Achieving care equity and reducing health outcome disparities are tightly woven into the missions of health plans and managed care organizations. Reaching those overarching goals, however, is challenging given the numerous social determinants of health (SDoH) obstacles that many members face across rural, urban and suburban communities.
These SDoH barriers may include difficulties obtaining transportation, affording out-of-pocket responsibilities, understanding their condition(s), or even something as simple as finding an available provider within their community. As a result, some members may travel to their nearest emergency department, even if their health complaint could have been resolved through a virtual visit with a primary care provider. Conversely, other members who truly require an emergency intervention may further delay seeking care until a catastrophic event occurs.
Filling Care Gaps
Generative AI offers an enormous opportunity to help solve these SDoH challenges – and improve care equity overall – by helping members access the right level of care when they need it, regardless of where they live. Evidence has shown that powerful AI tools gather and analyze more data faster than any clinician to safely help direct patients to the most appropriate level of care based on their personalized health profile and status.
By leveraging AI to stratify and triage patients to the proper acuity level when they need it, health plans can reduce the burden on their care management teams and enable them to devote more time to high-risk, high-need member outreach. More importantly, an evidence-based, AI-enabled triage system can help payers rapidly stratify their members and fill care gaps by directing them to access in-network providers that will help them achieve optimal outcomes and drive down care costs for all stakeholders.
Enabling Safe, Effective Care
Although more widely adopted in Europe and Southeast Asia, AI solutions have experienced a more gradual buy-in among provider and payer organizations across the U.S. to support direct patient care. So far, the early adopters of generative AI are using the technology for creating documentation and other content that supports revenue cycle processes or that creates physician notes and correspondence for patients’ charts.
Direct patient encounters with AI to date have mostly involved online chatbot assistants who guide patients through standardized clinical questionnaires before a telehealth appointment. This narrow utilization barely scratches the surface of what AI can deliver – not in a year or five years down the road – but today.
For example, a May 2024 study published in Frontiers in Public Health that analyzed over 3 million patient interviews with an AI-based virtual triage engine found that about 12,000 users exhibited symptoms highly indicative of conditions such as myocardial infarction, stroke, asthma, pneumonia, and pulmonary embolism. Concerningly, 76% of patients with myocardial infarction symptoms reported they had not planned to seek emergency treatment.
Overall, more than half (53.5%) of high-risk patients indicated they were unlikely to seek emergency care and one-third (33.5%) of patients overall had not planned to consult a healthcare professional at all before completing virtual triage. Authors concluded that 2 in every 1000 virtual triages performed could lead to life-saving decisions.
Surpassing the Gold Standard
However, achieving care equity and reducing outcome disparities will not result from simply expanding access to more high-acuity care. On the contrary, driving more patients to the emergency department could contribute to overcrowding, longer care delays, unnecessary hospitalizations, and potentially care avoidance if patients believe their complaint is not serious and want to avoid wasting hours of their day in a packed hospital.
This is where AI algorithm training is so crucial. A poorly trained algorithm could potentially drive unnecessary emergency department utilization or lead to a lack of urgency that could result in patient harm. To assess AI’s ability to accurately identify acuity levels, researchers tested a virtual triage tool’s performance against live clinicians using the Schmitt-Thompson rule-based protocols, considered the gold standard for patient triage.
Study results, published in the Journal of Hospital Administration, show that the AI-based tool performed better overall, resulting in accurate triage decisions nearly 77% of the time, compared to Schmitt-Thompson’s 72% rate. Although the virtual AI triage tool tended to err on the side of caution, (for example, recommending higher-acuity care for pediatric cases), its “over-triage” rate of 14% was lower than the 19% rate for Schmitt-Thompson. Researchers concluded the virtual triage tool’s ability to collect and analyze four times as much patient symptom information in less time supported this more accurate, data-driven triage advice.
“Sticky” Member Engagement
Health plan member-facing mobile apps and online portals typically have low adoption rates. Offering such an AI-based virtual triage tool, however, could help make payer apps more “sticky,” meaning that members may repeatedly use it to navigate their and their family’s health concerns, especially given the increasing out-of-pocket responsibilities for covered individuals.
Such a tool can also promote care equity by activating members in provider-shortage areas or who face SDoH obstacles associated with delayed or avoided care. An easy-to-use online symptom assessment tool that urges members to visit the emergency department or an urgent care clinic can drive action because the advice comes from a trusted source: the member’s health plan.
An AI-driven triage tool can likewise support payer outreach by quickly identifying the highest-risk members based on claims data as well as information shared with the triage application. Payer care managers can then help members facing SDoH challenges coordinate more convenient and/or affordable high-quality care that could end up preventing a critical and costly health event down the road.
Above all, while AI is still no replacement for care from a human clinician, it appears that it can help members better access the most appropriate care in less time and at a lower cost. Increasing access while saving members’ time and money are crucial factors in achieving care equity and helping all patients experience the optimal outcomes they deserve.
Amanda Bury
Amanda L. Bury, MS is the Chief Commercial Officer of Infermedica, an AI-powered healthcare platform that helps providers deliver efficient, safe, and reliable care to their patients. Bury brings more than 15 years of experience connecting health systems and insurers with cutting-edge technology. Most recently, Bury served as the Vice President of Global Strategic Alliances at TeraRecon, the leading provider of medical imaging visualization, AI development, and interoperability technologies. Prior to that, she spent nearly three years as Senior Director of Channel Development & Strategy Kyruus. Bury has also become a leading voice in the healthcare technology community, presenting at various industry conferences and events, including SHSMD, HCIC, and HIMSS.