Journal of the American College of Surgeons | 2019
Operative Mortality Prediction for Primary Rectal Cancer: Age Matters.
Abstract
BACKGROUND\nThe risk of colorectal cancer increases with age, and the number of older adults requiring operations has increased. The purpose of this study was to determine whether a current risk calculator can accurately predict operative mortality for rectal cancer and whether the predictive accuracy varied with age.\n\n\nMETHODS\nThe American College of Surgeons NSQIP database using ICD-9/10 codes for rectal cancer and CPT codes for proctectomy was accessed (2012 to 2015). The prognostic value of the risk calculator was evaluated using the predicted mortality variable code. Age categories were 18 to 64 years, 65 to 79 years, and 80 to 89 years. Analysis of variance was performed to assess differences between age categories in predicted and actual mortality and Pearson correlation coefficients were computed. Logistic regression models were constructed to evaluate associations adjusted for key covariates.\n\n\nRESULTS\nThere were 9,289 patients included, with age distribution as follows: 18 to 64 years (n\xa0= 5,674), 65 to 79 years (n\xa0= 2,899), and 80 to 89 years (n\xa0= 716). Both predicted and actual mortality increased with age, adjusting for functional status, comorbidity, and other covariates (p\xa0<\xa00.0001). The overall correlation between predicted and actual mortality was low (r\xa0=\xa00.20). The correlation was weakest from 18 to 64 years (r\xa0= 0.07), strongest from 65 to 79 years (r\xa0= 0.25), and in between from 80 to 89 years (r\xa0= 0.13). Predicted mortality was overestimated in the 18 to 64 years and underestimated in both the 65 to 79 years and 80 to 89 years age groups. Predicted mortality by age category interaction terms was also significantly associated with actual mortality in covariate-adjusted logistic regression models, providing additional evidence that the accuracy of predicted mortality varies by age.\n\n\nCONCLUSIONS\nThe American College of Surgeons NSQIP mortality risk estimates appear to be poorly associated with actual mortality and the accuracy might differ between younger and older patients with primary rectal cancer. Goals of care discussion with the older patient about outcomes are indicated, as there is an almost twice predicted mortality.