001 ⬍0.0001
Reference 1.13–2.09 1.68–3.12 0.47–0.80 25.6–49.5
0.006 ⬍0.0001 0.0004 ⬍0.0001
1.00 1.54 2.29 0.61 35.6
Note: Imputed variables were used for analysis. The use of multiple imputation methodology does not allow the calculation of an overall measure of fit of the model. OR: odds ratio; CI: (95%) confidence interval; BMI: body mass index.
Am J Geriatr Psychiatry 11:5, September-October 2003
573
Benzodiazepine Use in Elderly Persons prevalence of benzodiazepine use than previous studies that have used a similar methodology to determine benzodiazepine use (automated pharmacy files). The 1-year prevalence was 23% in a Swedish population2 and 29% in a Canadian population.22 The point prevalence of benzodiazepine use at a single interview has been reported at 10%–13%.5–7 The lower prevalence found in this study is likely because individuals were required to be independent in activities of daily living to be eligible for participation in the original trial and, thus, were generally healthier than the overall population. A major concern surrounding use of benzodiazepines is the potential for long-term use, overuse, and dependence. An important attribute of this study is that we were able to describe the intensity of benzodiazepine use over a 4-year period. Several positive findings should be highlighted. First, the intensity of use was minimal for new users during the index year. Half of new users had infrequent use that corresponded to using a benzodiazepine for 1 to 20 days (a single prescription) over the year, if usual geriatric doses were consumed. Second, new users were unlikely to continue use long-term; only 16% of new users continued use over the next 3 years, compared with 63% of past users. Third, of long-term users (e.g., use over 4 years), most (56%–71%) were using for less than 150 days of any individual year, if usual geriatric doses were consumed. Fourth, the proportion of long-term users in each intensity category remained stable over the 4 years, suggesting that overall use was not escalating over time in this sample. As expected, those with high use during the index year (e.g., daily) had a high probability of continued use over the following 3 years (72% of new users; 90% of previous users). Furthermore, 16.7%–20.8% of long-term users, 0.8%–1.0% of the sample, had evidence of benzodiazepine use almost daily during the 4-year period. Thus, chronic daily use of benzodiazepines was evident for only a small proportion of this healthy adult sample. Three variables emerged as predictors of benzodiazepine use that have not been reported previously: chronic disease score, pain, and BMI. The chronic disease score reflects overall number and severity of chronic conditions and was a better predictor of benzodiazepine use than self-rated health in this study. Self-rated fair or poor health has been found by previous investigators to be associated with benzodiaz-
574
epine use.5,6 Two explanations may account for the association between pain and benzodiazepine use. First, benzodiazepines may be used to treat pain conditions, especially pain secondary to muscle spasms. Simon et al.3 reported that 38% of new benzodiazepine users had pain (e.g., back, neck, headache) listed in the chart as the indication for use. Elderly patients were less likely to have psychological symptoms (anxiety, depression, and situational stress) listed in the chart as the indication for benzodiazepine initiation. Second, patients with chronic pain frequently have comorbid psychiatric disturbances, most often depression.23 Although this association was significant even when controlling for self-reported emotional disorder, it is likely that we were not able to fully adjust for depressive disorder, which is consistently associated with benzodiazepine use.5,7,8 Individuals who were overweight (BMI ⱖ25)15 were 39% less likely to use a benzodiazepine, even when controlling for previous benzodiazepine use. The explanation for this finding is unclear. Obesity is a risk factor for sleep difficulties, especially obstructive sleep apnea,24 which is common in elderly patients.1 Benzodiazepines should be avoided in individuals with sleep apnea because the CNSdepressant properties may interfere with resumption of breathing. Although avoidance of benzodiazepine use in individuals with sleep apnea may account partially for the lower risk of benzodiazepine use in overweight individuals, this is unlikely to be a major explanatory factor, given the strength of the association. Contrary to our findings, a positive correlation between BMI and anxiolytic use has been reported in middle-aged women and men.25,26 These findings are of concern in light of the evidence that women with low-to-normal BMI are at increased risk for hip fractures.27,28 Benzodiazepines, also associated with hip fractures,29,30 were most likely to be used by the segment of the sample at greatest risk for hip fractures, on the basis of body size (women with low–normal BMI). As expected, previous use of benzodiazepines was a strong predictor of future use. Isacson2 reported that previous benzodiazepine use predicted longterm use of these agents. Sex, education, and selfreported pain and stress were no longer associated with benzodiazepine use in the following year, once previous benzodiazepine use was accounted for. The intensity-of-exposure measure used in this
Am J Geriatr Psychiatry 11:5, September-October 2003
Gray et al. study, TSD, reflected both dose and duration. The TSD represents the number of days that a benzodiazepine was used over a 1-year period if used at the minimum effective dose for elderly patients. Although the TSD was useful for describing the general pattern of intensity of use, we recognize that actual daily usage may have been higher or lower, which would change the distribution among infrequent, occasional, frequent, and daily categories of use. The TSD should not be confused with the defined daily dose (DDD) methodology developed by the WHO Collaborating Center for Drug Statistics, which has been used in many European pharmacoepidemiology studies and is a technical unit reflecting the normal daily dose for an adult on the main indication for the drug.31 In general, the TSD conversion resulted in a higher number of days of use for a given period, when compared with DDD calculations, since the TSD conversions are smaller than the DDD conversions. As an example, 4 mg was used for calculating the TSD for diazepam, whereas 10 mg was used for DDD calculations. Our results are unlikely to be generalizable to the older adult population at-large. This study was conducted among predominately white, relatively healthy, middle-class older adults who were members of an HMO and had volunteered to participate in a health promotion trial. Selection bias is a concern, given the 36% participation rate, however, respondents were similar to nonrespondents with respect to health status and the chronic disease score;12 the latter was an important predictor of future benzodiazepine use. The sample studied is likely to explain in part the lower prevalence of use in this sample, compared with other studies, and may also have an impact on intensity of use. Another limitation is that misclassi-
fication of benzodiazepine use was possible. Information about prescription drug use was obtained from a computerized pharmacy database, and therefore, it is not known if the benzodiazepine was actually consumed. Although most medications prescribed by GHC physicians are filled in GHC pharmacies, it is possible that prescriptions were filled at outside pharmacies that would not be captured.
SUMMARY The use of benzodiazepines by older adults has received considerable attention. The results of this study are mainly encouraging, for many reasons. First, new users had low intensity of use and had a low probability of continuing use over the following 3 years. Of those who did have persistent use, intensity of use did not escalate over the 4 years, and most had just occasional use or less. Of concern is that benzodiazepine were used by the segment of the sample that are at greatest risk for hip fractures (women with low–normal BMI). Approximately 1% of the sample had evidence of daily use for 4 consecutive years. Benzodiazepines are recommended for short-term use for most indications. All individuals using benzodiazepines should be assessed periodically for possible reduction and discontinuation of therapy to minimize serious adverse effects (falls, hip fractures, impaired cognition, and depressive symptoms). Dr. Buchner is currently at the Centers for Disease Control. This work was supported by Grant K08AG00808-01 from the National Institute on Aging.
References 1. Folks DG, Burke WJ: Psychotherapeutic agents in older adults: sedative-hypnotics and sleep. Clin Geriatr Med 1998; 14: 67–86 2. Isacson D: Long-term benzodiazepine use: factors of importance and the development of individual use patterns over time: a 13year follow-up in a Swedish community. Soc Sci Med 1997; 44:1871–1880 3. Simon GE, VonKorff M, Barlow W, et al: Predictors of chronic benzodiazepine use in a health maintenance organization sample. J Clin Epidemiol 1996; 49:1067–1073 4. Egan M, Moride Y, Wolfson C, et al: Long-term continuous use of benzodiazepines by older adults in Quebec: prevalence, incidence, and risk factors. J Am Geriatr Soc 2000; 48:811–816 5. Blazer D, Hybels C, Simonsick E, et al: Sedative, hypnotic, and
Am J Geriatr Psychiatry 11:5, September-October 2003
antianxiety medication use in an aging cohort over ten years: a racial comparison. J Am Geriatr Soc 2000; 48:1073–1079 6. Gleason PP, Schulz R, Smith NL, et al: Correlates and prevalence of benzodiazepine use in community-dwelling elderly. J Gen Intern Med 1998; 13:243–250 7. Taylor S, McCracken CF, Wilson KC, et al: Extent and appropriateness of benzodiazepine use. results from an elderly urban community. Br J Psychiatry 1998; 173:433–438 8. Mayer-Oakes SA, Kelman G, Beers MH, et al: Benzodiazepine use in older, community-dwelling southern Californians: prevalence and clinical correlates. Ann Pharmacother 1993; 27:416–421 9. van Hulten R, Leufkens HG, Bakker A: Usage patterns of benzodiazepines in a Dutch community: a 10-year follow-up. Pharm World Sci 1998; 20:78–82
575
Benzodiazepine Use in Elderly Persons 10. Jorm AF, Grayson D, Creasey H, et al: Long-term benzodiazepine use by elderly people living in the community. Aust N Z J Public Health 2000; 24:7–10 11. Wagner EH, LaCroix AZ, Grothaus L, et al: Preventing disability and falls in older adults: a population-based randomized trial. Am J Public Health 1994; 84:1800–1806 12. Wagner EH, Grothaus L, Hecht JA, et al: Factors associated with participation in a senior health-promotion program. Gerontologist 1991; 31:598–602 13. Saunders KW, Stergachis A, VonKorff M: Group Health Cooperative of Puget Sound, in Pharmacoepidemiology. Edited by Strom BL. London, UK, Wiley, 1994 14. Kastrup EK (ed): Facts and comparisons, 2000, in Facts and Comparisons; 2000. Edited by Kastrup EK. St. Louis, MO, 2000 15. McEvoy GK (ed): AHFS Drug Information, 2000. Edited by McEvoy GK. Bethesda, MD, American Society for Healthcare Pharmacists; 2000 16. Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults: The Evidence Report: National Institutes of Health. Obes Res 1998; 6(suppl2):51S–209S 17. Veit CT, Ware JE: The structure of psychological distress and wellbeing in general populations. J Consult Clin Psychol 1985; 51:730–742 18. Scholes D, LaCroix AZ, Wagner EH, et al: Tracking progress toward national health objectives in the elderly: what do restricted activity days signify? Am J Public Health 1991; 81:485–488 19. Von Korff M, Wagner EH, Saunders K: A chronic disease score from automated pharmacy data. J Clin Epidemiol 1992; 45:197– 203 20. Little RJA, Rubin DB: Statistical Analysis with Missing Data. New York, Wiley; 1987 21. Statistical Solutions System for Data Analysis, v. 2.0. Saugus, MA, Statistical Solutions; 1999
576
22. Rojas-Fernandez CH, Carver D, Tonks R: Population trends in the prevalence of benzodiazepine use in the older population of Nova Scotia: a cause for concern? Can J Clin Pharmacol 1999; 6:149–156 23. Von Korff M, Simon G: The relationship between pain and depression. Br J Psychiatry 1996; (168suppl):101–108 24. National Task Force on the Prevention and Treatment of Obesity: Overweight, obesity, and health risk. Arch Intern Med 2000; 160:898–904 25. Rosmond R, Bjorntorp P: Psychiatric ill-health of women and its relationship to obesity and body fat distribution. Obes Res 1998; 6:338–345 26. Rosmond R, Lapidus L, Marin P, et al: Mental distress, obesity, and body fat distribution in middle-aged men. Obes Res 1996; 4:245–252 27. Ensrud KE, Lipschutz RC, Cauley JA, et al: Body size and hip fracture risk in older women: a prospective study. Study of Osteoporotic Fractures Research Group. Am J Med 1997; 103:274– 280 28. Farmer ME, Harris T, Madans JH, et al: Anthropometric indicators and hip fracture: The NHANES I Epidemiologic Follow-Up Study. J Am Geriatr Soc 1989; 37:9–16 29. Ray WA, Griffin MR, Downey W: Benzodiazepines of long and short elimination half-life and the risk of hip fracture. JAMA 1989; 262:3303–3307 30. Herings RM, Stricker BH, de Boer A, et al: Benzodiazepines and the risk of falling leading to femur fractures: dosage more important than elimination half-life. Arch Intern Med 1995; 155:1801–1807 31. Ronning M, Blix HS, Harbo BT, et al: Different versions of the anatomical therapeutic chemical classification system and the defined daily dose: are drug utilisation data comparable? Eur J Clin Pharmacol 2000; 56:723–727
Am J Geriatr Psychiatry 11:5, September-October 2003
Objectives: The authors examined patterns of benzodiazepine use in older adults. Specifically, they describe prevalence and incidence of benzodiazepine use during the index year, describe persistence and intensity of benzodiazepine use over a 4-year period; and examine factors associated with benzodiazepine use in the upcoming year. Methods: Authors performed a secondary analysis of data collected as part of a health promotion intervention trial conducted from 1986 to 1992 in older health maintenance organization enrollees (N⳱1,505). Benzodiazepine use was ascertained from computerized pharmacy records. Demographic characteristics, health status, and health behaviors were ascertained from mailed questionnaires. Results: During the index year, the prevalence and incidence of benzodiazepine use was 12.3% and 6.6%, respectively. Of those using during the index year, 16% of new users and 63% of previous users continued to use for the following 3 years. The factors significantly associated with benzodiazepine use in the following year were female gender, high school education, higher chronic disease score, higher levels of self-reported pain and stress, low-to-normal body mass index (BMI), and self-reported nervous disorder. Conclusions: New users had low intensity of use and a low probability of continuing use over the following 3 years. A very small percentage of this sample had evidence of daily use for 4 years. Of concern, benzodiazepines were used by the segment of the sample that were at greatest risk for hip fractures (women with normal/low BMI). Clinicians should assess the need for continued benzodiazepine use at regular intervals. (Am J Geriatr Psychiatry 2003; 11:568–576)
T
he use of benzodiazepines by older adults is common. A major concern with use of benzodiazepines is the potential for long-term use, overuse, and
dependence. Although these agents are effective for short-term management of insomnia and anxiety, long-term use is generally not recommended because
Received October 24, 2002; revised January 29, 2003; accepted February 6, 2003. From the University of Washington, School of Pharmacy, Seattle, WA (SLG,DB), Department of Pharmacoepidemiology and Pharmacy Practice, University of Tromsø, Norway (AEE), Department of Health Services, School of Public Health and Community Medicine, University of Washington, Seattle, WA (D.Buchner), and the Department of Epidemiology, University of Washington, Seattle, and Center for Health Studies, Group Health Cooperative of Puget Sound (AZL). Send correspondence to Dr. Shelly L. Gray, University of Washington, School of Pharmacy, Box 357630, Seattle, WA 98195. e-mail: [email protected] Copyright 䉷 2003 American Association for Geriatric Psychiatry
568
Am J Geriatr Psychiatry 11:5, September-October 2003
Gray et al. of risk of adverse events in elderly patients and limited efficacy for insomnia.1 The negative health effects of benzodiazepine use have been well described; some of these include increased risk for falls, hip fractures, depressive symptoms, and impaired cognition. Once starting on a benzodiazepine, older patients are more likely to use benzodiazepines chronically.2–4 Several studies have examined correlates of benzodiazepine use in population-based samples of older persons in the United States5,6 and Europe.7 Many studies have been cross-sectional and focused on correlates of prevalent use of benzodiazepines5,6 or had limited availability of patient-specific information.7 Use of benzodiazepines in older adults has been associated with several demographic factors, including white race,5,6,8 female gender,5,6 geographic location (Forsyth County, NC, or Washington County, MD),6 and higher level of education.8 Users of benzodiazepines tend to have poorer health than nonusers. For example, users of benzodiazepine were more likely to have coronary artery disease,5,6 have fair or poor self-reported health,5,6 report a nervous or emotional disorder,5,6 have depressive symptoms,5,8 have impaired functioning,5 report sleep problems,8 use over-the-counter sleep aids,6 and use more health services.5 Several investigators in Europe,2,7,9 Australia,10 and Canada4 have examined long-term benzodiazepine use, but little information exists regarding the extent and intensity of long-term benzodiazepine use among elderly individuals in the United States. For example, what is the likelihood that an older person using a benzodiazepine continues to have evidence of use over several years? The objectives of this study were to 1) describe prevalence and incidence of benzodiazepine use during the index year; 2) describe persistence and intensity of benzodiazepine use over a 4-year period; and 3) examine factors associated with benzodiazepine use in the upcoming year.
METHODS Study Population This was a secondary analysis of data from a disability-and-falls prevention trial conducted between 1986 and 1992,11 targeting multiple risk factors among elderly persons residing in the community.
Am J Geriatr Psychiatry 11:5, September-October 2003
The study sample was recruited from a random sample of men and women age 65 and older enrolled in Group Health Cooperative of Puget Sound (GHC). GHC is a large-staff model health maintenance organization (HMO) providing medical care to more than 370,000 people, including more than 40,000 older adults. To summarize, mailed questionnaires (up to two mailings) were sent to participants age 65 years and older, drawn from a random sample of 5,240, who received care from three large Seattle GHC clinics (after exclusion of those deemed too ill to participate by their primary care physicians [8%]). Of the original random sample, 36% returned completed consent forms and questionnaires; 13% refused participation; 2% were ineligible because they were institutionalized, seriously ill, or out of the area; and 41% failed to respond. Of those returning completed questionnaires, an additional 400 individuals were excluded because they reported difficulty with ambulation or with one or more self-care activities of daily living (ADL), thus yielding a sample of 1,559 for the original trial. The details of the eligibility statistics have been previously described.11 Participants were more educated, more affluent, and less likely to smoke than nonrespondents, but both groups had similar chronic disease scores and health status.12 The final sample for this analysis consisted of 1,505 subjects who were alive at the first follow-up interview and were continuously enrolled for the 4-year followup period or until the time they were censored. Data Collection Data for this analysis were taken from automated pharmacy files and from mailed questionnaires obtained at baseline and in Follow-Up Years 1 and 2. Participants were asked to provide information regarding demographics, health status, and health behaviors. A telephone interview was conducted if a participant failed to return a questionnaire, occurring in fewer than 5% of participants. The pharmacy files were available for 1 year before baseline data and for 4 years after baseline. Benzodiazepine exposure. Exposure to benzodiazepines was determined from GHC automated pharmacy files. The pharmacy files included information regarding the therapeutic class, drug form, strength, date dispensed, and quantity dispensed for each prescription filled at GHC’s outpatient pharmacies since
569
Benzodiazepine Use in Elderly Persons March 1977. Previous research indicates that most medications prescribed by GHC physicians are filled in GHC pharmacies.13 A survey of GHC enrollees treated for various types of pain during 1989 through 1990 found that 93% of prescriptions for diazepam were filled at GHC pharmacies. Each participant was classified as using or not using a benzodiazepine the year after the baseline interview (Year 1, index year), and for each of 3 years after the index year (Year 2, Year 3, and Year 4). Previous benzodiazepine use was defined as filling a benzodiazepine prescription in the 6 months before the baseline interview. The intensity of benzodiazepine use was calculated for each year according to the following three steps: First, the total number of milligrams for each benzodiazepine fill was calculated by multiplying the number of pills dispensed by the tablet strength. Second, this sum was standardized by dividing by the minimum effective dose per day according to standardized references:14,15 diazepam (4 mg), alprazolam (0.75 mg), chlorazepate (15 mg), lorazepam (2 mg), flurazepam (15 mg), oxazepam (30 mg), temazepam (15 mg), triazolam (0.125 mg), clonazepam (0.5 mg), and chlordiazepoxide (15 mg). Third, all standardized doses within the 1-year period were summed to create the total standardized dose (TSD). The TSD represents the number of days of use of the benzodiazepine over 1 year if used at the minimum effective dose. The TSD was categorized as: infrequent use (up to and including 20 TSD); occasional (20.1 through 150 TSD); frequent (150.1 through 300 TSD); and daily (⬎300 TSD). Explanatory variables. Several demographic, health status, and health behavior factors were examined that may be associated with benzodiazepine use. Demographic variables included age, sex, education (completed high school; did not complete high school), income (less than $15,000 annual income; at or above $15,000) and living status (alone; with others). Health behavior variables included smoking status at baseline (current, past, never) and alcohol use (non-drinker, up to 1 drink/day, ⬎1 drink/day). Body mass index ([BMI] weight in kilograms divided by height in meters squared) was calculated from self-reported weight and height; BMI was categorized as ⬍25 (low–normal weight) or ⱖ25 (overweight).16 Very few individuals (N⳱17) were considered underweight according to the NIH guidelines (BMI
570
⬍18.5); therefore, these individuals were not examined separately. Presence of chronic conditions (arthritis, cancer, diabetes, stroke, lung disease, heart condition, nervous/emotional problem) was determined by self-report. A measure of psychological well-being, was obtained from a 10- item positive affect scale with items adapted from Veit and Ware.17 This scale correlates moderately with depressive symptoms assessed by the Geriatric Depression Scale (r ⳱ –0.59).18 Subjects were asked to indicate, by means of a Likert scale, about degree of pain they had experienced on a daily basis (0 ⳱ no pain, to 10 ⳱ very severe pain). Subjects were asked about how much stress they had experienced for the past year (0⳱none; 10⳱a great deal). The chronic disease score (CDS) is a measure derived from computerized pharmacy data reflecting the number and severity of several chronic conditions (excluding symptomatic conditions, e.g., insomnia, depression). The CDS has been shown to be highly predictive of death, hospitalization, and ambulatory utilization in a variety of GHC study populations.19 Statistical Analysis Missing data. Multiple imputation was performed to accommodate missing values for 10 covariates by use of Statistical Solutions software.20,21 Table 1 identifies the variables that were imputed and the sample size for each variable. The proportion of data that was imputed for an individual variable did not exceed 16%. A limitation with including only individuals with complete data (complete case analysis) in the analyses is that the remaining sample may not be representative of the target population, and the estimates may be biased. Using the method of propensity scores and Bayesian predictive distributions, missing data were randomly imputed, and five complete datasets were obtained. Each model was estimated with each of the five datasets, and the results were combined. Multiple imputation takes into account the uncertainty in imputing missing values, resulting in conservative estimates of standard errors. Statistical analysis. SAS and STATA statistical software were used to perform all statistical analyses. The analysis proceeded in three phases. In the first phase, descriptive statistics were used to summarize the intensity of benzodiazepine use during the index year (year after baseline interview) according to new or
Am J Geriatr Psychiatry 11:5, September-October 2003
Gray et al. previous use. In the second phase, of those using in the index year, “survival” plots were used to examine continued use over the following 3 years. Participants who died were censored after the last interview. The actuarial method was used to obtain the plots, since the data are naturally grouped by interview year. In the third phase, univariate analyses were performed to examine the association of each variable with benzodiazepine use at baseline. In multivariate models, benzodiazepine use over 3 years was modeled as a binary outcome (use/no use), and correlaTABLE 1.
tions among outcomes measured on the same subject were accounted for by the use of generalized estimating equations (GEE). GEE is similar to a logisticregression analysis; however it allows for repeated measures on each subject, thereby greatly increasing the effective sample size. The following variables were included as time-varying covariates: stress, pain, positive affect score, living status, alcohol use, and previous benzodiazepine use. The multivariate models were constructed in two ways: Model 1 included demographic, health status, and health be-
Description of sample at baseline and univariate associations with benzodiazepine use Univariate Associationsa
Characteristics Demographics Age-group, years 65–69 70–74 75 and older Female Graduated from high schoolc (N⳱1,470) Income ⬍15,000c (N⳱1,262) Live alonec (N⳱1,453) Health status Cancer Diabetes Stroke Emphysema Heart disease Self-reported nervous disorderc (N⳱1,398) Positive affect,c,d mean (SD) (N⳱1,437) Stress,c,f mean (SD) (N⳱1,441) Pain,c,f mean (SD) (N⳱1,339) Fair/poor self-rated healthc (N⳱1,488) Chronic disease score 0 1–3 ⱖ4 Health behaviors Smoking statusd (N⳱1,473) Current Former Never Alcohol used (N⳱1,436) Non-drinker 1 drink/day ⱖ1 drink/day Overweight (BMI ⱖ25) (N⳱1,393)
b
N
%
Odds Ratio
95% Confidence Interval
p
530 493 482 889 1,238 432 507
35.2 32.8 32.0 59.0 84.2 34.0 34.9
1.00 1.06 1.18 1.78 1.45 1.43 1.39
Reference 0.76–1.47 0.85–1.63 1.35–2.35 0.96–2.20 1.06–1.93 1.07–1.80
0.727 0.331 ⬍0.0001 0.077 0.018 0.013
62 98 14 108 174 88 45.1 (8.2) 36.0 (26.5) 20.9 (20.1) 158
4.1 6.5 0.9 7.2 11.6 6.3
10.6
1.48 0.69 0.46 1.62 1.58 3.20 0.93e 1.12e 1.13e 1.86
0.84–2.60 0.40–1.18 0.12–1.83 1.01–2.60 1.14–2.20 2.08–4.93 0.88–0.98 1.08–1.17 1.07–1.18 1.35–2.57
0.171 0.173 0.271 0.047 0.006 ⬍0.0001 0.010 ⬍0.0001 ⬍0.0001 0.0002
604 539 362
40.1 35.8 24.1
1.00 1.82 3.02
Reference 1.30–2.56 2.14–4.27
0.0005 ⬍0.0001
104 546 823
7.1 37.1 55.9
1.00 1.09 0.66
Reference 0.82–1.46 0.35–1.24
0.555 0.195
438 620 378 700
30.5 43.2 26.3 50.3
1.00 0.90 0.97 0.58
Reference 0.64–1.27 0.71–1.31 0.45–0.76
0.563 0.823 ⬍0.0001
Note: SD: standard deviation; BMI: body mass index. a Logistic regression was used for univariate associations for each variable and benzodiazepine use. Imputed data were used for those variables with missing values. The use of multiple imputation methodology does not allow the calculation of an overall measure of fit of the model. b This column provides the distribution of each variable for observed data (not imputed). Total number of observations were specified for variables with missing values in the first column. c Variable was imputed. d Range: (0–60); higher score is better. e Per 10-unit increase. f Range: (0–100).
Am J Geriatr Psychiatry 11:5, September-October 2003
571
Benzodiazepine Use in Elderly Persons havior characteristics. A backward-selection method was used to determine which subset of predictors was significantly related to benzodiazepine use. Model 2 included the covariates from Model 1, but also included a variable indicating previous benzodiazepine use. Two-way interactions with sex were examined for both models.
RESULTS Benzodiazepine Use During the Index Year The distribution of the demographic, health status, and health behavior variables at baseline are presented in Table 1. The average age was 72.5 years (standard deviation [SD]: 5.5), 59% were women, and 10.6% rated their health as fair or poor. During the index year, the prevalence and incidence of benzodiazepine use was 12.3% (N⳱185) and 6.6% (N⳱94), respectively. One hundred and six individuals (7.0%) were using benzodiazepines classified as sedatives (temazepam, triazolam, flurazepam); 69 (4.6%) were using agents classified as anxiolytics (diazepam, alprazolam, lorazepam, chlordiazepoxide, clorazepate). and 10 (0.7%) used both. Benzodiazepines with a long elimination half-life (predominately diazepam and flurazepam) were used by 40.6% of users, but this pattern depended on whether an individual was a new (27.7%) or previous user (51.0%). The intensity of use varied according to whether an individual was a new or previous user during the index year (Figure 1), with infrequent use occurring in 50.0% and 8.8%, respectively. The median (range) TSD for new and previous users was 59.4 (2.5–666.6) and 100.0 (6.7– 900), respectively (Wilcoxon rank sum; z ⳱ –7.076; p⬍0.0001). Persistence of benzodiazepine use among users during the index year. Of those using during the index year, we examined the persistence of use for the next 3 years. Overall, 61.6%, 47.0%, and 38.9% of participants, respectively, used for 1, 2, and 3 consecutive years after the index year. Figure 1 shows the persistence of use according to new or previous use in the index year. Of those using during the index year, 16% of new users and 63% of previous users continued to use for the following 3 years. The probability of continuing use for the 3 years also varied according to the intensity of use during the index year. For new
572
users, the percent continuing use for 3 years after the index year was 6.3% for infrequent users, 18.4% for occasional users, 40.0% for frequent users, and 75.0% for daily users. For previous users, the percent continuing use for 3 years after the index year was 12.5%, 61.0%, 72.7%, and 92.3%, for infrequent, occasional, frequent, and daily users, respectively. Table 2 describes the intensity of use for the 72 individuals who filled at least one prescription for 4 consecutive years. The distribution of intensity of use appears to be stable over the 4 years. The majority of participants (50%–63%) had occasional benzodiazepine use (20–150 TSD) during each of the 4 years. Approximately 17%–21% of users or 0.8%–1.0% of the sample had evidence of daily use (⬎300 TSD). Use of benzodiazepines with a long elimination half-life were used by 50%–53% during each of the 4 years. Factors Associated With Use Over the Following Year Univariate analysis. The unadjusted odds ratios for use of benzodiazepines according to demographic, health status, and health behavior variables are presented in Table 1. Users were more likely to be female, have a high school education, live alone, and have an income of less than $15,000. Furthermore, users of benzodiazepines differed from non-users on FIGURE 1.
Distribution of benzodiazepine use for participants during the index year (Nⴔ185), according to frequency of use
Past
Daily
New
Frequent Occasional Infrequent
0
20
40
60
80
Percent of Users Note: Infrequent: 1–20 TSD; occasional: 20.1–150 TSD; frequent: 150.1–300 TSD; daily: ⬎300 TSD. TSD: total standardized dose.
Am J Geriatr Psychiatry 11:5, September-October 2003
Gray et al. many health-related variables: they were more likely to report fair or poor health, a nervous disorder, emphysema, and heart disease. Users showed higher average levels of pain and stress, lower positive affect scores, and higher chronic disease scores. Multivariate analysis. The factors that remained significantly associated with benzodiazpine use in the following year in the multivariate model included being female, having a high school education, higher chronic disease score, higher levels of self-reported pain and stress, and self-reported nervous disorder at baseline (Table 3, Model 1). Overweight participants were 28% less likely to use a benzodiazepine than low- or normal-weight participants (Table 3, Model 1). When including past use, only chronic disease score, BMI, and self-reported nervous disorder remained significantly associated with use (Table 3, Model 2). Previous benzodiazepine use was the strongest predictor for subsequent benzodiazepine use; those with evidence of benzodiazepine use in the 6 months prior to the interview were 36 times more likely to use in the following year (95% confidence interval [CI]: 25.6–49.5). The interactions between all TABLE 2.
variables and sex were examined for both models and were not significant. The most parsimonious model was derived from Model 2 and included the following variables: BMI, chronic disease score, self-reported nervous disorder, and past benzodiazepine use. We estimated the probability of using a benzodiazepine in the following year, based on the presence of all risk factors identified from this multivariate model (low BMI, chronic disease score [CDS] ⱖ4, self-reported nervous disorder), with and without previous benzodiazepine use. The probability of benzodiazepine use in the following year for an individual with previous use, with all risk factors, was 0.90 (95% CI: 0.84–0.93). However, the probability of use in the following year was only 0.18 (95% CI: 0.12–0.26) for an individual with all risk factors and no previous use.
DISCUSSION The prevalence and incidence of benzodiazepine use was 12.3% and 6.6%, respectively. We found a lower
Distribution (percentages) of benzodiazepine use for participants for 4 consecutive years (Nⴔ72) according to intensity of use
Infrequent Occasional Frequent Daily Total
Index(Nⴔ72)
Year 1(Nⴔ72)
Year 2(Nⴔ72)
Year 3(Nⴔ72)
5.6 59.7 13.9 20.8 100
5.6 50.0 25.0 19.4 100
4.2 62.5 16.7 16.7 100
8.3 62.5 11.1 18.1 100
Note: Infrequent (1–20 TSD); occasional (20.1–150 TSD); frequent (150.1–300 TSD); daily (⬎300 TSD); TSD: total standardized dose.
TABLE 3.
Multivariate model of characteristics associated with benzodiazepine use in Year 1 after the index year, using general estimating equations (GEE) Model 1
Model 2
Characteristics
OR
95% CI
p
OR
95% CI
p
Female Graduated from high school Self-reported nervous disorder (10-unit increase) Stress (10-unit increase) Pain (10-unit increase) Chronic disease score (CDS) 0 1–3 ⱖ4 Overweight (BMI ⱖ25) Previous benzodiazepine use
1.52 1.60 2.83 1.06 1.06
1.14–2.04 1.04–2.45 1.79–4.48 1.02–1.11 1.01–1.12
0.007 0.041 ⬍0.0001 0.005 0.022
1.15 1.37 1.82 1.0037 1.0024
0.87–1.50 0.87–2.16 1.19–2.77 1.00–1.01 1.00–1.01
0.326 0.170 0.006 0.128 0.358
1.00 1.83 3.11 0.72
Reference 1.30–2.58 2.20–4.39 0.54–0.97
0.003 ⬍0.0001 ⬍0.0001
Reference 1.13–2.09 1.68–3.12 0.47–0.80 25.6–49.5
0.006 ⬍0.0001 0.0004 ⬍0.0001
1.00 1.54 2.29 0.61 35.6
Note: Imputed variables were used for analysis. The use of multiple imputation methodology does not allow the calculation of an overall measure of fit of the model. OR: odds ratio; CI: (95%) confidence interval; BMI: body mass index.
Am J Geriatr Psychiatry 11:5, September-October 2003
573
Benzodiazepine Use in Elderly Persons prevalence of benzodiazepine use than previous studies that have used a similar methodology to determine benzodiazepine use (automated pharmacy files). The 1-year prevalence was 23% in a Swedish population2 and 29% in a Canadian population.22 The point prevalence of benzodiazepine use at a single interview has been reported at 10%–13%.5–7 The lower prevalence found in this study is likely because individuals were required to be independent in activities of daily living to be eligible for participation in the original trial and, thus, were generally healthier than the overall population. A major concern surrounding use of benzodiazepines is the potential for long-term use, overuse, and dependence. An important attribute of this study is that we were able to describe the intensity of benzodiazepine use over a 4-year period. Several positive findings should be highlighted. First, the intensity of use was minimal for new users during the index year. Half of new users had infrequent use that corresponded to using a benzodiazepine for 1 to 20 days (a single prescription) over the year, if usual geriatric doses were consumed. Second, new users were unlikely to continue use long-term; only 16% of new users continued use over the next 3 years, compared with 63% of past users. Third, of long-term users (e.g., use over 4 years), most (56%–71%) were using for less than 150 days of any individual year, if usual geriatric doses were consumed. Fourth, the proportion of long-term users in each intensity category remained stable over the 4 years, suggesting that overall use was not escalating over time in this sample. As expected, those with high use during the index year (e.g., daily) had a high probability of continued use over the following 3 years (72% of new users; 90% of previous users). Furthermore, 16.7%–20.8% of long-term users, 0.8%–1.0% of the sample, had evidence of benzodiazepine use almost daily during the 4-year period. Thus, chronic daily use of benzodiazepines was evident for only a small proportion of this healthy adult sample. Three variables emerged as predictors of benzodiazepine use that have not been reported previously: chronic disease score, pain, and BMI. The chronic disease score reflects overall number and severity of chronic conditions and was a better predictor of benzodiazepine use than self-rated health in this study. Self-rated fair or poor health has been found by previous investigators to be associated with benzodiaz-
574
epine use.5,6 Two explanations may account for the association between pain and benzodiazepine use. First, benzodiazepines may be used to treat pain conditions, especially pain secondary to muscle spasms. Simon et al.3 reported that 38% of new benzodiazepine users had pain (e.g., back, neck, headache) listed in the chart as the indication for use. Elderly patients were less likely to have psychological symptoms (anxiety, depression, and situational stress) listed in the chart as the indication for benzodiazepine initiation. Second, patients with chronic pain frequently have comorbid psychiatric disturbances, most often depression.23 Although this association was significant even when controlling for self-reported emotional disorder, it is likely that we were not able to fully adjust for depressive disorder, which is consistently associated with benzodiazepine use.5,7,8 Individuals who were overweight (BMI ⱖ25)15 were 39% less likely to use a benzodiazepine, even when controlling for previous benzodiazepine use. The explanation for this finding is unclear. Obesity is a risk factor for sleep difficulties, especially obstructive sleep apnea,24 which is common in elderly patients.1 Benzodiazepines should be avoided in individuals with sleep apnea because the CNSdepressant properties may interfere with resumption of breathing. Although avoidance of benzodiazepine use in individuals with sleep apnea may account partially for the lower risk of benzodiazepine use in overweight individuals, this is unlikely to be a major explanatory factor, given the strength of the association. Contrary to our findings, a positive correlation between BMI and anxiolytic use has been reported in middle-aged women and men.25,26 These findings are of concern in light of the evidence that women with low-to-normal BMI are at increased risk for hip fractures.27,28 Benzodiazepines, also associated with hip fractures,29,30 were most likely to be used by the segment of the sample at greatest risk for hip fractures, on the basis of body size (women with low–normal BMI). As expected, previous use of benzodiazepines was a strong predictor of future use. Isacson2 reported that previous benzodiazepine use predicted longterm use of these agents. Sex, education, and selfreported pain and stress were no longer associated with benzodiazepine use in the following year, once previous benzodiazepine use was accounted for. The intensity-of-exposure measure used in this
Am J Geriatr Psychiatry 11:5, September-October 2003
Gray et al. study, TSD, reflected both dose and duration. The TSD represents the number of days that a benzodiazepine was used over a 1-year period if used at the minimum effective dose for elderly patients. Although the TSD was useful for describing the general pattern of intensity of use, we recognize that actual daily usage may have been higher or lower, which would change the distribution among infrequent, occasional, frequent, and daily categories of use. The TSD should not be confused with the defined daily dose (DDD) methodology developed by the WHO Collaborating Center for Drug Statistics, which has been used in many European pharmacoepidemiology studies and is a technical unit reflecting the normal daily dose for an adult on the main indication for the drug.31 In general, the TSD conversion resulted in a higher number of days of use for a given period, when compared with DDD calculations, since the TSD conversions are smaller than the DDD conversions. As an example, 4 mg was used for calculating the TSD for diazepam, whereas 10 mg was used for DDD calculations. Our results are unlikely to be generalizable to the older adult population at-large. This study was conducted among predominately white, relatively healthy, middle-class older adults who were members of an HMO and had volunteered to participate in a health promotion trial. Selection bias is a concern, given the 36% participation rate, however, respondents were similar to nonrespondents with respect to health status and the chronic disease score;12 the latter was an important predictor of future benzodiazepine use. The sample studied is likely to explain in part the lower prevalence of use in this sample, compared with other studies, and may also have an impact on intensity of use. Another limitation is that misclassi-
fication of benzodiazepine use was possible. Information about prescription drug use was obtained from a computerized pharmacy database, and therefore, it is not known if the benzodiazepine was actually consumed. Although most medications prescribed by GHC physicians are filled in GHC pharmacies, it is possible that prescriptions were filled at outside pharmacies that would not be captured.
SUMMARY The use of benzodiazepines by older adults has received considerable attention. The results of this study are mainly encouraging, for many reasons. First, new users had low intensity of use and had a low probability of continuing use over the following 3 years. Of those who did have persistent use, intensity of use did not escalate over the 4 years, and most had just occasional use or less. Of concern is that benzodiazepine were used by the segment of the sample that are at greatest risk for hip fractures (women with low–normal BMI). Approximately 1% of the sample had evidence of daily use for 4 consecutive years. Benzodiazepines are recommended for short-term use for most indications. All individuals using benzodiazepines should be assessed periodically for possible reduction and discontinuation of therapy to minimize serious adverse effects (falls, hip fractures, impaired cognition, and depressive symptoms). Dr. Buchner is currently at the Centers for Disease Control. This work was supported by Grant K08AG00808-01 from the National Institute on Aging.
References 1. Folks DG, Burke WJ: Psychotherapeutic agents in older adults: sedative-hypnotics and sleep. Clin Geriatr Med 1998; 14: 67–86 2. Isacson D: Long-term benzodiazepine use: factors of importance and the development of individual use patterns over time: a 13year follow-up in a Swedish community. Soc Sci Med 1997; 44:1871–1880 3. Simon GE, VonKorff M, Barlow W, et al: Predictors of chronic benzodiazepine use in a health maintenance organization sample. J Clin Epidemiol 1996; 49:1067–1073 4. Egan M, Moride Y, Wolfson C, et al: Long-term continuous use of benzodiazepines by older adults in Quebec: prevalence, incidence, and risk factors. J Am Geriatr Soc 2000; 48:811–816 5. Blazer D, Hybels C, Simonsick E, et al: Sedative, hypnotic, and
Am J Geriatr Psychiatry 11:5, September-October 2003
antianxiety medication use in an aging cohort over ten years: a racial comparison. J Am Geriatr Soc 2000; 48:1073–1079 6. Gleason PP, Schulz R, Smith NL, et al: Correlates and prevalence of benzodiazepine use in community-dwelling elderly. J Gen Intern Med 1998; 13:243–250 7. Taylor S, McCracken CF, Wilson KC, et al: Extent and appropriateness of benzodiazepine use. results from an elderly urban community. Br J Psychiatry 1998; 173:433–438 8. Mayer-Oakes SA, Kelman G, Beers MH, et al: Benzodiazepine use in older, community-dwelling southern Californians: prevalence and clinical correlates. Ann Pharmacother 1993; 27:416–421 9. van Hulten R, Leufkens HG, Bakker A: Usage patterns of benzodiazepines in a Dutch community: a 10-year follow-up. Pharm World Sci 1998; 20:78–82
575
Benzodiazepine Use in Elderly Persons 10. Jorm AF, Grayson D, Creasey H, et al: Long-term benzodiazepine use by elderly people living in the community. Aust N Z J Public Health 2000; 24:7–10 11. Wagner EH, LaCroix AZ, Grothaus L, et al: Preventing disability and falls in older adults: a population-based randomized trial. Am J Public Health 1994; 84:1800–1806 12. Wagner EH, Grothaus L, Hecht JA, et al: Factors associated with participation in a senior health-promotion program. Gerontologist 1991; 31:598–602 13. Saunders KW, Stergachis A, VonKorff M: Group Health Cooperative of Puget Sound, in Pharmacoepidemiology. Edited by Strom BL. London, UK, Wiley, 1994 14. Kastrup EK (ed): Facts and comparisons, 2000, in Facts and Comparisons; 2000. Edited by Kastrup EK. St. Louis, MO, 2000 15. McEvoy GK (ed): AHFS Drug Information, 2000. Edited by McEvoy GK. Bethesda, MD, American Society for Healthcare Pharmacists; 2000 16. Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults: The Evidence Report: National Institutes of Health. Obes Res 1998; 6(suppl2):51S–209S 17. Veit CT, Ware JE: The structure of psychological distress and wellbeing in general populations. J Consult Clin Psychol 1985; 51:730–742 18. Scholes D, LaCroix AZ, Wagner EH, et al: Tracking progress toward national health objectives in the elderly: what do restricted activity days signify? Am J Public Health 1991; 81:485–488 19. Von Korff M, Wagner EH, Saunders K: A chronic disease score from automated pharmacy data. J Clin Epidemiol 1992; 45:197– 203 20. Little RJA, Rubin DB: Statistical Analysis with Missing Data. New York, Wiley; 1987 21. Statistical Solutions System for Data Analysis, v. 2.0. Saugus, MA, Statistical Solutions; 1999
576
22. Rojas-Fernandez CH, Carver D, Tonks R: Population trends in the prevalence of benzodiazepine use in the older population of Nova Scotia: a cause for concern? Can J Clin Pharmacol 1999; 6:149–156 23. Von Korff M, Simon G: The relationship between pain and depression. Br J Psychiatry 1996; (168suppl):101–108 24. National Task Force on the Prevention and Treatment of Obesity: Overweight, obesity, and health risk. Arch Intern Med 2000; 160:898–904 25. Rosmond R, Bjorntorp P: Psychiatric ill-health of women and its relationship to obesity and body fat distribution. Obes Res 1998; 6:338–345 26. Rosmond R, Lapidus L, Marin P, et al: Mental distress, obesity, and body fat distribution in middle-aged men. Obes Res 1996; 4:245–252 27. Ensrud KE, Lipschutz RC, Cauley JA, et al: Body size and hip fracture risk in older women: a prospective study. Study of Osteoporotic Fractures Research Group. Am J Med 1997; 103:274– 280 28. Farmer ME, Harris T, Madans JH, et al: Anthropometric indicators and hip fracture: The NHANES I Epidemiologic Follow-Up Study. J Am Geriatr Soc 1989; 37:9–16 29. Ray WA, Griffin MR, Downey W: Benzodiazepines of long and short elimination half-life and the risk of hip fracture. JAMA 1989; 262:3303–3307 30. Herings RM, Stricker BH, de Boer A, et al: Benzodiazepines and the risk of falling leading to femur fractures: dosage more important than elimination half-life. Arch Intern Med 1995; 155:1801–1807 31. Ronning M, Blix HS, Harbo BT, et al: Different versions of the anatomical therapeutic chemical classification system and the defined daily dose: are drug utilisation data comparable? Eur J Clin Pharmacol 2000; 56:723–727
Am J Geriatr Psychiatry 11:5, September-October 2003