ML flags overtreatment risk
A UBC Therapeutics Initiative study used machine learning to estimate the prevalence of glucose‑lowering overtreatment among older adults in long‑term care and community settings—findings published in BMJ Open highlight systematic overtreatment risk in seniors [] []. The work underlines how analytics can pinpoint populations where de‑intensifying therapy may be safer.
The cohort included 133,773 older adults with an A1C ≤7.0%, of whom 38,074 (28.5%) were classified as overtreated; the overtreated group had a mean age of 79.6 years and a median A1C of 6.4%. bmjopen.bmj.com The study a priori defined potential overtreatment as overlapping prescriptions for ≥2 glucose‑lowering medications or any insulin or any sulfonylurea dispensed within 90 days after the index A1C test. bmjopen.bmj.com A gradient‑boosting machine‑learning model that combined expert‑selected and data‑driven variables was the top performer, achieving an area under the curve (AUC) of 0.87, sensitivity 0.81 and a negative predictive value of 0.89 on a temporally distinct 2021–2023 test set. bmjopen.bmj.com The analysis used province‑wide British Columbia linked administrative health claims from 2016–2023 and was led by UBC Therapeutics Initiative authors (Carney, Burnett, Ambasta, Thompson, Lapp, Dormuth); the manuscript was received 17 June 2025 and accepted 18 February 2026, and supplemental code/materials were posted in a UBC Therapeutics Initiative GitHub repository. bmjopen.bmj.com
