A Multidisciplinary, AI-Supported Quality Improvement Intervention to Manage Polypharmacy in Aging People with HIV

Background Often driven by multimorbidity, polypharmacy may cause adverse drug reactions, drug–drug interactions (DDIs), reduced adherence, and functional decline, emphasizing the need for improved medication safety in people with HIV (PWH). This study describes the implementation of a multidisciplinary, AI-supported quality improvement (QI) intervention at the Modena HIV Metabolic Clinic (MHMC) to optimize medication management in aging PWH. Methods We conducted a cross-sectional, single-center QI project involving PWH aged ≥50 years. The intervention included four components: (1) AI-driven medication reconciliation: Before the visit, PWH received automated messages to submit photos of medications via WhatsApp. AI translated packaging into Anatomical Therapeutic Chemical (ATC) codes, enabling automated medication capture and reconciliation. (2) Clinical decision support via NavFarma®: This platform integrates medication data with comorbidities and flags risks like PIMs (based on STOPP/START and Beers Criteria), DDIs, anticholinergic burden, nephrotoxicity, and QTc prolongation. (3) Pharmacist review: Clinical pharmacists interpreted reports and generated therapeutic recommendations in the electronic patient chart (EPC). (4) Multidisciplinary decision-making: Clinicians reviewed these recommendations and incorporated them into care plans when appropriate. Results Between November 2024 and July 2025, 181 PWH submitted photographs of their medications. Median age was 63 years (IQR: 8); 131 (72%) were male. Median nadir CD4 count was 200 cells/μl (IQR: 208), and current CD4 count was 746.5 cells/μl (IQR: 395). Median time since HIV diagnosis was 31 years (IQR: 13). Complete agreement between medications in the EPC and those reported by PWH was observed in 111 (61.3%) individuals, while 70 (38.7%) showed discrepancies. Of these, 51 (28.2%) were prescribed medications recorded in the EPC, whereas 119 (34.3%) were taking medications not documented in the EPC. Additionally, 21 (11.6%) initiated new medications, and 19 (10.5%) had medications discontinued during MHMC visit. Among new prescriptions, 13 (61.9%) were lipid-lowering agents, 3 (14.3%) antidepressants, and 7 (33.3%) other classes. Among discontinued medications, 11 (57.9%) were lipid-lowering agents, 3 (15.8%) proton pump inhibitors, and 5 (26.3%) other classes. Overall, daily pill burden was 6 (IQR: 4.8), while total number of pharmacologically active agents was 10 (IQR: 5). Finally, 98 PWH were also evaluated by NavPharma and a clinical pharmacist. Pharmacist recommendations were recorded and reviewed during visits, and 135 (48.2%) were considered clinically relevant. However, medications were discontinued by the pharmacist in only 3 cases (1.7%). A total of 69 (70.4%) major DDIs were identified and most frequently involved CNS depressants, antiretrovirals (dolutegravir and cobicistat), and cardiovascular agents (statins, ACE inhibitors). Notably, 44 (44.9%) of PWH showed moderate-to-high risk for QTc prolongation, while 26 (26.9%) had significant ACB scores due to CNS drugs. Renal toxicity risks were flagged in 37 (38.3%) of reports, commonly associated with NSAIDs, ACE inhibitors, and tacrolimus. Conclusions In conclusion, preliminary findings from this quality improvement intervention suggest that integrating AI technology with clinical judgment can enhance the quality of polypharmacy management in aging PLWH. The process supports accurate medication reconciliation, improves the identification of inappropriate prescription and DDIs, and support the integration of AI and pharmacist-led reviews into routine HIV care to enhance medication safety.