Health equity - Emerging Technologies Genomics and Evaluation

Artificial Intelligence, Genomics, and Health Equity Introduction Health disparities—differences in health outcomes between population groups—persist across most conditions and regions. Artificial intelligence (AI) and genomic medicine offer powerful tools to identify and reduce these inequities, but only when they are developed and deployed thoughtfully. This section explores how AI and genomics can either worsen or improve health equity, and what strategies are essential to ensure these technologies benefit all populations equally. How Artificial Intelligence Can Address Health Disparities Identifying Disparities Through Data Analysis Artificial intelligence excels at analyzing large datasets to uncover patterns that humans might miss. By processing electronic health records, administrative data, and population health information, AI systems can identify which populations experience worse health outcomes and why. This is a critical first step—you cannot address a problem you haven't measured. For example, AI can reveal that a particular medication is prescribed at different rates across racial or ethnic groups, or that certain communities have less access to diagnostic tests. These insights provide the evidence needed to develop targeted interventions. Suggesting Interventions to Reduce Disparities Beyond identifying disparities, AI can propose solutions. Machine learning algorithms can help predict which interventions are most likely to improve outcomes for specific populations, optimize resource allocation to underserved areas, or identify patients at highest risk of poor outcomes who need additional support. In this way, AI moves from diagnosis to action. The Critical Role of Community Engagement in AI Development Why Community Engagement Matters Here's an important principle: bias in AI systems usually comes from the data used to train them. If your training data comes primarily from wealthy, majority-population groups, the resulting AI system will perform poorly for other populations. Engaging diverse communities during AI development is the most effective way to prevent this problem before it starts. When representatives from marginalized groups—whether defined by race, ethnicity, geography, income, disability status, or other factors—participate in designing and building AI health tools, they can: Identify potential biases early Provide crucial context about their communities' needs Ensure the system addresses real problems that matter to them Build trust with the communities that will use these tools Ongoing Bias Assessment After Deployment Community engagement doesn't end when the AI system launches. Continuous monitoring for bias is essential because biases can emerge over time as the AI system encounters new data or is used in ways developers didn't anticipate. Regular audits of AI algorithms should specifically measure whether the system performs equally well across different population groups, and whether it has any "disparate impact"—that is, whether it produces systematically worse outcomes for certain groups even if unintentionally. Best Practices Summary Implementing AI for health equity requires: Include diverse representatives in design teams from the beginning, not as an afterthought Conduct regular audits that specifically measure algorithm performance across different population groups Establish feedback mechanisms so communities can report problems Remain transparent about how the AI system works and what data it uses Recent scoping reviews have reinforced that these practices—systematic community engagement and bias assessment—are what separate harmful AI systems from those that genuinely advance equity. Genomic Research and Health Disparities The Problem of Underrepresentation Genomic research—studies that examine genetic variation across populations—has historically excluded many groups. Rural communities, ethnic minorities, indigenous populations, and low-income groups are significantly underrepresented in genomic databases. When most genetic data comes from European ancestry populations, researchers have a much deeper understanding of genetic variation in those groups than in others. This creates a serious problem: genomic medicine relies on understanding genetic variation to identify disease risk, predict treatment response, and develop personalized treatments. When genetic variation data is incomplete for certain ancestry groups, the precision medicine approaches developed are literally less precise for those populations. Consequences: Health Inequities in Genomic Medicine Underrepresentation in genomic research directly translates into health inequities when these advances are applied in clinical practice. Consider sickle cell disease as a concrete example. Sickle cell disease is a genetic blood disorder caused by a mutation in the hemoglobin gene. It affects approximately 1 in 365 Black births in the United States, making it a significant health burden in Black communities. Recent advances in gene therapy offer the potential to cure or dramatically improve sickle cell disease. However, if communities of color are not included in the clinical trials and development of these therapies, the treatments may not work as well for these populations, or access may be limited due to cost and geography. This is why the World Health Organization proposed its 2024 Genomics Program: to promote international collaboration and develop strategies to overcome barriers that prevent underserved populations from participating in and benefiting from genomic research. Strategies for Equitable Genomic Medicine Three Key Goals To ensure genomic medicine reduces rather than worsens health disparities, three interconnected strategies are essential: 1. Recruit diverse populations into genomic research This means intentionally partnering with communities that have been historically excluded. It requires going beyond simply "offering" participation—it means building genuine partnerships, addressing barriers like transportation and childcare, compensating participants fairly, and ensuring that research benefits are shared with communities, not just extracted from them. 2. Increase genetic counseling services in underserved groups Genetic counseling helps people understand their genetic risks and the implications for their health and family planning. These services are concentrated in wealthy areas and are unavailable in many rural and low-income communities. Expanding access to genetic counselors—including those who speak patients' languages and understand their cultural contexts—is essential. 3. Ensure equitable access to genetic information and personalized treatments Discovering new genetic treatments is only the first step. Novel therapies like gene therapy for sickle cell disease are expensive and not yet available everywhere. Policies must ensure that once these treatments are proven effective, they are accessible to all populations, not just those who can pay out-of-pocket or live near major medical centers. Measuring and Monitoring Health Equity Progress Data Collection: The Foundation You cannot improve what you don't measure. However, many public health surveillance systems have a critical gap: they lack comprehensive data on race, ethnicity, language, and nativity. Without this information, disparities may go unrecognized. Improving data collection is therefore a necessary first step. Additionally, addressing health disparities requires understanding not just health conditions, but the social factors that drive them. Standardized screening tools for social needs—such as questions about food security, housing stability, transportation access, and social isolation—help identify patients at highest risk of poor outcomes. Community-Based Implementation Evidence shows that addressing health disparities requires more than clinical interventions; it requires community partnerships. Community health workers (often from the same communities they serve) build trust and improve uptake of health services among minority groups Patient-centered medical homes that integrate behavioral health services address the intersection of physical and mental health—something that matters particularly in communities with high rates of trauma and chronic stress Evaluation Metrics To know whether equity initiatives are working, health systems must measure: Disparity ratios: How do health outcomes compare across population groups? Are gaps widening or narrowing? Process equity: Are all groups equally likely to receive evidence-based treatments? Outcomes by population: Longitudinal studies that follow populations over time to assess whether policy changes actually improve health outcomes for marginalized groups <extrainfo> Without longitudinal studies spanning multiple years, it's difficult to determine whether interventions have truly changed health outcomes or whether apparent improvements are due to other factors. This is why rigorous evaluation is essential—it prevents wasting resources on interventions that don't actually work, and it identifies what does work so it can be scaled and replicated. </extrainfo> Key Takeaways Artificial intelligence and genomic medicine are powerful tools, but they are not automatically equitable. Both technologies can either amplify existing inequities or reduce them, depending on how they are developed and deployed. The difference lies in: Intentional inclusion of marginalized communities from the design phase onward Systematic assessment for bias and disparate impact Equitable access to both the technology and its benefits Rigorous measurement to ensure progress Health equity in the age of AI and genomics is not an afterthought—it requires fundamental changes to how we design, test, and implement these technologies.