From the Departments of Anesthesiology & Pain Medicine, University of Ottawa and The Ottawa Hospital, Ottawa, Ontario, Canada.
School of Epidemiology & Public Health, University of Ottawa, Ottawa, Ontario, Canada.
Anesth Analg. 2021 Aug 1;133(2):366-373. doi: 10.1213/ANE.0000000000005290.
Frailty-a multidimensional syndrome related to age- and disease-related deficits-is a key risk factor for older surgical patients. However, it is unknown which frailty instrument most accurately predicts postoperative outcomes. Our objectives were to quantify the probability of association and relative predictive performance of 2 frailty instruments (ie, the risk analysis index-administrative [RAI-A] and 5-item modified frailty index [mFI-5]) with postoperative outcomes in National Surgical Quality Improvement Program (NSQIP) data.
Retrospective cohort study using Bayesian analysis of NSQIP hospitals. Adults having inpatient small or large bowel surgery 2010-2015 (derivation cohort) or intermediate to high risk mixed noncardiac surgery in 2016 (validation cohort) had preoperative frailty assigned using 2 unique approaches (RAI-A and mFI-5). Probabilities of association were calculated based on posterior distributions and relative predictive performance using posterior predictive distributions and Bayes factors for 30-day mortality (primary outcome) and serious complications (secondary outcome).
Of 50,630 participants, 7630 (14.0%) died and 19,545 (38.6%) had a serious complication. Without adjustment, the RAI-A and mFI-5 had >99% probability being associated with mortality with a ≥2.0 odds ratio (ie, large effect size). After adjustment for NSQIP risk calculator variables, only the RAI-A had ≥95% probability of a nonzero association with mortality. Similar results arose when predicting postoperative complications. The RAI-A provided better predictive accuracy for mortality than the mFI-5 (minimum Bayes factor 3.25 × 1014), and only the RAI-A improved predictive accuracy beyond that of the NSQIP risk calculator (minimum Bayes factor = 4.27 × 1013). Results were consistent in leave-one-out cross-validation.
Translation of frailty-related findings from research and quality improvement studies to clinical care and surgical planning will be aided by a consistent approach to measuring frailty with a multidimensional instrument like RAI-A, which appears to be superior to the mFI-5 when predicting outcomes for inpatient noncardiac surgery.
衰弱是一种与年龄和疾病相关的多种因素综合征，是老年手术患者的主要风险因素。然而，目前尚不清楚哪种衰弱工具最能准确预测术后结果。我们的目标是量化 2 种衰弱工具（即风险分析指数-行政[RAI-A]和 5 项改良衰弱指数[mFI-5]）与国家外科质量改进计划（NSQIP）数据中术后结果的关联概率和相对预测性能。
使用 NSQIP 医院的贝叶斯分析进行回顾性队列研究。2010-2015 年接受住院小肠或大肠手术的成年人（推导队列）或 2016 年接受中高危混合非心脏手术的成年人（验证队列）使用 2 种独特方法（RAI-A 和 mFI-5）术前分配衰弱。关联概率基于后验分布进行计算，使用后验预测分布和贝叶斯因子评估 30 天死亡率（主要结局）和严重并发症（次要结局）的相对预测性能。
在 50630 名参与者中，有 7630 人（14.0%）死亡，19545 人（38.6%）发生严重并发症。未经调整，RAI-A 和 mFI-5 与死亡率的关联概率均>99%，比值比（OR）≥2.0（即大效应量）。在调整 NSQIP 风险计算器变量后，只有 RAI-A 与死亡率的关联具有≥95%的非零概率。预测术后并发症时也出现了类似的结果。RAI-A 对死亡率的预测准确性优于 mFI-5（最小贝叶斯因子 3.25×1014），只有 RAI-A 能够提高预测准确性，超过 NSQIP 风险计算器（最小贝叶斯因子=4.27×1013）。留一交叉验证结果一致。
使用 RAI-A 等多维工具衡量衰弱，将衰弱相关研究和质量改进研究的发现转化为临床护理和手术计划，将有助于为住院非心脏手术患者预测结果提供一致的方法，而 RAI-A 似乎优于 mFI-5。
