A Novel Primary Care Planning Informatics Tool Informed by Data-Driven Multimorbidity Grouping: User-Centered Design and Feasibility Testing.

Background: Patients with multimorbidity have complex health care needs and are at high risk for adverse health outcomes. Primary care teams need tools to effectively and proactively plan care for these patients. We developed VET-PATHS (Veteran Panel Management Tool for High-Risk Subgroups), a novel care planning informatics tool for complex primary care patients. VET-PATHS groups patients by chronic condition via latent class analysis of electronic health record data, then jump-starts care planning by suggesting "care steps" based on data-driven high-priority care for the group. Objective: The study aimed to iteratively adapt VET-PATHS with user input, then test feasibility and acceptability by frontline primary care teams for empaneled patients at high risk. Methods: Three rounds of user-centered design sessions with 17 primary care providers and registered nurses at 5 sites from 2019 to 2021 were conducted to obtain feedback on the VET-PATHS layout, content, and user interface. Feedback was summarized into 4 user experience domains (useful, desirable, credible, and usable), leading to progressively updated prototypes. After the national tool release, we conducted a pilot intervention study in 2023-2024 with 6 primary care teams at 4 sites using VET-PATHS during asynchronous regular meetings. Tool use and resulting care plans were assessed by templated observation during meetings, postpilot chart review, and administrative data. Individual qualitative interviews were analyzed by rapid template analysis for feasibility, acceptability, and utility. Results: User-centered feedback led to updated tool content, context (eg, use in proactive panel management), targeted users (eg, focusing on primary care providers), and display layout. Pilot teams used VET-PATHS over 4 to 8 weekly meetings (mean length 24, range 16-49 min), actively reviewing 80% (280/351) of empaneled patients at high risk visible in the tool. Tool use prompted 127 new actions for 91 unique patients (33% of patients reviewed) and documentation of >1 new care plan for 19% of patients reviewed. Common actions included requests to return to the clinic (n=34, 27%), referrals (n=25, 20%), or vaccinations (n=24, 19%). Of the 127 actions planned, 53 (42%) were received by patients. Difference-in-difference trends for acute hospitalizations declined post pilot, while outpatient utilization was stable or increased for pilot team patients compared to all patients at high risk at pilot sites (per-patient counts: acute hospitalizations -0.15; primary care visits 0.00; mental health visits 0.99). Four generalist teams (n=11 interviews) described higher acceptability. Two "focused" teams with more homogenous panels, for example, substance use disorder (n=3 interviews), found care steps less useful. Teams described how VET-PATHS improved efficiency of care planning through automated patient grouping and identification of care gaps and increased multidisciplinary involvement. Conclusions: User-centered improvements to VET-PATHS were designed to help clinicians process and use complex information about patient multimorbidity to efficiently create new care plans. Subsequently, VET-PATHS was acceptable and feasible to frontline primary care teams, particularly with more general patient panels, and led to concrete changes to clinical care delivery.