Real-world evidence from an implemented, patient-centered artificial intelligence (AI) to address social determinants of health (SDOH) and Quality Oncology Practice Initiative (QOPI) measures to close care gaps.

Abstract

752 Background: Two ASCO frameworks anchor ambulatory oncology accountability: the 2024 ASCO SDOH policy statement endorses systematic SDOH screening with validated tools (e.g., PRAPARE), and QOPI requires ≥75% performance across ~25 chart-abstracted certification measures (core, symptom/toxicity, EOL, disease-specific). Both are retrospective data collections. We implemented an AI with guided conversations and analyzed real-time data collection and intervention. Methods: Retrospective analysis of de-identified interactions with HopeAI, Manta Cares' conversational oncology AI (11/27/25-6/2/26). Users self-identified as Patient (P) or Care Partner (CP). Parallel QOPI-aligned and PRAPARE 8-item SDOH codebooks were applied to conversation text. HopeAI messages were tagged for patient self-management (SM) suggestions (coping, symptom self-care, lifestyle, tracking, self-advocacy, support-seeking). Three platform behavioral endpoints were tabulated: user-logged medication adherence (doses taken/skipped); clicks on SDOH-tagged resource guides (financial, transport, lodging, mental health); and patient-advocacy-group (PAG) referrals. Results: 3,588 P+CP enrolled; 1,119 used HopeAI (P=1,047, CP=72) across all 50 US states; 3,055 conversations. 63% of users began their first conversation on the day they enrolled. 69% of conversations occurred outside business hours. HopeAI raised non-clinical (QOPI distress/survivorship) topics before the user did in 58% of conversations and returned SM suggestions in 66% of conversations overall: 72-86% within QOPI-coded and 61-81% within PRAPARE SDOH-coded subsets (Table). 556 SDOH-tagged resource guides by 166 unique users since HopeAI launch reflect SDOH interventions, and 756 referrals were made to 12 patient-service non-profits. Adults aged ≥70 years were 19% of age-disclosing users. Conclusions: HopeAI not only captured QOPI and PRAPARE SDOH measures, but intervened with individualized self-management suggestions. The AI is available at all hours, is adopted the same day users enroll, measures and closes care gaps, and is geographically scalable. HopeAI can augment clinician care team efforts to deliver quality care and reduce inequitable availability of oncology care. QOPI quality proxies and PRAPARE SDOH items among P+CP HopeAI users (n=1,119). QOPI proxy Prevalence, % of users SM % PRAPARE SDOH Prevalence, % of users SM % Nutrition/survivorship 33 83 Social isolation 34 79 Sleep/rest 32 86 Insurance 25 71 Body image 27 81 Transportation 11 61 Distress/mental health 22 84 Income 5.5 81 Financial/copay 21 72 Housing instability 3.3 69