Original articlesA Research Registry: Uses, Development, and Accuracy
Introduction
Although large databases are increasingly used for epidemiologic and health services research, the role of registries in facilitating such work has received little discussion. Population registries maintained by an insurer normally include information on each individual’s place of residence, insurance coverage, date of birth, and date of death. Information on place of residence enables the study both of mobility within a catchment area (such as a province) and of in-migration and out-migration for all people enrolled. Marital status and family information may also be provided. The use of one or more identification numbers allows the insurer to contact individuals and families as appropriate (names and addresses can be removed when data are released for research purposes).
The effectiveness of population registries generated for administrative purposes is multiplied when such registries (typically relevant for just a short duration) are combined into a research registry to define accurately the health insurance status for each individual over many years [1]. Administrative data on services and stays alone cannot distinguish between “well” populations (no contact with the health care system), loss to follow-up (ineligibility associated with leaving the insurance plan), loss of continuity (two or more registrations over time for the same person that remain unlinked), and mortality.
A carefully maintained registry is particularly helpful in dealing with the possibility of loss of continuity and associated subject misidentification. If an individual’s identifiers are incorrect, loss to follow-up and mortality may be misassigned; information on each enrolled individual is not appropriately updated. For a given cohort, loss of continuity leads to underestimating the frequencies of events based on hospital discharge abstracts or physician claims. To illustrate, in one Alberta study, a lack of adjustments for women who were not correctly identified, because they either left the province or changed their health care numbers, led to a serious underestimate of the risks of cancer following breast augmentation [2].
This article describes our experience at the Manitoba Centre for Health Policy and Evaluation (MCHPE) in developing a research registry from files used since 1970 by Manitoba Health. We present some uses of the research registry, discussing its development and quality. The centrality of the registry in our work is highlighted in Figure 1.
Section snippets
Provincial registries
Canadian provincial health departments maintain a comprehensive registry of all residents in their own particular provinces. Ineligible residents include inmates of federal penitentiaries, armed forces, and Royal Canadian Mounted Police (RCMP) personnel, foreign students, and wards of the province such as some residents of mental health facilities. Because residents of each province continue to receive benefits for 3 months after moving to another province, they are also included in the file.
Common Tasks
The Manitoba research registry has been designed to support several analytical, data retrieval, and record linkage tasks; the software described in Table 1 highlights many of these. Further restructuring of the data can meet other requirements without revisiting the original source files.
Being able to retrieve registration data at any point in time has enabled researchers to do the following:
- 1.
Track the movement of Manitoba residents within the province as well as migration to or from other
Development
The research registry was initially designed to resolve loss to follow-up and date-of-death problems; in carrying out this work, “loss of continuity” errors were found to affect more than 20% of the population from 1974 to 1977 [12]. Loss of continuity may occur when Manitoba Health issues new family REGNOs to individuals at age 18 or following a change in marital status. Because family REGNOs generally follow males (heads of household), tracing women using just REGNOs involves accurately
Confidentiality, security, and access
Taking the necessary steps to ensure confidentiality and security of data has been important since the registry began. As a data repository, the university (and MCHPE as part of the university) has assumed the delegated obligation to ensure the secure custody of these records. Confidentiality is maintained through a number of procedures and security measures. First, all records deposited in the registry have been processed by Manitoba Health to remove names and addresses. Second, file layouts
Checking for quality
Standardized systems for updating and reviewing the quality of registry data are critical. Because of the care in its construction, the Manitoba research registry is generally taken as a “gold standard” against which other data files can be checked. This registry is especially helpful for combining different files (e.g., physician, hospital, nursing home), identifying and resolving discrepancies that arise from file comparisons, and filling in missing information. Examples 1 to 3 (presented
Discussion
The research registry requires significant continuing investment but provides a tremendous overall saving when the information is used across many studies. Almost all the population health and clinical epidemiologic studies based on Manitoba Health data utilize the research registry 17, 18, 19. The evaluation of systems designed to provide defined services (such as immunizations) to populations is facilitated by such a registry [20]. Studies increasingly rely on 1 or more years of data for case
Acknowledgements
Results and conclusions are those of the authors, and no official endorsement by Manitoba Health is intended or should be inferred. This work was supported by HEALNet (Canadian Networks of Centres of Excellence Program), by a National Health Research and Development Program Career Scientist Award to Dr. L. L. Roos, by the Manitoba Centre for Health Policy and Evaluation, and by the Canadian Institute for Advanced Research. The authors gratefully acknowledge the assistance of the St. Boniface
References (27)
- et al.
Age-specific education and income gradients in morbidity and mortality in a Canadian province
Soc Sci Med
(1997) - et al.
A needs-based methodology for allocating health care resources in Ontario, CanadaDevelopment and an application
Soc Sci Med
(1991) - et al.
Using administrative data for longitudinal researchComparisons with primary data collection
J Chronic Dis
(1987) - et al.
The art and science of record linkageMethods that work with few identifiers
Comput Biol Med
(1986) - et al.
Surveillance of vaccine-related adverse events in the first year of lifeA Manitoba cohort study
J Clin Epidemiol
(1996) - et al.
Registries and administrative dataOrganization and accuracy
Med Care
(1993) - et al.
Breast implants and breast cancerRe-analysis of a linkage study
N Engl J Med
(1995) - et al.
Project on Matching Census 1986 Database and Manitoba Health Care FilesPrivate Households Component (Analytical Studies Branch Research Paper Series, No. 91)
(1996) - et al.
Treatment and health outcomes of women and men in a cohort with coronary artery disease
Arch Intern Med
(1997) - et al.
Managing health servicesHow administrative data and population-based analyses can focus the agenda
Health Serv Manage Res
(1998)
Deprivation and Health in Scotland
Proxies for health care need among populationsValidation of alternatives—a study in Quebec
J Epidemiol Community Health
Measuring the health of the population
Med Care
Cited by (206)
Metabolic Acidosis is Associated With Acute Kidney Injury in Patients With CKD
2022, Kidney International ReportsAssociation of Membership at a Medical Fitness Facility With Adverse Health Outcomes
2021, American Journal of Preventive MedicineAn unconditional prenatal cash benefit is associated with improved birth and early childhood outcomes for Metis families in Manitoba, Canada
2021, Children and Youth Services ReviewChanges in health indicator gaps between First Nations and other residents of Manitoba
2021, CMAJ. Canadian Medical Association JournalCitation Excerpt :Researchers at the Manitoba Centre for Health Policy partnered with the First Nations Health and Social Secretariat of Manitoba to access a deidentified research file from the First Nations and Inuit Health Branch that comprises data from First Nations individuals in Manitoba who are registered under the Indian Act15 as “status Indians.” Previous work has validated the Repository data and shown excellent linkage accuracy.16-18 The study population included nearly all residents of Manitoba registered for provincial health insurance in 2016.