Evaluating the impact of frailty in older adults on functional outcomes post-mechanical thrombectomy: a multicentre...

Frailty is associated with high risk of mortality. Our study evaluated frailty as a predictor of outcomes in older adults with acute ischemic stroke treated with endovascular thrombectomy utilizing machine learning models. We conducted a multinational observational cohort study. Patients > 70 years-old treated with endovascular thrombectomy were eligible. The primary outcome was 90-day functional independence(FI). 674 patients were included, with frailty defined as Clinical Frailty Score(CFS) > 3. Univariate and multivariate analyses were performed. Predictive machine learning models were developed and evaluated, while values were computed for variable importance. A P-value of < 0.05 was considered statistically significant. On univariate analysis, age, race, diabetes mellitus, albumin, National Institute of Health Stroke Scale(NIHSS), Alberta Stroke Programme Early CT Score(ASPECTS), number of thrombectomy attempts(> 3), recanalization(> TICI 2B), and frailty were significantly associated with 90-day-FI. On logistic regression, NIHSS(adjusted OR: 1.105, 95% CI:1.03-1.11,P = 0.005), recanalization > TICI 2B(adjusted OR: 0.049,95% CI:0.005-0.522,P = 0.012), and frailty(adjusted OR: 13.451,95% CI:2.572-70.3,P = 0.002) were independently associated with 90-day FI. The Random Forest model achieved an AUROC of 0.82 with a Brier score of 0.129. Youden's J statistic yielded a best threshold of 0.63. Confidence intervals were estimated using a non-parametric bootstrap method(95% CI:0.735-0.915). Frailty had the highest feature importance and the second-highest Shapley value. The CFS is an effective prognostication tool for older adults undergoing endovascular thrombectomy. Models incorporating frailty demonstrate good predictive value. Further study involving larger cohorts may refine models that can be applied pre-procedurally to identify patients at risk of poor outcomes.