Health administrative data are frequently used for health research in Canada and abroad. In the past two decades, such data have been widely employed by health services and population health researchers to study healthcare outcomes, effectiveness, appropriateness and utilization of healthcare services, and to investigate or monitor population health status and its determinants [1–11]. The varied and broad use of administrative data has been facilitated by important advantages of the data, including their accessibility, their wide geographic coverage and their relatively complete capture of contacts with the health system for a defined population [12, 13].
The use of health administrative data in health services research has been enabled by some key characteristics, notably the use of a standard system for coding diagnoses, the International Classification of Diseases (ICD). Established by the World Health Organization in 1893 to categorise causes of death, this system adopts a standardised format to code diagnoses, thereby enabling longitudinal and comparative studies [14]. The ninth revision, ICD-9, was expanded in 1977 to ICD-9-CM (Clinical Modification) to enable more precision in diagnostic codes, together with the addition of surgical intervention codes. In 1992, the 10th Revision of ICD (ICD-10) was introduced. ICD-10 has been used by many countries throughout the world for coding cause of death and for hospital diagnoses since 1994 [15–17]. It has been used for mortality data since 2000 in Canada, and provinces have adopted ICD-10 for coding hospital diagnoses in a phased approach, beginning in 2001.
One of the major advantages of ICD-10 is that it is far more detailed (there are a total of 12,420 codes in ICD-10 compared to 6,969 in ICD-9), permitting richer capture of clinical information. However, its implementation means that a number of established methodological tools applicable to ICD-9 or ICD-9-CM need to be redesigned for application in ICD-10. Another issue is that the structure of ICD-10 differs substantially from ICD-9. Furthermore, since each country licences the coding system individually from WHO and can create its own modifications, there may be more opportunity for discrepancies between countries. Finally, ICD-10 does not include procedure codes and so each country has developed its own coding system. The system used by Canada is the International Classification of Diseases, 10th revision, Canadian version, Canadian Classification of Health Interventions (ICD-10-CA/CCI).
Clearly the implementation of ICD-10 offers many benefits while also raising significant challenges for the international health services and population health research communities. In addition, research using ICD administrative data must address other limitations, largely stemming from the fact that the data were created not for research but for other purposes. Data quality is a concern; errors in the data can stem from inaccurate or missing information in the patient record, from the failure to abstract relevant data, or from incorrect coding of the abstracted data. Another concern is that administrative data lack clinical details. Even when data quality is good, the diagnoses that are coded do not reflect the severity of disease, diagnostic findings are not coded, and clinical sequence is not available.
This paper describes the origins and first symposium of a new international group that has come together to discuss how to take advantage of these potential benefits, and to address the new and ongoing challenges associated with using administrative data in health services and population health research. International collaborative research on health services has many advantages. From the methodological perspective, such research allows investigators to develop analytic tools that are more robust and more generalisable. It also allows those tools to be adopted in a systematic and uniform manner across countries, thereby fostering international exchange of research data and findings. From the policy perspective, it helps us to understand the strengths and weaknesses of various healthcare systems, and identifies opportunities for improvement in those systems.
The consortium
The consortium came together through a fortuitous set of circumstances. Australian researcher Vijaya Sundararajan contacted Canadian researchers William Ghali and Hude Quan because they were all doing similar work. While on sabbatical, William Ghali met Swiss researchers with similar interests: Patricia Halfon, Jean-Christophe Luthi and Bernard Burnand. These links led to two initial collaborative projects: new ICD-10 coding algorithms for two widely-used comorbidity measures, the Charlson index and the Elixhauser comorbidity categories [18].
Meanwhile the Canadian Institutes of Health Research (CIHR) announced a funding opportunity for workshops. A successful proposal by Ghali and Quan to the Institute for Health Services and Policy Research permitted a seminar and workshop held June 17 and 18, 2005 in Calgary and Banff, Alberta. The objectives of the workshop were to:
-
1)
solidify collaborative relationships through a face-to-face meeting of researchers;
-
2)
initiate dialogue around launching a set of collaborative research projects on methodological issues surrounding the use of administrative data; and
-
3)
stage a symposium in parallel to the workshop meetings at which the invited researchers would present their work to interested attendees.
Additional invitees to the seminar and workshop included representatives from two stakeholder organizations (Canadian Institute for Health Information (CIHI), and Statistics Canada), five Canadian collaborators, and investigators from the United States, the United Kingdom, Australia, Switzerland and China. The list of invited participants was a convenience sample whose selection was based on two criteria: they were bona fide experts in this area and/or they were known to the organisers.