A recent study by Amelia Haviland and colleagues found that consumer-directed health plans (CDHPs) offered by employers are associated with reductions in health care spending over three years. CDHPs are high deductible plans coupled with tax-advantaged personal medical spending accounts.
The study, published as an NBER working paper, examined changes in health spending for individuals at firms that began offering CDHPs, compared to those at firms that didn't offer CDHPs. It's an intent-to-treat design, as it considers individuals in the "treatment" group if they worked at a CDHP-offering firm, whether they enrolled in one or not. (The study also employed a second analysis using an instrumental variables design that I'm not covering in this post.)
The design eliminates selection bias at the individual level that could arise, for example, from healthier individuals opting into CDHPs. The authors also argue that it reduces firm-level selection bias because "offering a CDHP alongside other plans is a moderate step relative to full replacement with a CDHP." That is, perhaps firms that institute a full switch to a CDHP are systematically different from firms that do not. Arguably, firms that offer CDHPs alongside other plans are more similar to firms that do not offer one at all.
The data are from 54 large U.S. employers, span 2003-2007, and include about five million and eight million person-years of data from CDHP-offering and non-offering firms, respectively.
The investigators found that firms that offered CDHPs experienced 6.6%, 4.3%, and 3.4% lower annual spending in the first three years after offer, respectively, and relative to non-offering firms.* The spending reductions were concentrated in outpatient care and prescription drugs, not inpatient care or the ED.
Relative to non-offering firms, annualized spending growth on pharmaceuticals is 5 to 9.5 percentage points lower in the three years after firms offer CDHPs (p < 0.01) and spending growth on outpatient care is 3.0 to 6.8 percentage points lower in the first three years though the estimate loses statistical significance in the third year (p < 0.05 in first two years). In contrast, for inpatient cost growth, we have only marginally statistically significant evidence of lower spending relative to non-offering firms in the first two years of CDHP offer (p < 0.10) while the third year estimate is non-significant and very close to zero. Finally, we do not detect any differences in cost growth for emergency department (ED) care in any of the first three years of CDHP offer although, due to high variance in ED spending, estimates are imprecise.
The findings are consistent with some prior work, including that of Paul Fronstin, Christopher Roebuck, and Martín Sepúlveda. They examined one, large, midwestern manufacturer that fully replaced its PPO with a CDHP. Like the Haviland study, this one also found sustained reductions in health care spending concentrated in outpatient care and prescription drugs. Also like the Haviland et al. study, Fronstin and Roebuck found lower savings over time.
One can speculate why CDHP savings might erode over time. My own hypothesis is that perhaps early on, CDHP enrollees try relatively hard to forgo care, saving their money under the deductible for only things that seem urgent. This may also have the effect of putting off some care they needed, to the extent that people can't easily distinguish between necessary and unnecessary care. In turn, perhaps in subsequent years there was a relief of this self-imposed, if accidental, pent-up demand.
Because the study focused on workers for large employers, findings may not generalize to other insurance markets (like individual market exchange plans) or other populations (like non-working, Medicaid enrollees).
* When restricting to firms in the sample for all three years, savings also decrease over time, but the differences are not statistically significant. Thus, we can't completely rule out the possibility that compositional changes explain reduced savings over time.
Austin B. Frakt, PhD, is a health economist with the Department of Veterans Affairs and an associate professor at Boston University’s School of Medicine and School of Public Health. He blogs about health economics and policy at The Incidental Economist and tweets at @afrakt. The views expressed in this post are that of the author and do not necessarily reflect the position of the Department of Veterans Affairs or Boston University.