cover large populations. Integrated systems provide data on all levels of inpatient and outpatient care, from primary care through highly specialized tertiary care. In such systems, computerized information systems are extensively used for a variety of patient care and administrative functions. The availability of computerized information covering the entire spectrum of medical care for large populations provides opportunities for many types of health research, including public health surveillance, epidemiological studies, and health services research.
The continued growth of computerized medical information systems in managed care organizations should improve the efficiency with which such studies can be conducted. At present, however, computerized databases are not sufficiently evolved to negate the need to obtain additional confirmatory or supplemental information from hard-copy records or interviews with patients. The key elements of medical information systems for use in population-based health research include (1) mechanisms for the identification and progressive tracking of individuals in a defined population, (2) procedures that ensure the accuracy of the diagnostic and other clinical data entered into an electronic database, and (3) development of uniform data sets with information on demographic variables (e.g., race and ethnicity) and other risk factors, in addition to diagnostic and treatment information.
The Centers for Disease Control and Prevention (CDC) effort to study rare adverse events that occur after the receipt of vaccines serves as an example of how computerized databases of various large health maintenance organizations (HMOs) can be linked into a single research database. The Vaccine Safety Data Link (VSD) is a project of CDC's National Immunization Program, which began in 1991 as a collaboration between four federal agencies and four long-established staff model HMOs. VSD links together the computerized databases of the four HMOs to provide primary information for research on vaccine safety and other health issues. The data fields include population and vaccination status, as well as health outcomes (diagnostic codes from hospital discharges, emergency room visits, and clinic outpatient visits), laboratory test results, and pharmacy records.
Staff model HMOs are good partners for this project because they provide and maintain detailed records on all levels of patient care, including cost data. They also have a stable, identifiable patient population from which subjects can be selected, and they are able to implement interventions (and ensure follow-up) on a systemwide basis. One of the HMO partners, for example, is conducting a systemwide study of pneumococcal polysaccharide vaccine in which revaccinated patients are compared with patients being vaccinated for the first time for immunogenicity and reactogenicity.
The databases of managed care organizations, however, also pose challenges to clinical research because of a lack of uniformity and their orientation toward administrative and management purposes rather than clinical or research purposes. Thus, adaptation of these databases for use in research can be expensive and time-consuming. In addition, many HMOs do not routinely record data on important variables, such as race or ethnicity, consanguinity, socioeconomic