Equity in Diagnostic Imaging Trial (EDIT)

Investigators at our institution (Hwang & Richardson)* documented a racial disparity in the use of CT scans to diagnose subarachnoid hemorrhage, (SAH) a life-threatening form of stroke. Subsequent analyses found a similar finding in NHAMCS, a national ED Visit database*. The investigators will compare the effectiveness of 1) an audit-feedback intervention that provides data on race-specific CT rates and 2) a clinical decision support (CDS) message in the electronic "headache" charting template in eliminating the previously observed Black-White disparity in CT rates to rule out SAH.

Study Setting: The adult section of the Mount Sinai Emergency Department, an urban academic department with an annual volume of over 70,000 adult visits and a patient population representing a diverse cross-section of race and socioeconomic groups. The adult patients are cared for by a staff that includes: attending physicians, physician assistants, emergency medicine resident physicians, and rotating resident physicians from other specialties. The department has a dedicated CT scanner that is readily available around the clock. Epic Production is used for all physician and nurse documentation and for all order entry (including CT scans).

Over a 24-month period the investigators will compare the intervention and control groups' rates of testing for SAH across races, controlling for patient acuity, age and co-morbidities. Because the close working relationships among the attending and resident physicians in our department may lead to knowledge diffusion, the intervention may change the behavior of the control group as well as the intervention group, thus leading to an underestimation of the intervention effect as measured by the between- groups difference. The investigators will use a time series design to allow us to evaluate this effect. 10 months after the implementation of the first intervention, the clinical decision support message will be implemented. The impact of each intervention will be evaluated using a repeated measures design using a mixed linear model.