Risk adjustment and responding to shocks

Hydroxychloroquine (HCQ) is supposed to be a boring drug. It was a boring drug in 2018 and 2019 when it was inserted into the ACA risk adjustment formula as a marker for a category of high cost and complex auto-immune diseases like lupus (RxC-09). This category of risk adjustment has a pricing factor of just over 12 which means that an insurer that has a member taking this drug will be eligible to receive 12 times average state premium from the risk adjustment system as that is about the average incremental cost of insuring someone with lupus and other qualified auto-immune disorders. This translates out to be somewhere between $70,000 and $100,000 depending on the state. This is real money in a zero sum risk adjustment system.

According to a recent paper, the drug was not commonly searched for on Google in the first two months of 2020:

Search interest in HCQ in the United States and globally remained low through February 2020. The first major surge in search interest occurred in March 2020, following a White House press briefing in which the President described HCQ as a “game changer” for COVID‐19 treatment; global HCQ search interest reached its highest peak on March 20. On March 28, the FDA issued an EUA to open HCQ supplies from the Strategic National Stockpile for treatment in patients with COVID‐19 and, days later, declared a national shortage of HCQ.

Another spike in search interest occurred in late May, following an announcement that the President was reportedly taking HCQ prophylactically for COVID‐19. Search interest was highest in the United States in late July, following a viral video of a press conference held by a group of doctors critical of the pandemic response. The video featured depictions of HCQ as a “cure” for COVID‐19, which eliminated the need for masks and social distancing measures. The video was posted on news and social media platforms and shared by public figures, including the President.

There was a huge spike in demand for prescriptions for this drug that typically is not widely prescribed. These prescriptions were for off-label uses that did not correlate at all with the risk adjustment category of auto-immune diseases in the risk adjustment model.

This is a major problem. RxC-09 is big money on a per person basis. RxC-09 is a fairly rare to trigger under clinical and actuarial modeling assumptions. There is a massive off-label use of the drug as a response to an information and political shock. That off-label use is unlikely to be randomly distributed. It is likely to be concentrated in counties and insurers that are very Trumpy and less likely to be concentrated in counties that went heavily Biden.

This means that some insurers that cover a heavily Republican leaning population will get massive windfalls from risk adjustment for patients who are coding as incredibly predictably sick and expensive for a cohort that is likely to be fairly inexpensive. Given that the ACA uses revenue neutral risk adjustment, that means insurers that are covering populations that are less likely to have been prescribed HCQ for an off-label use are going to be paying in a lot for risk adjustment.

CMS knows that this is not right. Their first attempt to make it right was to exclude HCQ from risk adjustment. This eliminates all of the people who were taking HCQ off-label for COVID. It also eliminates a lot of people who have lupus from the risk adjustment system. That is bad. People who have lupus and other auto-immune disorders know that they have those conditions when they buy insurance. This knowledge shapes their buying choices. People with lupus and other auto-immune disorders will rationally select plans that are good at taking care of people like them. This is adverse selection and adverse selection is what risk adjustment is intended to mitigate.

CMS should still be trying to figure out how to resolve this real risk adjustment problem. But more broadly, this is a good illustration of a challenge of risk adjustment when dealing with shocks. Risk adjustment is usually based on previous years claims looking something like current year claims. However shocks do get introduced into the market. Sometimes they are technological shocks like the introduction of the Hep-C antivirals. And sometimes they are information and political shocks that are short run fads like HCQ. In either case, these are significant discontinuities that lead to really weird risk adjustment results that pervert incentives and reward behavior we want to penalize and penalize behavior that we want to encourage.