In The Journal of surgical research
INTRODUCTION : Emergency General Surgery (EGS) conditions in older patients constitutes a substantial public health burden due to high morbidity and mortality. We sought to utilize a supervised machine learning method to determine combinations of factors with the greatest influence on long-term survival in older EGS patients.
METHODS : We identified community dwelling participants admitted for EGS conditions from the Medicare Current Beneficiary Survey linked with claims (1992-2013). We categorized three binary domains of multimorbidity: chronic conditions, functional limitations, and geriatric syndromes (such as vision or hearing impairment, falls, incontinence). We also collected EGS disease type, age, and sex. We created a classification and regression tree (CART) model to identify groups of variables associated with our outcome of interest, three-year survival. We then performed Cox proportional hazards analysis to determine hazard ratios for each group with the lowest risk group as reference.
RESULTS : We identified 1960 patients (median age 79 [interquartile range [IQR]: 73, 85], 59.5% female). The CART model identified the presence of functional limitations as the primary splitting variable. The lowest risk group were patient aged ≤81 y with biliopancreatic disease and without functional limitations. The highest risk group was men aged ≥75 y with functional limitations (hazard ratio [HR] 11.09 (95% confidence interval [CI] 5.91-20.83)). Notably absent from the CART model were chronic conditions and geriatric syndromes.
CONCLUSIONS : More than the presence of chronic conditions or geriatric syndromes, functional limitations are an important predictor of long-term survival and must be included in presurgical assessment.
Ho Vanessa P, Bensken Wyatt P, Flippin J Alford, Santry Heena P, Claridge Jeffrey A, Towe Christopher W, Koroukian Siran M
2022-Nov-21
Emergency general surgery, Frailty, Functional status, Geriatric surgery, Multimorbidity
