Misclassification of current benzodiazepine exposure by use of a single baseline measurement and its effects upon studies of injuries.

Abstract

Measurement of drug exposure is a major methodologic challenge for pharmacoepidemiologic studies of acute effects of medications taken intermittently. If the effect is plausible only during periods of active drug use, daily (or even more frequent) exposure measurement is optimal. Benzodiazepines, episodically used hypnotics and anxiolytics, impair psychomotor function and some epidemiologic studies have reported users have increased risk of unintentional injuries. However, several prospective cohort studies of this question defined benzodiazepine exposure status from a single baseline measurement, and these have not consistently reported increased risk. We used data from an historical cohort study to demonstrate the extent of misclassification potentially induced by this practice. The cohort consisted of 2510 Tennessee nursing home residents 65 years of age or older identified in a prior study of antidepressants and falls. Both baseline users (any in 7 days preceding start of follow-up) and current use (use on a given day of follow-up) of benzodiazepines were determined from facility medication administration records, which record information on drugs given to the resident each day. Falls were ascertained from nursing home incident reports and medical records. The effect of benzodiazepine exposure for each ascertainment method was estimated from incidence rate ratios adjusted for multiple fall risk factors by Poisson regression. The 666 baseline benzodiazepine users had current use on 44.6% (95% CI, 40.2-49.2%) of follow-up person-days; baseline non-users had current use for 3.7% of days (2.8-4.7%). Misclassification increased with length of follow-up and with quintile of fall risk. The adjusted fall incidence rate ratios for the baseline user and current use definitions were 1.02 (95% CI, 0.95-1.10) and 1.44 (1.33-1.56), respectively. These findings strongly suggest that to avoid potentially serious misclassification, studies of the acute effects of benzodiazepines and other drugs taken intermittently need to track exposure on a day-by-day basis.