Exploring the Intersection of AI, Bias, and Value-Based Care
Moving to value-based care will require providers and payers to invest in high-quality data analytics, but not at the cost of fairness and equity.
Source: Getty Images
By - Many healthcare experts consider value-based care as a key step toward improved health outcomes and lower costs, but adoption has been slow.
Research from Insights by Xtelligent Healthcare Media indicates that organizations moving to value-based care benefit from robust data analytics capabilities and support to fuel the move away from fee-for-service reimbursement. These efforts can be bolstered through the use of artificial intelligence (AI) and machine learning (ML), according to Dr. Marzyeh Ghassemi, a principal investigator at MIT’s Jameel Clinic.
Listen to the full podcast to hear more details. And don’t forget to subscribe on iTunes, Spotify, or Google Podcasts.
Ghassemi leads the “Healthy ML” lab at MIT, and her work is focused on applying machine learning to improve healthcare in ways that are robust, private, and fair. AI and ML have seen increasing use across the healthcare industry in recent years, rising alongside conversations about health equity and value-based care goals. But are these goals at odds?
Ghessemi shared during an episode of the Healthcare Strategies podcast that AI has significant potential for use in value-based care without sacrificing health equity. However, providers, payers, and other stakeholders need to be aware of how they are using data and analytical models.
“In healthcare, we often use proxies for health that are not appropriate and can be problematic. When you use, for example, the level of patient visit, as a proxy for whether they may need a checkup in the next year, then you disadvantage minority patients who may have less access to care or are concerned about making a copay,” Ghassemi explained.
These proxies, she continued, can cause stakeholders to miss large portions of a patient’s story. If this is repeated across a population, it can introduce bias. Stakeholders can try to prevent and address bias in multiple ways, but Ghassemi noted that this varies largely by stakeholder.
These biases and harm can cause some of the same care gaps and health disparities that value-based care models seek, in part, to address. But Ghassemi argues that AI and ML can advance health equity and improve value-based care by giving providers a tool to combat burnout.
“[Clinicians] want to have some sort of stake in the outcome of their patient and want them to succeed. But they're overloaded and overwhelmed with all of this information and all of these different tasks that they have to do,” Ghassemi explained.
“If we can have systems that have been intelligently trained not to naively replicate the things that we've seen clinical systems do, but to recommend improved care—what we'd actually like to do—then I think that is the best of both worlds,” she concluded.
