The application of machine learning in the evaluation of urinary tract infections incidence in patients in a Nursing and Treatment Facility.

Abstract

Background: Urinary tract infection (UTI) is a serious problem in the healthcare system. It is caused by bacteria from the gastrointestinal tract. The risk factors that impact the UTI incidence include administration of certain drugs (flozins), sex, use of urinary catheter, and diabetes. This is a retrospective study of the records of 76 patients from a Nursing and Treatment Facility at County Hospital in Drezdenko (Poland) aimed to assess the factors that may have an impact on the incidence of UTI.

Methods: The following factors were taken into consideration: dapagliflozin administration (yes/no), diabetes (yes/no), sex (male/female), kidney failure (yes/no), and use of urinary catheter (yes/no). The impact of the above variables on the UTI incidence was estimated using multivariate regression analysis and machine learning, such as logistic regression, artificial neural networks (ANN), and decision trees (recursive partitioning).

Results: As revealed by the multivariate regression analysis, UTI was significantly affected only by dapagliflozin administration. The machine learning techniques showed greater sensitivity in detecting significant factors - dapagliflozin administration was identified as the most important one. Moreover, the logistic regression analysis also indicated sex (female). In the case of ANN and decision tree, the other significant factors, besides dapagliflozin intake, in the model were the use of a urinary catheter, sex (female), diabetes, and kidney failure (in descending importance). The variables were listed in the same order of descending importance for both the ANN and the decision tree.

Conclusions: In the case of catheterized patients, the administration of flozins should be cautiously approached, as should the catheterization of patients taking flozins.

Clinical trial number: Not applicable.