An overview of the methods and data used in the CCORT Canadian
Transcription
An overview of the methods and data used in the CCORT Canadian
KENNEDY.qxd 4/20/2006 3:14 PM Page 6 CCORT ATLAS PAPER Chapter 2: An overview of the methods and data used in the CCORT Canadian Cardiovascular Atlas project Courtney C Kennedy MSc1, Susan E Brien PhD1, Jack V Tu MD PhD1,2, for the Canadian Cardiovascular Outcomes Research Team CC Kennedy, SE Brien, JV Tu for the Canadian Cardiovascular Outcomes Research Team. An overview of the methods and data used in the CCORT Canadian Cardiovascular Atlas project. Originally published in Can J Cardiol 2003;19(6):655-663. The Canadian Cardiovascular Atlas project, an initiative of the Canadian Cardiovascular Outcomes Research Team (CCORT), will be published as a series of 24 articles in future issues of The Canadian Journal of Cardiology. Through a wide range of data sources and analyses from a number of collaborators across Canada, the CCORT Atlas will provide a comprehensive overview of the current state of cardiac care and disease in Canada. Administrative data, clinical registries and community survey data will be analyzed at the provincial and health region levels. The purposes of this article are to 1) provide an overview of the data types and sources used in the Atlas project, 2) give a general description of the methods and analyses used to report Atlas data and 3) describe how Atlas maps were created and how they can be interpreted. Aperçu des méthodes et des données utilisées dans l’Atlas cardiovasculaire canadien L’Atlas cardiovasculaire canadien, initiative de l’Équipe canadienne de recherche en services cardiovasculaire, sera publié en une série de 24 articles dans les prochains numéros du Journal canadien de cardiologie. S’appuyant sur un large éventail de sources de données et sur des analyses faites par un certain nombre de collaborateurs partout au Canada, l’Atlas dressera un tableau complet de la situation actuelle en ce qui concerne les soins et les maladies cardiovasculaires au Canada. Les données administratives, les registres cliniques et les données d’enquêtes communautaires seront analysés aux niveaux des provinces et des régions sanitaires. Le présent article vise trois objectifs : donner un aperçu des types et des sources de données utilisés dans l’Atlas; décrire globalement les méthodes et les analyses utilisées pour consigner les données de l’Atlas; expliquer la façon dont les cartes de l’Atlas ont été dressées et la façon de les interpréter. Key Words: Administrative data; Atlas; Canada; Cardiovascular disease; Small area variation he Canadian Cardiovascular Atlas project, an initiative of the Canadian Cardiovascular Outcomes Research Team (CCORT), is a comprehensive national report on the state of cardiovascular health and health services across Canada. The Atlas material will be published as a series of 24 articles in future issues of The Canadian Journal of Cardiology. An earlier paper in this Journal provided an overview of upcoming topics and described the rationale and objectives of the Atlas project (1). The Atlas project was produced through the combined effort of over 30 CCORT investigators, programmers, research coordinators and other collaborators from across Canada. The extensive clinical and research expertise of team members, combined with a willingness to share multiple data sources, enabled the Atlas team to conduct a more detailed and wider range of analyses than has ever been done before in cardiovascular outcomes research in Canada. While each article will T describe methods specific to that particular study, the purpose of this article is to 1) Describe the data sources used in the Atlas project; 2) Summarize the major types of analyses conducted; 3) Explain how health regions were defined and how maps for cardiovascular disease in Canada were created; 4) Outline data privacy and confidentiality procedures used throughout the Atlas project. DATA SOURCES The Atlas used three major types of data sources: administrative databases, clinical registries and population-based surveys. Table 1 provides a list of the major databases used in the Atlas project. A complete list of CCORT investigators is published at www.ccort.ca for Clinical Evaluative Sciences, Toronto; 2Division of General Internal Medicine and the Clinical Epidemiology and Health Care Research Program, Department of Medicine, Sunnybrook and Women’s College Health Sciences Centre, University of Toronto, Toronto, Ontario Correspondence and reprints: Dr Jack V Tu, Institute for Clinical Evaluative Sciences, G106-2075 Bayview Avenue, Toronto, Ontario M4N 3M5. Telephone 416-480-4700, fax 416-480-6048, e-mail [email protected] Received for publication March 11, 2003. Accepted March 25, 2003 1Institute 6 A Collection of Original Research Papers Published in The Canadian Journal of Cardiology KENNEDY.qxd 4/20/2006 3:14 PM Page 7 CCORT Atlas methods TABLE 1 Major databases used in the Canadian Cardiovascular Atlas project Database Type of data Years Description APPROACH Clinical registry or 1995/96 to Clinical data capturing cardiac catheterization patients in Alberta (since 1995), surveillance 1999/2000 BC Cardiac Registries Clinical registry 1995/96 to BC PharmaCare Plan Administrative 1996/97 to BC (since 1999) and Saskatchewan (since 2000) BC provincial registry for cardiac surgery, pacemaker and angioplasty procedures 1999/2000 and, as of May 1999, diagnostic catheterization procedures Data from the provincial drug insurance program for BC; eligible people are 1998/99 seniors (aged 65 and over), long-term care residents, social assistance recipients and families with costs exceeding a set deductible Canadian Community Survey September 2000 to Administrative 1995 to 1997 The national registry of deaths in Canada, consolidated from provincial vital Administrative 1994/95 to Data collected for all provinces except Quebec and parts of Manitoba from hospital Health Survey Canadian Mortality Population survey data capturing health determinants, health status and health October 2001 system use variables for all 136 Canadian health regions Database CIHI DAD statistics data 1999/2000 discharge summaries containing demographic information and diagnostic, procedural and comorbidity codes CIHI Hospital Morbidity Administrative Database 1994/95 to Data downloaded from the CIHI DAD (most provinces) or submitted by the provincial 1999/2000 ministry of health for all provinces in Canada; patient separations from a hospital, listed by primary diagnoses, for all provinces and territories ECHO Clinical or chart review 1997 to 1998 Clinical data (chart review) of a sample of first CHF admissions from 14 hospitals in EFFECT Clinical or chart review 1999 to 2001 Clinical data (chart review) of a sample of acute myocardial infarction and CHF FASTRAK II Clinical registry January 2000 to Clinical data (patients with acute myocardial infarction) from participating Canadian ICONS Clinical registry October 1997 to IMS Health Canada Administrative February 1996 to Ontario patients from 104 hospitals in Ontario 2001/02 coronary care units or cardiac wards Clinical data (primary chart audit) on all Nova Scotia residents hospitalized with April 2002 Compuscript Audit acute coronary syndrome, CHF and atrial fibrillation Data on the number of prescriptions dispensed by over 4400 Canadian pharmacies January 2002 (nearly two-thirds of all retail pharmacies in Canada) Database Maritime Heart Centre Clinical registry Cardiac Surgery Database March 1995 to Detailed prehospital, in-hospital and postoperative information on all patients under- 1999/2000 Med-Echo Administrative 1997/98 to ODB Administrative 1997/98 to going cardiac surgery at a Halifax, Nova Scotia, hospital Clinical and demographic data for acute care Quebec CVD inpatients 1999/2000 Data from the provincial drug insurance program for Ontario; eligible people are 1999/2000 seniors (aged 65 and over), long-term care residents, and home care and social assistance recipients RAMQ Administrative 1997/98 to Data from the provincial drug insurance program for Quebec; eligible people are 1999/2000 seniors (aged 65 and over), recipients of employment assistance or those without access to group insurance Statistics Canada Census Census 1991 Information about Canada’s demographic, social and economic characteristics APPROACH Alberta Provincial Project on Outcome Assessment in Coronary Heart Disease; BC British Columbia; CHF Congestive heart failure; CIHI Canadian Institute for Health Information; CVD Cardiovascular disease; DAD Discharge Abstract Database; ECHO Evaulation of Congestive Heart Failure Outcomes; EFFECT Enhanced Feedback for Effective Cardiac Treatment; ICONS Improving Cardiovascular Outcomes in Nova Scotia; IMS Intercontinental Medical Statistics; ODB Ontario Drug Benefits; RAMQ La Régie de l’assurance maladie du Québec Administrative data Government agencies collect administrative health data for monitoring and financial purposes, such as auditing, remuneration of physicians, resource allocation, total quality management and comparative analysis. Data elements generally include demographic data, admission and discharge information, and codes for patient diagnoses, treatments and medical procedures performed. One of the major shortcomings of administrative databases is the lack of detailed information on clinical risk variables. Furthermore, data are not collected primarily for the purpose of research and can include coding errors and missing or ambiguous information. In contrast, potential benefits of administrative data are relatively low cost compared with primary data collection, availability of populationbased information in many jurisdictions, large sample sizes, long-term follow-up, ability to link to other databases and ready availability because they are necessarily collected for administrative purposes (2,3). Furthermore, several previous studies suggest that administrative data, in general, are a reliable way for health services researchers to study the Canadian health care system (4). Since the late 1980s, several studies have specifically evaluated the accuracy of cardiac coding in Canadian administrative databases either through chart reabstraction or by comparison with another data source (eg, clinical registry). Table 2 summarizes this literature, including the study A Collection of Original Research Papers Published in The Canadian Journal of Cardiology 7 KENNEDY.qxd 4/20/2006 3:14 PM Page 8 Kennedy et al TABLE 2 Summary of Canadian studies that have validated the accuracy of cardiac diagnoses or procedures in administrative data First author, year (reference) province Comparison Construct data ICD-8/9 codes Diagnosis Specificity/ Agreement sensitivity (%) (%) Design of validation component Hospital discharge data compared with data from other sources Austin, 2002 (5) MRD Ontario Clinical 410 AMI registry 427 Arrhythmia – 97.8/60.7 92.8/88.8 Most responsible CIHI discharge diagnosis was compared with the FASTRAK II CCU discharge diagnosis 428 CHF 96.8/58.5 (n=58,816) for AMI (primary goal) and arrhythmia, CHF, 786.5 Chest pain 94.2/44.4 unstable angina and chest pain NYD (secondary goal) 411/413 Unstable 93.9/57.9 angina Jha, 1996 (7) Ontario Death Clinical rates trials 410 AMI – 93.6/100 Administrative data (Hospital Medical Records Institute) on admissions and mortality from AMI were linked to GUSTO I (n=1414) and LATE trial data (n=249); comparability of death rates was examined Rawson, 1995 (10) Discharge Saskatchewan Roos, 1989 (11) Manitoba Roos, 1991 (3) Manitoba diagnosis Physician claims 410 AMI 69.3 – IHD patients were identified in the administrative data file 411 Acute IHD 14.6 and compared with matching physician claims (n=4402) 413 Angina 70.0 and abstracted medical charts (n=200) (see below) 414 Chronic IHD 55.6 Discharge Physician 414, 411, CABG 82.1 diagnoses claims 413 61.4 424, 394, Heart valve stays 414 replacement CVD Clinical Not bidities database defined Validity of physician claims for a particular surgery were compared with hospital discharge abstracts for the stay for surgical Comor- – in which the surgery took place – 92.0/43.0 The validity of comorbidity data was examined for several diseases including CVD; Manitoba claims data (hospital discharge abstracts) were compared with a clinical database at a large Winnipeg hospital Reabstraction of hospital discharge data Rawson, 1995 (10) Discharge Saskatchewan – 410 AMI 96.9 – diagnosis IHD patients were identified in the administrative data file and compared with matching physician claims (n=4402) (see above) and abstracted medical charts (n=224) Cox, 1997 (6) MRD – 410 AMI 88.4 – Ontario 475 charts (randomly selected from 16 hospitals) related to AMI were abstracted twice using a checklist with simplified MONICA criteria and compared with CIHI diagnosis of AMI Humphries, Comor- 2000 (12) bidities – 410, 412 Prior MI 428 CHF – 94.8/66.1 88.3/82.9 British Columbia Levy, 1999 (8) Quebec Primary – 410 AMI 96 – discharge Primary 1990 (13) discharge Ontario diagnosis Nova Scotia Saskatchewan were compared with chart review data collected in PCI patients (n=817) Discharge diagnosis A random sample of records with a primary discharge diagnosis of MI (n=234) was compared with medical diagnosis van Walraven, Comorbidity data derived from an administrative database charts in six Montreal hospitals (patients 65 and older) – 410 AMI 79.4 – 209 consecutive charts with a primary diagnosis of 410 (on the fact sheet) were reviewed in a Toronto hospital – 410 AMI (ICD 8/9) CVD Epidemiology ≥86.7 Approximately 250 charts in each province (random sample) were reviewed to estimate the false-positive negative rate for an AMI discharge diagnosis; 500 charts with ≤8.4) Group, 1992 (9) – (false 411 to 414 diagnoses were reviewed to determine the AMI false-negative rate AMI Acute myocardial infarction; CABG Coronary artery bypass graft surgery; CCU Coronary care unit; CHF Congestive heart failure; CIHI Canadian Institute for Health Information; CVD Cardiovascular disease; GUSTO I Global Utilization of Streptokinase and t-PA for Occluded Coronary Arteries 1; ICD International Classification of Diseases; IHD Ischemic heart disease; LATE Late Assessment of Thrombolytic Efficacy; MI Myocardial infarction; MONICA Monitoring trends and determinants in Cardiovascular disease; MRD Most responsible diagnosis; NYD Not yet diagnosed; PCI Percutaneous coronary intervention design and the reported level of diagnostic agreement or specificity and sensitivity of the administrative data evaluated. In general, acute myocardial infarction was the most well studied and accurately coded diagnosis, producing diagnostic agreement or specificity and sensitivity estimates of 90% or more (5-13). 8 Hospital discharge data Hospital discharge data were the most common administrative data sources used in the Atlas project and were obtained from two sources: 1) the Canadian Institute for Health Information (CIHI) Discharge Abstract Database (DAD) and 2) the CIHI Hospital Morbidity Database (HMDB). A Collection of Original Research Papers Published in The Canadian Journal of Cardiology KENNEDY.qxd 4/20/2006 3:14 PM Page 9 CCORT Atlas methods TABLE 3 Standard ICD-9-CM codes used to identify cardiovascular hospitalizations in the CCORT Canadian Cardiovascular Atlas TABLE 5 Canadian population in 1991* by age and sex Age group (years) Males Females Combined % of total population ICD-9-CM codes (at least one of those below) 0 to 04 1,002,469 955,983 1,958,452 7.0 Disease category 05 to 09 991,363 942,483 1,933,846 6.9 Cardiovascular disease 390 to 459 10 to 14 973,036 924,828 1,897,864 6.8 Ischemic heart disease 410 to 414 15 to 19 991,216 936,776 1,927,992 6.9 Congestive heart failure 428.x (428.0 to 428.9) 20 to 24 1,062,179 1,025,803 2,087,982 7.4 Hypertensive disease 401 to 405 25 to 29 1,269,802 1,237,171 2,506,973 8.9 Cerebrovascular disease 430 to 438 30 to 34 1,298,896 1,283,628 2,582,524 9.2 Stroke 430 to 432, 434, 436 35 to 39 1,172,240 1,173,175 2,345,415 8.4 Valvular disease 394 to 397, 424.0 to 424.3 40 to 44 1,068,117 1,060,889 2,129,006 7.6 Acute myocardial infarction 410.x (410.0 to 410.9) 45 to 49 843,971 829,574 1,673,545 6.0 Angina 411, 413 50 to 54 674,865 667,664 1,342,529 4.8 Chest pain 786.5 55 to 59 616,524 618,932 1,235,456 4.4 60 to 64 580,049 612,026 1,192,075 4.3 65 to 69 498,715 588,125 1,086,840 3.9 70 to 74 364,884 470,852 835,736 3.0 75 to 79 255,989 367,465 623,454 2.2 80 to 84 142,435 240,600 383,035 1.4 85 to 89 62,274 130,485 192,759 0.07 0.03 CCORT Canadian Cardiovascular Outcomes Research Team; ICD-9-CM International Classification of Diseases, ninth revision, Clinical Modification (23) TABLE 4 Standard CCP codes used to identify cardiac procedures in the CCORT Canadian Cardiovascular Atlas Procedure CCP codes (at least one of those below) Percutaneous coronary intervention 48.02, 48.03, 48.09 CABG 48.11 to 48.19 Isolated valve surgery 47.12, 47.22 to 47.25, Combined valve or CABG one of 47.12, 47.22 to 47.25, ≥90 25,468 69,913 95,381 Total 13,894,492 14,136,372 28,030,864 *This standard population was used for age and sex adjustments in many Atlas analyses. Source: Statistics Canada 47.31 to 47.33 47.31 to 47.33 and one of 48.11 to 48.19 CABG Coronary artery bypass graft surgery; CCORT Canadian Cardiovascular Outcomes Research Team; CCP Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures (24) The DAD contains information about inpatient hospital discharges and same-day surgeries from all hospitals across Canada (except in Quebec and parts of Manitoba) (14,15). Approximately 75% of all hospital inpatient discharges in Canada (about 4.3 million records annually) are represented in this database (15). The HMDB has a smaller subset of the information contained in the DAD, but includes information from all 10 Canadian provinces. The HMDB provides a count of patient separations (through discharge or death) from each hospital, categorized by the primary disease diagnosis (15). Clinical registries Clinical registries or databases contain precise and detailed information about diagnoses and clinical risk variables; however, their major drawback is the expense of development and maintenance (16). Although clinical registries vary in the scope of their data collection, most prospectively or retrospectively collect detailed information on patients, including preexisting medical conditions, current diagnoses and demographic information. By definition, a clinical database contains well defined, discrete and continuous data elements that are routinely recorded and coupled with outcome descriptors (17). Some registries are population-based (eg, all patients in a province undergoing catheterization), while others are voluntary at either the patient level or the hospital level (ie, the decision to participate in data collection is made by the individual or, in the case of provincial or national registries, by a hospital). Clinical databases used by the Atlas project include the Alberta Provincial Project on Outcome Assessment in Coronary Heart Disease database (18), the Improving Cardiovascular Outcomes in Nova Scotia database (19) and the Enhanced Feedback for Effective Cardiac Treatment database. Clinical databases were not available for every province; thus, in several cases, Atlas studies compared the characteristics of patients in only a few provinces. Population health surveys Population-based health surveys ask specific questions of health system users and are useful in health services research to establish the health needs and practices of a population at a given point in time (ie, they provide a ‘snapshot’ of a population’s health experience) (20). Population-based surveys are subject to bias because they rely on self or proxy report and the voluntary participation of randomly selected participants. The Canadian Community Health Survey (CCHS) Cycle 1.1 (21) was used in the population health section of the CCORT Atlas. Conducted by Statistics Canada, the CCHS aims to provide ongoing estimates of health determinants, health status and health system use across the country. Covering approximately 98% of the Canadian population aged 12 years or older, the CCHS randomly selected one, and in some cases two, respondents per household. The total sample size for Cycle 1.1 was 131,535 household respondents, representing a response rate of 84.7% (22). The CCHS data used in the Atlas were from the first collection cycle, September 2000 to November 2001. The complete survey instrument is available on the Statistics Canada Web site (www.statcan.ca) (21). A Collection of Original Research Papers Published in The Canadian Journal of Cardiology 9 KENNEDY.qxd 4/20/2006 3:14 PM Page 10 Kennedy et al TABLE 6 Definition of health regions for Atlas time period (28,30) Based on provincial administrative boundary? Number of health regions Health region changes since March 2000* Province or territory How health region was defined Newfoundland/Labrador Health and community services regions Yes (implemented 1994) 6 None Prince Edward Island Urban and rural division No† 2 None Nova Scotia Health zones Yes†‡ 6 None New Brunswick Regional health corporations Yes (1992 to April 1, 2002) 7 Same boundaries, one additional Quebec Regional health authorities/Régies Yes (implemented 1989) Ontario Public health units No§ 37 None Manitoba Regional health authorities Yes (1997 to June 30, 2002) 12 Amalgamation of Marquette and South Saskatchewan Service areas (groupings of No† 11 April 1, 2002: 12 regional health regional health authority board 18 None régionales de la santé et des services sociaux Westman to Assiniboine the health districts) authorities replace 32 health district boards Alberta Regional health authorities Yes (established 1994) 17 May 2002: announced regional health authorities will be reduced to 5 to 7 by 2004 British Columbia Regional health boards Yes (established 1993) 20 Yukon Territory Not regionalized – 1 Northwest Territories Not regionalized – 1 Nunavut Not regionalized – 1 In 2002, boundaries changed to 16 health service delivery areas *Most data used in the Canadian Cardiovascular Outcomes Research Team Atlas project did not go beyond March 31, 2000; †Statistics Canada created their own health regions for Prince Edward Island, Nova Scotia and Saskatchewan because legislated health regions in these provinces were relatively small and may not have been comparable with other Canadian health regions. New health regions were related, but not identical, to existing provincial administrative regions; ‡In Nova Scotia, three district health authorities were amalgamated with a neighbouring authority; §Health care in Ontario is administered and delivered by the province (ie, not by regional health authorities) METHODOLOGICAL OVERVIEW AND TECHNICAL TERMS Atlas cohorts were generated using a subset of selected International Classification of Diseases, ninth revision, Clinical Modification (ICD-9-CM) codes (23) and Canadian Classification of Diagnostic, Therapeutic and Surgical Procedures (CCP) codes (24). Tables 3 and 4 show the standardized CCORT ICD-9-CM and CCP codes used to identify cardiovascular hospitalizations and procedures. All data in the Atlas were analyzed according to the fiscal year. For example, the fiscal year for 1998/99 is from April 1, 1998, to March 31, 1999. Most analyses were conducted using data from 1997/98 to 1999/2000, but databases were queried back to 1992 when required data elements were available. Standardized analyses Utilization analyses (ie, hospitalization rates, procedure rates) counted all relevant procedures or hospitalizations during a study period, even if there were two or more events per patient. Outcome analyses (eg, mortality rates, readmission rates) counted patients only once in the numerator and once in the denominator. The majority of Atlas articles report regional rates based on patient province of residence, not province of treatment (eg, a British Columbia resident treated in Alberta was counted in the British Columbia rate). The exception to this rule was for cardiac procedures: outcomes for patients treated out of province were attributed to the province of treatment, not residence. Hospitalizations were regarded as ‘episodes of care’, ie, if an admission took place less than 24 h after the previous discharge, 10 the two hospital stays were considered to be the same episode of care. This was done to ensure an accurate reflection of the follow-up period and to prevent the counting of transfers as readmissions. TYPES OF ANALYSES Observed or crude rates An observed or crude rate represents the ‘actual’ experience of a population and is not adjusted for any factors that may affect the outcome, such as age, sex and disease comorbidities. Crude rates are summary measures and are useful for health planning and resource allocation purposes. Age-sex-standardized versus risk-adjusted rates To make comparisons across regions in Canada, most rates were also adjusted for potentially confounding factors. Some Atlas analyses used only age-sex adjustment, while others adjusted for several potential confounders, such as comorbid conditions (termed risk-adjusted). In general, there was a very high correlation between age-sex-standardized and riskadjusted rates at the health region level. Use of age-sex adjustment alone allowed for inclusion of Quebec and Manitoba data in many Atlas studies even though their administrative data did not distinguish comorbidities from complications. One method used for adjustment of rates was direct standardization. The purpose of direct standardization is to compare rates between regions as if they had similar population compositions. Rates calculated by direct standardization do not necessarily reflect the actual or crude rate in a population. A Collection of Original Research Papers Published in The Canadian Journal of Cardiology KENNEDY.qxd 4/20/2006 3:14 PM Page 11 CCORT Atlas methods 1001 1002 1003 1004 1005 1006 * Newfoundland and Labrador (10) Health and Community Services St. John’s Region Health and Community Services Eastern Region Health and Community Services Central Region Health and Community Services Western Region Grenfell Regional Health Services Board Health Labrador Corporation 1101 1102 Prince Edward Island (11)* Urban Health Region Rural Health Region 1201 1202 1203 1204 1205 1206 Nova Scotia (12)* Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 1301 1302 1303 1304 1305 1306 1307 New Brunswick (13)* Region 1 Region 2 Region 3 Region 4 Region 5 Region 6 Region 7 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 Quebec (24)* Région du Bas-Saint-Laurent Région du Saguenay - Lac-Saint-Jean Région de Québec Région de la Mauricie et Centre-du-Québec Région de l’Estrie Région de Montréal-Centre Région de l’Outaouais Région de l’Abitibi-Témiscamingue Région de la Côte-Nord Région du Nord-du-Québec Région de la Gaspésie-Îles-de-la-Madeleine Région de la Chaudière-Appalaches Région de Laval Région de Lanaudière Région des Laurentides Région de la Montérégie Région du Nunavik Région des Terres-Cries-de-la-Baie-James 3526 3527 3530 3531 Ontario (35)* Algoma Public Health Unit Brant Public Health Unit Durham Public Health Unit Elgin-St Thomas Public Health Unit 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3549 3551 3552 3553 3554 3555 3556 3557 3558 3560 3561 3562 3563 3565 3566 3568 3570 3595 Bruce-Grey-Owen Sound Public Health Unit Haldimand-Norfolk Public Health Unit Haliburton-Kawartha-Pine Ridge Public Health Unit Halton Public Health Unit Hamilton-Wentworth Public Health Unit Hastings and Prince Edward Public Health Unit Huron Public Health Unit Kent-Chatham Public Health Unit Kingston-Frontenac-Lennox and Addington Public Health Unit Lambton Public Health Unit Leeds-Grenville-Lanark Public Health Unit Middlesex-London Public Health Unit Muskoka-Parry Sound Public Health Unit Niagara Public Health Unit North Bay Public Health Unit Northwestern Public Health Unit Ottawa Carleton Public Health Unit Oxford Public Health Unit Peel Public Health Unit Perth Public Health Unit Peterborough Public Health Unit Porcupine Public Health Unit Renfrew Public Health Unit Eastern Ontario Public Health Unit Simcoe Public Health Unit Sudbury Public Health Unit Thunder Bay Public Health Unit Timiskaming Public Health Unit Waterloo Public Health Unit Wellington-Dufferin-Guelph Public Health Unit Windsor-Essex Public Health Unit York Public Health Unit City of Toronto Public Health Unit 4610 4615 4620 4625 4630 4640 4650 4655 4660 4670 4680 4690 Manitoba (46)* Winnipeg Brandon North Eastman South Eastman Interlake Central Marquette South Westman Parkland Norman Burntwood Churchill 4701 4702 4703 4704 Saskatchewan (47)* Weyburn (A) Service Area Moose Jaw (B) Service Area Swift Current (C) Service Area Regina (D) Service Area 4705 4706 4707 4708 4709 4710 4711 Yorkton (E) Service Area Saskatoon (F) Service Area Rosetown (G) Service Area Melfort (H) Service Area Prince Albert (I) Service Area North Battleford (J) Service Area Northern Health Services Branch (K) Service Area 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 Alberta (48)* Chinook Regional Health Authority Palliser Health Authority Headwaters Health Authority Calgary Regional Health Authority Health Authority #5 David Thompson Regional Health Authority East Central Health Authority WestView Regional Health Authority Crossroads Regional Health Authority Capital Health Authority Aspen Regional Health Authority Lakeland Regional Health Authority Mistahia Regional Health Authority Peace Regional Health Authority Keeweetinok Lakes Regional Health Authority Northern Lights Regional Health Authority Northwestern Regional Health Authority 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 British Columbia (59)* East Kootenay West Kootenay-Boundary North Okanagan South Okanagan Similkameen Thompson Fraser Valley South Fraser Valley Simon Fraser Coast Garibaldi Central Vancouver Island Upper Island/Central Coast Cariboo North West Peace Liard Northern Interior Vancouver Burnaby North Shore Richmond Capital 6001 Yukon Territory 6101 Northwest Territories 6201 Nunavut *This is the province prefix, in some geographically small health regions just the last 2 digits of the health region number appear (e.g. In Quebec, health region 2416 is labeled 16). Figure 1) Health regions (corresponding to numbers above) used in the Canadian Cardiovascular Outcomes Research Team Canadian Cardiovascular Atlas by population size, 1996 Most direct standardized rates in the Atlas were calculated using the 1991 Canadian population as the standard population (see Table 5 for the 1991 population breakdown by age and sex group). This is consistent with publications by other organizations (25) and is the standard currently used by Health Canada (26). For certain outcome analyses, the age-sex distribution in the overall study cohort was used as the standard population. Another method used for rate adjustment was model-based indirect standardization, using a regression model. First, expected rates for regions were calculated using a model that best predicted the outcome of interest. Then for each region, the crude rates were divided by expected rates and multiplied by the overall Canadian average. Resulting rates are considered ‘risk-adjusted’ and can be interpreted as the rates that would occur if all the time periods or regions of interest had similar case mixes. Hospitalization rates A hospitalization rate is a measure of the number of patient admissions into hospital for a designated condition, such as a myocardial infarction. Hospitalization rates are useful for comparing regional use of resources and as a proxy measure of disease burden in a region. In-hospital mortality rates In-hospital mortality is a measure of the number of in-hospital deaths that occurred within 30 days of admission to hospital for the condition of interest (index admission). In-hospital mortality is often used as a quality of care marker for the acute treatment of myocardial infarction or other condition. Readmission rates Readmissions rates relate to patients who survived their initial hospitalization but returned to hospital at least once within 30 days, 90 days or one year of their index admission. To calculate Atlas readmission rates, only patients who survived their initial hospitalization were included in the denominator. A readmission rate is a good indicator of whether the efforts to prevent secondary complications or another event, such as myocardial infarction, ongoing chest pain and heart failure, were successful. A Collection of Original Research Papers Published in The Canadian Journal of Cardiology 11 KENNEDY.qxd 4/20/2006 3:14 PM Page 12 Kennedy et al 12 A Collection of Original Research Papers Published in The Canadian Journal of Cardiology KENNEDY.qxd 4/20/2006 3:14 PM Page 13 CCORT Atlas methods Cardiac procedure rates Cardiac procedure (ie, utilization) rates represent the total number of people in a health region or province undergoing a procedure (cardiac catheterization, coronary artery bypass graft surgery, percutaneous coronary intervention, valve surgery) per 100,000 population. The purpose of calculating procedure rates is to understand a region’s rate of use of a particular procedure and determine variations in clinical practice or available resources. Median wait time The median wait time was the time required for half the patients to receive a procedure (coronary artery bypass graft surgery or percutaneous coronary intervention). Median wait times (in days) were calculated for acute myocardial infarction patients from the point of their initial hospitalization for acute myocardial infarction to the point of the procedure. Median wait times are good indicators of the availability of resources and patient access to care in a region. Drug utilization rates Drug utilization rates were reported for both in-hospital and outpatient treatment phases. These rates provide feedback to practitioners on whether treatment benchmarks are being met and reflect the extent of regional variation in Canada. GEOGRAPHY OF CARDIAC CARE IN CANADA The Canadian health care system provides comprehensive coverage for medically necessary hospital inpatient and outpatient physician services (27). With financial assistance from the federal government, each of the 10 provinces and three territories runs its own health care system and is responsible for managing and delivering health services in its respective region. However, in the 1990s, with the objective of streamlining the delivery system, most provinces partially devolved their responsibilities to subprovincial regional health authorities (RHAs) (28). RHAs typically have appointed or elected boards of governance and are responsible for funding and delivering community and institutional health services within their defined geographical region. Depending on the province, RHAs have varied scopes of responsibilities and differ in terms of geographical and population size (28). Regional level analyses in the CCORT Canadian Cardiovascular Atlas (including maps) used the 139 Statistics Canada-defined ‘health regions’ from December 2001, which, for most provinces, are based on the geographical boundaries of RHAs (29). Health regions for British Columbia were the 20 health service delivery areas in existence in January 2002 (before the health service redesign later in 2002). Health regions in Ontario were defined as public health units (PHUs), not district health councils, in order to provide as much detail as possible (district health councils have greater geographical areas than PHUs). Recent consultations with Ontario stakeholders have indicated a preference for the lowest unit of aggregation in constructing maps for Institute for Clinical Evaluative Sciences atlases (30). PHUs and counties have similar geographical boundaries in Ontario. Table 6 provides an overview of all provincial health regions in existence during the Atlas study period and changes made since that time. A master legend of all health region names and boundaries is provided in Figure 1. This map is also shaded according to the approximate population size of each particular region. Atlas maps CCORT Atlas maps were created using MapInfo software (MapInfo Corp, USA). Maps were subdivided by health region and, where possible, rate information was displayed for every health region in Canada. This was not possible in all cases because of necessary exclusion of regions with known errors in coding (eg, acute myocardial infarction cases in Newfoundland) and regions with small sample sizes (eg, the territories). For all provinces except Quebec and Ontario, patients were assigned to health regions by converting their enumeration area (EA) code (assigned by CIHI) to a health region code, using the Statistics Canada Postal Code Conversion file. For people who were missing an EA code, and for all patients of Ontario, the first three digits of their postal code were converted to an EA code by using a postal code conversion program. Health region codes were directly provided for Quebec patients. Health regions were colour coded on the maps according to how their regional rate compared with the Canadian average. First, the regional rate was divided by the overall Canadian rate to obtain a rate ratio. Then each health region was assigned a colour based on which of the following fixed ratio groups it fell into: minimum to less than 0.75, 0.75 to less than 0.90, 0.90 to less than 1.10, 1.10 to less than 1.30, or 1.30 or greater. For example, a region that had a rate that was at least 30% higher than the Canadian average fell within the highest ratio category (at least 1.30) and appeared on the map as dark red. A region that had a rate at least 25% less than the Canadian average fell into the lowest ratio category (less than 0.75) and appeared on the map as yellow. Regions with rates that fell into the middle categories (0.75 to less than 0.90, 0.90 to less than 1.10, 1.10 to less than 1.30) were coloured in sequential shades of orange. Regions shaded in grey indicated that data were not available or were suppressed because of confidentiality issues. In the map legend, the actual rates per 100 patients or 100,000 population were presented instead of the fixed ratio categories (ie, the range of rates that corresponded to the cutpoints in the ratio category) to provide readers with more usable information. However, it should be noted that, although the actual rates were different for each map, the underlying fixed ratio categories were identical in every map. Fixed ratio categories used in the CCORT Atlas were first used by the Dartmouth Atlas of Cardiovascular Health Care (31) in the United States. Instead of using quintiles, the Dartmouth group chose cutpoints based on clinical significance and symmetry about 1.0. Rates that vary less than 10% of expected are not clinically meaningful variations, and those that are higher than 30% of expected values are of potential concern. The chosen corresponding lower bounds (ie, 0.75, 0.90) are approximate reciprocals of the upper bounds. Because the variability in the colours on a map should reflect the variability of the data, a major benefit of the Dartmouth approach is that if the data are tightly clustered around the mean, the two extreme colours will not appear on the map. PRIVACY AND CONFIDENTIALITY CCORT investigators are firmly committed to protecting the privacy of all patients whose health information was used in the production of the Atlas project. To prevent any possibility that patients could be identified, the Atlas does not report on 1) cell sizes less than five, 2) health regions with fewer than 50 cases, 3) health regions with fewer than two hospitals serving a A Collection of Original Research Papers Published in The Canadian Journal of Cardiology 13 KENNEDY.qxd 4/20/2006 3:14 PM Page 14 Kennedy et al region, and 4) health regions with fewer than two physicians serving patients in that region. Research ethics approval for the CCORT Atlas project was received from the Sunnybrook and Women’s College Health Sciences Research Ethics Board and the University of Toronto. All participating investigators signed a data privacy and confidentiality agreement as part of the Atlas project. CONCLUSION The CCORT Canadian Cardiovascular Atlas used data from the CIHI DAD, clinical registries and the CCHS. Standardized ICD-9-CM and CCP codes were used to identify cardiovascular hospitalizations and procedures. Major types of analyses in the Atlas included crude and age-sex or risk-adjusted rates for in-hospital mortality, readmission, hospitalization, cardiac procedures and drug use. Where possible, rates were calculated for every health region in Canada and were represented geographically in colour-coded national maps. We hope that the methods used in the Atlas will serve as a template for future research endeavours of this nature in Canada. ACKNOWLEDGEMENTS: The authors would like to acknowledge Woganee A Filate for her assistance with preparing the manuscript. The CCORT Canadian Cardiovascular Atlas project was supported by operating grants to the Canadian Cardiovascular Outcomes Research Team from the Canadian Institutes for Health Research Interdisciplinary Health Research Team program and the Heart and Stroke Foundation. Dr Tu is supported by a Canada Research Chair in Health Services Research. REFERENCES 1. Tu JV, Brien SE, Kennedy CC, Pilote L, Ghali WA An introduction to the Canadian Cardiovascular Outcomes Research Team’s (CCORT) Canadian Cardiovascular Atlas project. Can J Cardiol 2003;19:225-9. 2. Iezzoni LI. Assessing quality using administrative data. Ann Intern Med 1997;127(Suppl 8):666-74. 3. Roos LL, Sharp SM, Cohen MM. Comparing clinical information with claims data: Some similarities and differences. J Clin Epidemiol 1991;44:881-8. 4. Williams JI, Young W. Inventory of Studies on the Accuracy of Canadian Health Administrative Databases. Working Paper. Toronto: Institute for Clinical Evaluative Sciences, 1996. 5. Austin PC, Daly PA, Tu JV. Multicenter study of the coding accuracy of hospital discharge administrative data for patients admitted to cardiac care units in Ontario. Am Heart J 2002;144:290-6. 6. Cox JL, Melady MP, Chen E, Naylor CD. Towards improved coding of acute myocardial infarction in hospital discharge abstracts: A pilot project. Can J Cardiol 1997;13:351-8. 7. Jha P, Deboer D, Sykora K, Naylor CD. Characteristics and mortality outcomes of thrombolysis trial participants and nonparticipants: A population-based comparison. J Am Coll Cardiol 1996;27:1335-42. 8. Levy AR, Tamblyn RM, Fitchett D, McLeod PJ, Hanley JA. Coding accuracy of hospital discharge data for elderly survivors of myocardial infarction. Can J Cardiol 1999;15:1277-82. 9. The Nova Scotia-Saskatchewan Cardiovascular Disease Epidemiology Group. Trends in incidence and mortality from acute myocardial infarction in Nova Scotia and Saskatchewan 1974 to 1985. Can J Cardiol 1992;8:253-8. 10. Rawson SB, Malcolm E. Validity of the recording of ischaemic heart disease and chronic obstructive pulmonary disease in the saskatchewan health care datafiles. Stat Med 1995;14:2627-43. 11. Roos LL, Sharp SM, Wajda A. Assessing data quality: A computerized approach. Soc Sci Med 1989;28:175-82. 12. Humphries KH, Rankin JM, Carere RG, Buller CE, Kiely FM, Spinelli JJ. Co-morbidity data in outcomes research: Are clinical data derived from administrative databases a reliable alternative to chart review? J Clin Epidemiol 2000;53:343-9. 13. van Walraven C, Wang B, Ugnat AM, Naylor CD. False-positive coding for acute myocardial infarction on hospital discharge records: Chart audit results from a tertiary centre. Can J Cardiol 1990;6:383-6. 14. Richards J, Brown A, Homan C. The data quality study of the Canadian Discharge Abstract Database. In: Proceedings of Statistics Canada Symposium 2001 – Achieving Data Quality in a Statistical Agency: A methodological perspective. Quebec: Statistics Canada, 2001. 14 15. Canadian Institute for Health Information. Health Services Databases. http://www.cihi.ca (August 2002). 16. Hannan EL, Kilburn H, Lindsey ML, Lewis R. Clinical versus administrative databases for CABG surgery: Does it matter? Med Care 1992;30:892-907. 17. Pryor DB, Califf RM, Harrell FE, et al. Clinical data bases: Accomplishments and unrealized potential. Med Care 1985;23:623-47. 18. Ghali WA, Knudtson ML for the APPROACH Investigators. Overview of ‘APPROACH’ theAlberta provincial project for outcome assessment in coronary heart disease. Can J Cardiol 2000;16:1225-30. 19. Optimizing disease management at a health care system level: The rationale and methods of the improving cardiovascular outcomes in Nova Scotia (ICONS) study. Can J Cardiol 1999;15:787-96. 20. Hennekens CH, Buring JE, eds. Epidemiology in Medicine. Toronto: Little, Brown and Company, 1987. 21. Statistics Canada. The Canadian Community Health Survey (CCHS) – Cycle 1.1. http://www.statcan.ca (January 8, 2003). 22. Béland Y, Bailie L, Catlin G, Singh MP. CCHS and NPHS: An improved health survey program at Statistics Canada. Proceedings of the section on survey research methods. Indianapolis: American Statistical Association, 2000. 23. International Classification of Disease, 9th rev (Clinical Modification). Washington: Public Health Service, USA Department of Health and Human Services, 1988. 24. Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures. Cat no 82-562E. Ottawa: Statistics Canada, 1986. 25. Changing Face of Heart Disease and Stroke in Canada 2000. Ottawa: Heart and Stroke Foundation of Canada, 1999. 26. Health Canada. Cardiovascular Disease Surveillance On-Line, Glossary. http://www.hc-sc.gc.ca (August 26, 2002). 27. Health Canada. Health Care. http://www.hc-sc.gc.ca (September 25, 2002). 28. Canadian Centre for Analysis of Regionalization and Health (CCARH). What is regionalization? http://www.regionalization.org/Index.html (December 14, 2002). 29. Statistics Canada. About Health Indicators. http://www.statcan.ca (January 8, 2003). 30. Naylor CD, Slaughter PM, eds. Cardiovascular Health and Services in Ontario. Toronto: Institute for Clinical Evaluative Sciences, 1999. 31. Dartmouth Medical School. Center for the Evaluative Clinical Sciences. The Dartmouth Atlas of Cardiovascular Health Care. Hanover: The Center for the Evaluative Clinical Science, Dartmouth Medical College, 1999. A Collection of Original Research Papers Published in The Canadian Journal of Cardiology