Technology Should Help Avoid, Not Amplify, Bias in Healthcare

A recent study in Science found that an algorithm widely used in healthcare to determine which patients require more care reflected racial bias. The model relies on individuals’ previous health care spending as a factor to predict which people have more complex health needs and will need more care in the future. As a result, the algorithm assumes that people who spend less on healthcare must be healthier, when they actually may be spending less on healthcare because they cannot access care or afford to prioritize healthcare spending over more immediate household expenses.

At Castlight we recognize that data and predictive analytics have the power to transform healthcare, making it easier for everyone—regardless of race or socioeconomic factors—to get the right care, at the right provider, at the right price, at the right time. Unfortunately, the Science research demonstrates just how challenging it is to do that in a thoughtful way. By basing predictions of future healthcare needs on past use, this algorithm amplified the existing disparities built into our healthcare system.

Conversely, Castlight’s predictive analytics model is based on identifying gaps in care according to the best evidence, rather than previous spending. We are able to avoid the issues highlighted in the Science article by using evidence-based, clinical guidelines for care.

When Castlight sees a user is not getting care that meets those guidelines, we make a recommendation based on that gap. For example, according to the U.S. Preventive Services Task Force, women over the age of 50 should get a mammogram every two years. Castlight’s platform can identify women who have not had a biennial mammogram, help them find a convenient and high-quality provider, and even assist them with booking an appointment.

It’s essential to consider the role that social determinants of health can play in a patient’s decision-making and eventual health outcomes, and we are constantly innovating to ensure our predictive models take these factors into account as well.

At Castlight, we have invested in building products that identify users who need care and provide them targeted recommendations that enable them to get the highest quality care at the best price for them. By leveraging our multifaceted dataset that effectively encompasses the whole person’s healthcare journey, we are able to help more people access and afford the care they need.