The Hidden Reasons Why Preventive Healthcare Requires AI
Preventive healthcare technologies, powered by AI ML, can change the face of our existing medical systems. Just weeks after he’s released from the hospital, a man is rushed back to the emergency room in an ambulance. The reason? He doesn’t have a car. He lives alone, couldn’t pick up his prescription from the pharmacy, and didn’t have a ride to his follow-up appointments. No one knew.
When his condition inevitably deteriorated, he had no choice but to call 911.
Healthcare systems once had little reason to prevent readmissions like these. Historically, under the traditional fee-for-service model, providers are paid for every service they provide — meaning there’s little incentive for keeping folks out of the hospital. Planning for potential barriers to health isn’t a priority in this environment. Fortunately, that’s quickly changing as the industry shifts to value-based care.
The Challenges of Patient Engagement in Value-based Care
Under value-based care, reimbursement is directly tied to patient outcomes. Providers now lose money if they fail to improve patient outcomes or prevent avoidable patient harm. This paradigm shift is tipping the economics of healthcare in favor of preventive care over-reactive care. Under value-based care, health systems now have a financial incentive to proactively engage patients at risk of poor health outcomes — before they need more intensive care.
Unfortunately, the proactive engagement part is where health systems are still struggling.
Over the past 18 months we have seen a flurry of patient engagement, communication and outreach platforms come to market. All of them with the intention of enabling better member engagement. The new buzzword is “care orchestration,” also intended to enable value-based care and patient engagement.
The challenge is, a platform alone does not help a case manager who is likely responsible for managing the care of hundreds of patients, and who every day must decide which individuals they should spend their limited time checking in on. Predictive models can help, but their usefulness only goes so far.
A recent McKinsey report found that care managers are missing up to 95% of their potential impact. The problems are many: care management programs often fail to target the right individuals, and even when they do, they often fail to successfully contact them or engage them in a way that changes their behavior.
Consider unplanned hospital readmissions, a key quality metric that determines how much hospitals are reimbursed under value-based care models. Last year, the Centers for Medicare & Medicaid Services (CMS) fined half of America’s hospitals for readmitting too many patients.
Yet according to McKinsey’s report, only 6% of patients at risk for readmissions changed their behavior after engaging with care managers. What’s more, as many as 80% of patients at risk are never even identified due to constrained resources and limited predictive models.
The good news, however, is that advances in machine learning and prescriptive artificial intelligence (AI) are both improving the accuracy of predictive models and enabling care managers to use their time with patients more efficiently.
Prescriptive AI Focuses Patient Engagement on the Right Questions
One of the biggest limitations of predictive models is that they can only predict. They rank patients by their level of risk for adverse outcomes, usually based on historical data. However, they provide care managers no guidance on why these patients are at risk, or more importantly, the specific levers that can be pulled to change their risk trajectory.
As a result, care managers are either spending much of their time poring over the patient record themselves to identify the most relevant factors to discuss with the patient — or making decisions on which programs to align the patient with based on minimal context.
To save care managers time, prescriptive AI can analyze patients’ data to identify the factors driving their holistic risk — overlaying their medical data with an expanded data set including behavioral, environmental, and other social determinants. This is probably the greatest value of AI: its ability to consider nearly unlimited data sets quickly and with purpose.
A patient’s risk trajectory is influenced by more than the medical data stored in the electronic health record: It’s estimated that up to 80% of health outcomes are linked to social determinants of health. Social determinants of health (SDOH) are all the ways our lifestyle and environment influence our health. Not having a car, as with the patient I mentioned earlier in the article, certainly counts as a social determinant. So does living alone, living in a food desert, or having low medical literacy.
And yet, care teams have almost no visibility on these risk factors outside of what their patients tell them. By analyzing external data, such as the databases compiled by the US Census and other government agencies, AI can provide insight on these SDOH risk factors, empowering care managers to focus SDOH into their discussions with patients as well as their care plans. Here too, AI can provide guidance.
Going one step further than identifying risk factors, AI can analyze historical data and connect the dots on what interventions and engagement strategies worked to improve outcomes for patients with similar risk factors, and provide care managers with patient-centric and clinically-validated recommendations. These insights enable care managers to focus their time with patients on the right questions, resulting in more informed — and ultimately effective — conversations.
To return to my initial example, had care managers known their patient might have struggled with access to transportation, or that he lived alone, could they have prevented his readmission? If care managers had these insights at their fingertips, perhaps they could have helped him enroll in mail-order pharmacy deliveries, or scheduled him for telehealth follow-up appointments. Maybe they could have spent more time educating the patient on the importance of medication adherence, knowing he lacked the social support to help with this essential routine.
Taking a more holistic view of the patient, and all the internal and external factors driving their risk, is the key to unlocking the preventive approach to healthcare necessary for success with value-based care. We can no longer wait until patients are having a medical emergency to take action. With its vast data processing power, AI can provide care management teams with the guidance they need to proactively engage patients at risk — in a way that effectively prevents them from needing more intensive (and costly) care in the future.