Accuracy of symptom checker for the diagnosis of sexually transmitted infections using machine learning and Bayesian network algorithms
2024

Using Machine Learning to Diagnose STIs

Sample size: 10680 publication 10 minutes Evidence: moderate

Author Information

Author(s): Soe Nyi Nyi, Towns Janet M, Latt Phyu Mon, Woodberry Owen, Chung Mark, Lee David, Ong Jason J, Chow Eric P.F., Zhang Lei, Fairley Christopher K.

Primary Institution: Melbourne Sexual Health Centre, Alfred Health

Hypothesis

Can machine learning models outperform Bayesian networks in diagnosing sexually transmitted infections (STIs) using symptom data?

Conclusion

Machine learning models can predict STI diagnoses with reasonable accuracy, outperforming Bayesian models for certain conditions.

Supporting Evidence

  • Machine learning models achieved AUC scores ranging from 0.81 to 0.95 for male conditions.
  • The models performed significantly better than Bayesian models for male balanitis, molluscum contagiosum, and genital warts.
  • The study included a large sample size of 10,680 participants, enhancing the reliability of the findings.

Takeaway

This study created a computer program that helps people figure out if they might have a sexually transmitted infection based on their symptoms, and it works pretty well!

Methodology

The study used prospectively collected data from STI patients to develop and evaluate machine learning models against Bayesian networks.

Potential Biases

Self-reported data may introduce bias based on users' understanding and interpretation of symptoms.

Limitations

The study was conducted at a single center, may not be generalizable, and excluded certain populations and conditions due to insufficient data.

Participant Demographics

The study included 6,162 men (median age 30) and 4,358 women (median age 27) presenting with STI symptoms.

Statistical Information

P-Value

p<0.05

Confidence Interval

95% confidence intervals reported for AUC scores

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1186/s12879-024-10285-4

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication