Better health care through algorithms

Predictive algorithm that identified patients for outreach programs reduced health care use and costs
August 18, 2020

A new study conducted by Texas A&M University suggests that using technology to identify high risk patients may aid in reducing unnecessary health care use and medical costs.

At the heart of efforts to improve quality of care and reduce costs are interventions such as health coaching and disease management programs. Past intervention efforts have been less effective than hoped, possibly because the interventions did not effectively target the small subset of patients most likely to benefit from additional treatment. However, algorithms analyzing holistically all health records to identify individuals who may benefit from an intervention may make it possible to improve quality of care.

A study published in the journal Health Economics investigated whether an intervention conducted by a large health insurer that was aided by such algorithm could be more effective at reducing costs and unnecessary care. Benjamin Ukert, PhD, assistant professor in the Health Policy and Management department at the Texas A&M University School of Public Health, together with researchers from the University of Pennsylvania and Independence Blue Cross, analyzed the effects of an intervention on high-risk Medicare Advantage patients with congestive heart failure. The intervention relied on a predictive algorithm that identified patients with the highest risks of hospitalization in the near future. Ukert and colleagues then analyzed the targeted intervention’s effectiveness in reducing long-term hospitalizations and other outcomes.

The intervention’s predictive algorithm used a health insurer’s data including, claims clinical information, and member demographics for patients diagnosed with congestive heart failure between 2013 and 2017. Congestive heart failure causes more than 1 million hospitalizations each year, more than 20 percent of which are likely avoidable. The intervention consisted, among other things, of conventional health coaching to educate patients about their condition and how to manage it, and specialized condition management aimed at getting patients to enroll in care management programs.

The researchers analyzed outcomes using claims data on inpatient and outpatient health care use, and spending. They evaluated hospitalizations, such as number of admissions, total days admitted and average length of stay as well as information on visits to primary care providers and specialists. The intervention program consisted of 10 waves of interventions and each wave reached different percentages of members. This means that some of the members identified as high risk were not contacted for the intervention that served as comparison members. Additionally, not all patients who were identified and contacted participated in the intervention.

Ukert and colleagues found that the targeted intervention decreased the probability of being admitted to a hospital and the number of hospital admissions for high risk patients compared to high risk patients who did not received the intervention. There was also a similar decrease in emergency room visits. High risk patients who received the intervention also visited primary care providers and specialists with less delay and had an average monthly lower medical cost of more than 700 dollars per month in the first six months after the intervention. These findings indicate that unlike previous intervention efforts, an intervention that targets high risk patients can reduce avoidable health care use and decrease medical costs.

Although these findings are promising, the researchers note a few limitations and highlight areas for further study. Ukert and colleagues found decreases in health care use and costs, but the factors that led to the success of this intervention are still unclear. Thus, any attempt to characterize the mechanisms at work would be speculation at this point. Further analysis of outreach programs will be needed to better identify the aspects of the program that led to success. Additionally, the algorithm used to obtain patient risk score is proprietary, so the inner workings are unclear. Despite these limitations, it is clear that technology can play a decisive role in reducing unnecessary care and related medical expenses and improve quality of care for many people.

— Rae Lynn Mitchell

You may also like
Guiding efforts to improve physical activity resources along the Texas-Mexico border
Public Health professor continues to make her mark, receives prestigious award
college students wearing masks ride bus
POV: Colleges and the Thanksgiving COVID-19 risk
Emergency department visits for dental conditions take a bite out of everyone’s wallet