{
  "abstract": "Background Lateral ankle sprains are common and often recur. While most recover well, a significant minority have poor outcomes. The existing SPRAINED model uses predictors—age, BMI, pain at rest and on weight-bearing, time to assessment, weight-bearing ability, and history of sprain—to identify those at risk of poor 9-month recovery. External validation and an online prediction tool would enhance usability and support targeted early interventions.Methods SALI study - multicentre cohort study recruited ankle sprain patients from 11 UK emergency departments. Self-reported data were collected at injury and 3 and 12 months after injury. A SuperLearner ensemble model combined logistic regression, random forest, SVM, gradient boosting, and elastic net. Sample size (n=662) was calculated a priori. Hyperparameters were tuned via grid search with 10-fold cross-validation. Brier score decomposition assessed model contributions. Internal validation used 2,000 bootstraps. Performance was evaluated using discrimination, calibration, Brier score, and R 2.Results SALI study showed excellent external validation with model showed strong predictive performance for poor - calibration was excellent (slope 0.91, Hosmer-Lemeshow p=0.45). Discrimination was good as well, with sensitivity 0.77, specificity 0.73, and AUC 0.79. These findings support its use for early identification of high-risk patients and targeted intervention.Conclusion External validation demonstrated that the predictive model accurately identifies patients at risk of poor recovery, supporting its use in clinical settings and as a stratification tool for future targeted RCTs. Development of an online calculator and Shiny application will further enhance accessibility and implementation in practice and research",
  "authors": [
    {
      "affiliations": [
        "Queen’s Medical Centre Nottingham University Hospitals, United Kingdom"
      ],
      "name": "Thomas Bestwick-Stevenson"
    },
    {
      "affiliations": [
        "Queen’s Medical Centre Nottingham University Hospitals, United Kingdom"
      ],
      "name": "Kim Edwards"
    },
    {
      "affiliations": [
        "Queen’s Medical Centre Nottingham University Hospitals, United Kingdom"
      ],
      "name": "Mark Batt"
    },
    {
      "affiliations": [
        "The Wake Forest University School of Medicine, United States of America"
      ],
      "name": "Garrett Bullock"
    },
    {
      "affiliations": [
        "Queen’s Medical Centre Nottingham University Hospitals, United Kingdom"
      ],
      "name": "Brigitte Scammell"
    },
    {
      "affiliations": [
        "Queen’s Medical Centre Nottingham University Hospitals, United Kingdom",
        "The Wake Forest University School of Medicine, United States of America",
        "University of Oxford, United Kingdom"
      ],
      "name": "Stefan Kluzek"
    }
  ],
  "title": "49 Predicting poor recovery after ankle injury for targeted RCTs: external validation of an existing tool using a large prospective cohort",
  "uid": "9e18221c-01c7-5c43-922a-ef5046eb01e4"
}