The abstract is read at the publisher; the summary is JClub's.
Among 872 inpatients with sickle cell disease at the University of Ghana Hospital (2019–2024), 15.7% had prolonged hospitalisation, predicted best by logistic regression (AUC-ROC 0.692).
- Why it matters: Prolonged stays create major healthcare and caregiver burden, and clinicians need earlier risk signals to guide timely management and resource planning.
- What they did: The study retrospectively modeled prolonged hospitalisation using multivariable logistic regression and five supervised machine-learning algorithms with five-fold stratified cross-validation, applying class weighting and SMOTE within training folds. Performance was assessed with AUC-ROC, AUC-PR, sensitivity, specificity, PPV, NPV, F1, and balanced accuracy.
- The result: Acute chest syndrome (aOR=4.28), malaria (aOR=3.77), pneumonia (aOR=3.45), and anaemia (aOR=2.71) were independently associated, while higher haemoglobin was protective (aOR=0.86); ML performance was modest with overlapping AUC-ROC confidence intervals and no clear superiority.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.