After you define your health equity goals, secure them by eliminating hidden bias in risk stratification.
ACO REACH is breaking new ground by making the improvement of health equity an explicit goal for a value-based care program. But the greatest threat to improving health equity isn’t even mentioned once in the ACO REACH Request for Applications: bias.
Despite any participant’s best intentions, bias is pervasive in the risk stratification tools that are a hallmark of value-based care.
If undetected, hidden bias will exacerbate your population’s health disparities—exactly the wrong outcome when implementing your Health Equity Plan.
Meeting your health equity goals will require identifying and eliminating bias in your risk stratification tools to ensure disparities are not inadvertently worsened.
For examples of hidden algorithmic bias affecting health outcomes, look no further than Obermeyer et al. (2019), a seminal paper on algorithmic bias in healthcare. It uncovered that an algorithm managing 70 million lives inappropriately used healthcare utilization costs as a proxy for future health needs and systematically produced risk scores that disadvantaged Black people.
To make matters worse, many risk stratification tools are proprietary “black box” algorithms that prevent users from addressing the presence of bias, assuming bias can be detected.
To secure your health equity goals, you must use tools designed to detect and mitigate bias in your risk stratification approach before it’s deployed live and affects resource allocation decisions.
Optimal solutions will enable you to analyze model performance across subgroups, provide dashboards and reports that detail sources of bias, and be deeply explainable and auditable.
While defining health equity goals is a prerequisite to reducing health disparities, the single most important step you can to secure those goals is to ensure that hidden bias in your risk stratification methods is identified and eliminated.
