# Machine Learning Used in Communicable Disease Control: A Scoping Review

## TL;DR

This paper reviews how machine learning is used to control communicable diseases and highlights the need to address algorithmic biases in these models.

## Contribution

The study systematically identifies ML applications in communicable disease control and emphasizes the under-addressed issue of algorithmic bias.

## Key findings

- Most studies focused on SARS-CoV-2, with fewer on malaria, HIV, and tuberculosis.
- Only a small percentage of studies addressed or mitigated algorithmic bias in their models.
- ML is increasingly used for disease surveillance, but bias mitigation remains a priority for equitable outcomes.

## Abstract

Communicable diseases continue to threaten global health, with COVID-19 as a recent example. Rapid data analysis using machine learning (ML) is crucial for detecting and controlling outbreaks. We aimed to identify how ML approaches have been applied to achieve public health objectives in communicable disease control and to explore algorithmic biases in model design, training, and implementation, and strategies to mitigate these biases.

We searched MEDLINE, Embase, Cochrane Central, Scopus, ACM DL, INSPEC, and Web of Science to identify peer-reviewed studies from 1 January 2000, to 15 July 2022. Included studies applied ML models in population and public health to address ten communicable diseases with high prevalence.

28,378 citations were retrieved, and 209 met our inclusion criteria. ML for communicable diseases has risen since 2020, particularly for SARS-CoV-2 (n = 177), followed by malaria, HIV, and tuberculosis. Eighteen studies (8.61%) considered bias, and only eleven implemented mitigation strategies.

A growing number of studies used ML for disease surveillance. Addressing biases in model design should be prioritized in future research to improve reliability and equity in public health outcomes.