Frailty, antipsychotics and mortality among community-based older adults with impaired cognition

Background: Potentially inappropriate antipsychotic use among older adults with cognitive impairment remains an important concern. The identification of frailty may offer an opportunity to better understand the risks and benefits of antipsychotics in older populations across care settings. Objective(s): We examined the association between new antipsychotic use and mortality over 6 months among older home care clients and explored variation in mortality risk by frailty level. Method(s): We conducted a retrospective cohort study of long-stay older (aged 66+) clients in Ontario, Canada by linking their Resident Assessment Instrument for Home Care [RAI-HC] clinical data with health administrative databases. Included were clients with a diagnosis of dementia and/or cognitive impairment assessed between April 1, 2008 and March 31, 2013. Frailty was defined using a validated 72-item frailty index (FI) derived from RAI-HC data assessed at cohort entry. Exposed clients were defined as having newly received an antipsychotic in the 6 months post-cohort entry (with no claims for antipsychotics in the year prior to drug index date). We used a 2-stage matching process to define unexposed clients and their index date which included matching on age, sex, frailty group (robust, pre-frail, frail), year of RAIHC, and a derived propensity-score. Outcome was time to death assessed during 6 months following clients' index date. Cause-specific hazards models (stratified by FI group) were utilized. As the assumption of proportional hazards was not supported, we estimated time-specific hazard ratios at 1-, 3- and 6 months. From cumulative incidence function curves, we estimated the number needed to harm (NNH) at 6 months. Result(s): Among 5,853 matched exposed-unexposed pairs, new antipsychotic users showed a significant hazard of mortality at 1, 3 and 6 months relative to unexposed, with the highest risk observed in the first month (HR = 3.48, 95%CI 3.10-3.91). At 1-month (but not 6 months), estimates were significantly higher (p = 0.02) for robust (HR = 5.24, 3.71-7.41) vs. pre-frail (HR = 3.27, 2.69-3.97) and frail (3.36, 2.85- 3.96) clients. The estimated NNH at 6 months was 7.5 (95%CI 6.8-8.2) for the total sample and varied by frailty level: NNH of 9.9 (8.3-12.5), 9.1 (7.8-11.1) and 5.2 (4.6-6.1) for robust, pre-frail and frail, respectively. Conclusion(s): An easily derived frailty index may help clarify the relative and absolute risks of mortality associated with antipsychotics in older adults with cognitive impairment.

AM