Increase Trichomonas vaginalis detection based on urine routine analysis through a machine learning approach.

Area of Science:

• Urology
• Infectious Diseases
• Medical Diagnostics

Background:

• Accurate detection of Trichomonas vaginalis (T. vaginalis) using automated urinalysis remains a challenge.
• T. vaginalis infection is a significant public health concern requiring improved diagnostic methods.

Purpose of the Study:

• To propose and evaluate a machine learning (ML)-based strategy to enhance T. vaginalis detection rates in urine.
• To assess the cost-effectiveness of the ML-driven approach for T. vaginalis diagnosis.

Main Methods:

• Utilized urinalysis data from 2009-2013, including over 410,000 men and 428,000 women.
• Applied logistic regression, support vector machine, and random forest models to identify high-risk specimens.
• Confirmed T. vaginalis infection through microscopic examination of selected specimens.

Main Results:

• ML models achieved an area under the receiver operating characteristic curve >0.87 for women and >0.83 for men.
• The top 5% of high-risk specimens identified by the models showed Lift values above eight.
• Incremental cost-effectiveness ratios were $170.1 for men and $29.7 for women, indicating cost-effectiveness.

Conclusions:

• The proposed ML-based strategy significantly increases T. vaginalis detection rates from urinalysis data.
• This approach offers a cost-effective method for improving the diagnosis of T. vaginalis infections in both men and women.