Linking Electronic Medical Record Decision Support to Prior Authorization

Americans are still waiting for a return on their investment in electronic medical records. Digitizing health care has yet to facilitate either the continuous quality improvement expected by exposing inefficiencies in operational workflows within healthcare settings or transformational change from introducing new innovations to achieve healthcare improvements in novel ways. One area where health information technology could provide short-term process improvements and create new opportunities to transform the process might be linking decision support and prior authorization within an electronic medical record workflow (First disclaimer: I’m currently the co-chair of the Prior Authorization work group within the Da Vinci Project, a private sector initiative that addresses the needs of the Value Based Care Community by leveraging the HL7 FHIR platform.).

In this case, the government has provided the trigger to consider doing prior authorization differently. On January 1, 2020, all providers who order advanced diagnostic imaging services for Medicare beneficiaries must consult approved appropriate use criteria at the time of order placement for eight priority clinical areas. Unfortunately, the final report of the Medicare Imaging Demonstration project that this statute was based upon found no significant effect on advanced imaging use with a modest improvement in the appropriateness rating of advanced imaging orders. I pulled this content from a recent Annals of Internal Medicine article that included two approaches for providers and health organizations to address this mandate: 1) electronic health record integration with computerized physician order entry and 2) use a standalone clinical decision support portal.

If a medical center or physician group decided to try to integrate the appropriate use criteria within their electronic health records, alert fatigue may jeopardize Medicare reimbursement for advanced imaging orders. Many clinicians may not even log into an electronic medical record during a clinical encounter. For those providers who log into the electronic medical record in front of the patient, the circumstances of the specific encounter (e.g., patient would prefer more time talking to the provider rather than watching the provider type into a keyboard), or the provider’s day (e.g., running 90 minutes behind with a family event scheduled in an hour) may impact the likelihood that the provider will actually read, consider and implement any electronically-generated decision support recommendations regardless of the alerts’ value to the patient’s long-term health or financial stability. If the radiology appropriate use criteria is embedded with other decision support alerts, these alerts may be ignored if they pop-up alongside lower-value suggestions.

The challenge of real-time clinical decision support could be framed as an electronic intrusion intended to alert providers about diagnostic and therapeutic opportunities ranging from making safer decisions, choosing high-value options over lower-value ones, and prior authorization requirements (considering the January 1, 2020 advanced imaging mandate a prior authorization requirement). If providers understood that specific decision support aids were intended to reduce patient out-of-pocket payments or reduced barriers to obtain downstream services (i.e., tests, referrals and procedures), they may be less likely to ignore them. Ideally, decision support systems would be constructed with redundancies so a clinical team could recover if the alert was inadvertently ignored during the clinical encounter.

In this framework, the decision support tool would have multiple owners: practice or medical center-specific recommendations (local medical leadership team), evidence-based recommendations to suggest higher-value options and avoid unsafe behaviors (a federal agency, evidence-based medicine vendor), steps that may lead to a prior authorization (patient’s payer). To complicate matters further, providers who do not practice at Kaiser or the VA would need to accommodate multiple payers within a clinic session or day in the hospital. Rather than rely on every health information technology team supporting each electronic medical record installation across the country deploy decision support with the same commitment to this approach, it might be more effective to support an infrastructure that enables multiple entities to engage the provider and the clinical team in the decision support.

CDS Hooks is a technical platform that could support such a decentralized decision support model. The local medical leadership team would need to establish a governance process to determine what decision support is presented to the user when multiple CDS responses might be returned after a trigger (e.g., how to address conflicting recommendations, determining the alert sequence). A national or global vendor for evidence-based guidelines would be easier to debug and validate. Depending on specific diagnostic and treatment decisions, a payer might suggest options that reduce the burden of prior authorization for the provider or the provider’s office staff. The current consensus around CDS Hooks is an expectation that providers will interact with pop-up screens with a patient in the exam room. In the absence of that level of provider engagement, payers may need to consider the possibility of working with multiple provider offices using Direct messaging rather than developing custom interfaces with each instance of an electronic medical record (Second disclaimer: I serve as a board member for DirectTrust, the non-profit trade association that develops the policies, standards and practices to support trusted and effective use of Direct messaging.).

It may be too optimistic to consider combining clinical decision support with prior authorization given the suggested cooperation between evidence-based medicine providers and payers. Linking these two activities may increase provider engagement with pop-ups in the electronic medical record, but that hypothesis needs to be tested in multiple settings. Although the technical standards to allow for this decentralized alerting process are currently being developed, each market individual will need to see value in combining these workflows without placing an undue burden on the provider during the clinical encounter or their office staff either before or after the clinical encounter.