A formula for predicting postoperative functional decline using routine medical data in elderly patients after hip fracture surgery

Background: If functional decline after hip surgery can be predicted without special assessment, the effects of new treatments and rehabilitation practices can be easily compared with previous cases or those in other countries. The purpose of this study was to develop and examine a formula for such prediction. Methods: Data of 3,120 patients older than 65 yr with hip fracture were analyzed. The Barthel Index was used for evaluating activities of daily living (ADL). Low ADL was defined as patients with a lower score at discharge than the score at admission and patients with complete dependence at admission that did not change until discharge. Three models were developed in a training sample: Basic, Comorbidity, and Laboratory & Vital Signs models were created by inputting basic patient data, the basic data plus comorbidities, the basic data and comorbidities plus 8 laboratory test results and 5 vital signs, respectively. All potential variables with statistical significance < 0.2 on univariate analyses and some variables that may be clinically meaningful were included in multivariable models. The final model was developed by stepwise logistic regression. Results: The c-statistic of the Laboratory & Vital Signs formula was 0.701 and the predictive value was 76.9%. The c-statistics of the Basic and Comorbidity formulas were 0.643 and 0.664, respectively. Applying the Laboratory & Vital Signs formula to the validation sample, the c-statistic was 0.663. Conclusions: The formula developed from the medical data collected routinely before surgery could predict low ADL following hip fracture surgery in elderly patients. Level of Evidence: IVb