Much has been written about the demise of the value based care movement in light of the cancellation of certain CMS bundled payments and the perilous state of the ACA. It will take more than that to halt or reverse the steady progress of recent years. Over 90% of large employers with greater than 5,000 employees are now estimated to be self-insured. Such self-insured employer “payers” partner with healthcare services companies (Third Party Administrators) and provider groups to limit exposure through some form of shared risk. It is not just the Fortune 500. Smaller employers are also seeing savings opportunities with self-insured plans and emerging innovations such as level-funded plans.
What underlies these risk sharing plans is an incredibly complex set of interactions, incentives, penalties and variability. In short – it is a problem for artificial intelligence.
To manage risk in this world an entity must balance the economics of care with the contract they have with their employees and/or members. This is requires the risk-bearing entities to execute across multiple dimensions – often far removed from their core competencies as a company. The payoff, however, for success is considerable:
- They can gain a holistic, unbiased and prospective assessment of the at-risk population across multiple risk vectors: cost trajectories, utilization trajectories, disease trajectories
- They can develop a clear understanding of the nuanced clinical and non-clinical drivers of risk for each subpopulation – thereby informing targeted interventions
- They can match patients’ predicted journeys to targeted treatment regimens informed by the drivers of risk and appropriate for that stage of the member’s journey. Examples include localized clinical pathways for pre-acute, acute and post-acute episodes of care
- They can eliminate unnecessary care variation for expensive and highly variable episodes of care such as joint replacements, cardiac surgery, congestive heart failure admissions/readmissions
From the Population Risk Stratification perspective, AI solutions can examine a virtually unlimited set of member attributes contributing to the health of individuals, detect systematic patterns predictive of health degradation and provide high resolution visibility stratification of at-risk populations.
From a Clinical Variation Management perspective, AI solutions can examine years of local provider clinical data to surface treatment regimens that have yielded the best outcomes at the best cost within specific member/employee populations.
Let’s look at how this works in practice:
An AI powered population risk stratification application would be employed to discover a subpopulation of members with recurring and escalating orthopedic issues like back and joint disorders.
The application would also be able to predict trajectories towards joint replacement surgery. By proactively identifying such a subpopulation, the risk bearing entity can act to staff a program designed to provide pre-emptive therapy to stave off surgery.
For those patients that ultimately need elective surgery, the healthcare organization can guide patients to local surgical facilities adherent to localized surgical pathways which the AI solution has discovered to have empirically yielded the lowest length of stay and the fastest return to normal ambulation.
This simple example can easily be extended.
For example, by analyzing data across the continuum of care, this AI-generated clinical pathway can identify, prior to admission, characteristics predictive of opioid addiction and proactively identify post-acute interventions to reduce the possibility of addiction.
Going further still, the AI solution could, for more chronic conditions, distinguish a subpopulation of early, relatively healthy chronic condition patients from sub-populations of complex co-morbidities.
Relatively healthy subpopulations would be matched to proactive wellness regimens – likely powered by our ever expanding arsenal of digital health tools. Complex chronic patients, on the other hand, would be managed with higher touch, care manager driven regimens focused on ensuring adherence to medication and regular exams.
Over time, these AI systems would continue to monitor the population’s clinical and non-clinical data to learn emergent patterns of risk escalation / de-escalation and emergent innovations in clinical practice as the local healthcare ecosystem adapts to the evolving risk landscape while adjusting its projections on the evolving patient population.
Powerful stuff – and available today.
This is not to suggest this is easy, it is not. The technology is but a single piece, there are organizational requirements, organizational change, patient education, regulation and a myriad of other steps.
Still the opportunity for healthcare is massive – to both see risk and to manage it. This is the intersection of clinical variation management and population risk stratification.
With increasing financial risk and the opportunity offered to manage the risk through better, more informed understanding of the vast amounts of digital health data available on members, risk bearing entities are already starting to employ intelligent systems to drive intelligent healthcare investments. Join us on that journey.