When analytics falls short
July 24, 2014 in Medical Technology
The joys of unintended consequences never end. When the Patient Protection and Affordable Care Act required hospitals to get paid based on how much they improved their patients’ health rather than however tests and procedures were completed, the intent was to improve patient care.
But some data brokers saw dollar signs in those changes and have started mining patients’ payment card data, public records and loyalty programs and using that to create risk profiles, which they are then selling to hospitals and insurance companies. For asthma, they look at pollen counts associated with the current residence as well as cigarette purchases, plus pharmacy records indicating whether prescriptions are being refilled regularly and on time, suggesting compliance, according to a Business Week report.
[See also: Healthcare analytics enters new age.]
Although there are absolutely privacy issues raised, there are some initial limits. However, it’s unclear how long those limits will remain in place. For example, the individual information is analyzed and massaged by the data broker, who turns it into a series of risk numbers, but the hospital isn’t told the purchase/public record facts that went into those crunched numbers. In other words, the hospital won’t be told that a gym membership lapsed or that 20 pizzas are being ordered a week, but will solely see that weight gain risk increased.
The biggest problem with such data collection is simply that the analysis could be highly inaccurate, mostly because a purchase made is not necessarily for the patient to personally use.
“It’s important to remember that credit card and customer loyalty program data are not soundly indicative nor solely attributable to an individual’s behavior. For one thing, individuals may be purchasing items for family members or even an elderly or disabled neighbor,” said Pam Baker, a business data analyst who has a new book out on the topic: Data Divination: Big Data Strategies.
“If you attribute these purchase behaviors to the card holder, you’ll get a highly misleading read on that person’s health practices and conditions, she says. “For example, the card holder may be buying snacks at the grocery store and making fast food purchases for children in the house and not consuming any of it themselves. Or they may be buying supplements, protein drinks, or adult diapers for an elderly family member, again not using any of these items themselves.
“And those regular alcohol purchases you see on record may be just part of entertaining clients and colleagues and not an indication of an alcohol problem in that individual,” Baker adds. “The conclusions you draw on any given individual will therefore be flawed and no set conclusions should be made. If you fall into that trap, you will only alienate your customer-base and fall short of the profit goals you seek because you will be frequently reacting to incorrect information.”
[See also: Clinical data analytics next big thing.]
Michael Dulin, chief clinical officer for analytics and outcomes research at Carolinas HealthCare, one of the health groups that is purchasing this kind of data, is quoted in the BusinessWeek piece, arguing that “information on consumer spending can provide a more complete picture than the glimpse doctors get during an office visit or through lab results.” That’s certainly true, but even if the data were accurate — which is far from a given — how much good is likely to come from it?
Given that much of this information would become apparent in lab tests and a general examination (lapsed gym memberships and increased donut purchases mean nothing if weight is good and the blood tests are favorable), the theoretical benefits would materialize earlier. If a patient is not refilling a key prescription, is he or she forgetting and would benefit from a reminder call? Or perhaps fallen on bad economic times and could use a financial plan to fund the refills? The problem is that there is a disconnect. The brokers are promising to only deliver risk numbers and not specifics. Without those specifics, this prescription refill problem wouldn’t be disclosed to the doctors.
Then we get into even murkier waters. Are pharmacists now supposed to initiate calls to the doctor saying that a certain patient isn’t pickup or—or paying for – refills? Should that pharmacist call the patient? And would such a call coming from a pharmacist seem more like a sales pitch?
Data analytics is generally a wonderful thing, but I’m not seeing how this program could really deliver much of a benefit to anyone involved, other than the data brokers who sell it.
Article source: http://www.healthcareitnews.com/news/when-analytics-falls-short