اولین همایش بین المللی تحول خدمات سلامت: تلفیق مراقبت، درمان، بازتوانی و امیدبخشی

رفتن به صفحه اصلی

ارائه مجازی مقاله

کد مقاله 1129

عنوان An Explainable AI for Personalized Polypharmacy Management in Elderly Individuals

نویسندگان Farshid Nematzadeh - Samin Mirzapour - Sina Alizadeh Tabrizi

چکیده Background: The problem of polypharmacy among older people has become increasingly prevalent and has resulted in several drug-related events, drug-drug interactions, and increased health care utilization. The current clinical decision support systems are more concerned about detection than prevention and do not have much transparency, thus affecting clinicians' confidence in such solutions. Method: In this research, the researchers designed and validated an explainable AI-based solution that combines deterministic clinical rules (the Beers Criteria and STOPP/START) with machine learning risk models (Random Forest, XGBoost) for deprescribing and individualized pharmacotherapy management. The researchers used the multiphase mixed-methodology approach to develop this solution using electronic health records and insurance claims data. The software uses a modular service-oriented architecture with HL7 FHIR interoperability. Finding: There was great agreement between the system and experts (kappa = 0.82); the sensitivity was very high (91%) along with specificity (88%) and positive predictive value (86%). The usability test received excellent results (SUS = 82.4). In the pilot application, there was a decrease in ADEs (−18%), serious drug interactions (−22%), hospitalization from medications (−11%) as well as the average number of medications (from 8.7 to 7.9). Attributions on feature level and references to guidelines considerably increased clinicians' confidence and quick decision making. Conclusion: The explainable AI system integrated into clinical practice can support polypharmacy management in elderly patients.

فایل ها

فایل پوستر مقاله

محتوای ارائه مجازی مقاله

نظرات