Background: Despite strong performance on Venous Thromboembolism (VTE) Core Measures, our institutional approach to VTE prophylaxis placed significant burden on providers through frequent automated alerts. Within our Electronic Medical Record (EMR), provider alerts triggered when a patient with at least one identified risk factor for VTE did not have prophylactic medication ordered. However, only 11% of patient encounters with an alert resulted in a prophylaxis order. Ordering providers frequently chose an action that delayed the alert, resulting in retriggering the alert to the same or other provider at a later time. Despite the low efficiency of the alert system, our performance on VTE prophylaxis measures exceeded 90% compliance. The low efficiency of the alert was concerning for increasing the risk of alert fatigue. We were also concerned that alterations to the alert could jeopardize Core Measure performance and patient safety.

Methods: Our aim was to reduce the firing of VTE prophylaxis alerts by 50% within 6 months without impacting patient safety or VTE-1 performance. Our secondary measure was to decrease the frequency of the alert and decrease provider time addressing the alert. We estimated provider time addressing the alert as “dwell time:” the time between the triggered BPA and the provider action to address the alert. Using AHRQ’s Five Right to Clinical Decision Support framework, we revised the VTE-1 alert three iterative cycles: 1) Excluding some provider types from triggering the alert, including the emergency department and procedural areas. Additionally, for the consulting only providers, postponing the re-firing of the alert from one hour to 24 hours. 2) The second cycle limited the alert to adult patients. 3) The third cycle removed two risk factors (coronary artery disease and dehydration) and modified two others (chronic kidney disease and warfarin use).

Performance on eCQM data was captured in a report populated from discrete data elements captured within provider EMR documentation. The results were reviewed weekly during the active improvement phase. We generated and validated the algorithm for the report in collaboration with clinical experts, information technology, and by manual abstraction performed by our existing Core Measures auditing group.

Results: At baseline, providers received 1194 VTE alerts per week. After the first revision, alerts dropped to 722 (P < 0.001). They dropped to 615 per week after the second cycle, but this drop was not statistically significant (P = 0.168). Performance on the eCQM VTE-1 measure was unchanged before and after our improvement cycles (79 vs. 78%, P = 0.628). Mean dwell time after a VTE provider alert was 6 seconds. This translates to 104 provider-hour per year spent responding to VTE alerts. If the observed trends hold after our improvement cycle, the net savings to the institution would be 50 provider-hours per year.

Conclusions: Refining the firing criteria for our VTE alert achieved our goal of reducing the firing rate while maintaining our compliance on VTE-1. In addition to that it can help the institution to save physician’s time that can be directed to patient care.