6.8 Assumptions and Interpretive Cautions

6.8.1 Data source: Canadian Community Health Survey (CCHS)

6.8.1.1 Included population

The CCHSs used to both develop the mortality algorithm (MPoRT) employed in the Project Big Life Planning Tool and to assess burden, only included individuals living in the community setting. Excluded from the sampling frame were people living on First Nation Reserves and Crown Land, people living in long-term care facilities, incarcerated individuals, and full-time members of the Canadian Forces. The algorithm also excluded people younger than 20 years of age.

The exclusion of these individuals may lead to a small under-estimation of the health behaviour attributable burden given the excluded populations often have a higher prevalence of unhealthy behaviours. For example, current smoker prevalence is higher on First Nations reserves and among the incarcerated compared to the general population {Marrett and Chaudhry (2003)} {Baybutt, Ritter, and Stover (2012)}. That stated, this under-estimation is likely small as the excluded population accounts for approximately 2% of the Canadian population {Beland (2002)}.

6.8.1.2 Reporting of health behaviours

The CCHSs collect self-reported responses for exposure to health risks. With self-reported surveys there is the possibility of social desirability bias, where respondents may over-report what they perceive to be healthy behaviours and under-report their unhealthy behaviours. For example, in Ontario the sum of self-reported alcohol consumption is about half the volume of alcohol sold {Rehm J (2006)}.

This reporting inaccuracy results in underestimation of burden. That stated, burden estimates are mostly affected when people report that they have the healthiest risk exposure, when they are actually in an unhealthy category (e.g., heavy smokers reporting that they are non-smokers).

Additionally, the full spectrum of the behaviour may not have been captured due to limited questions about the health behaviour in the CCHSs. For example, our measure of diet was based on fruit and vegetable frequency of consumption. Information on diet factors like quantity, sodium intake, trans fat, or other aspects of healthy and unhealthy eating were not captured. This will likely result in underestimation of the burden of poor diet.

Although more accurate risk factor ascertainment could improve discrimination and calibration of the algorithms used in the Project Big Life Planning Tool, these algorithms already have high discrimination and calibration.

6.8.2 Multivariable predictive risk algorithms

6.8.2.1 Hazard ratios and exposures are assumed constant

The hazard ratios in the algorithms of the Project Big Life Planning Tool and an individual’s health behaviour exposures are assumed to be constant. Constant hazard ratios do not account for period effects which could lead to an over-estimation of the burden of a health behaviour. For example, smoking intensity (number of cigarettes smoked per day) and duration (number of years a person has smoked) have been decreasing over time, therefore the differential mortality between smokers and non-smokers may be greater for older periods.

6.8.2.2 Uncertainty

While the original MPoRT algorithm estimated uncertainty intervals through a stepwise approach to model building that assessed confounding and mediation, the Project Big Life Planning Tool does not currently incorporate these uncertainty intervals.

D Bibilography

Baybutt, Michelle, Catherine Ritter, and Heino Stover. 2012. “Tobacco Use in Prison Settings: A Need for Policy Implementation.”

Beland, Y. 2002. “Canadian Community Health Survey - Methodological Overview.” Journal Article. Health Reports 13 (2): 9–14.

Marrett, L. D., and M. Chaudhry. 2003. “Cancer Incidence and Mortality in Ontario First Nations, 1968-1991 (Canada).” Journal Article. Cancer Causes Control 14 (3): 259–68. https://doi.org/10.1023/a:1023632518568.

Rehm J, et al. 2006. “Alcohol-Attributable Mortality and Potential Years of Life Lost in Canada 2001: Implications for Prevention and Policy.” Addiction 101: 373–84. https://doi.org/10.1111/j.1360-0443.2005.01338.x.