Infrastructure for Medication Adherence Prediction & Intervention

Medication adherence prediction operates against a partially formalized knowledge base: prescription fill policies and refill protocols exist as documented procedures, but the clinical decision logic for intervention type matching (when to recommend pill packaging vs. reminder calls vs. pharmacist counseling for a specific patient profile) is largely undocumented, reflecting individualized pharmacist judgment. Documentation practice exists for adherence measurement methodology (PDC calculations) but intervention protocols and personalization criteria are not systematically documented.

Adherence prediction requires systematic capture of prescription fill events, refill request timing, PDC calculations, prior intervention attempts, and outcomes of previous outreach. EHR and pharmacy management systems capture fill history through defined workflows. Template-driven documentation ensures prior intervention attempts (call attempts, outcomes, patient responses) are logged with consistent fields needed to train the prediction model and avoid duplicate outreach to the same patient.

Adherence prediction modeling requires consistent schema across patient records: medication list with NDC, fill dates, days supply, PDC calculations, diagnosis codes, demographic fields, and prior intervention history with outcomes. RxNorm and ICD-10 codes provide structured taxonomy for medications and conditions. All patient records must share these fields to enable comparative risk scoring—identifying which patient profiles correlate with non-adherence across the population.

The adherence prediction system requires API access to pharmacy fill data (internal and Surescripts external pharmacy claims), EHR (diagnoses, medication complexity, prior hospitalizations), and care management platforms (prior intervention history). These connections enable the model to assemble a complete adherence risk profile per patient. The existing EHR-to-pharmacy integration supports internal data access; Surescripts APIs enable external fill data that reveals adherence gaps invisible within the health system.

Adherence prediction models and intervention protocols are updated on scheduled cycles—typically when adherence program reviews occur or when drug formulary changes affect the medication classes being monitored. The intervention matching logic (which patient profile gets which intervention type) isn't updated by clinical events but by periodic program evaluations. P&T committee quarterly meetings provide the governance mechanism, but adherence protocol updates lag evidence, consistent with an L2 maintenance posture.

Medication adherence prediction requires API-based connections between the pharmacy fill system, EHR (diagnosis and medication complexity), care management platform (prior intervention tracking), and outreach systems (phone, portal, messaging). These connections enable the system to generate risk scores and trigger intervention workflows without manual data assembly. The existing closed-loop medication use process integration, supplemented by Surescripts API access for external fill data, provides the required connectivity.