Administrative Data Systems
There is growing use of administrative data in social science and epidemiological research, particularly on topics related to health and aging. Critics of the uses of these data rightly point out that excessive reliance on administrative data lead to studies asking questions that are limited to where the data are. They bemoan the fact that investigators are only looking underneath the street lights when the more informative studies must, of necessity, force us to examine issues that have been, until recently, in the dark. While this is clearly true and by no means should primary data collection efforts be abandoned, it is the case that studies using administrative data have unique advantages and can be powerful complements to primary data collection efforts.
Qualitative studies can be greatly enhanced by placing them into context with the use of population-based administrative data, where administrative data are used to identify sources of variation that beg further investigation using surveys and case studies. Insights from the latter generate hypotheses that can then be tested using administrative data.
Investigators’ consideration of the best approach to address a research question inevitably requires balancing design issues such as the representativeness of the sample, the ability to recruit, and the cost of data collection given the size of the sample needed to test the study hypotheses. Because both response rates to household surveys and the representativeness of telephone surveys are declining, the trade-offs between the advantages and disadvantages of surveys versus administrative data may be changing. Notwithstanding the complexities of securing permission from survey respondents to link their responses to their administrative records, increasingly the advantages of administrative data are being added to the advantages of surveys by linking these two powerful sources of data. The power of these combined data sets to address a broad array of issues is evident in the large number of publications that have emanated from such studies.
What is unique about the data systems being assembled by investigators at Dartmouth and Brown is their longitudinal and hierarchical character. Since providers are located in space and Medicare beneficiaries’ zip codes are known, studies linking surveys to Medicare claims could also be used to examine individuals’ choice of provider and whether the forces that appear to shape that choice vary from market to market or from state to state. This ability to marry the “demand” side of health care utilization (individuals’ wealth, education, insurance and even preferences derived from surveys) with “supply” side information on provider quality, the medical care “culture” prevalent in the market, and relevant state policies will be the new frontier in social science research in health and aging.