30-Second Takeaway
- Preoperative metabolomics plus machine learning predicted postoperative delirium with **AUC 0.855**.
- Dexmedetomidine monotherapy reduced postoperative delirium with moderate-certainty evidence.
- MySurgeryRisk accurately predicts multiple major postoperative complications across centers (AUROCs ≥0.92).
Latest - Week ending May 2, 2026
Grand Rounds: Predictive models and preventive strategies for perioperative complications in older and high-risk patients
Observed shared decision-making remains low by OPTION instruments since 2015
This systematic review of 174 studies (≈20,000 consultations) reports mean OPTION-12 25.1 and OPTION-5 31.8 for routine encounters. Postintervention studies showed higher scores (OPTION-12 38.4, OPTION-5 47.7), suggesting measurable improvements after targeted interventions. Longer consultation length and clinical setting influenced scores, but overall SDM change since 2015 is small.
Preoperative metabolomics with machine learning predicts delirium after hemiarthroplasty
In a prospective multicenter cohort of 260 older adults undergoing hemiarthroplasty, 201 metabolites were quantified and 41 differed by outcome. Machine-learning selection yielded 16 candidate biomarkers and a logistic model with AUC 0.855 (95% CI 0.800–0.910). Model performance was stable on 70/30 train/test splits (AUCs 0.844 and 0.856), supporting preoperative risk stratification for delirium.
Postdischarge prognostic model identifies diabetes patients at risk for foot complications
Among 107,830 discharges, 2.7% developed a foot complication within one year and 12.1% died. A Fine-Gray competing-risks model achieved pooled AUROC ~0.71 and showed better calibration than a random survival forest. In a pilot, median predicted 1-year risk was 1.1%, and 41% of reached patients accepted community foot-care referral.
References
Numbered in order of appearance. Click any reference to view details.
Additional Reads
Optional additional studies from this edition.