In a rapidly evolving digital age, machine learning (ML) is transforming public health, particularly in addressing non-communicable diseases (NCDs) such as diabetes, heart disease, and cancers. A new scoping review sheds light on the current applications of ML in population health and the critical issue of algorithmic bias, raising important questions about equity in healthcare technology.

Key Findings

• Prevalence of Studies: Out of over 27,000 studies on ML in public health, only 65 met the criteria for this review, revealing a significant gap in research focused on NCDs.
• Focus on Type 2 Diabetes: The majority of studies concentrated on type 2 diabetes, highlighting its prominence among NCDs.
• Risk Modeling Dominance: Risk modeling was the most common application of ML, suggesting a focus on predicting health outcomes rather than addressing biases in data collection and model training.
• Limited Bias Mitigation: Very few studies addressed algorithmic bias comprehensively, with the majority focusing on mitigating sex-related biases.

"Algorithmic biases can enter models through incomplete training data or datasets that are not representative, leading to inaccurate predictions for diverse populations," - Lead Author.

---

Why It Matters

The increasing reliance on ML in healthcare systems raises significant concerns about equity and fairness. As ML models are trained on historical data, they may perpetuate systemic biases present in that data, potentially disadvantaging marginalized communities. Understanding and addressing these biases is crucial for ensuring that technological advancements in healthcare benefit all populations, not just a select few.

The review highlights the need for equitable ML practices, particularly as NCDs disproportionately affect individuals in low- and middle-income countries (LMICs). According to the research, NCDs are responsible for a staggering number of deaths worldwide, making it imperative to deploy ML tools that do not exacerbate health disparities.

---

Research Details

The scoping review, conducted by a team of researchers from various institutions, mapped the landscape of ML applications in public health, specifically targeting NCDs. The methodology involved a comprehensive search of indexed literature across multiple databases, including Medline and Embase, up to March 2022.

Methodological Highlights:

• Comprehensive Database Search: Over 27,310 studies were initially identified, but only 65 were included for detailed analysis based on specific criteria related to NCDs.
• Diverse NCD Focus: The review categorized studies based on the type of NCD addressed, data sources, technical approaches, and potential biases.
• Algorithmic Bias Examination: The research sought to assess not only the application of ML but also the biases that may arise during the design, training, and implementation of ML models.

---

Looking Ahead

The implications of this research are significant for future studies and the application of ML in healthcare. The authors emphasize the need for:

• Broader Research Scope: Future research should expand beyond NCDs to include communicable diseases and the transferability of ML models in LMICs.
• Comprehensive Bias Mitigation Strategies: More robust methods should be developed to identify and mitigate biases beyond just sex-related issues, including those related to race and socioeconomic status.
• Guidelines for Equitable ML Use: Establishing guidelines for the equitable use of ML in public health could enhance health outcomes across diverse populations.

"Our findings can guide the development of guidelines for the equitable use of ML to improve population health outcomes," - Co-Author.

As we progress further into the era of data-driven public health, addressing these challenges will be essential. Prioritizing equity in ML applications will not only improve health outcomes but also foster trust in technology that is increasingly shaping our healthcare landscape.

By understanding the current landscape and potential pitfalls of ML in public health, we can harness its power for good while ensuring that it serves all segments of society fairly and justly.