Is There an Ideal Method of Predicting Postoperative Complications Following Major Head and Neck Procedures?: A Comparative Analysis of Three Frailty Indices

BACKGROUND: Frailty is a critical determinant of surgical outcomes in head and neck surgery, given that patients often present with multiple comorbidities. Identifying the optimal frailty index for preoperative risk stratification can enhance perioperative management and postoperative outcomes. The objective of this study was to compare the predictive accuracy of the modified frailty index-5 (mFI-5), the risk analysis index for administrative data (RAI-A), and the revised risk analysis index (RAI-Rev) in forecasting postoperative complications and nonhome discharge following soft tissue free-flap reconstruction, total laryngectomy (TL), and parotidectomy. METHODS: Retrospective cohort study querying a multicenter, national surgical database (ACS-NSQIP) from 2015 to 2020 for patients aged 18 or older who underwent soft tissue reconstruction of the head and neck, TL, or parotidectomy. Univariate and multivariate logistic regression models assessed predictive value, with discriminative ability evaluated using receiver operating characteristic (ROC) curve analysis. RESULTS: A total of 3633 soft tissue free flaps, 1115 TL, and 10 678 parotidectomy patients were identified. The mFI-5 most reliably stratified patients into predictive frailty tiers, showing strong associations with major complications and Clavien-Dindo Grade IV events, particularly in the free flap (OR = 4.06, AUC = 0.62) and TL cohorts (OR = 7.23). The RAI-A "severely frail" group demonstrated the highest specificity for nonhome discharge across all procedures, especially in the free flap (OR = 10.75) and parotidectomy (OR = 83.81) cohorts. RAI-Rev outperformed both mFI-5 and RAI-A in predicting mortality following TL (AUC = 0.76, p = 0.029) and showed superior AUCs for mortality (0.84) and reoperation (0.64) in parotidectomy. All indices demonstrated their strongest overall performance in the parotidectomy cohort. CONCLUSIONS: The frailty indices varied in predictive utility by procedure and outcome. RAI-A and RAI-Rev offered greater granularity, with both indices excelling in nonhome discharge prediction and the RAI-Rev demonstrating superior sensitivity for rare outcomes like mortality. The predictive strength of each index was best within the context of parotidectomy, suggesting particular value in ambulatory settings. Index selection should be tailored to the surgical context and the specific complication of interest.