Concept: Population Denominators
Last Updated: 2001-05-01
Researchers should be aware that the population used in a denominator might have a significant effect on the rate for beginning and end of life time periods:
Where there are deaths before, or births after, the population count.
Migration might also affect population counts but this was tested, and is not a significant problem.
The problem may be exacerbated when looking between groups where there are differential death or birth rates.
- E.g. When comparing income quintiles or healthy/non-healthy areas.
The populations that MCHP uses are created from a June snapshot of the Manitoba Health registry and represent the prior December 31 population.
Region / Location of Residence
In both cases a combination of municipal code (and postal code in some cases) are used to determine the region or location of residence. If survey or other sources of data are being used that are based on postal code alone or census areas (such as enumeration areas) then the MCHP or Manitoba Health populations should not be used.
Data that have only postal codes should use populations generated specifically for postal code data. There is a lookup table from Manitoba Health and population files already developed.
The following two graphs illustrate the age-sex specific hospital separation rates for AMI & Hip Fracture over 16 years when two different denominators are used:
December population files excluding anyone who died during that calendar year.
- December population files including any deaths during that calendar year.
The graphs indicate that for age 85+ the rate can be about 10% lower when we include deaths during the year in the denominator. There was no change in the trend or relationship of the points.
Similar results were previously found for rates of ambulatory physician visits per 1,000 population aged 85+ over the past nine years when the two different denominators are used.
Currently, most research projects look at utilization or events that fall within the fiscal year (April to March). The denominator or population used by most MCHP research is as of December 31 (i.e., an approximate method). This date is generally used because it is a convenient approximation of the overall population and there are limited issues with regard to calculating ages at a specific time.
Since the late 1980s, Manitoba Health has been using June 1 populations. Starting in 2002, MCHP will get a copy of the Manitoba Health population file. This file can be recreated for earlier years from the registry snapshots within the MCHP data repository. A mid-year population is generally the correct method to use for rates, but not always convenient to create.
In both cases populations are defined from the Manitoba Health registry at the next snapshot date (e.g. December 2001 population uses the June 2002 registry).
- Manitoba Health uses the "live" registry in December to create the June population file. This allows changes in coverage, death and births to be captured appropriately.
Suggestions for making corrections to the population:
Suggestions range from a complete accounting of the person-years at risk (PYAR) for the denominator, which would count fractional years for those with less than full coverage during a study period.
Using the December (MCHP) population and add all of the deaths back into the population.
- This may not be appropriate since it will overcount the population.
Taking population as of September (1/2 of the fiscal year), add 1/2 of deaths during the year back in to the data.
- There are problems with this method because of delays in getting the mortality data.
Taking population at beginning and end of year and averaging the two. Alternatively to this is to take the June and December populations and average those.
- This cannot be done "across the board" because of the number of different regions that various research projects use. Along with the different regional definitions researchers will need to deal with the problem of who is the individual and where do (did) they live? In this case it is possible to get portions of individuals.
Leave as is, just footnote in a report that we are aware of this issue.
- Over the whole population, even in small areas, there is limited error - only elderly and newborns.
- This problem is not an issue for cohort studies where the population is based on a selected cohort such as all newborns during the year.
Recommendations from Researcher Meeting (April 24, 2002)
When focus of study is on seniors (65+) or age groups over 85+, the population used should be an average of the start and end of year.
Infant (age 0-1) should use a birth cohort for the population.
- If neither of the above then use the December MCHP population.
- Roos LL, Nicol JP. A research registry: uses, development, and accuracy. J Clin Epidemiol 1999;52(1):39-47. [Abstract] (View)
- Roos LL, Mustard CA, Nicol JP, McLerran DF, Malenka DJ, Young TK, Cohen MM. Registries and administrative data: organization and accuracy. Med Care 1993;31(3):201-212. [Abstract] (View)
Manitoba Centre for Health Policy
Community Health Sciences, Max Rady College of Medicine,
Rady Faculty of Health Sciences,
Room 408-727 McDermot Ave.
University of Manitoba
Winnipeg, MB R3E 3P5 Canada