Policy analysis in an information-rich environment

Abstract

Population-based insurance systems using longitudinal administrative data and record linkage techniques have helped create “information-rich” environments in several sites around the world. The output of five research groups using administrative data (Oxford, Western Australia, and three Canadian centres: Manitoba, Ontario and British Columbia) was analysed from contacts with the research groups and through use of the National Library of Medicine's PubMed and Medical Subject Headings (MeSH) categories. MeSH words “utilization”, “economics”, “physicians”, and “physician practice patterns” more frequently characterized the research by the three Canadian centres than that of the other sites. With core funding for deliverables negotiated with the provincial health ministries, Canadian researchers have been more likely to use linked databases for policy analyses.

Manitoba examples highlight the capabilities associated with these information-rich environments. They include the ability to analyse interventions longitudinally; to compare regions, areas and hospitals in defined populations; to combine information on patients and physicians; to add up expenditures for different services within the Canadian health-care system; and to examine population health issues in areas such as education and family services. Well-organized data and the capability for rapid response have been critical for timely policy analysis in Manitoba. A number of successes are mentioned; less successful efforts to influence practice patterns and to modify the internal workings of hospitals are noted. Investments in filling gaps in data collection and in enriching existing data would facilitate additional research. Planning and managing health care for an entire population has benefited greatly from the development of an information-rich environment.

Introduction

Population-based insurance systems using longitudinal administrative data and techniques linking records across multiple files have helped create “information-rich” environments in several sites around the world. Working in Western Australia to develop a linked database (Holman, Bass, Rouse, & Hobbs, 1999) identified the Oxford Record Linkage Study, the Scottish Record Linkage System, the Rochester Epidemiology Project, and the Manitoba Centre for Health Policy as “centres known internationally for the conduct of research using record linkage.” Two other Canadian groups, the Centre for Health Services and Policy Research in British Columbia and the Institute for Clinical and Evaluative Sciences in Ontario, have also been using administrative data for both provincially funded policy work and investigator-initiated research. Routine updates of administrative data (building on consistent formats) facilitate the study of change; analyses of variation in population-based rates over time and cohort studies are often performed (Goldacre, Griffith, Gill, & Mackintosh, 2002; Kendrick, Douglas, Gardner, & Hucker, 1998; Melton, 1996).

Lists of publications, projects, and investigators were obtained for five centres: Oxford, Western Australia, Manitoba, British Columbia, and Ontario. Some research groups (such as in Manitoba) concentrated more on administrative data than others. Substantial efforts were made to identify all relevant publications. Each publication with a title suggesting the use of administrative data was looked up using the National Library of Medicine's PubMed. Abstracts were read by a research associate and by the senior author (LLR) to make sure that administrative data were used. Common exclusion rules were applied across the centres: only studies published from 1994 to 2002 were included. Commentaries and letters were excluded, as was work based on a single disease or surgical registry. Several Canadian studies included authors from more than one centre; the centre providing the data was credited for the research. In multi-centre studies, all participating centres were credited. Each relevant abstract and attached Medical Subject Headings (MeSH) categories were entered into a centre-specific Reference Manager database. MeSH categories were analysed using SAS version 8.2.

Despite PubMed's independent coding and economy, important research might be missed for several reasons: because the work is never published in an academic format (a problem with many projects conducted for government agencies) and because book chapters and certain journals (such as the Canadian Journal on Aging) are not included in PubMed. Further details on the methodology are available from the senior author.

The MeSH word “epidemiology” was most used in reference to the Oxford and Western Australia centres. The terms “utilization”, “economics”, “socioeconomic factors”, “physicians”, and “physician practice patterns” were more frequently used to characterize the research of the three Canadian groups than that of the other sites. Canadian research centres oriented towards local policy problems receive much of their funding from provincial Ministries of Health. This thrust seems to have pushed the Canadian centres in roughly similar directions.

Studies of specific diseases and diagnoses have been popular among all centres; the coding for “disease” in Oxford and “incidence” in Western Australia appears to reflect a broad range of outcomes research at those sites (Table 1). “Utilization” and “socioeconomic factors” were particularly frequent for Manitoba. Published studies with a defined geographic focus, those involving rural populations and others analysing urban populations, were noted more often in Manitoba than elsewhere. The relatively low frequency of “cohort studies” and “case control” in Manitoba reflects less of an emphasis on clinical epidemiology and a greater concern with policy analysis.

The Manitoba Centre has negotiated a series of annual deliverables; almost all of the projects explicitly take advantage of the available databases. As categorized by Manitoba Health, the great majority of projects over the last 13 years have focused on: health and quality of health care (9), regional health authorities—both urban (14) and rural/northern (5), physicians and human resources (5), and financial and cost issues (8). One set of deliverables (11) concerned development of the database and information system. These general topics have all been noted as of national concern (Iglehart, 2000).

In Manitoba the capabilities for policy analysis include the ability to:

• study interventions longitudinally,

• compare regions, areas, and hospitals,

• combine information on patients and physicians,

• add up expenditures for different services within the Canadian health-care system, and

• examine the determinants of health using education and family services data in conjunction with health-related information.

Given the great interest in health policy among OECD countries, these information-rich environments should be more extensively used for policy analysis. This paper provides examples from recent Manitoba work, highlighting successful uses of the database which build on its special capabilities. Limitations—both in terms of less successful efforts to influence policy and gaps in available data—are also discussed.