Administrative Data Systems
6. Advantages and Disadvantages
Administrative health data are clearly more like demographic data available from the census than the rich, nuanced information that comes from direct observation. On the one hand, administrative health data contain clinical detail, chronological sequencing, powerful linkages between clinical and geographic detail, and health care utilization data that are situated in place, time, and person.
On the other hand, virtually completely missing is information about the social roles and functions that so often influence patients' decisions about health care utilization, regardless of clinical situation.
Advantages. The primary advantage of administrative data systems is that in many cases they are population-based or at a minimum based upon a clearly defined population. With the correct linkages, population surveys and even population-based registries can greatly enhance the available information about clinical or social characteristics of the population. Here are some examples.
- The National Long Term Care Survey was conducted periodically over two decades using the Medicare/Medicaid population base as a sampling frame (Manton, Vertrees et al., 1990; Manton, Woodbury et al., 1993).
- The Longitudinal Survey on Aging, conducted during the 1980’s, matched individual respondents’ Medicare beneficiary numbers to their surveys (Mor, Wilcox et al., 1994; Wolinsky, Krygiel et al., 2002).
- The SEER/Medicare match files, created by the National Cancer Institute in collaboration with CMS, has repeatedly shown the value of integrating two population based data sources designed for totally different purposes (Nattinger, Gottlieb et al., 1992; Polsky, Armstrong et al., 2006).
- The Medicare Beneficiary Survey (MCBS). Integrating primary survey data collection with Medicare beneficiary administrative records and claims, the MCBS identified beneficiaries who’d switched to Managed Care companies that weren’t required to submit administrative medical claims data, which has allowed numerous comparisons of fee for service and managed care (Chulis, Eppig et al., 1993; Eppig and Chulis, 1997).
In the “run up” to the introduction of Medicare’s Part D prescription drug act, the MCBS was one of the only sources of population-based data which could be used to estimate how beneficiaries would behave with universal drug coverage (Gilman and Kautter, 2008).
With the proper arrangements and assurances, these data can also often be encoded as belonging in markets which can be easily located in states or other political boundaries which influence the manner in which health care is delivered (Roos, Mustard et al., 1993). This ability to aggregate individuals’ utilization experience makes it possible to estimate the effect of being served by a provider, or residing in an area, characterized by high rates of aggressive medical care, for example (Wennberg, Fisher et al., 2004b). To the extent that individuals’ health care experience varies as a function of geographic region, as Wennberg and his colleagues have repeatedly shown, having “only” a nationally representative sample of individuals, even if one is able to locate them geographically, displaces the ability to construct contextual variables which are only possible if individual level data about the population of events, for example hospitalizations, are aggregated to the level of the area or the provider (Wennberg, Freeman et al., 1989; Wennberg, Fisher et al., 2004b).