Health Indicators - Technical Notes

82-221-XIE

Health Indicators

April 2001

Technical Notes

Health status indicators based on vital statistics (Statistics Canada - STC)

Rates are based on place of residence for indicators derived from birth and death events.

Vital statistics data in this product are based on three years of birth and death data (1995-1997) averaged over three years of births (1995-1997) and the 1996 population estimate, respectively. Table titles associated with these data reflect the mid-point of the three-year period being averaged (1996).

All data presented have an associated 95% confidence interval (CI). The confidence interval illustrates the degree of variability associated with a rate. Wide confidence intervals indicate high variability, thus, these rates should be interpreted and compared with due caution. Some age-standardized rates were suppressed due to both a very small underlying count plus extremely high variability. Confidence intervals can also be used to determine whether a rate in one health region is statistically below, above or no different than the rate for the same indicator in another health region.

The confidence intervals for the age-standardized rates were produced via the Spiegelman method. Reference: Spiegelman M. Introduction to Demography, Revised Edition. Cambridge MA, Harvard U Press, 1968. Formula 4.29, p 113

The confidence intervals for the crude count, crude rate and birth-related data were produced via the Fleiss method. Reference: Fleiss JL, Statistical Methods for Rates and Proportions, 2nd Ed, Wiley and Sons, NY, 1981

Due to the small population of Churchill health region (4690), Manitoba (pop. 1,110 in 1996) and the number of deaths, virtually all vital statistics data for this health region would, in the absence of any adjustment, need to be suppressed. As such, in this product all vital statistics data presented for region 4680 (Burntwood) are an aggregate of Burntwood and Churchill regions. For census-related data, however Burntwood and Churchill are presented separately.

Mortality rates, with the exception of crude rates and infant and perinatal mortality, are age-standardized using the direct method, and the 1991 Canadian Census population structure. The use of a standard population results in more meaningful mortality rate comparisons, because it adjusts for variations in population age distributions over time and across different geographic areas.

Birth and death data for 1996 and 1997 (and 1995 death data from Alberta) have been linked to health regions using postal codes reported with place of residence and converted to enumeration area (EA) using the automated geo-coding system developed in Health Statistics Division. These data were then aggregated to health region based on the EA level correspondence developed in Health Statistics Division with the cooperation of provincial Ministries of Health, Alberta Treasury and BC Stats.

Birth and death data (except Alberta) from 1995 were linked to health regions using two methods:

For most records (where health regions are comprised of complete census subdivisions) the standard geographical classification codes recorded on the vital statistics database were used to link records to health region.

Selected records (those linked to census subdivisions which are associated with more than one health region) were extracted from the data base. Using the registration numbers of these events, birth registration records and death certificates were accessed and the postal codes for place of residence were captured. These records were then geo-coded to enumeration areas (same as for data years 1996 and 1997), linked to health regions using the EA-to-HR correspondence, then merged with the remaining data for that year to get the most accurate health region link.

Life expectancy: This variable was calculated using the Chiang methodology for abridged life tables. The estimates are based on three years (1995-1997) of mortality data and 1996 population estimates, as described above. Abridged life tables use five-year age groupings of both population and mortality rate inputs (as opposed to single year age breakdown). Since there is more variability in the number of events by age in smaller geographic areas, abridged life tables are more suitable for the adaptation to a sub-provincial level (health region). Chiang's method in particular was chosen because it was relatively easy to adapt to the health region level data and included the calculation of standard error (in this case, addressing the variability of deaths from one year to the next).

Disability-free life expectancy: Estimates of disability-free life expectancy are calculated using Sullivan's method (Sullivan, DF. A single index of mortality and morbidity. HSMHA Health Reports 86 (April 1971): 347-354). Essentially, the latter generalizes Chiang's method (Chiang, CL. The Life Table and its Applications. Robert E. Krieger Publishing Company, Malabar, Florida, 1984: 316).

Sullivan's method is based on activity limitation rates within a population, according to sex and age group, in the calculation of life expectancy with disability. In the case of people living in health institutions, it was assumed that everyone had at least one activity limitation. For people living in other types of institutions, the hypothesis established is that the activity limitation rate by age group and sex was identical to the population in private households.

Disability-free life expectancy represents the difference between life expectancy and life expectancy with disability. The standard deviations of disability-free life expectancy estimates (and consequently the upper and lower limits of the confidence intervals associated with these estimates) are based on Colin Mathers' method (Mather, C. Health Expectancies in Australia 1981 and 1988. Australian Government Publishing Service, Canberra, 1991: 117). This method takes into account both the stochastic fluctuations in the observed death rates and the sampling variability of the activity limitation rates.

NOTE: The disability data for DFLE came from the 1996 Census of Population. Questions on disability in the Census of Population are generally used to capture the sample of post-censal Health and Activity Limitations Survey. Because of the decision not to conduct this survey in 1996, data on disability from the Census of population of 1996 were neither verified nor imputed. More precisely, no validation was undertaken to check the completeness or consistency of the data, and as a result, no corrections to the data were made. In addition, the data were nor adjusted to account for population undercounts.

DFLE estimates will vary according to both the concepts from which they are based and the surveys from which the data are extracted.

Birth statistics: Birth data on our Vital Statistics Database for Ontario are underestimated for data years 1996 and 1997 due to incomplete files. At least 1,500 Ontario births from 1997 were not available for geocoding. Birth data for those same years for some other provinces may also be incomplete. Thus, birth-related data (low birth weight, infant mortality, perinatal mortality), particularly for Ontario, should be interpreted with caution.

Low birth weight: Due to a high number of missing birth weight values from the original 1996 and 1997 birth files from Quebec, we received supplementary data files that matched birth records with hospital records. These files allowed us to analyze a larger group of Quebec births for low birth weight data. Thus, our present analysis may differ from previous Statistics Canada publications for low birth weight that present data for those same data years.