Report 4: Economic Burden of Illness in Canada, 2005–2008 – EBIC value of lost production due to premature mortality, 2004–2008

Report 4: EBIC Value of Lost Production due to Premature Mortality, 2004-2008

1. Background

Traditionally, the human capital method has been used to estimate the value of lost production due to premature mortality caused by illness and/or injury. This method estimates the value of lost production as the present value of an individual’s future earnings stream lost due to premature death. It is based on the assumption of zero involuntary unemployment or, in other words, it implicitly assumes that when a person dies he or she cannot be replaced. This assumption is likely to be untrue in today’s labour market, as evidenced by an unemployment rate ranging from 6% to 10% across the years 1996-2010 (48-50).

A more recent method, the friction cost method, was developed by a group of Dutch economists in the 1990s (51-54). The friction cost method allows for non-zero involuntary unemployment, which is closer to the real life situation of today’s economy. The friction cost method states that when a person dies he or she will be replaced by a worker who was previously unemployed. Of course, it will take some time to replace the worker, including the time required for job training.  This method estimates the loss in production only for the period it takes to replace the deceased worker, referred to as the friction period.

In previous EBIC reports, the value of lost production due to premature mortality was estimated using the human capital method. However, on the basis of the outcomes of the 2009 and 2010 EBIC workshops (organized by PHAC), it was recommended that the friction cost method be used to estimate mortality costs (55,56).

Although the current edition of EBIC focuses on the years 2005-2008, the 2004 results are also presented, as the data required to produce 2004 estimates were available. The EBIC 2004-2008 value of lost production due to premature mortality associated with labour market activities was estimated by diagnostic category/subcategory, age, sex, and province/territory using the friction cost method and a prevalence-based approach. This report describes the data sources and methods used to derive the 2004-2008 mortality cost estimates. Additionally it presents and discusses the results and the data and methods limitations.

2. Data Sources

Statistics Canada's Vital Statistics Death Database (2003-2008) was used in the estimation of 2004-2008 mortality costs. This database contains information on all deaths that occurred in Canada, including day/month/year of death, cause of death (coded using ICD version 10 codes), age, sex, province of residence, province of occurrence and other variables.

Additional data inputs were obtained from Statistics Canada, including average annual earnings specific to sex, age and province; annual provincial unemployment duration (in consecutive weeks); and average annual employment rate specific to sex, age and province (57,58,69).

3. Methods

The EBIC 2004-2008 estimates for the value of lost production due to mortality were derived by diagnostic category/subcategory (see Appendix C), age group, sex and province/territory.^{Footnote 66},^{Footnote 67} The estimates were derived by multiplying the period of lost production by the dollar value of production, more specifically the age-sex-province-specific earnings.

Following the methods of Koopmanschap et al. the value of lost production was estimated for the working age population comprising individuals aged 15-64 years (53).^{Footnote 68} As mentioned earlier, the 2009 and 2010 EBIC Workshop participants recommended considering premature deaths that occurred in the year of study as well as those that had occurred in previous years if the lost production fell in the year of study. The length of the friction period determined the required timeframe considered in order to estimate the value of lost production.

Following van Ours and Ridder's model, Koopmanschap et al. estimated the length of the friction period for the Netherlands in the years 1988 and 1990 using quarterly data on uncompleted vacancy durations and the number of vacancies from a large sample of Dutch firms (53,59). Given that such data were not available for Canada, provincial unemployment duration was used as a proxy for the length of the friction period.^{Footnote 69} Unemployment duration data were not available for the territories, and so the national average unemployment duration was used as a proxy. Mortality costs were estimated to the nearest half month, thus in 2004-2008 the unemployment duration used in the analysis ranged from 2 to 4.5 months (58).^{Footnote 70}

Since the unemployment duration ranged from 2 to 4.5 months, it was only necessary to consider premature deaths that occurred in the year of study (year t) and in the previous year (year t–1).  For example, if the friction period in year t was 4 months, then the analysis required going back 3.5 months in year t–1 (to September 16th) to capture the premature deaths considered for analysis (the minimum lost production of 0.5 months would have fallen in year t andthe remaining 3.5 months in year t-1). Then, the total value of lost production in year t is the sum of the value of lost production due to all premature deaths that occurred from September 16 in year t–1 to December 31 in year t. Thus, the number of deaths and period of lost production valued in the analysis depended on the length of the friction period.

Once the period of lost production for each death had been determined, lost production was valued using the appropriate age-sex-province-specific earnings. Average annual sex-age-province-specific earnings were used to value lost production for each person group (57).^{Footnote 71},^{Footnote 72} As earnings data for the territories were not available, corresponding national averages were used. For the data marked “use with caution”, which were mainly earnings data for the youngest age group, the corresponding national average was used.^{Footnote 73} Additionally, earnings data for individuals less than 20 years of age were used to value lost production for individuals aged 15-19 years, as age-specific earnings data exclusive to individuals in this age group were not available.

Considering that individuals who died may have been unemployed or not in the labour force, the value of lost production associated with each premature death was multiplied by the appropriate sex-age-province-specific employment rate (69).^{Footnote 74} Please refer to Appendix 1 in this report for a mathematical representation of the mortality methods and to Appendix 2 for hypothetical examples. For each year of analysis, the value of lost production due to mortality by diagnostic category/subcategory, sex, age group and province/territory was found by aggregating the costs into the appropriate categories.

4. Results

4.1 Costs by Diagnostic Category

Table 12 illustrates the EBIC 2004-2008 national cost estimates of the value of lost production due to mortality by diagnostic category. In 2008, total national mortality costs were $454.0 million. The three diagnostic categories with the largest costs were malignant neoplasms ($166.0 million, 36.6%), cardiovascular diseases ($92.4 million, 20.4%) and injuries ($84.6 million, 18.6%). All mortality costs were attributable across EBIC categories.

4.2 Costs by Diagnostic Category and Sex

Table 13 illustrates the EBIC 2008 national cost estimates of the value of lost production due to mortality by diagnostic category and sex. Total mortality costs were higher for males ($336.0 million, 74.0%) than for females ($118.0 million, 26.0%). The three diagnostic categories with the highest costs for males and females were malignant neoplasms ($106.2 million males, $59.8 million females), cardiovascular diseases ($77.2 million males, $15.3 million females) and injuries ($70.0 million males, $14.6 million females).

4.3 Costs by Diagnostic Category and Age

Figure 20 illustrates EBIC 2008 mortality costs for each age group. Individuals aged 15-34 years incurred the lowest percentage of mortality costs (6.3%) and individuals aged 35-54 years the highest (51.7%).

Figure 21 illustrates EBIC 2008 mortality costs by diagnostic category and age group for the five most costly diagnostic categories. Costs for the highlighted diagnostic categories were highest for individuals aged 35-54 years, except in the malignant neoplasms and cardiovascular diseases categories, in which costs were highest for individuals aged 55-64 years.

4.4 Costs Across the Years 2004-2008

Table 14 illustrates the EBIC 2004-2008 national estimates of the total value of lost production due to mortality in 2010 constant dollars. The value of lost production was lowest in 2004 ($446.8 million) and highest in 2006 ($470.7 million).

5. Discussion and Limitations

The main limitation is that unemployment duration is used as a proxy for the friction period. Koopmanschap et al. estimated the length of the friction period using vacancy duration data (53). As mentioned earlier, such data were not available for Canada for 2004-2008; instead, unemployment duration was used as a proxy for the friction period.

Conceptually, the friction period is the period of time it takes to replace a worker who has died. In general, it might take longer to replace a highly skilled worker than an unskilled/low-skilled worker. This is because highly skilled workers may be in short supply and also because of the length and complexity of the training required to replace these workers.

Unemployment duration is the time it takes for an unemployed person to find a suitable job. One might argue that it would be easier to find an unskilled/low-skilled job than a highly skilled one; if so, the unemployment duration would be shorter for the unskilled/low-skilled worker.

Given the differences between the human capital method and the friction cost method, estimates from EBIC 2004-2008 should not be compared with estimates from previous EBIC editions. Koopmanschap et al. estimated 1988 mortality costs for the Netherlands using both methods and found that mortality costs were 53 times higher using the human capital method (53).

Mortality cost totals by sex and age group are influenced by the total number of deaths that occurred and the earnings used to value lost production for each sex-age group. When the earnings data were unavailable or the coefficient of variation was high (greater than 16%), the corresponding national average was used; this may have misrepresented the value of production for certain person groups.

There is no clear trend for EBIC mortality costs over the years 2004-2008, the lowest and highest costs being seen in 2004 and 2006 respectively. Although the national unemployment duration (for both sexes) is approximately 4 months across the years 2004-2008, the province-level results (not discussed in this report) show some evidence that the friction period had a considerable impact on costs, as provincial mortality costs followed the same trend as the friction period.^{Footnote 75}

The 2004-2008 results excluded mortality cost estimates for residents of other provinces/territories who died in Quebec. However, the magnitude of the effect is expected to be small as the majority of individuals died in their province of residence. Mortality costs may be overestimated since individuals may have been employed but off work at the time of their death. If an individual died after being away from work for 3 consecutive months or longer due to an illness or injury, he or she would have been considered replaced according to the friction cost method. The value of lost production for these individuals should be included in the morbidity component, and any costs included in the mortality component would be considered double counting. Finally, unlike previous editions of EBIC, the 2004-2008 estimates do not include mortality costs associated with non-labour market activities.

6. Conclusion

The 2004-2008 value of lost production due to mortality associated with labour market activities was estimated for premature deaths using a prevalence-based approach and the friction cost method. Mortality cost estimates from previous EBIC editions were derived using the human capital method and thus cannot be compared with 2004-2008 estimates. The value of lost production for non-labour market activities was not considered in this edition of EBIC.