Dissecting Racial Bias in Health Management Algorithms

Examining Racial Bias in Health Algorithms

Introduction to Algorithmic Bias in Healthcare

• Core Issue: Commercial prediction algorithms used in health systems to manage population health are found to exhibit significant racial bias.
  • These algorithms are deployed to identify patients who would benefit most from intensive care management programs.
  • The bias leads to systemic disadvantages for minority groups, particularly Black patients.
• Specific Manifestation: At any given algorithmic risk score, Black patients are considerably sicker than White patients.
  • This means that for the same risk score, Black patients have more chronic conditions, higher disease burdens, and often more urgent health needs than their White counterparts.
  • The algorithm misjudges their actual health needs, leading to delayed or insufficient intervention.
• Quantifiable Impact: Remedying this bias could dramatically increase the percentage of Black patients receiving additional care management support, from $17.7\%$ to $46.5\%$.
  • This suggests a substantial unmet need for care among Black patients that is currently overlooked by biased algorithmic systems.
  • Such an adjustment would lead to more equitable distribution of healthcare resources.
• Root Cause: The algorithm predicts health care costs rather than actual illness.
  • Due to unequal access to care, historical and systemic inequities, and biases within the healthcare system, less money is spent on Black patients for similar health needs.
  • This flaw in predicting costs as a proxy for health outcomes inherently penalizes groups that historically receive less healthcare spending.
  • The algorithm learns from past data, which reflects existing disparities, thus perpetuating them.
• Broader Principle: The choice of convenient but potentially flawed proxies for desired outcomes can embed and amplify existing societal biases.
  • In this case, using cost as a proxy for health risk is problematic because costs are not solely determined by illness severity but also by healthcare utilization, which is influenced by socioeconomic factors and systemic biases.
  • This highlights the importance of carefully selecting metrics and understanding their downstream ethical and social implications to avoid perpetuating disparities.