This is a guest post from Adrianna McIntyre, a graduate student studying health policy at the University of Michigan. Adrianna currently blogs at Project Millennial, tweets at @onceuponA, and will be joining TIE soon as the team’s first intern.
Can insurance be used to drive a more efficient health care system? One feature of health coverage commonly touted as a “flaw” is moral hazard, the tendency for patients to overuse low-value care when insured, adding unnecessary expenses to our system.
But that isn’t the whole story, according to Katherine Baicker and colleagues in an NBER working paper (ungated). The authors propose a new kind of hazard—“behavioral hazard”—to describe underutilization of high-value care caused by behavioral biases and mistakes. When properly accounted for, behavioral hazard suggests that optimal copayments for some high-value treatments should be zero (or even negative) to offset efficiency reductions caused by exaggerated underuse.
The areas in which we observe substantial underutilization — such as management of chronic diseases like diabetes, high blood pressure, asthma, and high cholesterol — are responsible for a substantial share of total health care utilization. Much of the cost of such diseases occurs late in the progression of the condition and likely involves overuse (moral hazard or positive behavioral hazard) following underuse early in the progression of the condition. That many of these domains of care seem sensitive both to small changes in copayments and potentially to behavioral nudges — and that many of the treatments affected seem to be of high health value — suggests that incorporating not only moral hazard but behavioral hazard into our models of optimal insurance design may have large-scale implications for public policy.
Various psychological biases that drive behavioral hazard. “Present bias” refers to our tendency to discount benefits in the future when we have to pay costs of care up front. This may contribute to underuse of prenatal care; despite very low cost barriers, research suggests that 30% of Medicaid-eligible women don’t take advantage of coverage during the first trimester of pregnancy, and many enroll instead at point of birth.
“Salience” is another issue that traditional insurance doesn’t account for. At the same price point, we’re more inclined to observe adherence with pain medication than antidiabetics—patients value the pain medication more highly because it’s treating a wildly more salient symptom. If we want to encourage drug adherence with the antidiabetics, the authors suggest “it is possible that co-pays for drugs treating chronic conditions (e.g., antidiabetics) should be lower than co-pays for symptom-alleviating drugs.”
The model is applied to empirical work regarding the effects drug copays have on recent heart attack victims. In the original study, Choudhry and colleagues found that eliminating copays resulted, on average, in patients using $106 more in cardiovascular prescriptions. An analysis of the moral hazard associated with this uptick suggested that only $26.50 of that is added health value (the remainder being excess utilization). However, Baicker et al. suggest that this analysis may not have properly valued health gains, due to behavioral hazard:
The increase in prescription drug use was associated with significantly improved clinical outcomes … We apply the commonly used estimate of a $1 million value of a statistical life to the reduction in the mortality to get a crude measure of the dollar value of health improvements. This implies that the elimination of copays leading to a .3 percentage point reduction in mortality generates a value of $3,000. This $3,000 improvement substantially exceeds the standard model’s prediction of $26.50, suggesting large negative behavioral hazard.
Incorporating behavioral hazard could be a key component in value-based insurance, but it comes with challenges. Medication adherence is one of the more straightforward treatment issues to quantify, because costs (including side-effects) and benefits are relatively easy to measure. Unobservable effects and preferences complicate the model. Still, considering the widespread implications for chronic disease—and related costs as boomers age into Medicare—it’s worth keeping on the radar.
