A model for identifying potentially inappropriate medication used in older people with dementia: a machine learning study

• DOI: 10.1007/s11096-024-01730-0
• Publication Date: 2024-07-09T11:01:59Z
• Authors: Qiaozhi Hu; Mengnan Zhao; Fei Teng; Gongchao Lin; Zhaohui Jin; Ting Xu
• Keywords: Categorical variable

Abstract Background Older adults with dementia often face the risk of potentially inappropriate medication (PIM) use. The quality PIM evaluation is hindered by researchers' unfamiliarity criteria for drug While traditional machine learning algorithms can enhance quality, they struggle multilabel nature prescription data. Aim This study aimed to combine six and three classification models identify correlations in information develop an optimal model PIMs older dementia. Method was conducted from January 1, 2020, December 31, 2020. We used cluster sampling obtain data patients 65 years assessed using 2019 Beers criteria, most authoritative widely recognized standard detection. Our modeling process problem transformation methods (binary relevance, label powerset, classifier chain) algorithms. Results identified 18,338 36 types. chain + categorical boosting (CatBoost) demonstrated superior performance, highest accuracy (97.93%), precision (95.39%), recall (94.07%), F1 score (95.69%), subset values (97.41%), along lowest Hamming loss value (0.0011) acceptable duration operation (371s). Conclusion research introduces a pioneering CC CatBoost warning patients, utilizing machine-learning techniques. enables quick precise identification PIMs, simplifying manual process.