Abstract
To control the upward spiral of healthcare costs, hospitals seek to implement efficiency interventions whose benefits are frequently assessed by reductions in average inpatient length of stay (LOS). However, average hospital LOS is a crude metric when trying to assess the utility of an intervention focussed on a particular service or over a specific time window. It cannot isolate the time or place of the intervention from the full duration of a patient's hospital visit, which may include more than 1 hospital service or extend beyond the intervention's time window. At Montefiore Medical Center, a new analytic method was used to describe a month-long effort to improve care efficiency in a hospital teaching service. Using an extension of the Cox proportional hazard model (S-plus), we were able to analyze the contribution of only those patient-days that took place during the time window of interest on the service of interest, eliminating the contamination of the "non intervention days." Having built the appropriate model, we were then able to graph the behavior of the groups with and without the intervention and calculate the model's expected average LOS, controlling for the appropriate variables. By comparing this method with a conventional average LOS analysis, we demonstrate the superiority of using this "time slice" method over the conventional analysis of LOS.