The Magnitude of Time-Dependent Bias in the Estimation of Excess Length of Stay Attributable to Healthcare-Associated Infections

Abstract

BACKGROUNDEstimates of the excess length of stay (LOS) attributable to healthcare-associated infections (HAIs) in which total LOS of patients with and without HAIs are biased because of failure to account for the timing of infection. Alternate methods that appropriately treat HAI as a time-varying exposure are multistate models and cohort studies, which match regarding the time of infection. We examined the magnitude of this time-dependent bias in published studies that compared different methodological approaches.METHODSWe conducted a systematic review of the published literature to identify studies that report attributable LOS estimates using both total LOS (time-fixed) methods and either multistate models or matching patients with and without HAIs using the timing of infection.RESULTSOf the 7 studies that compared time-fixed methods to multistate models, conventional methods resulted in estimates of the LOS to HAIs that were, on average, 9.4 days longer or 238% greater than those generated using multistate models. Of the 5 studies that compared time-fixed methods to matching on timing of infection, conventional methods resulted in estimates of the LOS to HAIs that were, on average, 12.6 days longer or 139% greater than those generated by matching on timing of infection.CONCLUSIONOur results suggest that estimates of the attributable LOS due to HAIs depend heavily on the methods used to generate those estimates. Overestimation of this effect can lead to incorrect assumptions of the likely cost savings from HAI prevention measures.Infect. Control Hosp. Epidemiol. 2015;36(9):1089–1094