Fall risk score in community-dwelling older people: a derivation and validation study

Abstract

Abstract Background : Early clinical detection of patients at high risk of falling is crucial. This study aimed to derive internally validate the novel risk score to classify patients at high risk of falls. Methods: Study data from 334 older people from the fall clinic in medical center. We developed logistic regression model on the entire dataset using screened-out variables by Least absolute shrinkage and selection operator(LASSO) regression, with 1-year falls prediction. Using bootstrap method, we evaluated the model performance with measures of overall predictive performance, discrimination, and calibration. The clinical usefulness and net benefit were estimated with decision curve analysis. To streamline assessment process, we created a scoring system for predicting 1-year fall risk. Results: Model 1 showed a C-index of 0.803 [95% CI 0.7491–0.8569]. The internal bootstrap validation indicated only minimal overfitting (optimism-corrected C-index 0.802). Model 2 showed a C-index of 0.752 (95% CI 0.6981–0.8059). The internal bootstrap validation indicated only minimal overfitting (optimism-corrected the C-index 0.747). Compared with model 2, model 1 also significantly improved continuous net reclassification improvement (NRI) [0.4682 (0.3141–0.6223), P < 0.01], categorical NRI [0.507 (0.2905–0.7235), P < 0.01, cutoff: 0.2–0.8), and integrated discrimination improvement [0.2047 (0.1473–0.2621), P < 0.01]. Conclusions: This study developed and internally verified a score for predicting fall risk of community-dwelling older people. The score could predict falls in elderly people with relatively higher accuracy compared with timed up and go test and 30-s chair sit-stand test.