: To the extent that high health care spending persists over time, problems of adverse selection would be accentuated in private insurance markets based on annual contracts, exacerbating concerns about the availability of affordable coverage. Despite the importance of understanding the persistence of medical spending, surprisingly little is known. Several studies are based on Medicare data, limiting generalizability to the under 65 population. Studies representing younger persons have limitations, including the lack of recent data, short follow-up, single employer or insurer samples, and lack of attention to attrition from employer-based coverage. New knowledge regarding spending persistence can improve the evidence base underlying many policy options for covering the uninsured, including subsidized reinsurance markets, Medicare buy-ins, and incentives to expand the role of individual insurance contracts and consumer-driven health plans. We will assess the persistence of health care costs using the 2002-2007 Thomson Healthcare MarketScan Database, which includes claims data for individuals insured through over 100 large firms, and the 2001-2006 waves of the MEPS. Spending categories will be based on cut points in the annual spending distribution (e.g., top 5%). The primary analyses will focus on individuals under age 65. Subanalyses will be done by age/gender group and for selected clinical conditions and Medicare-eligible retirees. Predictors of cost patterns will include demographics, health plan design, and comorbid conditions. Because the MEPS has a limited sample size and follows households for only two years, the primary analyses will use MarketScan data to provide a longer panel on a large sample. Over 3 million people can be followed for at least 5 years. The MEPS will be used to assess the generalizability of the MarketScan findings and the effects of attrition from employer coverage on estimates of spending persistence based on private insurer claims. The analytic approach is to characterize the run-up to high spending periods and the persistence of high spending over time by 1) describing the cost distribution, transition probabilities, and spending correlations across categories over time (e.g., from top 50% to top 5%), and 2) modeling the predictors of transitioning between distributional categories as a function of employee and household characteristics, and the predictors of attrition from the MarketScan database. The resulting estimates can inform policy simulations addressing issues such as the cost of reinsurance under different stop-loss thresholds, and the potential magnitude of adverse selection in expanded individual insurance markets or Medicare buy-ins. We will provide an example of how such estimates can inform the evaluation of risk-adjustment and reinsurance as mechanisms to improve the functioning of small group health insurance markets.
New knowledge regarding persistence of health spending can improve the evidence base underlying most of the policy options for covering more of the uninsured, including subsidized reinsurance markets, Medicare buy-ins, and incentives to expand the role of individual insurance contracts. To provide such evidence, we will assess the persistence of health care costs using the 2002-2007 Thomson Reuters MarketScan Database, which includes claims data for individuals insured through about 100 large firms, and the 2001-2006 waves of the MEPS. 1
| Hirth, Richard A; Gibson, Teresa B; Levy, Helen G et al. (2015) New evidence on the persistence of health spending. Med Care Res Rev 72:277-97 |
