New Evidence on the Persistence of High Health Spending
While the degree of concentration in health care spending has been well-documented by previous research, most studies have had short follow-ups and/or small sample sizes. A better understanding of how spending patterns may be both concentrated and persistent over time in the privately-insured population can help inform the design of insurance plans. This is particularly salient for insurance companies tailoring plans to market in health insurance exchanges, as patterns of spending could render some designs more efficacious than others. Towards this end, we examine trends in medical expenditures in the privately insured population over six years to better understand the concentration and persistence of spending.
Methods:
Using the TruvenHealth MarketScan database of administrative claims from employers, we characterize the concentration and persistence of medical expenditures in the privately-insured under-65 population from 2003-2008. For each year, we classify individuals based on their spending percentile as high (top 10%), moderate (10-30%), or low (bottom 70%). We track movement across each of these categorizations to understand the degree of persistence in spending. Individuals are labeled as usually low, low/moderate, sometimes high, often high, and usually high based on the number of years they are labeled as either low or high spenders. Logistic regressions are done to determine which baseline characteristics significantly predict future spending levels and overall six year trends.
Findings:
In this sample of over 2.1 million privately insured individuals who continuously remain in the database, there is a great deal of persistence at the tails of the spending distribution. 69.8% of the sample is never characterized as high spenders, while 34.4% of those who were top spenders in 2003 had the same relative spending level in 2008. While more than 75% of individuals who were labeled as high spenders at least once did not exhibit persistence of high spending, the level of continuous high spending was substantial for the rest. Overall, the degree of concentration decreases only modestly as the time frame expands. The top 10% of spenders in 2008 accounted for 62% of the expenses, while from 2003-2008, the top 10% of spenders accounted for 49% of all expenditures. The percentage increase in expenditures over the six years was greatest for the often high(3-4 years of high spending) spending group. Baseline characteristics associated with increased likelihood of being a high spender in subsequent years include enrollment in a PPO, a range of comorbidities, age, female gender, and location.
Conclusions:
We find evidence of a great deal of concentration and persistence over six years in this sample of over 2.1 million privately insured individuals from 2003-2008. While many individuals are excluded from our sample because they do not remain continuously enrolled, analysis of sample attrition suggests that the resulting bias may be limited. Although the MarketScan database is not nationally representative, we are able to develop a better understanding of spending patterns for privately insured individuals across the country. This furthers knowledge regarding the distribution of health care spending over time and could improve insurance plan design.