Duke Uses Records Data, Analytics for Personalized Care

The Duke Institute for Health Innovation (DIHI) is embarking on new initiatives that use past records from patient encounters to develop personalized care models for patients.


The Duke Institute for Health Innovation (DIHI) is embarking on new initiatives that use past records from patient encounters to develop personalized care models for patients.

DIHI, based in Durham, N.C., and housed within Duke University, began the project in April and is focusing on certain disease states where individualized approaches to patients could both improve outcomes and patient satisfaction, as well as reduce overall healthcare costs, says Suresh Balu, Program Director for the Institute.

The approach uses Duke’s own care delivery system to provide a “living lab for innovation,” says Will ElLaissi, who heads partnership development. “We’re not creating a brick and mortar infrastructure where you can model new types of care. We are using the clinical infrastructure at Duke, but providing them with capabilities that will enable them to innovate.”

Data lie at the core of efforts to drive innovation through the living lab, Balu says. “When it comes to analytics and visualization, data are complex and large, and without proper analytics, it becomes difficult to communicate the complexity of specific types of patient encounters.” The effort goes beyond just having guidelines to creating predictive models that are tailored to individual types of patients.

Also See: Precision Medicine Requires Unlocking Data from EHRs, Other Sources

Some early efforts target patients with chronic kidney disease and how to work with them to avoid emergency treatment and delay progression into end-stage renal dialysis (ESRD).

Part of the solution to the puzzle involves invoking early interventions for patients with the appropriate care provider, and being aware of the first indications that patients test results warrant attention. With chronic kidney disease, for example, patients typically are referred to a specialist when their glomerular filtration rate (GFR) drops below 30, and a patient is at risk for severe loss of kidney function. However, Duke’s initiative seeks to identify patients that show declining GFR results before critical levels are reached, before patients are likely to need dialysis.

As GFR values start to decline, a visit to a primary care physician or an early consultation with nephrologists and care managers and other types of patient education may prove valuable, Balu says. The Duke study is looking at what kinds of factors could improve early interventions; for example, even something as simple as providing transportation for at-risk patients might improve the likelihood of keeping physician or dialysis appointments.

“We need to understand the socioeconomic factors,” Balu says. “Something like that might help to lower the overall cost of care. The types of visualization tools we’re using can help us develop the right sets of care for each of these patients.”

Current predictive models for the potential for progression of chronic kidney disease are about 80 percent accurate, Balu says. The Duke study hopes to look at information in longitudinal patient records to increase accuracy, and it hopes to fine-tune predictive capabilities over the course of the year-long study.

In addition to the work on chronic kidney disease, the Duke study will aim to improve post-surgery regimens that are used to reduce complications after colorectal surgery, trying to determine factors that could tailor the regimens to further reduce complications and aim towards patient-centered approaches to care.

More for you

Loading data for hdm_tax_topic #better-outcomes...