Vascular Surgery

Unhoused patients have worse crude survival after infrarenal AAA repair, driven by urgent presentations.

In a VQI retrospective cohort of 99,733 patients, 127 were unhoused and had shorter mean postoperative survival than housed patients. Unhoused patients were younger but had more symptomatic or ruptured presentations (44.1% vs 17.7%). After adjustment for age, procedure type, preoperative characteristics, and urgency, the survival difference lost statistical significance. Among unhoused patients ≥65 years, 1-year survival was markedly lower after open repair (~45%) versus EVAR (~85%). 1

Meta-analysis: ML models moderately predict postoperative complications and early cancer recurrence.

Across 31 studies, pooled performance for postoperative-complication prediction was sensitivity 0.75, specificity 0.78, and AUC 0.83. For early recurrence, pooled sensitivity was 0.74, specificity 0.73, and AUC 0.80, with top proposed models reaching AUC ~0.88. Performance varied by model type, tumor type, sample size, and geography, and no publication bias was detected by Deeks' test. Authors conclude ML is promising but needs larger, standardized studies for reliable clinical application. 2

SLIM method translates treatment effects between correlated endpoints with low bias.

The authors present SLIM, a stacked multivariate random-effects model that estimates translational coefficients between correlated endpoints. Simulations showed near-zero bias and nominal confidence coverage compared with biased Daniels-Hughes estimates. Applied to an AAA dataset, SLIM quantified associations between growth rate and rupture risk and enabled biomarker–outcome what-if analyses. This is a methodological advance for meta-analysts and translational researchers, not an immediate bedside tool. 3