A personalized prediction model for distinguishing between asymptomatic bacteriuria and symptomatic urinary tract infections in patients with type 2 diabetes mellitus using machine learning.

BACKGROUND

Patients with type 2 diabetes mellitus (T2DM) have an increased susceptibility to urinary tract infections (UTIs), caused by uropathogenic Escherichia coli (UPEC). Asymptomatic bacteriuria (ASB) is a significant contributor, but lots of patients are difficult to distinguish. Distinguishing between ASB and symptomatic UTIs can greatly assist clinicians in rational use of antimicrobials.

METHODS

Patients with T2DM and UTIs caused exclusively by UPEC were recruited from the Second Hospital of Tianjin Medical University between 2018 and 2023. Demographic and clinical data were systematically collected for these patients through a retrospective electronic chart review, in accordance with the inclusion and exclusion criteria. We utilized this dataset as training set to develop an ASB predictive model called ASBPredictor.

RESULTS

A total of 337 cases were collected, comprising 158 cases (46.9%) of ASB and 179 cases (53.1%) of symptomatic UTIs. Based on the optimal predictive model, ASBPredictor exhibited a remarkable level of precision, achieving an area under the curve score of 0.82. The identification of ASB is influenced by several crucial factors, including urinary bacteria, urinary white blood cell clusters, C-reactive protein, alanine aminotransferase, glucose, gamma-glutamyl transpeptidase, sodium ions (Na+), and eosinophils.

CONCLUSION

The ASBPredictor is an accurate, efficient, and reliable tool that helps doctors differentiate between ASB and symptomatic UTIs. This precise differential diagnosis has the potential to enhance the quality of antimicrobial prescribing.