Using Analytics to Track an EHR Deployment

By any standard, the electronic health records rollout at Duke Medicine is big.


By any standard, the electronic health records rollout at Duke Medicine is big. Located in Durham, N.C., the academic medical center is in the midst of a 4-year deployment of an Epic system across its ambulatory and inpatient settings. Epic will replace multiple legacy systems and offer an integrated package of clinical and revenue cycle applications.

At Health Data Management’s Healthcare Analytics Symposium, Monica Horvath, Duke’s team lead in health intelligence and research services, described how data analytics can be used to monitor the effectiveness of the roll-out. The findings were not always pleasant, as Horvath discussed how some physicians wound up returning to a legacy documentation system rather than use the new Epic module because of cumbersome documentation requirements.

Horvath’s team, part of Duke Health Technology Solutions, the medical center’s technology arm, surveyed clinics about the ease of doing various transactions both before and after go-live. That helped track clinics having trouble with adoption, leading to site visits, additional retraining and even software redesign. The survey was broken down by role, with nurses, physicians and even front desk staff weighing in on the merits of the EHR.

Among the go-live measures tracked was transcription utilization. Duke wanted to move providers to an embedded voice recognition module in the Epic system, rather than use a variety of outsourced transcription services. To date, Duke has reduced the use of transcription nearly a third, Horvath said. But some clinics continued to pay for outside transcription. Upon analysis, Horvath’s group determined that these were physicians who had returned to their legacy EHR as a workaround to cumbersome documentation requirements.  “That was exactly what we didn’t want to happen,” she said.

And while the initial deployment did not go entirely as planned, the data gathered by user surveys, transcription utilization measures, and other billing metrics offered the staff responsible for the EHR design valuable insight, she said. For example, some clinics saw big spikes in their billing lag times, as it proved difficult to close out charts, apply codes and dispatch claims. Revised workarounds and system improvements have resulted in those lag times returning back to the pre-EHR levels.