Chronic Diseases and Injuries in Canada

Volume 31 · Supplement 1 · Fall 2011
Patterns of Health Services Utilization in Rural Canada


Defining “rural”

In Canada, there is not a universally adopted or officially sanctioned definition of “rural,” and various definitions have been adopted for different reasons. For this research project, we used the Statistics Canada definitionFootnote 20a20a of “rural and small town (RST)” when analyzing secondary data. RST refers to populations living outside the commuting zones of larger urban centres, specifically outside census metropolitan areas (CMA) and census agglomerations (CA). To a large extent, the RST definition was chosen because of the heterogeneity of rural areas, enabling these to be divided into four different degrees of rurality expressed as Metropolitan Influenced Zones (MIZ). It also takes into account distance to urban cities, where many of the specialized health services are located.

An extension of the RST concept, the MIZ approach was developed by Statistics Canada “to better show the effects of metropolitan accessibility on non-metropolitan areas”: Footnote 20b,2120b,21 MIZ commuting flows are calculated using data about place of work from the census. This method recognizes the possibility of “multiple centres of attraction”: flows of commuters from an RST community to any urban centre with a population of 10 000 or more for employment reasons are combined to determine the degree of metropolitan influence (i.e. strong, moderate, weak or no influence) of one or more urban centres on that community.Footnote 2222 The classification and its methodology have been extensively validated by Statistics Canada.Footnote 21b21b

The MIZ method distinguishes rural populations with less access to the labour markets of larger urban centres from those with greater access; distance between urban and rural communities is one of the major determinants of such access. Labour force commuter flow is used as a proxy for the access to such services as health, education, financial services, shopping, government services, and cultural and sports activities. This reflects the relative influence of one or more urban centres on a rural area. Although all rural communities have, by definition, a population of fewer than 10 000 people, the MIZ method does not reflect differences in population size between types of rural community.

The following geographic categories are used in this study, and the distribution of the Canadian population according to these geographic categories is shown in Table 1:

  • Census Metropolitan Areas (CMA) have a population of 100 000 or more in the urban core and include all the neighbouring towns and municipalities where 50% or more of the labour force commutes to the urban core.
  • Census Agglomerations (CA) have a population of 10 000 to 99 999 in the urban core and include all neighbouring towns and municipalities where 50% or more of the labour force commutes to the urban core.
  • Strong Metropolitan Influenced Zones (MIZ) are areas where 30% to 50% of the labour force commutes to work in any CMA.
  • Moderate MIZ are areas where at least 5% but less than 30% of the labour force commutes to work in any CMA or CA.
  • Weak MIZ are areas where less than 5% of the labour force commutes to work in any CMA or CA.
  • No MIZ are areas with a small labour force (i.e. fewer than 40 people) or with nobody commuting to work in a CMA or CA.
Table 1 - Canadian population by degree of rurality, 1996 and 2001
  Population and percent distribution (within 2001 boundaries) Percent change within MIZ groups between 1996 and 2001
1996 % 2001 %

Data source: Statistics Canada. Census of Population, 1996 and 2001. Abbreviations: CA, Census Agglomeration; CMA, Census Metropolitan Area; MIZ, Metropolitan Influenced Zone; RST, Rural and Small Town.

Urban (CMA/CA) 22 654 692 78.5 23 839 086 79.4 +5.2
All RST areas 6 192 069 21.5 6 168 008 20.6 −0.4
Strong MIZ 1 470 493 5.1 1 524 579 5.1 +3.7
Moderate MIZ 2 307 387 8.0 2 285 538 7.6 −0.9
Weak MIZ 2 027 488 7.0 1 969 211 6.6 −2.9
No MIZ 386 701 1.3 388 680 1.3 +0.5
Total 28 846 761 30 007 094 +4.0

Data sources and analytical methods

This study used data from Canadian Community Health Survey (CCHS), Health Services Access Survey (HSAS), physician claim files and The Hospital Morbidity Database (HMDB). The data produced several indicators of health services utilization. For each indicator, the urban (CMA/CA) group was used as the reference group and compared with the different rural groups (MIZ categories).

Canadian Community Health Survey and Health Services Access Survey

Data from the Canadian Community Health Survey (CCHS) Cycle 1.1, 2000–2001, and the Health Services Access Survey (HSAS) was analyzed in two stages. We first performed bivariate analyses to examine the differences in self-reported use of health services between urban and rural communities. Age-standardized rates for several indicators were calculated by sex in the urban (CMA/CA) and rural (all MIZ categories) groups. Rates for this analysis were standardized to the 2001 Census population. Data were weighted to take into account the complex sampling design and to adjust for non-response. The Bootstrap procedure was used to calculate 95% confidence intervals (CI).

We then performed multivariate logistic regression analyses to ascertain the relation between place of residence and self-reported no family doctor and between place of residence and hospitalization. The goal of this analysis was to assess whether place of residence has an independent effect on specific outcomes after controlling for various health determinant variables. The choice of such health determinants for analysis was based on Anderson’s theoretical framework,Footnote 2d2d but was restricted by the availability of these factors in the databases used. As with the bivariate analysis, the data were weighted to take into account the complex sampling design and to adjust for non-response. We used the Bootstrap procedure to calculate 95% confidence intervals.

The CCHS and the HSAS data were obtained from a sample of all census subdivisions (CSD) in Canada. Both surveys were administered to individuals 12 or 15 years of age or over, respectively, but excluded persons living in First Nations reserves or on Crown lands, those in institutions (e.g. prisons), full-time members of the Canadian Armed Forces and residents of certain remote regions. Consequently, if the rural or remote areas sampled in the two national surveys showed smaller numbers of these subpopulations than they actually had, such as some sampled No MIZ areas with very small First Nations on-reserve populations, the results might not be representative. Generalizing CCHS or HSAS results to these subpopulations should be done with caution.

Physician claims files

The national-level analysis was complemented by an analysis of rural health services utilization patterns at the provincial level. We used billing data from Nova Scotia, Ontario and British Columbia to examine the use of physician services for three biennial periods: 1997–1998, 1999–2000 and 2001–2002. In each of the three provinces, individuals were classified, using the MIZ method described above, into one of the five geographic categories (using either 1996 or 2001 Census boundaries) with place of residence defined as the CSD in which each individual lived at the beginning of each biennial period examined. For example, if an individual living in a No MIZ area visited a physician in a Weak MIZ area, this individual would be represented in the No MIZ group in these analyses.

We calculated cross-sectional physician-visit rates (i.e. office visits, visits to hospital outpatient departments and visits to hospital emergency departments) as well as physician-visit rates for family physicians (FPs), medical specialists and surgical specialists (Table 2). The “family physician” category includes general practitioners and family physicians providing emergency care in emergency departments. The calculations excluded hospital in-patient care, home care, nursing home care and laboratory- or hospital-associated services.

Table 2 - Groupings of physician specialties used in the analysis of physician claims
Family physicians Medical specialties Surgical specialties

NOTE: Medical scientists and laboratory specialties were excluded from the analyses.
For the British Columbia analyses, the visits based on the laboratory specialties of pathology and medical microbiology were excluded from the study population.

General practice, family practice Rheumatology Orthopedic surgery
Internal medicine Cardiovascular surgery
Cardiology General surgery
Clinical immunology General thoracic surgery
Dermatology Neurosurgery
Gastroenterology Obstetrics
Genetics Ophthalmology
Geriatrics Otolaryngology
Haematology Plastic surgery
Pathology Urology
Nuclear medicine
Physical medicine
Diagnostic radiology
Therapeutic radiology
Respiratory disease

A physician visit was defined as one patient-doctor encounter per day. Only one visit would be attributed to an individual who had multiple billing records from the same physician on the same day if the same ICD-9Footnote *  chapter diagnosis code was used. Multiple visits would be attributed to an individual who had multiple billing records from different physicians in one day. Shadow billings, where available, were also included in the analysis to account for some physicians, especially in northern or remote areas, who were not reimbursed on a fee-for-service basis but were on alternative payment plans. Excluded were provincial residents seeing physicians outside their home province and services used by out-of-province patients.

Physician-claims data are administrative in nature and, thus, are limited to physician services that are fee-for-service or shadow-billed. If physician services are not reported, such as those under alternative payment plans, or if the administrative data codes do not distinguish between different types of services, such as mental health services, the results may not be representative of what actually occurs.

The following indicators were calculated for three biennial periods by sex:

  • physician-visit rates, or the average number of visits to all physicians per 1000 residents of the area;
  • FP-visit rates, or the average number of visits to FPs per 1000 residents of the area;
  • specialist-visit rates, or the average number of visits to medical specialists and surgical specialists per 1000 residents of the area; and
  • physician-visit rates by disease group, or the average number of physician visits by ICD-9 chapters code and sex per 1000 residents of the area.

All rates were age-standardized using the 1991 Census population age structure. Rate ratios were calculated using the age- standardized rate of CMA/CA as the reference rate. Finally, we also examined specific diagnoses, chosen according to their relevance to rural populations and impact on population health, as well as the availability of data (Table 3).

Table 3 - Diagnoses used to compare relative risks of physician visits and hospitalizations for urban and rural populations
ICD-9 chapter Disease category Diagnostic code Specific diagnosis Diagnostic code

Data source:World Health Organization.
Abbreviations: ICD-9, International Statistical Classification of Diseases and Related Health Problems, 9th Revision.

II Neoplasms 140–239 Breast cancer 174 (female)
Lung cancer 162
V Mental disorders 290–319 Depression 296.2, 296.3, 300.4, 311
VI Diseases of the nervous system and sense organs 320–389 Alzheimer’s/dementia disorders 331
VII Diseases of the circulatory system 390–459 Coronary heart disease 410–414
Stroke 430–434
VIII Diseases of the respiratory system 460–519 Asthma 493
Chronic obstructive pulmonary disease 490–492, 496
XIII Diseases of the musculoskeletal system and connective tissue 710–739 Osteoarthritis 715
Rheumatoid arthritis 714
XVII Injuries and poisoning 800–999
XVIII Endocrine, nutritional and metabolic diseases, and immunity disorders 240–279 Diabetes 250

Hospital Morbidity Database

The Hospital Morbidity Database (HMDB), maintained by the CIHI, provides national data on acute-care hospitalization by diagnosis and procedure. Data are reported according to the region of the patient’s residence, not the region of the hospital. Consequently, these figures represent how frequently residents of a given area received hospital care, rather than the volume of services provided by hospitals. Data in the HMDB are based on discharges from (rather than admissions to) a hospital so only people who are alive at the time of discharge are included in the analysis. Stillborn infants and cadaveric donor “discharges” are excluded, and day procedures (such as day surgeries) and emergency department visits are also not captured in the database. For the purposes of this analysis, discharge data for newborns were also excluded.

The HMDB contains data from fiscal years 1994/1995 to 2000/2001 in the ICD-9/CCPFootnote **  classification system.Footnote   In 2001/2002, Yukon, Nova Scotia, Prince Edward Island, Newfoundland and Labrador, and some facilities in Saskatchewan implemented the ICD-10-CAFootnote   and CCIFootnote §  systems. As a result, the database from fiscal year 2001/2002 contains data in both classification systems.

Discharge data from the HMDB were extracted for Nova Scotia, Ontario and British Columbia, as well as for Canada as a whole, except QuebecFootnote ||  . Hospital data were analyzed using a historical cohort design. Data were extracted in the classification system (either ICD-9 or ICD-10-CA) in which they were originally submitted. The effect of using different ICD-9 and ICD-10-CA diagnostic classifications on national rural-urban patterns is unknown at this time. ICD-10-CA differs from ICD-9 in several respects, with the former being more detailed. Data for fiscal year 2001/ 2002 (April 1, 2001, to March 31, 2002) were extracted and translated into 1996 Census boundaries. Rural place of residence was defined according to the MIZ approach.

Discharge rates and length-of-stay figures were based on the number of discharges from an acute-care facility in Canada in fiscal year 2001/2002. If an individual was admitted and discharged from an acute-care facility more than once, that individual would be counted more than once. In addition, cross-sectional discharge rates for different disease groupings were calculated for ICD-9 codes. These ICD-9 diagnosis chapters were created on the basis of the “most responsible diagnosis” extraction criteria. To be included in these criteria, the particular diagnosis had to be listed on the discharge abstract as describing the most significant condition of the patient’s stay in hospital. All indicators were sex- and age-standardized using the 1991 Census population. The statistical significance of discharge rates from acute-care facilities was tested using Byar’s method and was based on the assumption of a Poisson distribution.Footnote 2323

General statistical notes

Throughout this report, estimates are provided with 95% confidence intervals. Reported statistics are taken to be significantly different if the 95% confidence intervals do not overlap. Rates described as “significantly different” mean that they are statistically different at the 95% confidence interval level. The small population in some Weak MIZ or No MIZ sometimes restricts the amount of data that can be used to calculate the rates. The level of uncertainty associated with rates calculated for these areas is greater than for areas with larger populations, such as CMA and CA. Consequently, confidence intervals have been calculated and rates presented so that the level of uncertainty is clearly expressed. These confidence intervals do not describe the uncertainty associated with potential bias, such as the uncertainty in proper CSD identification.

The primary boundaries chosen for the analysis were those of the 1996 Census because these were the boundaries available at the beginning of the research program. Differences in census boundaries for the different analyses in the report emerged as a result of the lag time between analyses. For example, requests to access physician claims data in British Columbia and discharge data from the HMDB were made at the beginning of this project. Comparing the use of the Nova Scotian census boundaries from 1996 and from 2001 revealed much greater efficiency in assigning geographic location by applying the more recent 2001 Census boundaries (2% of CSDs were unassigned) versus the earlier 1996 Census boundaries (approximately 18% of CSDs were unassigned). This was attributed to the creation of new postal codes in the period after the 1996 Census boundaries had been identified. Subsequent data extractions and analyses incorporated 2001 Census boundaries where possible.

Finally, although different data sources are presented in this publication, comparisons of results between data sources should be made with caution since they may measure different concepts within similar topic areas.

Report a problem or mistake on this page
Please select all that apply:

Thank you for your help!

You will not receive a reply. For enquiries, contact us.

Date modified: