BACKGROUND: In Massachusetts we have developed and implemented a distributed network of informatics systems accessing clinical practices’ electronic health record (EHR) data for public health surveillance. We created an open-source, freely available system called Electronic Medical Record Support for Public Health (ESP, www.esphealth.org). ESP has two components: electronic case reporting for notifiable diseases and aggregate-level querying for non-reportable conditions (e.g., diabetes, smoking, opioid prescribing). We describe how the system is implemented, its features, and lessons learned over the last decade.
METHODS: ESP was created by the Department of Population Medicine (DPM) at Harvard Medical School in collaboration with the Massachusetts Department of Public Health (MDPH) and clinical partners. We depend upon strong historic and current partnerships between MDPH, DPM, and clinical partners to get buy-in to create, maintain, and enhance the system. DPM acts as the coordinating center for the network on behalf of MDPH. DPM’s tasks including liaising with MDPH, clinical partners, and a private informatics vendor to implement and maintain ESP installations at all partners sites as well as build new capabilities. ESP runs on servers within clinical partners’ firewalls. Clinical partners export data on every encounter every day to their local ESP instance, building a reporting data mart with full patient medical histories. Once in ESP, relevant data are mapped to standard terminology values and analyzed for conditions of interest using algorithms that query vital signs, diagnosis codes, laboratory tests, and prescriptions to yield summative diagnoses. Samples of cases are validated. Notifiable disease cases are reported to MDPH using encrypted hl7 messages. Summary aggregate data are available in each ESP data mart for MDPH to query via a query distribution network. Partners can approve or reject queries on a case-by-case basis. Clinical partners can elect to participate in notifiable disease reporting, aggregate querying, or both.
RESULTS: ESP currently has 5 clinical partners in Massachusetts: all participate in notifiable disease reporting, 3 permit aggregate queries. Algorithms have been developed for 13 notifiable diseases and 11 chronic conditions. The practices that permit aggregate queries collectively represent ~20% of the state population. Coverage is concentrated towards the eastern part of the state. Ongoing challenges include validation of data completeness and algorithm performance as well as finding ways to engage new clinical partners to join the network.
CONCLUSIONS: Establishing and nurturing collaborative relationships between all participating entities has been the cornerstone of this work.