Administrative Data

Administrative data are collected by organizations and agencies expressly for the purpose of conducting administrative tasks and meeting administrative responsibilities for that organization or agency (e.g., evaluate program performance, agency accountability). These tasks and responsibilities may include contacting individuals within the system, tracking resource utilization, reporting counts to an oversight agency, and projecting trends for resource allocation. Examples of administrative data sources include health maintenance organizations, Medicare and Medicaid programs, vital records administrations, school health systems, hospitals, and [Page 4]health insurance providers. Although administrative data are collected through a system designed for nonresearch purposes, they can be very useful in epidemiologic research and have been used extensively for this purpose. There are, however, limitations to using administrative data in epidemiologic studies.

The most apparent advantage of administrative data is their availability. Abstracting medical records or conducting surveys to collect the same data in a comparable sample size may not be possible due to time and financial constraints. Administrative data sets often contain records spanning decades. Such data are gathered prospectively by the organization or agency and, therefore, can then become a source for nonconcurrent prospective investigations by the epidemiologist. Access to these data is also relatively easy, since most administrative systems are now computerized and data can be transferred onto electronic media for sharing. Administrative data generally are not for public use and require a research proposal submitted to the agency or organization that collects and oversees use of the data, and sometimes a fee is associated with requests. Administrative databases also tend to be relatively large, allowing for subgroup analysis by various factors such as geographic subdivision, gender, ethnicity, and age. The size of a data set is due mainly to the inclusiveness of the administrative records—everyone in the system is included in an administrative data set. This inclusiveness is another advantage of administrative data as it minimizes the prospect of selection bias.

Although there are no data collection or data entry costs associated with administrative data, data cleaning can still be a time-consuming task on a very large administrative data set. In addition, an administrative data set will seldom have all the data elements needed for a specific epidemiologic study, raising concerns over residual confounding, and variables often need to be recoded. Another disadvantage can be the presence of missing data on key items, although this varies by data source. Finally, as most administrative data sets are not population based, study results from analyses of administrative data may not be generalizable to the larger population or community.

An archetypical example of a population-based administrative data resource that is commonly used in perinatal and pediatric epidemiology is state or provincial birth records. In addition to being able to identify and select a cohort from birth records, birth certificates have a variety of data elements that can be useful in studies of adverse neonatal and childhood outcomes. The content of birth records varies by state or province but usually includes the essential elements of name, date and time of birth, sex, race, parents’ names, and place of birth. Some will also include birthweight, gestational age, and age, occupation, and race of the parents. Most epidemiologic studies using birth records for secondary data analysis are cross-sectional or case-control in design. Some have linked birth records of siblings to prospectively examine birth outcomes that may be associated with a previous adverse birth outcome, such as low birthweight. The quality and completeness of data items included on birth records are known to vary widely.

...