the large “grey literature” of reports issued by advocacy 12 organizations, relying instead on academic, government, and foundation reports that meet accepted standards of scientific rigor. For studies evaluating the effects of tort reforms implemented in the states, this meant exclusion of study findings based solely on univariate or bivariate analysis, rather than a wellcontrolled multivariate regression analysis. A summary of data sources consulted is presented in Exhibit 3. For this update to the 2010 report, we reviewed an additional 109 studies and other papers on traditional and innovative tort reforms. Exhibit 3. Scope of Literature Search Source Sources and Scope of Search Legal literature Westlaw’s “Journals & Law Reviews” combined library, Jan 1990-Mar 2016; Social Science Research Network (unrestricted date search) Health and medical literature PubMed, Jan 1990-Mar 2016 Economics literature EconLit, Jan 1990- Mar 2016; Social Science Research Network (unrestricted date search); older articles cited in more recent works Government reports Reports collected on websites of the Congressional Budget Office, General Accountability Office, Office of the Assistant Secretary for Planning and Evaluation, Office of Technology Assessment (archive), and Agency for Healthcare Research and Quality, Jan 1990-Mar 2016; websites of the Washington State Task Force on Noneconomic Damages and Oregon Patient Safety Commission Foundation reports Reports collected on websites of the National Academy of Medicine, Robert Wood Johnson Foundation, Kaiser Family Foundation, Commonwealth Fund, and Pew Charitable Trusts (unrestricted date search) Other RAND Compare Dashboard9 2. Analysis of Paid Malpractice Claims Involving Elderly Patients 2.1. Methods 2.1.1. Data and Limitations Using data in the National Practitioner Data Bank (NPDB) Public Use File (PUF),10 we analyzed paid malpractice claims involving elderly patients that were reported to the NPDB between January 1, 2005 and December 31, 2015. Pursuant to the federal Health Care Quality Improvement Act of 1986, the Health Resources and Services Administration (HRSA) requires that liability insurers and certain other entities report to the NPDB all payments made on malpractice claims in the name of a healthcare practitioner. Reports must be made within 30 days of a payment. A restricted set of data fields is made available to the public in the PUF. The claims listed in the PUF for each year generally represent claims that closed with a payment in that year, although claims with a January reporting date may have closed in December of the previous year. The filing dates for these claims may have been as short as a few months earlier, or as long as a few years earlier. The dates of the incidents of alleged malpractice are still earlier: among the claims we analyzed, the mean lag time between the incident and reporting dates was 4.9 years (median: 4.0 years). 13 For simplicity, we confined our analytical sample to claims against MDs or DOs (PUF practitioner codes 010, 015, 020, and 025). Both trainees and non-trainees were included. We adjusted indemnity payments to 2015 dollars using the annual Consumer Price Index.11 Though it is the standard dataset used in the field of medical liability research for analysis of trends in malpractice claims, the PUF has several limitations. One set of shortcomings relates to what is collected by HRSA. The NPDB elicits information about claims only when they result a payment, so it cannot be used to explore trends in the total number of claims filed (including those dropped, dismissed, tried to verdict for defendant, and settled without an indemnity payment). Further, the NPDB does not collect information about payments made in the name of healthcare facilities. There is also believed to be some degree of underreporting to the NPDB, stemming in part from the practice of “corporate shielding,” in which an institutional defendant (for example, a large medical center) pays a claim in its own name only, dropping the physician so as not to trigger the NPDB reporting requirement for the physician.12 The magnitude of the underreporting problem is not known. However, there is no theoretical reason to think it is either disproportionately high or disproportionately low for claims involving elderly patients. Other limitations relate to how data are recorded in the PUF. First, in order to reduce the likelihood that a particular claim can be identified, indemnity payments are reported in the PUF as the midpoint of a dollar value range near their actual amount. For example, a payment of $318,000 would be recorded as the midpoint of the nearest $10,000 increment, or $315,000. This introduces measurement error into calculations of mean, median, and total payments—probably only a modest amount, because the dollar value ranges are fairly narrow. Second, some of the critical fields for this analysis, including patient age, were not made available in the PUF until end of January 2004. For that reason, our analysis is limited to 2005-2015. Third, patient age is reported in ranges rather than precise years. This complicates estimates of the volume of claims brought by Medicare beneficiaries, because the age band 60-70 spans the onset of Medicare eligibility on the basis of age. Our strategies for addressing this challenge are described