Development of an electronic patient navigation platform to enhance adjuvant therapy coordination after cancer surgery.

Abstract

e13685 Background: For many solid tumors, adjuvant therapy (AT) after curative surgery can reduce recurrence risk and improve survival and is supported by practice guidelines. Implementation of guideline-based care faces several challenges, including comprehensive and efficient identification of patients who may be candidates for AT to optimize care coordination. This study assessed the feasibility and accuracy of an electronic patient navigation platform that identifies patients eligible for AT. Methods: In a single health system, this platform was piloted for renal cell carcinoma and urothelial carcinoma. Clinical databases were searched for required criteria, and sources of inaccuracies were documented. Initially, the EHR was used to identify patients by querying for surgery codes, but it couldn’t filter for cancer cases or specific pathologic data because relevant variables were in unstructured text. The laboratory information system (LIS), in contrast, stored structured data for cancer specimens using the College of American Pathologists (CAP) Cancer Protocol templates, so an LIS report was generated by selecting templates, querying pathologic features, and calculating AT eligibility using rule-based logic. To validate, an EHR report was created by querying the EHR for surgery codes, focusing on cases from 9/2025 to 11/2025, and a genitourinary oncologist manually verified eligibility for AT per established guidelines and whether the patient was appropriately referred to a medical oncologist. The LIS report was compared with the EHR validation report to determine accuracy. Results: The LIS report identified 127 surgical pathology reports, of which 117 were for cancers and 19 were flagged as requiring AT using rule-based logic. The EHR report identified 262 surgical cases. Manul review verified that 123 were for cancers and 18 met criteria for AT (of which 6 were not appropriately referred). The LIS report missed 7 cancers from the EHR report because a CAP template was not used; all were early-stage and did not meet criteria for AT. The EHR report missed 4 cancers in the LIS report, one of which met criteria for AT, due to inefficiencies in surgical coding. For determining AT eligibility, using denominators of patients identified in either report, the rule-based logic had 95% sensitivity and 99% specificity, with 1 false-positive (metastatic cancer not eligible for AT protocol) and 1 false-negative (histology type in the CAP template was incorrect). Conclusions: A report querying CAP Cancer Protocols enabled automated determination of AT eligibility. However, clinical documentation errors can lead to inaccurate reporting. In the next phase, a large language model (LLM) approach will be used to refine the report. For implementation, a coordinator will review the report monthly and facilitate referrals. The program's success will be determined by adherence to AT guidelines.