The nursing process and total health cost variability: an analysis using machine learning

Aims: To find out whether the information that the nursing process provides (functional patterns and the NANDA-NIC-NOC taxonomy), presented through clinical histories, influences predictions of total healthcare costs. Background: The nursing process, is not included in the systems that calculate expenditure in the Spanish healthcare system. Such an omission can result in suboptimal resource allocation. Methods: Analytical and retrospective observational study of a population of 1,691,075 people over the age of 15. The explanatory variables were age, sex and nursing process data, with total healthcare cost as the outcome variable. A bivariate analysis and a multiple regression were performed for the multivariate analysis. To improve prediction accuracy and account for non-linear relationships, the analysis was completed using two machine learning models. Results: 58% (n = 980,437) of the population presented some data from the nursing process, for individuals with an assessed pattern, the average cost was €2304.17 compared with €950.93 for those who had none; with a nursing diagnosis, the average cost was €1,666 versus €840 without it. Having created the best model for the analysis using neural networks and XGBOOST, an average coefficient of determination of R ²  = 21.45% was obtained. Conclusions: The variability in total healthcare costs can be explained in more than 21% of cases by the model created, including sex, age, and the information related to the nursing process. Implications for health policy: Demonstrating the influence of nursing care on total patient costs will facilitate its inclusion in management programs, promoting the use of nursing data in risk adjustment models and healthcare planning.