Digging for Empathy in Big Data

A study of diabetes diagnoses and long-term health points to the numbers' softer side.

The concept of big data in healthcare can seem impersonal and absolute, with thousands of de-identified patient records crunched through various analytics to produce endpoints. At the Big Data and Analytics for Pharma summit in Philadelphia this week, Victoria Gamerman discussed how data analytics can deal in both empathy and complexity.

The head of health informatics and analytics for the pharma company Boehringer Ingelheim pointed to her team’s ongoing IntroDia study. The study intends to determine how the initial delivery of a type 2 diabetes (T2D) diagnosis can impact a patient’s disease management and health in the long term. The researchers hypothesize that it does have an impact. If they’re right, physicians may need to find different ways to discuss the diagnosis.

The study takes place across 26 countries, thousands of physicians and patients, and multiple specialties. Gamerman said it’s a big data problem. First, they must gauge physician diagnosis-delivery and patient response using self-reported questionnaires. Then they need to follow disease progression over time.

The study contains more than 6700 physicians and is cross-sectional to include how primary care physicians, psychologists, and endocrinologists believe a T2D diagnosis and treatment is best delivered. The patient side of the equation is also split, with more than 5900 patients given an initial T2D diagnosis and 4200 who have received news that they may require an add-on treatment for the disease.

From the sets of patient and physician surveys, the IntroDia study must draw qualitative data that can be compared: messages, themes, and sentiments from all sides. It uses an exploratory factor analysis (EFA) to group 43 conversation elements and deliver 4 patient-perceived diagnosis dimensions: encouraging, collaborative, recommending of other resources, and discouraging.

So far, the team has found that patients believe doctors should use more encouraging and collaborative language, which is a natural inference. Resource recommendations had no significant impact on patient feelings about their new diagnoses. While discouraging language may seem negative, the study found that physicians most often felt they were being “honest and upfront” with their patients.

That may not validate the study’s hypothesis. But it is an early spark for a complex study, which excited Gamerman and her colleagues.

“Hiding within those mounds of data is knowledge that could change the life of a patient, or change the world,” she said, quoting Stanford University’s Atul Butte, MD. “This is why I get up every morning. This is why I work.”