How health plans use predictive analytics to manage diabetes, hypertension and heart failure
Data Science

Chronic disease care: Essential AI for health plans

April 27, 2021
Predicting Diabetes, Heart Failure and Hypertension Complications.
Lead Principal Data Scientist

Chances are you know chronic disease is a tremendous burden to the healthcare system and to far too many patients. Perhaps you or a loved are among the approximately two-thirds of American adults who have one or more chronic health conditions.

The Cost and Prevalence of Chronic Disease

Nearly 150 million Americans had one chronic condition, as of 2014. About 100 million people had more than one, and nearly 30 percent had five or more chronic conditions.

The risk and prevalence of chronic illness increases as people age, and the proportion of people age 65 and older is expected to more than double by 2060. More than a quarter of children have a chronic condition compared to 60 percent of adults. “Among those 60 or older, at least 80 percent have one chronic illness and 50 percent have two.”

Estimated direct costs of chronic disease totaled $1.1 trillion in 2016, according to the Milken Institute. Nearly 90 percent of total health costs are associated with treating chronically-ill patients, according to the Medical Expenditure Panel Survey.

The Geneia Data Intelligence Lab

That’s why the Geneia Data Intelligence Lab (GDI Lab) is eager to apply the robust tools of artificial intelligence and machine learning to mitigating chronic disease. The GDI Lab data scientists prioritize projects that address major cost drivers and enable health plans, hospitals and physicians to identify, stratify and predict high-cost patients and chronic conditions.

The models developed by the GDI Lab help healthcare organizations to intervene earlier with patients whose risk is expected to rise and/or who are at risk for major, expensive conditions, allowing healthcare organizations to remedy future costs while improving the health of those they serve. In collaboration with Geneia clinicians, the GDI Lab develops and continuously refines algorithms to help healthcare organizations move from insights to increasingly sophisticated targeted actions.

How AI is Helping

The GDI Lab primarily creates models using administrative claims since health plans have ready access to this information, as well as models using a limited number of features or variables. For example, the GDI Lab’s model to predict opioid abuse and overdose*+ uses 22 variables whereas others use as many as 200 data points to achieve comparable predictive accuracy. Fewer variables make it easier to implement the models; they also improve model interpretability, enabling physicians and patients to more easily understand the output.

Given the burden of chronic disease, two primary foci for the GDI Lab are:

  • Predicting the onset of chronic disease for those that don’t have it yet; and
  • Predicting the complications of chronic disease for those that already have it.

Hypertension: Predicting Complications

Take hypertension, for example. Forty-five percent of adults have hypertension, and it is increasingly common with age:

  • 20-44 years:     26 percent
  • 45-64 years:     59 percent
  • 65+ years:       78 percent

Hypertension is often symptomless, and, if poorly managed, can contribute to devastating complications, including heart attack, stroke, end-stage renal disease and more.

The GDI Lab hypertension complications model*+  uses 28 variables such as total medical costs, the number of electrocardiograms, and a diabetes and/or hyperlipidemia diagnosis to predict, for those with hypertension, who is likely to experience a complication across three complication stages in the next 12 months. Members can meet the criteria for one, two or all three. As the stage number increases, so does the severity.

  • Stage 1 includes often treatable, if not reversible, damage.
  • Stage 2 reflects progressive chronic conditions that, if not treated and self-managed, will lead to increased emergency department visits and inpatient stays.
  • Stage 3 identifies severe and potentially irreversible complications that may lead to death.

Within each stage, a patient is classified as high, medium or low risk.

Hypertension complication stages

Given the prevalence of hypertension, there are almost always too many patients to intervene with all of them in a timely manner. That’s the reason innovative health plans use the hypertension complications model to identify where to start and to align their interventional and care management strategies to the risk level.


The GDI Lab uses a similar approach to identify people at risk of diabetes onset in the next 12 months as well as diabetics and heart failure members at risk of complications in the coming year. Looking ahead, the GDI Lab has initiated the creation of an impactability model - that is, those whose cost can be most affected by care management intervention. Targeting only the riskiest members or those expected to develop a chronic illness or experience complications means lost opportunities. For example, the chart below compares a high-risk patient with an impactable one, and shows the associated savings opportunity.

Impactable patient versus high-risk patient

Combining the lab’s impactability model with its series of chronic disease analytics will offer health plans an even more diverse toolset to reduce the burden of chronic illness for the plan and its members.

To learn more about the Geneia Data Intelligence Lab, download the white paper, Chronic Disease Care: Essential AI for Health Plans.

Read more

*Certain data used in this study were supplied by International Business Machines Corporation. Any analysis, interpretation, or conclusion based on these data is solely that of the authors and not International Business Machines Corporation.

+Predictive models, by their very nature, contain certain assumptions. This is not an attempt to practice medicine or provide specific medical advice, and it should not be used to make a diagnosis or to replace or overrule a qualified healthcare provider's judgment.