AHIP: Use predictive analytics to drive decisions, control costs
For health plans, the case for robust analytics comes down to the numbers.
Medication to treat hepatitis C costs $1,000 a dose, and patients who miss a dose need to start over, Dan Walczak, assistant director of healthcare informatics at Minnesota-based UCare, noted at the America's Health Insurance Plans (AHIP) Institute this week.
The cost of hepatitis C treatments has forced insurers into financial and ethical dilemmas, FierceHealthPayer previously reported. However, the treatments work--and the cost pales in comparison to a liver transplant, Walczak said, which comes with a price tag of $500,000 and a laundry list of complications.
Payers that want members with chronic conditions to receive the best care they can while also controlling the cost of that care cannot simply rely on historical data, said Somesh Nigam, chief informatics officer with Independence Blue Cross (IBC).
"We are quickly realizing that significant decisions require predictive analytics," Nigam said, adding that the role of data and analytics is now central to how the Philadelphia-based payer operates and interacts with the healthcare system at large.
IBC's data warehouse, initially developed to build better reports, now includes 265 million medical claims, along with pharmacy claims, lab data, eligibility and benefits records and even Blue Button data, Nigam said. He described the warehouse as holding "five Wikipedias' worth of data" and added that it only includes partial electronic health records.
IBC uses this data to create predictive models on a variety of factors, such as who in a particular population is likely to be hospitalized in the next 90 days, which diabetes patients are likely to develop the most complex and expensive comorbidities and who is least likely to adhere to a medication regimen. UCare, meanwhile, has tracked member leakage and developed benchmarks to improve efficiency ratings for particular episodes of care.
Analytics often points out the obvious, Walczak said. For example, if a patient's length of stay for a certain condition matches the standard, expected length of stay, he or she is an unlikely candidate for readmission. But having the data to back up these presumptions can help stimulate institutional change.