Key requirements for a new era of emergency department operations research.

In the 2006 reports on the future of emergency care, the Institute of Medicine called for adoption of hospital-wide operations management principles to improve patient care and flow, eliminating emergency department (ED) boarding of inpatients. Further emphasizing the need, multiple studies have documented associations between ED crowding and adverse patient-oriented outcomes. Despite these calls and data, the application of operations research methods in ED settings has been limited. Furthermore, published studies in the field often have limitations in scope or methodology, which makes interpretation and generalizability difficult. We believe it is time to define the key objectives needed to have effective operations research in emergency care. We will focus on 3 general areas: study designs, outcome measures, and how to facilitate operations research that can change practice in a positive manner. In this issue of Annals, Retezar et al reported the effect of diagnostic triage standing orders on ED treatment time. Using data from a single institution, the investigators conducted a retrospective nested cohort study comparing treatment time for patients who received either full or partial triage standing orders with those who received room orders. To control for systematic differences among treatment groups, the authors calculated propensity scores to reflect the probability of receiving triage standing orders. This technique creates comparison groups that are similarly matched on known characteristics such as clinical factors (eg, age, sex) and system conditions (eg, time of day, day of week, and ED occupancy). In the adjusted analysis, the use of full or partial triage standing orders was associated with a 16% reduction in median ED treatment time (defined as the time from placement in an ED treatment room until physical departure for discharged patients and disposition decision time for admitted patients). We commend the authors for the sophistication of their approach and their commitment to formal evaluation of a common operational intervention. Nonetheless, limits exist, well noted by the authors. Because patients were not randomized to their treatment group, it is impossible to eliminate unmeasured confounders that could have