30-Second Takeaway
- Preoperative risk scores can stratify TORS patients for severe 30-day morbidity.
- ML models show moderate discrimination for in-hospital mortality on ECMO (**AUC 0.759**).
Week ending June 20, 2026
Five recent studies relevant to cardiothoracic surgeons: risk scores, ECMO prognostication, postoperative symptom trajectories, hybrid care trial design, and simulation training evidence
TSOS: preoperative score identifies predictors of severe 30-day morbidity after TORS for OPSCC
In 1,592 TORS patients with OPSCC, 15.7% experienced severe postoperative events within 30 days. Independent preoperative predictors included advanced T stage, BMI <22, smoking, preoperative dysphagia, and hyperlipidemia. A point-based TSOS threshold of ≥24 showed the best discrimination for the composite severe outcome. This is a derivation cohort from TriNetX; external validation is required before clinical deployment.
Machine learning predicts in-hospital mortality for post-cardiotomy ECMO with moderate accuracy
Using 5,982 ELSO registry patients, six ML algorithms predicted in-hospital mortality during ECMO support. The boosting model achieved the highest discrimination (AUC = 0.759); other models performed worse. Key predictors included age, on-support lactate, arterial pH, and BMI, but on-support variables reflect trajectory. Authors advise external validation and view models as tools for dynamic risk assessment during ECMO.
Older adults show delayed cough recovery after VATS lung cancer resection compared with dyspnea
In 292 patients aged 60–80 undergoing VATS, dyspnea largely normalized by 6 months (95.2% mMRC 0–1). Cough remained below preoperative LCQ-MC scores at 6 months and had median recovery time 75 days. VATS lobectomy was associated with slower cough recovery versus sublobectomy (adjusted HR 0.47, 95% CI 0.34–0.66). Use these data to set expectations and plan symptom follow-up in older surgical patients.
References
Numbered in order of appearance. Click any reference to view details.
Additional Reads
Optional additional studies from this edition.