((B-205)) (B-205) Narrative and Analytical Methods to Assess Bias and Pool Results of Studies Associated with Fall-Related Injuries

Background: Medications are a modifiable risk factor for fall-related injuries (FRIs). Pharmacoepidemiologic trials and observational studies of older adults could benefit from a measure of the burden of medications associated with FRIs, especially in nursing home (NH) residents, but no such evidence-based, composite outcome measure exists. An index is needed that synthesizes the published evidence on medications and fall outcomes while correcting for potential bias, such that decreases in the index reflect decreases in the global burden of medications associated with FRIs.

Objectives: To develop a novel index for the burden of medications associated with FRIs by synthesizing the evidence quantifying the effects of medications on FRIs in NH residents while correcting for potential bias in the evidence.

Methods: A pragmatic literature review was conducted to identify studies that aimed to estimate the effects of medications on falls and FRIs in NHs from January 1990 to present. Full research articles were then assessed for bias using either the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) or the Revised Cochrane risk-of-bias tool for randomized trials. Effect estimates for each study were then corrected for the degree of potential bias assessed since bias was expected to result in estimates that were further from the null than the actual effect. Each degree of bias represented a specific reduction in the magnitude of the effect estimate (e.g., moderate bias = 20% reduction). Corrected effect estimates were standardized to the same scale, then pooled into a single summary estimate for each medication class that represented the effect of the class on falls and FRIs.

Results: In total, 106 articles were identified. Of those, 80 were eligible for assessment of the risk of bias, including 75 observational studies and 5 randomized controlled trials (RCTs). We found 67 (89.3%) observational studies to be of serious risk of bias and 8 (10.7%) of moderate risk of bias. No observational study was found to have low or critical risks of bias. Of RCTs, 4 (80%) were found to have some concerns regarding the risk of bias, and 1 (20%) to have a low risk of bias. No RCT had a high risk of bias. For example, among 10 studies of selective serotonin reuptake inhibitors (SSRIs) with falls and FRIs, SSRIs increased the risk of falls and FRIs by 98% before and 67% after bias correction.

Conclusions: The successful correction of effect measures for potential bias will ensure our medication index is robust despite heterogeneous studies. Furthermore, our approach to developing an outcome measure for medication burden associated with FRIs could be extended to assess other critical outcomes like major bleeds and cardiovascular events.