JMIR Publications

ABSTRACT

Background:

The early prediction of antibiotic resistance in patients with urinary tract infection is important to guide appropriate antibiotic therapy selection.

Objective:

In the present study, we aimed to predict antibiotic resistance in patients with urinary tract infection. Additionally, we aimed to interpret the machine learning models we developed.

Methods:

We used admission, diagnosis, prescription, and microbiology records of patients who underwent urine culture tests in Yongin Severance Hospital, South Korea. We developed 5 sub-models to classify urinary tract infection pathogens as either sensitive or resistant to cephalosporin, piperacillin/tazobactam, trimethoprim/sulfamethoxazole, fluoroquinolone, and carbapenem. To analyze how each variable contributed to the machine learning model’s predictions of antibiotic resistance, we used the SHapley Additive exPlanations method. Finally, we proposed a prototype machine learning based clinical decision support system to provide clinicians the resistance probabilities for each antibiotic.

Results:

The area under the curve values ranged from 0.710 to 0.826 in the training set and 0.642 to 0.812 in the test set for predicting antibiotic resistance. The administration of drugs before infection and exposure time to these drugs were found to be important variables for predicting antibiotic resistance.

Conclusions:

The study results demonstrated the potential of machine learning to predict antibiotic resistance in patients with urinary tract infection. Machine learning can assist clinicians in making decisions regarding the selection of appropriate antibiotic therapy in patients with urinary tract infection.