What impact do consumer-directed health plans really have?

Do “Consumer-Directed” Health Plans Bend the Cost Curve Over Time?

By Amelia M. Haviland, Matthew D. Eisenberg, Ateev Mehrotra, Peter J. Huckfeldt, and Neeraj Sood
National Bureau of Economic Research, March 2015, NBER Working Paper 21031


“Consumer-Directed” Health Plans (CDHPs) combine high deductibles with personal medical accounts and are intended to reduce health care spending through greater patient cost sharing. Prior research shows that CDHPs reduce spending in the first year. However, there is little research on the impact of CDHPs over the longer term. We add to this literature by using data from 13 million individuals in 54 large US firms to estimate the effects of a firm offering CDHPs on health care spending up to three years post offer. We use a difference-in-differences analysis and to further strengthen identification, we balance observables within firm, over time by developing weights through a machine learning algorithm. We find that spending is reduced for those in firms offering CDHPs in all three years post. The reductions are driven by spending decreases in outpatient care and pharmaceuticals, with no evidence of increases in emergency department or inpatient care.

From the Introduction

At the firm level, we find that CDHP offer is associated with an approximately 5 percent reduction in total health care spending in each of the three years after CDHPs were introduced relative to cost growth observed for non-offering employers. The long term decreases in spending are focused in outpatient care and drugs and there is little impact on inpatient or emergency department spending. If these effects are due only to changes in health care spending among those enrolled in CDHPs, they imply local average treatment effects for those enrolled in CDHPs of an approximately 15 percent reduction in total spending in each the first three years. Differences in impacts by CDHP plan structure are not statistically significant. However, consistent with our hypotheses, the pattern of the point estimates suggests that the impact of CDHPs is greater when paired with HSAs (versus HRAs) and when employers make smaller account contributions.

From the Summary and Discussion

This study substantially adds to our knowledge on the long term cost impacts of CDHPs. We estimated spending trends for three years across over 13 million people across the country in an analysis estimating CDHP impacts without the threat of individual level selection bias. We find that health care cost growth among firms offering a CDHP is significantly lower in each of the first three years after offer. This result suggests that, at least at large employers, the impact of CDHPs persists and is not just a one-time reduction in spending. However, an important caveat is that the decrease in spending may be smaller in year 3 compared to year 1 post-offer. Recognizing that the differences are not statistically significant, these results are suggestive and consistent with a decreasing impact of CDHPs over time.

The decreases in total spending growth observed are primarily due to reductions in spending on outpatient care and pharmaceuticals. In contrast, by the third year there are no differences in either emergency department or inpatient spending.

The results presented here are limited to large employers and therefore may not extend to Medicaid beneficiaries, the individual or small group market, or to the health insurance exchanges where, on average, deductibles and out of pocket maximums are higher and/or enrollees have fewer financial resources. While the firms in this study were specifically selected to have lower income employees, all families had at least one adult working full time with benefits so they are typically better off than families not offered employer sponsored insurance.

In summary, in the first large multi-employer study to investigate long term CDHP spending impacts we find reductions in health care cost growth in all three years post CDHP offer and do not detect increases in any component of health care spending. These findings do not support either the concern that decreases in spending will be a one-time occurrence or that short-term decreases in spending with a CDHP will result in increases in spending in the long term due to complications of forgone care. We cannot rule out either of these concerns developing over an even longer time frame.



By Don McCanne, MD

This study will no doubt be used to claim that high deductible health plans with health savings accounts (CDHPs - consumer-directed health plans) are effective in reducing health care spending without causing any harm. However, the conclusions that can be drawn are far more limited.

The observed reductions of spending by those offered CDHPs by their employers were in outpatient care and drugs. The nature of these services may enable shopping for lower prices, but they also are services that frequently are used for medical problems of lower acuity which enables patients to make decisions as to whether or not they will forgo the medical services and/or prescriptions offered. These services may be for important interventions that could improve quality of life or even longevity, or they could be for interventions that would have no significant impact on health, or they could be for interventions in between these extremes that might have only a modest beneficial impact such as transient symptom relief. From other studies it is known that patients decline not only care that they perceive to be of little value, but when faced with deductibles, they often do decline care that is clearly beneficial.

Those who would claim that this study shows that no harm was done after enrolling in a CDHP point to the observation that being in a CDHP did not increase hospitalizations nor increase the use of emergency departments during the first three years of enrollment.

The reason this conclusion should be challenged is based on the fact that the population studied was the relatively healthy workforce and their healthy families in healthy years of their lives. It is highly unlikely that a very modest decline in outpatient visits and prescriptions would have directly resulted in crisis care requiring emergency department visits or hospitalizations during the first three years on the program, and thus these outcomes were an insensitive indicator of harm. The rate of these interventions in the relatively healthy control group was the same, as would be expected. Also, by limiting the outcomes studied to only these two, the study remained insensitive to other potentially beneficial results of obtaining health care, if nothing more than relief on receiving reassurance over concerns that patients may have had about their health.

This study has the same limitation of the oft-cited RAND Health Insurance Experiment which also studied a healthy population for a limited time in healthy years of their lives. These studies may have intrinsic validity for the populations studied, but they do not have extrinsic validity, particularly for an older, sicker population that also has been shown to forgo care when faced with deductibles.

This study asked if CDHPs bend the cost curve over time (title). The study showed that spending by employers was reduced 5 percent in each of the three years following the introduction of CDHPs. Since not all employees were enrolled in CDHPs, they theorized that the reduction in spending for those in CDHPs was about 15 percent. Even there, a 15 percent reduction in spending on a population that has minimal need for hospitalization and emergency department visits - where much of our total health care spending lies - certainly does not equate to anywhere near a 15 percent reduction in our total health care spending.

Remember that 80 percent of health care is used by the 20 percent of people with serious health problems. Almost all of that spending is well in excess of the deductibles and thus is not sensitive to consumer shopping.

Further, since employers insure the healthiest and least costly sector of our population, the 5 percent savings that they gained is only a drop in the bucket of our total national health expenditures. You won’t see much bend in this cost curve, especially if you alter policies so that patients do receive the beneficial services that they should have, but might otherwise forgo.

If we are going to bend the cost curve, let’s not do it through methods that reduce beneficial health care services. Let’s do it though ways that eliminate wasteful spending, such as the administrative excesses imposed on us by the private insurers and public payers operating in a fragmented, dysfunctional system. Let’s enact a single payer national health program instead.