Assessing residual confounding of the association between antipsychotic medications and risk of death using survey data

Abstract

Background: Nonrandomised studies on the causal effects of psychotropic medications may be biased by patient characteristics that are not fully adjusted.

Objective: Studies using linked claims databases found that typical antipsychotic medications were associated with increased short-term mortality compared with atypical antipsychotics. It has been suggested that such results may be due to residual confounding by factors that cannot be measured in claims databases. Using detailed survey data we identified the direction and magnitude of such residual confounding.

Design: Cross-sectional survey data.

Participants: 17 776 participants aged > or =65 years from the Medicare Current Beneficiary Survey (MCBS).

Measurements: To determine the association between typical antipsychotic use and potential confounding factors we assessed five factors not measured in Medicare claims data but in the MCBS, i.e. body mass index, smoking, activities of daily living (ADL) score, cognitive impairment and Rosow-Breslau physical impairment scale. We estimated adjusted associations between these factors and antipsychotic use. Combined with literature estimates of the independent effect of confounders on death, we computed the extent of residual confounding caused by a failure to adjust for these factors.

Results: Comparing typical antipsychotic users with atypical antipsychotic users, we found that not adjusting for impairments in the ADL score led to an underestimation of the association with death (-13%), as did a failure to adjust for cognitive impairment (-7%). The combination of all five unmeasured confounders resulted in a net confounding of -5% (range -19% to +2%). After correction, the reported association between typical antipsychotic use and death compared with atypical antipsychotic use was slightly increased from a relative risk (RR) of 1.37 to 1.44 (95% CI 1.33, 1.56). Comparing any antipsychotic use with non-users would result in overestimations of >50% if cognitive impairment remained unadjusted.

Conclusion: Claims data studies tend to underestimate the association of typical antipsychotics with death compared with atypical antipsychotics because of residual confounding by measures of frailty. Studies comparing antipsychotic use with non-users may substantially overestimate harmful effects of antipsychotics.

Figure 1

Sensitivity of bias estimates towards misspecification of selected confounder-disease association (RR_CD).