Contraception Clusterfuck

Every risk adjustment system which drives money that I know needs a claim based event to trigger an action. Previous history of chronic conditions is the easiest to access predictor of current chronic conditions. When I worked at UPMC, I spent three years figuring out how to optimize the risk adjustment revenue for the Medicaid line of business. UPMC Health Plan is a multi-line insurer with products in Medicaid, CHIP, Exchange, Medicare and Employer Groups. It is not at all unusual for people to bounce between Medicaid, CHIP, Exchange and Employer coverage over time. One of my major projects that I was very happy to have completed was building an integrated data model that mined the entire UPMC claims universe instead of just the Medicaid claims universe. That increased the total revenue haul and decreased the number of false positives.

I had it easy. Data geeks working for insurers with either low market share or shallow data had a much harder time optimizing their risk adjustment revenue.

Aetna has a kick-ass data team. They have huge and deep data sets that they control. It is quite likely that a significant chunk of their risk adjusted covered lives in 2018 have shown up in some point in their data bases in the past decade. An individual who is now insured by Aetna Medicare Advantage in Texas may have had an amputation claim from Aetna Medicaid in Pennsylvania that is dated in 2009. That is valuable information to build and curate a risk adjustment optimization list.

However there are always serious holes in the Aetna list. Either someone has never been on Aetna before or there was a major change in health status when that person was covered by someone else. This is where CVS comes in. There is a good chance that CVS has filled some prescriptions for people who do not show up in Aetna’s data banks. Newly covered lives by Aetna can have a risk profile built off of CVS prescription data to minimize the number of surprises and optimize risk adjustment strategies.

This is the most obvious play from my days as a risk adjustment data geek. The other side of the far more complete pre-enrollment data universe for Aetna via the CVS pharmacy data is that Aetna will have far more granular level information on their markets. This will influence plan design, it will influence marketing materials, it will influence whether or not Aetna enters or leaves a market or bids for certain contracts.

Finally, the biggest data bonanza from my point of view is the CVS non-prescription data that is tied to the loyalty card that almost everyone carries on their keychain. This should give a massive predictive edge to the Aetna data geeks. Let me share way too much personal information to illustrate.

Our two children were extremely planned children. My wife used oral contraception until we started trying for our first child. After our daughter’s birth, we switched to condoms as our birth control method as she felt better off the pill and for the most part, we could handle a happy accident or a baby one year premature. I felt that I was tempting fate if I bought condoms from Costco. I walked past a CVS at least twice a day to and from the bus-stop I used for work. If we were running low, I would pick up condoms and a gallon of milk.

If an insurer could see the non-prescription purchases tied to the customer loyalty card, they had an excellent idea of when my wife and I started trying for Kid #2. If this was an insurer that sought to be socially productive and useful, we could expect to get mailings and outreach calls on pre-natal and perhaps pre-conception health enhancers. If the insurer was run by cynical bastards and the time of the year was right, they might try to be enough of a pain in the ass to get us to switch insurers so that someone else could pay for labor and delivery.

That is the most obvious data play that I can think of based on personal experience. I can think of using the CVS retail data as population health monitoring service, I can think of using the over the counter sales data tied to individuals to fuel predictive models for future opioid issues, or arthritis flares, or pulmonary hospital admissions or one hundred other things.

So from my former point of view as an insurance data geek, this merger offers an incredibly rich vein of data that can be mined and minted. This makes a lot of sense to me without even thinking about how the entire pharmacy benefit management function is a messed up situation.