5 ways to use big data

February 1, 2014

Choose your goals before crunching numbers

The age of big data is here and many health plans have been building and leveraging their data analytics capabilities for some time. Plans that have not invested in the technology infrastructure and data building necessary to maximize the benefits of data analytics could soon find themselves lagging behind. Here are some ways health plans can start differentiating themselves by using big data.

1/ Build a foundation

Data analytics is only as powerful as the underlying data. That is why many health plans are investing heavily in upgrading their technology infrastructure and cleaning up and standardizing their data.

“Some plans are relatively sophisticated in using their data and others are struggling,” says Pamela Peele, chief analytics officer for UPMC Insurance Services Division in Pittsburgh. “The constraining factor is how much investment health plans have in their IT infrastructure and that varies widely across health plans.”

UPMC Health Plan, which has invested some $1.5 billion in its IT infrastructure, has created a large data source of what Peele calls “a harmonized, groomed layer of information holdings and data from multiple disparate sources.”

UPMC Health Plan is using data analytics in a number of areas. For example, it focuses on reducing hospital readmissions before a member is even admitted rather than waiting until the patient is discharged. The plan has developed data models that calculate the probability of readmission among its entire health plan membership.

“Every month, we are predicting readmission probability based on whether a plan member who is admitted to the hospital today would be readmitted to the hospital within 30 days after discharge,” says Peele. “When someone is admitted to one of our hospitals, that readmission risk is displayed on the opening screen.” At that point, the hospital creates the authorization for the admission and also begins the work on reducing that readmission risk as much as possible.

2/ Set guidelines

Using data analytics to bolster existing priorities may be tempting, but doing so will not allow health plans to maximize their return on their investments. Ken Park, vice president of payer and provider solutions at WellPoint in Indianapolis, offers three suggestions that can serve
as broad guidelines when using
data analytics:

  • Don’t bend the data in order to prove an ongoing hypothesis. Look at what the data is actually showing you.

  • The focus should be on ways to deliver the highest quality healthcare at the most affordable prices rather than ways to provide the lowest cost healthcare regardless of the quality.

  • The most effective data analytics focus on a valid clinical question that is not already answered by the academic literature, are relevant to the business and promise a significant business impact, and begin with a clear idea of how the organization will use the resulting information.

3/ Learn from other industries

As health plans shift to more consumer-oriented business models, data analytics will become more important.

“Health plans need to learn to use data the same way that American Express, Disney, Harrah’s and others have,” says Jack Newsom, vice president of marketing analytics at Silverlink Communications, Inc. “This means understanding what motivates individuals and learning how to communicate with them in order to build trust and loyalty, and ultimately change behavior.”

For example, UnitedHealthcare has leveraged its data analytics in an effort to increase colorectal cancer screening rates among minority populations. This effort included analyzing the screening rates among 500,000 plan members in different ethnic groups to identify barriers to screening and to determine the most effective methods of encouraging specific groups to complete recommended screenings. Based on the results, UnitedHealthcare created customized outreach programs to increase screening rates. The analysis found that a phone call from a plan representative to one group of men increased cancer screening nearly 11% compared to another group who received a recorded call.

 

4/ Leverage multiple data sources

The more strong data in the system, the more powerful data analytics will be. Core claims data, member-provided information from health risk assessments and general marketing data can all support data analytics. For example, plans can use general marketing data to get information on household size, whether members use mail order and other standard marketing information. Plans can use this data to get a clearer picture of each member that can be important when trying to coordinate and improve access to care.

UPMC Health Plan relies on health risk assessment data to get a sense of the potential plan usage among new enrollees in Medicare Advantage plans during their first 12 months with the plan.

“With no usage data among newly eligible Medicare Advantage enrollees, we would have to wait and see who is going to require care coordination,” says Peele. “We identify those members using a specific combination of answers on the health assessment survey and assign a care coordinator to that member before they have their first doctor’s appointment.”

UPMC Health Plan also receives daily data feeds from all of its vendors, including the provider of health risk assessments, labs and pharmacy benefit managers so that the plan does not have to wait for claims data before acting on that data.

Peele expects clinical outcomes data to be the next frontier in data gathering. Claims data can show productivity measures, such as how often members see a doctor and the exams or tests done. However, “we want to know the outcome of that care,” she says. “If a diabetic patient sees a physician for a hemoglobin A1c test, outcomes-based data will show how effective that care has been in terms of actual clinical outcomes.”

5/ Combine clinical and claims data

Data analytics do not have to focus solely on operations. WellPoint uses both claims and clinical data to evaluate medical policies and its drug formulary to see whether coverage for a certain drug is appropriate and to evaluate the effectiveness of different benefit designs and programs.

Another use for data analytics is to test and disseminate information on clinical care. For example, WellPoint’s analytics can evaluate treatment patterns for children with chronic headaches to determine the prevalence of CT scans. A key concern is that use of CT scans in treatment and diagnosis exposes patients to unnecessary radiation at a young age.

“We examined the data to see if we needed to change policies or programs to avoid scans,” says Park. “Upwards of one-quarter of children who identified with headaches had some type of imaging scan, most commonly a CT scan.”

When the analytics identified emergency rooms as the setting where scans for this population are most likely to occur, WellPoint was able to adjust its policies and disease management interventions to reduce that number. In addition to using this information in its own operations, WellPoint also shares its findings in conjunction with groups like the American Academy of Pediatrics as a way to support more evidence-based medicine.