Background: To compare machine learning (ML) methods beyond traditional regression analysis using a real dataset on nurse staffing and falls in nursing homes.
Methods:We used a total of 6 ML algorithms to compare the superiority of ML over traditional regression methods. We applied three representative ML algorithms-random forest (RF), logistics regression, support vector machine (SVM): linear, polynomial, radial, and sigmoid-to the preprocessed dataset to develop a prediction model (N = 60). We used an accuracy measure to evaluate prediction models.Results: RF was the most accurate model (0.883), followed by the logistic regression model (0.867).Using RF procedures, we identified a total of 6 variables as predictors of falls including hours per resident day of administrative staff, proportion of nursing home residents with psychiatric medications, aggressive behaviors and cognitive dysfunction, urinary incontinence, current number of residents in each nursing home, and the maximum capacity of each nursing home.
Conclusions:The appropriate choice of prediction model is quite challenging for nursing researchers.For effective fall management, researchers should consider organizational characteristics as well as personal factors. The examination of related factors on falls is quite meaningful in that preventing falls contributes to improving the quality of life and care of residents and decreases health care costs.
BackgroundThe most frequently reported adverse event among nursing home (NH) residents, falls are associated with increased morbidity and mortality, and with reduced functioning [1-3]. The U.S. fall prevalence is 1.7 falls per resident yearly in NHs [1]; about 50% of NH residents in developed countries fall eachyear [4]. Accurate prediction of factors associated with a fall in NHs is important because nurses, health care professionals, researchers in practice, administrative staff, researchers, and politicians address fall issues. These stakeholders develop targeted fall-prevention management and assess residents based on the factors associated with a fall [5]. Prior studies identified a number of risk factors for falls, including age, sex, visual deficits, psychotropic medications, cognitive dysfunction [6][7][8], range of motion, urinary incontinence [9-12], hours per residents day (HPRD) of registered