CHAMPAIGN, Ill. — An analytics-driven "decision framework" that accounts for the socioeconomic and demographic factors of patients can promote more equitable health care delivery and potentially improve chronic disease care outcomes, according to new research co-written by a University of Illinois Urbana-Champaign business scholar who studies technology adoption in health care.
A data-informed approach to scheduling patient encounters with clinicians can reduce risks associated with diabetes management by up to 19.4%, especially for underserved populations, says Ujjal Kumar Mukherjee , a professor of business administration at Illinois.
"Managing chronic medical conditions such as diabetes is a major challenge for health care organizations because it requires both committing resources over a long timeline and high levels of patient engagement in the care process," he said. "The diversity of the patient population also plays a role in terms of health risks across patient populations, which can influence care outcomes. But if you apply a strategic approach and essentially customize the care to the patient's demographics, you can drive improvements in health outcomes."
Mukherjee's co-authors are Dilip Chhajed of Purdue University and Han Ye of Lehigh University.
The study aimed to improve diabetes care by developing a predictive and prescriptive framework for allocating health care encounters more effectively, especially for socioeconomically and demographically diverse populations.
"Many high-risk patients receive fewer health care encounters than needed, which means it's important to tailor chronic care treatment to improve outcomes," Mukherjee said.
A mismatch leads to inequitable and inefficient resource use, especially for managing chronic health conditions.
"It's well known that there is pervasive health inequity in the U.S. that's coupled with a limited capacity to deliver chronic care — chronic being a progressive kind of a disease such as diabetes, COPD, cancer or heart disease," he said. "These are all progressive diseases that, if left untreated, advance to costlier stages of care — which, of course, makes it more expensive and time-consuming for patients and practitioners."
But if you treat it early enough and often enough, "then you can manage it and bend the cost curve down," Mukherjee said. "Chronic diseases don't get cured. Rather, it's a question of managing both the disease progression and the overall risk of it."
The researchers analyzed data from more than 10,000 diabetes patients collected from a multifacility clinic in the U.S., with population-level socioeconomic and demographic data culled from the U.S. census.
Mukherjee and his co-authors employed machine learning to determine if future diabetes risk for individual patients could be predicted using past clinical measures and population-level socioeconomic variables such as income and education.
The researchers found striking disparities in care access: Patients from low-income, less-educated or predominantly minority communities were significantly less likely to have regular health care encounters — despite having higher average glucose levels.
The finding underscores the need for risk-sensitive decision frameworks to support and augment clinical decision-making processes, Mukherjee said.
"Many of the chronic care diabetes patients were from underserved, under-resourced communities, and they don't have regular contact with medical professionals," he said. "They don't go through preventative or primary care processes. Which means they'll often end up in the emergency department with some kind of adverse health event, whether it's a heart attack, kidney failure, retinal problems or a liver dysfunction, all from untreated diabetes."
And when patients end up in an emergency room, both the patient and the health care system end up spending a lot more time and money than they would if the disease had been managed from the outset.
"If you have regular contact with clinicians, you can avoid unnecessary emergency hospitalizations," Mukherjee said. "In that way, patients from disadvantaged backgrounds benefit the most from an optimized health care allocation strategy."
The implications of the research point to how health care providers can use analytics to ensure that limited clinical resources such as appointment slots are equitably and efficiently distributed, Mukherjee said.
"The approach supports fairer access to chronic care and has the potential to reduce health disparities on a population level," he said.
The paper was published by the Journal of Operations Management.
