Modeling Impactability with Predictive Analytics Can Drive Better Outcomes in Whole-Person Care

The following is the guest article by Michele Kratz, Vice President of Operations at HGS AxisPoint Health.

Humans are complex systems, but we’re no more than the sum of our parts — all of our parts. Our overall health is determined not only by the health of our physical systems, but also by our mental and emotional health, and by a multitude of outside variables that shape our daily lives: the environmental, cultural, and social factors known as social determinants of health (SDoH).

To more effectively address the multifaceted aspects of an individual’s health, as well as their SDoH, payers and providers alike are beginning to embrace the holistic approach of a whole-person care model. Whole-person interventions can improve health outcomes, reduce costs, and increase patient satisfaction by identifying and addressing the root causes of noncompliance and unhealthy behaviors — which often stem from undiagnosed or underdiagnosed behavioral health, or from substance-use issues driven by SDoH.

So how can payers and providers identify and account for these root causes? One key strategy is the use of predictive analytics to model what is known as “impactability.”

Modeling Impactability for Better Whole-Person Care

The concept of impactability comprises how an individual is likely to respond to interventions, and which care-management strategies will effectively engage them, and measuring their impactability can help determine what a clinician prioritizes during whole-person interventions. To properly understand and account for impactability, it is useful to come up with a score or number that shows at a glance where an individual sits on the impactability scale. That’s where analytics comes in.

Feeding data into a predictive analytics model can generate propensity scores or predictive values for individuals within a member population, giving payers and providers an impactability measurement for a patient or member. As more data goes into the predictive analytics engine, an impactability measurement becomes more accurate, and more multidimensional.

A base-level impactability measurement uses clinical marker data, which can include:

  • Diagnosis data
  • Payer claims data for physical and behavioral health care, including prescriptions claims
  • Self-reported data from health risk assessments
  • Admit, discharge, transfer (ADT) data from health care facilities
  • Prior authorization data
  • Electronic medical record (EMR) data
  • Care cost data

The next dimension brings in SDoH data: education level, marital status, income, family size/number of children, and more. Using predictive analytics, this data is layered on top of the clinical marker data and mapped against real-world data on other individuals within a health care system or member population who have similar SDoH. This creates a more refined view of a patient or member’s impactability, based on how others of similar background, in similar circumstances, and with similar conditions have behaved, and/or how they have been effectively engaged by payers or providers.

On another dimension, predictive analytics can be used to determine an individual’s receptivity toward being engaged and taking action — in sales terms, is the individual a cold, warm, or red-hot lead? — and preferred engagement methods, e.g., whether they prefer communication via text, phone call, messaging app, or video chat; whether they respond to small incentives like gift cards for taking certain actions on their health; and so on. For health care systems or payers that have mapped out the personas of their patient/member bases, using predictive analytics to layer clinical marker data, SDoH data, and persona data together can often determine these answers and create a well-rounded impactability score.

These three dimensions will inform initial identification, and then hierarchy of stratification, of the patients or members for whom modified care-management approaches, such as whole-person interventions, will produce the greatest gains.

Don’t Forget the Human Element

These three dimensions of data inform the initial engagement modality by helping clinicians infer some of an individual’s existing challenges and barriers even before a clinical intervention session.

Yet up-front data from clinical markers, SDoH, and even personas can only go so far when it comes to effective interventions; in the whole-person care model, relationships — and trust — matter. Human-to-human interactions help clinicians establish trust by listening, offering empathic responses, and using tools like motivational interviewing to better understand an individual’s unique circumstances. They also yield more data.

As clinicians establish a relationship with an individual, they not only gain a better understanding of what to prioritize during future sessions; they also gather new insights from the real-time information they hear. This data creates an additional layer of intelligence that can be put back into the predictive analytics engine to further guide care-management and whole-person intervention strategies for an individual, creating a multidimensional, ever-evolving measurement of their impactability and needs.

Better Tools, Better Outcomes

Using predictive analytics to model impactability is just one tool to leverage in the whole-person care model — but it’s a powerful one. When paired with the expertise of talented clinicians and other supporting health care platforms and technologies, it can help providers, payers, and other stakeholders design more effective whole-person interventions — and thus achieve bigger impacts on population health and operating models.

   

Categories