Doctors Demand More Data About Patients

The medical profession - as well as the employers and insurers who pay the bills - want solutions that are predictive, not palliative, in order to improve outcomes and reduce costs. JL

Melanie Evans reports in the Wall Street Journal:

Spurred by employers and insurers that want health-care providers to prevent illness—and not merely treat it—ever more complex algorithms forecast patients’ medical future. They’re searching for new kinds of data to make predictions as accurate as possible, mining behavioral, consumer and financial data. Two algorithms that used nonmedical information, such as housing instability or the ease with which patients can dress and feed themselves, improved results.
The race is on to develop better ways to predict if patients will develop diabetes, heart disease or other critical conditions.
Spurred by employers and insurers that want health-care providers to prevent illness—and not merely treat it—doctors and hospitals are creating ever more complex algorithms to forecast their patients’ medical future. And they’re searching for new kinds of data to make those predictions as accurate as possible, mining behavioral, consumer and financial data for potential clues.
Some hospitals are collecting new information from patients directly, while others have sought data from companies that sell consumer and financial information, or federal agencies that provide statistics on poverty, housing density and unemployment.
Knowing more about how people live—from their interests to their income—could prove useful as doctors look for clues to poor health and tailor interventions to address patients’ needs, potentially preventing illness and saving money, proponents of this approach say.
“So much of what determines a person’s health and well-being is independent of medical care,” says Rishi Sikka, senior vice president of clinical operations for 12-hospital Advocate Health Care in Downers Grove, Ill.

Casting a wider net

Advocate’s efforts offer a window into these new missions to get better insight into patients’ health. In 2012, Advocate introduced an algorithm to predict which hospital patients would come back less than a month after leaving. The algorithm relies on a history of patients’ ailments, prescriptions and laboratory tests, vital signs and other medical care to single out who is at risk. In its first nine months, the predictive tool slashed readmissions for high-risk patients by 20% by allowing Advocate to better direct aid to those in need, says Tina Esposito, vice president of Advocate’s Center for Health Information Services.
Now the provider wants to make its algorithm even stronger by mining new information about its patients. By the end of the year, Advocate says it will acquire consumer data from a company that mines purchasing and demographic information for retail chains and marketers. Advocate’s team of data scientists will analyze the new trove of data—a slew of statistics on marital status, hobbies and household income—for its predictive potential.
The data could fill gaps in what doctors know about financial or lifestyle factors that influence patients’ health. Doctors are “largely blind” to details that are not included in patients’ medical records, says Dr. Sikka.

A need for information

So far, the verdict is that the more-complex algorithms need stronger data to be truly effective. One study looked at more than 70 predictive models that largely relied on patients’ medical history to identify people at risk for an unexpected hospital admission. The result: “inconsistent performance,” according to results published online by the British Medical Journal in June. Two that targeted avoidable admissions did well enough to be considered strong.

A separate study, meanwhile, provided support for using wider types of data. In the study, two algorithms that used nonmedical information, such as housing instability or the ease with which patients can dress and feed themselves, improved results, with performance that would be considered modest to strong, Devan Kansagara and colleagues found in their research, published in the Journal of the American Medical Association in 2011.

Tough to collect

But Dr. Kansagara argues that the movement to tap more data has to overcome a big obstacle: access to the data. Growing interest in nonmedical data “doesn’t solve the problem that it’s practically hard to get that information, he says.
Doctors and nurses have limited time to collect new data and patients bombarded with questions about their lives may suffer “interview fatigue,” he says.
That is an additional burden on patients who are already grappling with illness and it could make doctors reluctant to collect the data, he says.
Nonetheless, one Minneapolis hospital system decided to collect new data itself, with promising results.
Primary-care patients who visit Fairview Health Services complete a survey of their self-confidence and capacity to manage an illness, called the Patient Activation Measure. Fairview first added the measure to its clinic appointments in 2010. Patients describe if they feel assured that they can solve problems, seek help when necessary and keep up healthy habits “even during times of stress.”Research suggests patients’ confidence in their ability to manage their health or condition levels may be a predictor of future health. An analysis of responses between 2011 and 2012 found those without the know-how and confidence to manage their disease were 56% more likely to land in the hospital than the most assured and knowledgeable patients. The study, by researchers at Fairview, the University of Oregon and George Washington University, looked at hospitalizations that research suggests could be avoided with good primary care.

After Fairview identifies high-risk diabetic patients, nurses help the ones who lack confidence to set achievable goals, which may help improve health and build self-assurance. Goals can be as simple as getting patients to keep their next scheduled clinic appointment, says Carmen Parrotta, a performance improvement consultant for Fairview. “Getting them there is the first step,” he says.
More sophisticated patients set more complex or challenging goals, such as weight loss or better management of blood sugar. Unpublished results for this program show improvement in measures of blood sugar and cholesterol.

Some doctors are mining behavioral, consumer and financial data for potential clues about their patients’ health. Photo: iStockphoto/Getty Images