Population health has been an elusive goal in the transformation to value-based care. What, after all, does population health look like? How will we know when we get there? The answer, of course, is that population health is not a destination. It’s a moving target whose goals, once achieved, will be quickly replaced with new ones, just out of reach.
If ever there was a problem tailor-made for Artificial Intelligence (AI), it is population health management. Artificial Intelligence is uniquely capable of deriving actionable insights from the large, complex datasets required for population health management. It uncovers unseen patterns as well as reveals subtle predictive trends that traditional analytics platforms may miss.
Where we are now
Population health management is about identifying individual patients and groups of patients who are most likely to require interventions to stay healthy, and then targeting outreach to them at the optimal time to achieve long-term favorable outcomes. An ideal approach would include a continuous closed-loop process that gathers data from a vast number of disparate sources, assesses and updates care plans, records and reports outcomes, and generates new goals.
However, current approaches to population health management are typically piecemeal, with significant breakpoints and silos between different initiatives, that enable little transparency and visibility across patients’ holistic journeys.
Many analytic solutions tackle components of population health, but have not created an operational system of population health management.
But seeing the big picture, uncovering insights and deploying them efficiently and uniformly across an organization—these are core competencies of embedded AI solutions. In fact, population health may represent the biggest (largely untapped) AI opportunity within healthcare.
Capabilities that matter
Ask five people to define AI, and you’re likely to get five different answers. To understand the unique ways in which AI can address persistent population health challenges, we must define the essential capabilities necessary to find the hidden patterns within vast datasets that traditional analytics miss.
Discovery
Most healthcare analytics applications depend on queries from clinicians seeking to validate hypotheses about their patients and treatment options. Across many different types of data that make up population health, there are countless patterns and trends that won’t be uncovered because clinicians failed to ask the right question. However, AI solutions that include “unsupervised learning” open a whole new avenue to discover unseen gaps in care and pockets of best practices. Unsupervised learning considers all the data and all the possibilities within that dataset to detect the patterns, groups or anomalies that elude traditional approaches. Using their own systems of records, including EMRs, financial data, patient-generated data, and socio-economic data, healthcare organizations that deploy AI can automatically discover groups of patients that share unique combinations of characteristics.
Predictions
Key to the promise of AI for population health is the ability to predict the future with high accuracy. Which patients’ health status will worsen the quickest? Which interventions are most likely to slow the progression of this type of patient’s disease? These are the questions clinicians ask themselves every day, and the predictive power of AI can help them answer those questions with significantly better accuracy, based on existing data within the hospital or health system, across clinical and non-clinical data.
Justification
Effective population health management solutions require the ability to peel back layers of the onion to identify drivers of risk and which patients are impactable. What drives risk differences may not be a particular clinical diagnosis, but could be a combination of factors, such as how well the patient received treatment, lifestyle factors, and the quality of their health insurance. AI must justify its predictions, discoveries, and actions so clinicians feel confident to act upon its recommendations. AI solutions must always provide a “why” to back up decisions.
Action
For population health to keep moving forward toward more ambitious goals, any AI solution must provide actionable information that guides and augments human decision-making. Whether it is a recommended care path or a detailed risk profile, the patterns uncovered by AI must direct clinicians toward better outcomes. These actions may redefine patient “populations” in new ways, according to patterns uncovered by AI – whether by their capacity to engage, their dietary habits or their network of psycho-social support.
Learning
The biggest differentiator between AI and traditional analytics solutions is that AI platforms “learn” to improve predictions over time. As more and more data is analyzed, the technology learns from these complex data points to improve predictions over time. Whether it be claims, medical records, or socio-economic data, AI taps into these data points to generate more accurate predictions that continuously improve.
AI’s Promise for Population Health
By surfacing new trends and patterns, AI can help healthcare organizations design interventions and tailor programs to engage unique patient sub-populations. These care paths, with a holistic understanding of the patient gleaned from disparate datasets, will inform clinical pathways that can improve the health trajectory of the riskiest patients with high touch efforts while simultaneously informing less expensive, lower touch preventive programs targeted at rising risk sub populations. AI also can uncover patterns of best practices to allow organizations to develop and continuously refine strategies to keep patients with good outcomes on the right track.
Prashanth Kini is vice president of product management for healthcare at Ayasdi (www.ayasdi.com), a leader in machine intelligence software and a pioneer in enterprise-class intelligent applications.
###