Wait times for uncomplicated care in emergency departments (EDs) have risen dramatically for more than a decade (Horwitz, Green, and Bradley 2010), a reflection of increased demand for ED services.  This ED crowding has been repeatedly linked to adverse outcomes, including mortality.  A lack of access to primary care has been thought to be one of the primary drivers of ED crowding (Newton et al. 2008), but little data has been available to study the relationship between the two.  This study uses a novel methodology to obtain hospital-level ED wait times.  Because patients without a longitudinal relationship with a primary care provider (PCP) may be the most likely to use an ED in lieu of primary care (Medicaid Access Study Group 1994), we then examine the association of these ED wait times with directly-assessed new appointment availability at nearby primary care clinics.

The National Hospital Ambulatory Medical Care Survey (NHAMCS) has long been used as the authoritative source on ED crowding, including wait times.  However, the finest geography publicly available divides the country into only four regions.  This severely hampers the ability to link this critical resource to other datasets and hence learn about the impact of other parts of the healthcare system on ED crowding.  To obtain hospital-level wait times, we utilized the indirect estimation strategy utilized by Murray et al. (2008).  We established the wait time vs. visit duration relationship in the NHAMCS, then used that relationship to predict wait times in the State Emergency Department Database (SEDD), which provides hospital identifiers.

We identified 63 variables (comprising a total of 1708 diagnosis and procedure codes, the duration of the visit, and ED characteristics) with consistent definitions in both the SEDD and NHAMCS.  We then estimated a regression of these characteristics on log wait time in NHAMCS.  The estimation predicts with a high degree of confidence (R^2 of 0.53 at the visit level).  Intuitively, this regression strategy essentially works by predicting the total treatment time given detailed clinical information and ED characteristics, then subtracting the estimate of treatment times from the measured length-of-stay. 

When the SEDD is delivered in late April, we will compare ED wait times to the appointment availability and days-to-appointment for the clinics surrounding each ED via censored normal regression, controlling for area socio-demographic, health system, and provider workforce characteristics.  The study involves approximately 4 million visits in 125 EDs, and 4,196 audited clinics.

This study promises to bring new evidence to the interrelationship between PCP and ED access.  Insurance expansions such as the impending implementation of the ACA may divert unscheduled care from EDs to PCPs (Miller 2011) through a variety of mechanisms.  For this to happen, however, PCPs must have appointments available for patients.  If PCP capacity is not available near crowded EDs, it will blunt the potential of the ACA to reduce ED utilization.  Thus primary care capacity should be monitored closely, not just for overall PCP utilization but specifically for those who will be newly seeking an ongoing relationship with their PCP.