{
  "id": 3790630,
  "title": "A clinical decision support tool for accurate hip fracture prediction: A nationwide cohort study",
  "url": "https://urgent.news/2026/08/27/a-clinical-decision-support-tool-for-accurate-hip-fracture-prediction",
  "topic": "health",
  "section": "Health & Medicine",
  "published": "2026-08-27T14:00:00.000Z",
  "source": {
    "name": "PLOS Medicine",
    "slug": "plos-medicine",
    "url": "https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1005190"
  },
  "original_language": "en",
  "account": "A study conducted across Sweden aimed to create an accurate clinical decision support tool for predicting hip fractures, without the need for in-person assessments. The research focused on individuals aged 50 or older who did not use osteoporosis medication in the past two years. Over a period of follow-up, 142,327 hip fractures were recorded. The study used multiple data sources, including diagnoses, medications, procedures, demographics, and socioeconomic information.\n\nResearchers employed traditional Cox models, as well as machine learning algorithms XGBoost and DeepSurv, to evaluate the risk of hip fractures. The resulting clinical decision support tool, FRACTURE-ML, based on DeepSurv and utilizing 2,500 predictors, demonstrated an area under the curve (AUC) of 0.89 at one year and 0.88 at two years. A reduced version of the tool with 35 predictors maintained similar AUCs at two and five years.\n\nCox models with 35 and 400 predictors also achieved comparable AUCs. Both DeepSurv and Cox models performed well at the individual level, according to calibration plot analysis. Screening using the FRACTURE-ML tool within Sweden's fracture liaison services (FLS) approach yielded an AUC of 0.55 at two years. Compared to FLS, FRACTURE-ML identified nearly seven times more individuals at risk for a hip fracture in the two-year prediction timeframe (sensitivity of 0.84 vs. 0.12). However, specificity remained relatively high for both methods (0.79 vs. 0.98).\n\nThe study's limitations include the lack of external validation and implementation studies, which are necessary to determine the clinical usefulness of FRACTURE-ML. Overall, FRACTURE-ML proves effective in predicting hip fractures and could serve as a cost-effective solution for population screening, ultimately improving hip fracture prevention.",
  "summary": "by Kristian F. Axelsson, Henrik Litsne, Konstantinos Konstantinou, Hussnain Khalid, Aldina Pivodic, Mattias Lorentzon Background Although hip fractures are commonly associated with functional decline, increased morbidity, and mortality, accurate models for both short- and long-term prediction that do not rely on in-person assessment remain lacking. The aim was to develop and validate a…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}