Truace
Subscribe

Development of a machine learning-based risk warning model for potentially inappropriate medication use among older patients in a hematology service

IntroductionOlder hematology patients often have multiple comorbidities and receive complex medication therapies; this can increase the risk of potentially inappropriate medication (PIM) use. A machine learning model could potentially serve as an automated screening and warning tool to facilitate the identification of PIM and support timely medication review in these patients. The aim of this study is to develop machine learning-based model that provide warnings regarding PIMs use in elderly patients hospitalize…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-2056· cite & verify
Development of a machine learning-based risk warning model for potentially inappropriate medication use among older patients in a hematology service
Assessment of adequacy of reimbursement rates to pharmacies and its impact on the access to medication and pharmacy services by Medicaid recipients by Adams, E. Kathleen Gavin, Norma I., 1950- Kreling, David Henry Gondek, Kathleen Holahan, John United States. Health Care Financing Administration SysteMetrics, Inc Urban Institute. Public domain

In brief

Researchers built and compared seven classification algorithms to predict which hospitalized patients over 65 in a university hospital hematology service in Türkiye were receiving potentially inappropriate medications. Inappropriate use was defined using the 2023 Beers Criteria, and the data covered admissions from 2018 to 2020.

Older hematology patients often have several conditions and complex drug regimens, so automated screening could help prompt timely medication review. Because the work is a small retrospective pilot at a single center, it shows technical promise but not yet proof of effects on patient care.

Main points

  1. The study was retrospective and covered patients over 65 admitted to a hematology service between 2018 and 2020.
  2. It included 190 patients, and 45.8% of them had potentially inappropriate medication use.
  3. Seven classification algorithms were compared using accuracy, precision, recall, F1 score, ROC curve, AUC, and Brier score.
  4. Logistic regression performed best, with 0.87 accuracy, 0.83 recall, 0.86 F1 score, and 0.91 AUC.

The gain

A logistic regression model trained on 190 older hematology inpatients identified potentially inappropriate medication use with 0.91 AUC, suggesting a usable early-warning screen for pharmacists and physicians.

The rundown

The model was designed as a patient-based automated screening tool to support clinical pharmacists and physicians. Its target is the identification of potentially inappropriate medication use so that medication review can happen sooner.

Methodologically, the team used Python version 3.12.7 and stratified 5-fold cross-validation to estimate performance. Potentially inappropriate medications were labeled according to the American Geriatrics Society's 2023 Beers Criteria.

Nearly half of the cohort, 45.8%, had at least one potentially inappropriate medication, which indicates the issue is common in this population and the screening need is substantial.

Sources

  1. Peer-reviewedJournal of Oncology Pharmacy Practice2026-10-19

The debate