Every now and then we write a post regarding research methods. I believe this is one of the most important areas of health research, because it helps validate and guide all the other health research. The good people at the Bureau of Economic Analysis have published a paper in which they compared survey data with claims records as a method to estimate spending for persons covered by employer health plans. (BEA Paper) The survey data was from the well-known Medical Expenditure Panel Survey and the claims database was the one compiled by what is now Truven, consisting of all claims for persons covered by some health plans and some large employers. The MEPS data is self-reported with some verification of provider prices, but not quantities, and covers about 15,000 individuals. The commercial database covered about 28 million. The researchers looked at total spending, both by a payer and out-of-pocket.
The total spending estimate for the entire commercial population was about 10% higher using the Truven database than the MEPS data, and the per person average was higher as well, at around $2740 for the commercial database and $2470 for MEPS. This appears to be in part due to the fact that the MEPS data was not picking up outliers with very high spending and in part to general underreporting and consequent underestimation of utilization in the survey data. Some of this could be due to institutionalization of high-cost individuals or end-of-life care. In any event, it appears that large commercial databases may do a better job of estimating total health spending for the employer health plan-covered population. As there are now some even larger aggregations of data from these commercial health plans, it might be possible to achieve very high degrees of accuracy. And since MEPS data is often used in the National Health Expenditure Accounts and in AHRQ statistical briefs, the underreporting of spending is a concern in regard to the accuracy of those reports.