Competition through wait times in the market for unscheduled care

Americans spent 13,400 person-years waiting in emergency departments (EDs) in 2009 alone. Wait times in emergency departments have been increasing at a compounded rate of 3.5% per year since at least the early 1990s when data first became available. Furthermore, the quantity of emergency department services demanded has increased by 3.1% annually, but the supply of ED services has not increased concomitantly (Becker and Friedman 2013).

This article develops a theoretical model which explains this lack of supply response in the context of a 1986 law which prevents hospitals from denying patients ED care even if they cannot pay. In the model, consistent with anecdotal and cross-sectional evidence, hospitals are constrained from setting individual wait times based on non-clinical factors. However, the hospital chooses an overall set of policies (staffing levels, adoption of operations management innovations, etc.) which produces a hospital-wide baseline wait time. The hospital's choice of wait time is endogenous to the mix of patient profitabilities. We model demand as depending on the time price of services. The model predicts that higher wait times result from increased proportions of Medicaid and uninsured patients, who appear to be unprofitable even when only variable costs are taken into account (Hennepin et al. 2013). However, if the total number of ED visits declines simultaneously, as is the case with the entry of cherry-picking competitors, the model yields amgibuous predictions.

We then test the assumptions and predictions of the model using a novel dataset of emergency department wait times in seven states (MA, NJ, AZ, NE, FL, NY, RI). This dataset is a census consisting of 120 million visits to more than 500 hospitals. First, the model's assumption that hospitals are constrained in setting individual wait times based on profitability is supported by cross-sectional regression coefficients: hospitals with 50 percentage point greater uninsurance rates have 26.0 minute longer wait times (p<.01; national mean wait time is 58 minutes), whereas conditional on hospital uninsurance rate individuals who are uninsured are not shown to have longer wait times (coefficient of 0.86 minutes, p=0.13). Next, we test the model's predictions in three ways: cross-sectional models which instrument for area uninsurance/Medicaid rates, models assessing the effect of entry of urgent care clinics into the market (since these clinics see predominantly insured, less severely injured patients), and triple-difference estimates of the differential effect of Massachusetts' insurance expansion across the change in hospital insurance mix. Preliminary results support the theoretical model's conclusions.

Because hospital emergency departments seem to be unable to offer shorter wait times only to insured patients, significant negative spillovers appear to exist. The recent national expansion of insurance may mitigate this negative externality, providing a substantial welfare gain to the higher-income currently-insured who do not otherwise benefit from the Patient Protection and Affordable Care Act. Given the uncertainty as to whether the marginal costs of ED care are higher (Bameza and Melnick 2006) or lower (Williams 1996) than other settings, however, the full welfare implications are unknown.