اثر گسترش دسترسی به بیمه درمانی بر سلامت و مرگ و میر برای بهره برداران بیمه از کار افتادگی تامین اجتماعی

We use a social experiment to estimate the impact of expanding health insurance coverage on the health and mortality of newly entitled SSDI beneficiaries who lacked health insurance. Our intent-to-treat estimates show that expanding health insurance has significant effects on self-reported health at one year following health insurance enrollment, positive effects on mental health and physical health at one year following health insurance enrollment, and no significant effects on mortality within our observation period. About 30% of control group members obtained health insurance coverage within one year following enrollment. We use two-stage least square models to estimate the impact on those who would not have obtained health insurance coverage without the demonstration project and find larger effects on self-reported health and mental health among these SSDI beneficiaries. Finally, we examine distributional impacts to show that mean impacts miss meaningful effects.

Several recent studies conclude that the information on the impact of health insurance on health is limited, and have argued for social experiments as a means of gathering reliable data to inform policy.1 We use a social experiment to estimate the impact of expanding health insurance coverage on the health and mortality of newly entitled Social Security Disability Insurance (SSDI) beneficiaries who did not have health insurance when we contacted them to participate in the study. The data for our analysis is from the recently completed Accelerated Benefits (AB) demonstration project funded by the Social Security Administration (SSA) and conducted by MDRC. The AB demonstration used a random assignment design to estimate the costs and benefits of providing SSDI beneficiaries who did not have health insurance coverage with a health benefits package during the SSDI program's 24-month Medicare waiting period. Early results from the demonstration show that the AB health benefits package increased health care utilization and reduced unmet health care needs among beneficiaries. The AB health insurance package led to a 22 percentage point (46%) increase in beneficiaries who received a diagnostic test; a 10 percentage point (50%) increase in beneficiaries who underwent surgery; an 18 percentage point (25%) reduction in beneficiaries reporting any unmet medical need; and a 40 percentage point (53%) reduction in beneficiaries reporting an unmet need for a prescription drug (Michalopoulos et al., 2011). We examine whether these early findings translate into improvements in health within the first year and reduced mortality among beneficiaries within the first few years following random assignment. We find evidence that offering the health benefits package led to improvements in health among beneficiaries we targeted for the study. We find at least a 10 percentage point (nearly 30%) reduction in the percentage of beneficiaries who reported poor health in response to a commonly used health question. We find a 16 percentage point (88%) increase in the percentage of beneficiaries reporting their health as somewhat better or much better after the first year of receiving the health benefits package. Our analyses of the Qualitymetrics Short Form 36 (SF-36) health indices show small mean impacts for both the physical health measure and the mental health measure (1.5 and 2.4 points respectively).2 While the mean impacts appear small, we find statistically significant differences in the distributions of the SF-36 physical and mental health measures. This result is driven by the differences between the 30th and 50th percentiles of the distributions for the SF-36 physical health scores, and the 50th and 80th percentiles for the SF-36 mental health scores, rather than differences in the tails. The relatively larger impacts in these areas of the health distribution are along two meaningful health margins:mental health scores indicative of clinical depression, and physical and mental health scores that are indicative of meeting the SSDI disability standard. Our intent-to-treat (ITT) estimates show a 9 percentage point (14%) reduction in SF-36 mental health scores that are indicative of clinical depression and a 7.2 percentage point (13%) reduction in SF-36 scores indicative of an SSDI disability. Thus, we conclude that solely focusing on mean outcomes using SF-36 physical and mental health measures could potentially understate the benefits of expanding health insurance coverage to newly entitled SSDI beneficiaries. Approximately 30% of control group members reported that they had obtained some type of health insurance coverage in the 12-month follow-up survey. Thus, our estimate of offering the AB health benefits package might understate the impact of the health benefits package among those who are unable to obtain health insurance during the first year of the Medicare waiting period. We use a two stage least squares (2SLS) model to estimate the effect of the AB health benefits package among those unable to access health insurance within a year of the waiting period. When compared to our ITT estimates, our 2SLS estimates are substantially larger for our self-reported health outcome measures, the mental health score, the percentage of people with a mental health score indicative of clinical depression, and the percentage of individuals with a combination of physical and mental health scores indicative of an SSA-defined disability. We do not find evidence that expanding access to health insurance coverage led to a statistically significant reduction in mortality within the 2–3 years following enrollment in the AB project. These findings appear to be consistent with those presented in the summary of the literature in Levy and Meltzer (2008). We conclude that if there were impacts on mortality within this time period, the impacts were too small to detect given the sample size for our study. Our analysis is limited to newly entitled SSDI beneficiaries who do not have health insurance at the time of recruitment, and they are not representative of broader populations. Specifically, they have a health condition that is severe enough to qualify for SSDI benefits. While our sample is not representative of the entire population of SSDI beneficiaries or even uninsured individuals, our results provide rigorous evidence for a well-defined subgroup that would likely be affected by changes in health policy envisioned in the Affordable Care Act (ACA). In addition, our sample size may not be large enough to detect some dimensions of health insurance effects that are important to some policymakers. However, our sample size is large enough to show significant and substantive impacts of expanding access to health insurance on the health of beneficiaries.

The AB demonstration project provides a starting point for an assessment of the potential effects of expanding health insurance coverage under the ACA. Our baseline data on household income show that about 20% have household income below 133% of the poverty line for a single-family household, and over 70% have household income below 400% of the poverty level (Weathers et al., 2010). These individuals will qualify for health insurance subsidies under the ACA, and our data show that expanding access to health insurance may lead to substantial improvements in mental health in the short-run. When we supplement our evidence on mental health improvements with evidence suggesting that poor mental health may contribute to a number of secondary health conditions, we conclude that our short-term findings on mental health may lead to longer-term reductions in other health impairments and may lead to longer-term improvements in mortality. Our findings on the impact of health insurance on the reduction in SF-36 scores indicative of an SSDI disability suggest that the ACA may also have an impact on reducing the SSDI benefit rolls. There are caveats to our study. Several of our measures are based on survey responses, which are subjective in nature. The self-reported health measures come directly from questions asked in the baseline and 12-month surveys. Our PCS and MCS scores are computed based on responses to a subset of questions asked in the 12-month survey; however, we do not have a baseline comparison for these two measures of health from which to compare the results at 12 months. Additionally, our sample size may be too small to detect significant impacts on certain outcome measures such as mortality. Lastly, while we did not find any evidence of the AB health benefits package lowering mortality rates, this may be due to the short time frame (three years) of observation. The effects of providing health insurance and additional services during the waiting period may only show up over a longer time frame. As additional Centers for Medicare and Medicaid Services (CMS) and SSA data become available over time, we plan to examine long-term costs and benefits of the project. For example, we will use CMS Medicare expenditure data to examine whether increased access to health insurance during the 24-month Medicare waiting period results in reductions in Medicare expenditures and assess whether there are Medicare expenditure patterns that are consistent with mental health related secondary health conditions. We will test whether expansions in access to health insurance will have a long-term impact on mortality similar to the positive short-term impact on the health of participants. Finally, we will use SSA administrative data to estimate the impact of health insurance coverage on exits from the SSDI program and savings to the Social Security trust funds.