• Daniel I McIsaac, Coralie A Wong, Allen Huang, Husein Moloo, and Carl van Walraven.
• Department of Anesthesiology and Pain Medicine, University of Ottawa and The Ottawa Hospital, Ottawa, ON, Canada.
• Ann. Surg. 2019 Jul 1; 270 (1): 102-108.

ObjectiveTo develop and validate a preoperative frailty index (pFI) for use in population-based health administrative (HA) data.Summary Background DataFrailty is a robust predictor of adverse postoperative outcomes. Population-level frailty measures used in surgical studies have significant methodological limitations. Frailty indices (FIs) are a well-defined approach to measuring frailty with well-described methods for development and evaluation. An appropriate preoperative FI in HA data has not been derived or evaluated.MethodsRetrospective cohort study using linked HA data in Canada. We identified people >65 years (2002-2015) who had major elective or emergency surgery. Standardized methods were used to construct a 30-variable pFI. Unadjusted and multilevel, multivariable adjusted models were used to measure the association of the pFI with 1-year mortality and institutional discharge. Elective patients were the derivation cohort, emergency patients were the validation cohort. Prespecified sensitivity analyses were performed.ResultsWe identified 415,704 elective, and 95,581 emergency patients. The elective 1-year mortality rate was 4.7%. Thirty percent of population-level deaths occurred in people with frailty. Every 0.1-unit increase in the pFI was associated with a 2.20-fold increase in the adjusted odds of mortality (95% CI 2.15-2.26; c-statistic 0.81), and a 1.70-fold increase in institutional discharge (95% CI 1.59-1.80; c-statistic 0.71). pFI performance was similar in emergency patients, and was robust to changes in index composition.ConclusionsA preoperative FI derived from HA data is a robust method to measure frailty in elective and emergency patients. Generalizable FIs should be considered a standard approach to population-level study of surgical frailty.

21 keywords...

hide…