june-cjp

The Valuation of Productivity Costs Due to Premature Mortality: A Comparison of the Human-Capital and Friction-Cost Methods for Schizophrenia

Ron Goeree, MA¹, Bernie J O’Brien, PhD², Gordon Blackhouse, MBA³, Karen Agro, PharmD⁴,
Paula Goering, RN, PhD⁵

Objective: To compare productivity-cost estimates for schizophrenia-related premature mortality in Canada in 1996 using the human-capital (HC) approach and friction-cost (FC) method.

Methods: The number of deaths directly attributed to schizophrenia was combined with the estimated number of deaths attributable to schizophrenia from suicide. These premature deaths were valued using 2 methods: 1) the traditional HC approach, based on “potential” lost output to normal age of retirement, and 2) the FC method, based on finding a replacement worker.

Results: In 1996, there were 342 male and female preretirement deaths attributed to schizophrenia, directly or indirectly by suicide, in Canada. Most deaths were in males (78%) and by suicide (97%). The productivity cost of these deaths was estimated to be $105 million using the HC approach but only $1.53 million using the FC method.

Conclusions: Productivity-cost estimates from the HC approach are substantially higher than those obtained from the FC method (69 times higher). In circumstances of unemployment, the HC approach is an overestimate of future productivity losses for premature mortality.

(Can J Psychiatry 1999;44:455–463)

Key Words: schizophrenia, productivity cost, indirect cost, human-capital approach, friction-cost method

The use of economic appraisal to evaluate health care technologies has increased substantially over the past decade (1,2). Economic evaluations have been less frequent regarding mental health than other areas of health care; however, their use is increasing. For example, all new mental health drug submissions in Ontario and most other provinces require that an economic evaluation accompany the submission (3,4). It is important, therefore, that we know what these evaluations are, how they differ, what methods and assumptions are used in their conduct, and how to interpret the study findings and conclusions.

Despite the increased use of economic evaluations, numerous books and articles discussing theoretical and practical issues (1,2), and the development of guidelines on how to conduct and report these evaluations (3–5), there is still debate on several contentious methodological issues. Since the basic principles of economic appraisal apply equally to all evaluations, these issues and debates are as important to mental health as they are to other areas of health care. This paper addresses one of the more contentious and heavily debated issues; namely, the measurement and valuation of permanent work loss due to disability or premature mortality. For illustrative purposes, the paper focuses specifically on premature death resulting from disease, disability, or illness. However, the issue is equally applicable to permanent work loss due to disability.

Productivity Losses in Economic Evaluations

In addition to direct health care costs, economic evaluations commonly include cost estimates for time off work and reduced productivity on the job. If they were not included, the analysis would overlook an important financial burden to patients, their families, and society in general. These productivity costs can arise for several reasons. Time off work can be temporary or permanent and can result from illness or disability, from treatment and rehabilitation, or from premature mortality. Permanent work loss may arise when an employee simply can no longer return to work, when the employee cannot perform his or her previous job and cannot find another suitable job, or through premature mortality as a result of disease, disability, illness, or suicide.

The inclusion and valuation of permanent work loss is particularly important for diseases and interventions that have high rates of premature mortality. In prevalence-based burden-of-illness studies, for example, the total direct and indirect cost of a disease, disability, or illness is typically estimated for 1 year. All premature deaths attributed to the disease for 1 year are estimated, and the present and future work absences as a result of these deaths are included in the 1-year burden cost estimate. As can be expected with this method of assessment, diseases or treatments with excessive rates of premature mortality (from disease, treatment, or suicide) can appear to be very expensive. For example, in an assessment of the economic burden of schizophrenia in the United States (US), Rice and Miller calculated the total cost of the disease at US$32.5 billion in 1990, of which 47% ($15 billion) was due to lost productivity from mortality and morbidity alone (6).

Human-Capital Approach

The traditional approach to measuring and valuating production that is lost due to temporary work absences, reduced productivity at work, and permanent work absence from morbidity or premature mortality is the so-called human-capital (HC) approach. This term derives from the observation that variation in earnings over a person’s lifetime is due to investment in HC through education, on-the-job training, and work experience. Valuing life by means of HC has a long history, dating back to the 1700s (7,8). However, it wasn’t until the late 1950s (9,10) and the 1960s (10–14) that the approach gained in popularity. The HC approach is based on the concept of “potential” productive output that may be lost. Morbidity and mortality destroys potential output by causing persons to lose time and effectiveness from work, forcing them out of the labour force completely, due to disability, or bringing about premature death.

Although measuring and valuating short-term work absence and reduced job productivity are problematic, the more contentious calculations with the HC approach concern permanent work absence and premature mortality. In measuring the value of a premature death at age 35 years for example, the HC approach estimates the annual earnings for each year of potential lost employment (that is, from age 35 years to 65 years) and adds these earnings together. This calculation results in an estimate, at age 35 years, of the potential future income that may be lost because of premature death. An example of the productivity-loss estimate for a 35-year-old male is represented by the large light shaded area under the earning curve in Figure 1. This potential loss in productivity is usually valued using market wage rates, and earnings in the future are discounted at a constant annual rate (for example, 3%) to account for a positive time rate of preference (that is, we value costs in the future less than costs today). Some researchers make a further adjustment to annual earnings to account for labour-force participation rates by age and gender.

A particular limitation of the HC approach to measuring productivity losses for permanent work absence is that it implicitly assumes that labour markets are in equilibrium with no or little unemployment (that is, a worker cannot be replaced). This approach further assumes that if a person leaves the labour force they will not be replaced, with their production being lost until the age of retirement. Clearly this is a questionable assumption for most industrialized countries, where an excess supply of labour exists. The rate of unemployment in Canada, for example, currently is around 8% to 10%. The question is whether it is reasonable to assume that a worker dying today will not and cannot be replaced by another worker from the unemployed pool.

With current levels of unemployment, many researchers have questioned whether the burden estimates provided by the HC approach are “real” (that is, what society would actually realize). This issue of valuation is important because it has been suggested (especially by special interest groups) that the magnitude of these burden estimates should be used, at least in part, for setting priorities for health care spending. In light of the ongoing debate and controversy around their calculation, some researchers have simply chosen to exclude productivity costs from their analysis, while others have decided to report direct costs and productivity costs separately. However, a new method recently has been proposed that offers researchers a different way to measure and value productivity losses from permanent work absence due to disability or premature mortality.

Friction-Cost Method

In an attempt to measure “actual” rather than “potential” production loss, Koopmanschap and others have developed the friction-cost (FC) method (15–18). This method modifies the HC approach by allowing for worker replacement by other workers or by those in the unemployed pool. In its simplest form, the FC method adjusts cost calculations by stating that, when a person is absent from work or terminates employment altogether, the actual productivity loss from that job continues only until the time at which other workers assume that job or when another worker from the unemployed pool fills that vacant position. The period of time required for worker replacement is called the “friction period.”

The FC adjustment to HC estimates reduces the value of productivity loss, recognizing that workers can and will be replaced. The friction period will vary not only by type of job but also by local employment circumstances at the time of worker replacement. Across all employed males, the productivity loss for the 35-year-old male using the FC method is shown as the thin dark shaded area in Figure 1. The magnitude of the difference between the 2 shaded areas in Figure 1 illustrates the conceptual differences between these 2 approaches to productivity-loss estimation.

Figure 1. Average annual earning profile for males in Canada in 1996 and the productivity loss for a premature death at age 35 years, comparing the human-capital (HC) approach and friction-cost (FC) method, excluding adjustments for labour-force participation and not discounting future costs

Objective

Overall, this study compares productivity-cost estimates for schizophrenia-related premature mortality in Canada for the year 1996 using the unadjusted HC approach with estimates obtained from the adjusted HC approach, known as the FC method. Only productivity costs due to premature mortality are included in this analysis. Direct costs and productivity costs due to schizophrenia morbidity are examined in another paper in this issue.

Methods

To estimate the productivity cost of premature mortality due to schizophrenia in Canada, the following 5 data inputs, by age and gender, were required:

1. Labour-force participation (employment ratios) for the general population and for persons with schizophrenia.

2. Average annual earnings for the general population and for persons with schizophrenia.

3. Deaths directly attributable to schizophrenia and estimates of the proportion of all suicide deaths that can be attributed to schizophrenia.

4. Estimates of friction periods for worker replacement.

5. Normal age of worker retirement.

Employment and Earnings Data

Employment rates are not a true indicator of the proportion of the population in paid employ because these calculations exclude individuals who are not actively seeking employment. Therefore, information from Statistics Canada publications on persons in paid employment (19) and population (20) were combined to calculate the proportion of the general population, by age and gender, in paid employment. Similar estimates for persons with schizophrenia were obtained from a study by Bland in Alberta (21).

Average annual earnings data for persons in paid employment, by age and gender, were also obtained from a Statistics Canada publication (22). Canadian estimates of annual earnings for persons with schizophrenia are not available. Therefore, earnings for persons with schizophrenia were estimated by adjusting the general-population earnings by applying productivity weights from a study by Rice and Miller (6), which reported the ratio of earnings of persons with schizophrenia to earnings of age–gender matched persons of the general population. These annual earnings for the general population and for persons with schizophrenia were adjusted using employment ratios to account for the probability that someone of a particular age and gender would be in the labour force.

Premature Mortality

Deaths due to schizophrenia were obtained from the most recent Statistics Canada publication on the number of deaths by cause (23). Code 295 from the International Classification of Disease—Clinical Modification (ICD-9-CM) (24) was used to identify the deaths directly attributed to schizophrenia. Since several patients suffering from schizophrenia commit suicide, the schizophrenia cause-of-death code by itself would underestimate the total deaths that result from the disease every year. Therefore, 10% of deaths reported as due to suicide (ICD-9-CM E950-E959) were assumed to result from schizophrenia (25). Deaths reported as directly due to schizophrenia were added to suicide deaths attributable to schizophrenia for estimates of total schizophrenia deaths, by age and gender, in Canada in 1996.

Productivity-Cost Estimates

The conceptual differences between the HC and FC methods of calculating productivity loss were discussed previously. The algebraic formulas for each calculation method are presented in the Appendix. Estimates of annual earnings and labour-force participation for persons with schizophrenia were used for calculating productivity losses under both the HC and FC methods. The impact of using general-population earnings and employment ratios (as opposed to schizophrenia-specific earnings) on the results for the HC approach was explored in a sensitivity analysis. The friction period required for worker replacement in the event of premature death was assumed to be 3 months (15–17). Alternative friction periods, ranging from 1 month to 30 years, were used as variants for sensitivity analyses. The normal age of retirement for both males and females was assumed to be 65 years of age. Therefore, schizophrenia-related deaths beyond age 65 years were not included in the productivity-cost calculations.

Results

Population and employment data, by age and gender, for the general population are presented in Table 1. The general-population employment ratios are smallest for both males and females in the youngest age-group (0.338 and 0.356 respectively). These ratios increase substantially for both males and females in later ages and then drop again just before normal age of retirement at age 65. Throughout all age-groups, the employment ratios were higher for males than for females. Also presented in Table 1 are the employment ratios for persons with schizophrenia from the study by Bland (21). The employment ratios for persons with schizophrenia have similar age and gender relationships as for the general population. As might be expected, the employment ratios for the general population are higher than the ratios for age–gender-matched persons with schizophrenia.

Table 1. Employment data for the general population and for persons with schizophrenia in Canada in 1996, by age and gender
	Population (thousands)^a		Persons in paid employment (thousands)^b		General-population employment ratios		Schizophrenia employment ratios^c
Age-group	Male	Female	Male	Female	Male	Female	Male	Female
15–19 years	1 008	961	341	342	0.338	0.356	0.280	0.340
20–24 years	1 005	984	643	614	0.640	0.624	0.469	0.546
25–34 years	2 364	2 355	1 945	1 671	0.823	0.709	0.756	0.670
35–44 years	2 498	2 509	2 120	1 815	0.849	0.723	0.785	0.701
45–54 years	1 920	1 916	1 582	1 278	0.824	0.667	0.775	0.688
55–64 years	1 247	1 281	689	429	0.552	0.335	0.529	0.344
Total	10 042	10 006	7 320	6 149	0.729	0.615	—	—

^aSource: Statistics Canada “Annual Demographic Statistics” 1995 (20).
^bSource: Statistics Canada “The Labour Force” 1996 (19).
^cSource: Bland (21).

Average annual earnings, by age and gender, for persons in paid employment are presented in Table 2. For both males and females, earnings increase sharply between ages 15 and 34 years and then level off until retirement. Throughout all age-groups, annual earnings are higher for males than for females. Also presented in Table 2 are the schizophrenia productivity weights, showing the percentage of general-population earnings that persons with schizophrenia earn, from the US study by Rice and Miller (6). Based on these weights and using Canadian earnings data, the estimated earnings for persons with schizophrenia in Canada are presented in the last 2 columns of Table 2. A notable drop in annual earnings is apparent for males aged 55 to 64 years, reflecting a productivity weight of only 22.6% from the US study.

Table 2. Earnings data for the general population and productivity-weighted imputed earnings for persons with schizophrenia in Canada in 1996, by age and gender
	Average annual earnings for employed general population^a ($)		Schizophrenia productivity weights^b (%)		Imputed annual earnings for employed persons with schizophrenia^c($)
Age-group	Male	Female	Male	Female	Male	Female
15–19 years	5 205	4 663	92.86	97.05	4 883	4 525
20–24 years	24 180	20 113	92.86	97.05	22 454	19 520
25–34 years	35 234	28 390	92.51	93.87	32 595	26 650
35–44 years	43 935	32 088	78.00	92.00	34 269	29 521
45–54 years	47 156	33 071	55.00	89.00	25 936	29 433
55–64 years	43 973	29 543	22.60	84.70	9 938	25 023

^aSource: Statistics Canada “Earnings of Men and Women” 1995 (22).
^bSource: Rice and Miller (6).
^cCalculated by multiplying general population earnings by schizophrenia productivity weights (6).

The number of preretirement deaths directly due to schizophrenia and due to schizophrenia by suicide are presented in Table 3 for males and in Table 4 for females. In total, there were an estimated 268 male and 74 female preretirement deaths due to schizophrenia in Canada in 1996. Listing schizophrenia as the cause of death was not common (6 males and 5 females). Most deaths (98% for males and 93% for females) were estimated from total suicide deaths in Canada. There were slightly more suicide deaths for persons aged 30–45 years; however, deaths were distributed through all age categories.

Table 3. Value of productivity loss for males with schizophrenia, using the human-capital approach and friction-cost method, in Canada in 1996
					Value of production lost for all male deaths ($, millions)
Age-group	Deaths from schizophrenia^a(ICD-295)	Deaths from suicide^a (ICD E950-E959) attributable to schizophrenia^b	Present value of future production lost per death^c ($)	Friction cost per death^d ($)	Human-capital approach	Friction-cost method
10–14 years	0	4	391 789	0	1.45	0
15–19 years	0	21	451 361	338	9.25	0.01
20–24 years	0	30	496 661	2 633	14.95	0.08
25–29 years	0	28	488 675	6 160	13.83	0.17
30–34 years	0	35	431 756	6 160	15.24	0.22
35–39 years	1	39	361 048	6 725	14.55	0.27
40–44 years	2	31	271 447	6 725	8.79	0.22
45–49 years	1	25	181 795	5 025	4.76	0.13
50–54 years	0	20	100 833	5 025	2.03	0.10
55–59 years	0	15	38 011	1 314	0.57	0.02
60–64 years	2	14	15 317	1 314	0.24	0.02
Total	6	262	—	—	85.67	1.24

^aSource: Statistics Canada “Causes of Death” 1994 (23).
^bAssumes 10% of all suicides attributable to schizophrenia (25).
^cCalculated as the sum of earnings lost for each year to retirement � employment ratio for each age-group and discounted to present value at 3%.
^dCalculated as annual earnings/12 � age-specific employment ratio � 3-month friction period.

The present value of productivity loss for each premature death using the HC approach is presented in Tables 3 and 4 by age-group. The present value of productivity loss per death initially increases with age, reflecting higher earnings and labour-force participation. After age 30 years, however, the present value per death decreases, reflecting fewer potential years of work lost due to the premature death. The FC for each premature death is also presented in Tables 3 and 4 by age-group. The FC per death increases sharply to age 25 years, levels off between the ages of 25 and 54 years, and then drops until age 65 years. For both males and females, the FC per death is highest between the ages of 35 and 44 years, reflecting higher wages and labour-force participation during these ages.

Table 4. Value of productivity loss for females with schizophrenia, using the human-capital approach and friction-cost method, in Canada in 1996
					Value of production lost for all female deaths ($, millions)
Age-group	Deaths from schizophrenia^a (ICD-295)	Deaths from suicide^a (ICD E950-E959) attributable to schizophrenia^b	Present value of future production lost per death^c ($)	Friction cost per death^d ($)	Human-capital approach	Friction-cost method
10–14 years	0	1	333 581	0	0.40	0
15–19 years	0	5	383 495	385	1.80	< 0.01
20–24 years	0	6	417 095	2 665	2.42	0.02
25–29 years	0	6	410 196	4 464	2.34	0.03
30–34 years	0	8	377 889	4 464	2,95	0.03
35–39 years	0	11	334 500	5 174	3.55	0.06
40–44 years	1	9	274 612	5 174	2.77	0.05
45–49 years	0	8	206 116	5 062	1.65	0.04
50–54 years	1	6	128 210	5 062	0.92	0.04
55–59 years	0	6	62 238	2 152	0.37	0.01
60–64 years	3	3	25 079	2 152	0.16	0.01
Total	5	69	—	—	19.33	0.29

The difference between the HC and FC methods of calculation is quite apparent when the 2 cost-per-death columns in Tables 3 and 4 are compared. The difference in the cost per death is largest for earlier ages; however, a large difference is maintained until age 55 years. For a male premature death at age 35 years, for example, the HC productivity-cost estimate is $361 048, whereas the FC estimate is only $6725. This suggests that the larger the number of deaths and greater the proportion of deaths for younger ages, the larger will be the difference in total productivity-cost estimation between the 2 approaches.

Based on these productivity costs per death and number of deaths by age, the total productivity costs for schizophrenia in Canada in 1996 using the HC and FC methods are presented in the last 2 columns of Tables 3 and 4. Using the HC approach, the total productivity-cost estimate for males and females combined is $105 million, of which 82% was for males. Two-thirds of the total cost estimate were for deaths between 20 and 40 years of age. Using the FC method, the total productivity-cost estimate for males and females combined is only $1.53 million. As with the HC estimate, most of the FC estimate was for males (81%), and two-thirds of the total cost were for deaths between 25 and 44 years of age. However, this is where the similarities end. The cost estimate using the HC approach is 69 times higher than the cost estimate from the FC method ($105 versus $1.53 million).

The productivity-cost estimates from the HC and FC methods for males and females combined are presented in Table 5. Also in Table 5 are 2 sensitivity variants, 1 for the HC approach and 1 for the FC method. The FC productivity-cost estimate, assuming a 1-month job vacancy, is essentially one-third the estimate for a 3-month period ($0.51 versus $1.53 million). When earnings and employment ratios from the general population are used, the productivity-cost estimate from the HC approach increases 66% to $174.51 million. This HC cost estimate is 114 times larger than the 3-month job-vacancy FC estimate and 342 times larger than the 1-month job-vacancy FC estimate. The results in Table 5 show that the FC estimates are sensitive to the assumed vacancy duration and that the HC cost estimates are sensitive to the earnings and employment ratios used (that is, general population or schizophrenia-specific). In Figure 2 we demonstrate that the productivity-cost estimate using the FC method is equivalent to the HC estimate only in the extreme case where a friction period of 30 or more years is assumed.

Figure 2. Sensitivity analysis of productivity-loss estimates for schizophrenia-related premature deaths in Canada, by length of friction period required for worker replacement

Table 5. Comparison of estimates of lost future productivity for male and female schizophrenia-related premature deaths in Canada using the human-capital and friction-cost methods
		Value of future productivity loss ($, millions)
		Human-capital approach		Friction-cost method
Age-group	Number of deaths	Schizophrenia earnings and employment ratios	General population earnings and employment ratios	3-month vacancy	1-month vacancy
10–14 years	5	1.85	2.65	0	0
15–19 years	26	11.06	16.05	0.01	< 0.01
20–24 years	36	17.37	25.44	0.10	0.03
25–29 years	34	16.17	24.18	0.20	0.07
30–34 years	43	18.19	28.88	0.25	0.08
35–39 years	51	18.09	30.72	0.33	0.11
40–44 years	43	11.57	21.42	0.27	0.09
45–49 years	34	6.41	13.60	0.17	0.06
50–54 years	27	2.95	7.23	0.14	0.05
55–59 years	21	0.94	3.04	0.03	0.01
60–64 years	22	0.40	1.31	0.03	0.01
Total	342	105.00	174.51	1.53	0.51

Discussion

The traditional HC approach for measuring and valuing productivity losses due to permanent work absence and premature mortality has come under increasing criticism. The HC approach assumes that a worker will not and cannot be replaced by other workers or by unemployed persons. The concept of potential productive output that may be lost is central to the HC approach. However, many have questioned whether these potential losses are realistic when there is an excess supply of labour (that is, unemployment). The HC approach has been criticized for assuming that a vacant position will never be filled and that society will continue to lose the value of a person’s output up until that person’s retirement. The FC method is an attempt to estimate actual productivity losses to society by acknowledging that workers can and will be replaced. Given current rates of unemployment and the potential for worker replacement, we believe that the cost estimates produced by the HC approach are not realistic.

We found that, in the case of schizophrenia, the HC approach produced productivity-cost estimates that are substantially higher than those arrived at by the FC method. For the estimated 342 schizophrenia-related deaths in Canada in 1996, our base-case productivity-cost estimate was $105 million using the HC approach, compared with $1.53 million with the FC method. In our base-case analysis, the HC approach resulted in an estimate that was 69 times higher than the FC method. Sensitivity analysis demonstrated a productivity-cost estimate for the HC approach that was 114 times higher when general-population earnings and employment ratios were used and up to 342 times higher when a 1-month friction period was assumed. Only in the unlikely situation where the friction period was 30 years or longer were the estimates from the HC and FC methods equivalent.

The differences in cost estimates between the HC and FC methods in our base-case analysis are similar to those found in other studies. In a study on cardiovascular disease, Koopmanschap found the HC estimate for mortality was 35 times higher than the FC estimate (15). In studies of all-cause mortality, HC estimates have been found to be between 53 (17) and 70 (26) times higher than those obtained by the FC method. Although it is difficult to generalize from these few studies, some observations are apparent. The difference in productivity-cost estimates for premature mortality between the HC and FC methods will be larger with more deaths; a greater proportion of younger deaths; higher earnings and labour-force participation rates; shorter friction periods (including due to higher unemployment rates); and a lower discount rate for future costs.

The productivity-cost estimates in this analysis depend on the proportion of suicide cases in Canada that may be due to schizophrenia. Although alternative proportions will affect the absolute cost estimates, alternative proportions will not have a significant impact on the relative cost differences and, therefore, will not change the conclusion of the paper. Nevertheless, there is a need for further research to confirm the proportion of suicide cases in Canada that may be attributable to the disease. Similarly, Canadian earnings or productivity weights for persons with schizophrenia are required. Once again, however, this information will only impact on the absolute, not relative, cost results and conclusions from this paper.

The FC method is a recent development and is starting to be adopted in health care economic evaluations. The major difficulty with using the FC method is the need for data on disease-specific employment and job-vacancy duration (that is, the friction period). This information is not routinely collected by federal or provincial governments in Canada or by employment-statistics agencies. Studies in other countries suggest that the vacancy duration may be as long as 3 months, depending on the occupation and local employment circumstances. For example, Koopmanschap estimated a friction period of 2.8 months for persons in jobs requiring basic education, compared with 3.5 months for persons in jobs requiring university education (17). The same authors calculated the friction period to be 2.8 months in the Netherlands in 1988, when the unemployment rate was 10.2% (15). However, when the unemployment rate decreased to 8.2% in 1990, the estimated friction period increased to 3.2 months (16). This suggests the friction period may be sensitive to changes in the unemployment rate. In Canada, technological labour-saving capital investments and immigration of new workers from other countries likely will mean that unemployment rates will not fall substantially in the future.

This paper focused on calculating productivity costs resulting from premature mortality, but the same issues apply to calculating permanent work absence due to disability. There are several important methodological issues with both the HC and FC methods concerning short-term and temporary work absence, which were not considered in this analysis. In addition, we have not attempted to impute values for non–labour-market activities such as housekeeping, house maintenance, childcare, child education, or leisure activities. Imputations for these activities may be particularly important concerning persons not in the labour force. As noted by proponents of both the HC and FC methods, resource-allocation decisions based on cost calculations from labour-market activities only may lead to undesirable policies and programs that favour white middle-aged, well-educated men, simply because they have higher earnings and labour-force participation rates.

Health care professionals and mental health care administrators need to be aware of the method researchers use when calculating productivity losses and of the assumptions inherent in that method. Health care policies and resource-allocation decisions should be based on actual not potential cost consequences of disease, illness, or treatment. For high levels of unemployment, the FC method may offer more realistic productivity-cost calculations for premature mortality and permanent work absence due to disability.

Clinical Implications

Estimation methods are important for diseases like schizophrenia that have high rates of premature mortality.
The value of production loss should be based on labour-market conditions and the potential for worker replacement.

Limitations

The proportion of suicide deaths due to schizophrenia is uncertain.
Earnings of persons with schizophrenia are not well-studied.
The value for nonlabour activities was excluded.

Acknowledgements

This study was funded by the Mental Health Division of Health Canada. The authors thank Dr Roger Bland (University of Alberta), Dr Sam Sussman (London Psychiatric Hospital), Mr Bert van den Berg (Schizophrenia Society of Canada), and Ms Carol Silcoff (Health Canada) for their review and comments. Dr O’Brien is supported by a career award in health sciences from the Medical Research Council and Prescription Drug Manufacturers of Canada. The authors thank Christine Henderson for her valuable research assistance.

References

1. Drummond MF, O’Brien BJ, Stoddart GL, Torrance GW. Methods for economic evaluation of health care programmes. 2nd ed. Oxford: Oxford University Press; 1997.

2. Gold MR, Siegel JE, Russell LB, Weinstein MC. Cost-effectiveness in health and medicine. Oxford: Oxford University Press; 1996.

3. Ontario Ministry of Health. Ontario guidelines for economic analysis of pharmaceutical products. Toronto: Ontario Ministry of Health; 1994.

4. Canadian Coordinating Office for Health Technology Assessment. Guidelines for economic evaluation of pharmaceuticals: Canada. 1st ed. Ottawa: CCOHTA; 1994.

5. Commonwealth Department of Human Services and Health. Guidelines for the pharmaceutical industry on preparation of submissions to the pharmaceutical benefits advisory committee (including major submissions involving economic analyses). Canberra (Australia): Australian Government Publishing Service; 1995.

6. Rice DP, Miller LS. The economic burden of schizophrenia: conceptual and methodological issues, and cost estimates. In: Moscarelli M, Rupp A, Sartorius N, editors. West Sussex: John Wiley & Sons; 1996.

7. Petty W. Political Arithmetick, or a disclosure concerning the extent and value of lands, people, buildings, etc. London: Robert Caluel; 1699.

8. Farr W. Contribution to the 39th annual report of the Registrar General of births, marriages and deaths for England and Wales. London; 1876.

9. Fein R. Economics of Mental Illness. New York: Basic Books; 1958.

10. Mushkin SJ, Collings F. Economic costs of disease and injury. Public Health Rep 1959;74:795–809.

11. Weisbrod BA. Economics of public health. Philadelphia: University of Pennsylvania Press; 1961.

12. Klarman HE. Syphilis control programs. In: Dorfman R, editor. Measuring the benefits of government investments. Washington (DC): Brookings Institution; 1965. p 364–414.

13. Rice DP. Estimating the costs of illness. Am J.Public Health 1967;68:424–40.

14. Rice DP. Estimating the cost of illness. Health Economics Series #6. Publication #947-6. Washington (DC): US Government Printing Office; 1966.

15. Koopmanschap MA, van Ivenveld BM. Towards a new approach for estimating indirect costs of disease. Soc Sci Med 1992;34:1005–10.

16. Koopmanschap MA, Rutten FFH. The impact of indirect costs on outcomes of health care programs. Health Econ 1994;3:385–93.

17. Koopmanschap MA, Rutten FFH, van Ivenveld BM, van Roijen L. The friction cost method for measuring indirect cost of disease. Journal of Health Economics 1995;14:171–89.

18. Koopmanschap MA, Rutten FFH. A practical guide for calculating indirect costs of disease. Pharmacoeconomics 1996;10:460–6.

19. Statistics Canada. The labour force. Ottawa: Statistics Canada; 1996.

20. Statistics Canada. Annual demographic statistics, 1995. Ottawa: Statistics Canada; 1995.

21. Bland RC, Stebelsky G, Orn H, Newman SC. Psychiatric disorders and unemployment in Edmonton. Acta Psychiatr Scand 1988;77(Suppl 338):72–80.

22. Statistics Canada. Earnings of men and women, 1995. Ottawa: Statistics Canada; 1995.

23. Statistics Canada. Causes of death. Ottawa: Statistics Canada; 1994.

24. Hart A, Schmidt K, Aaron WS. St Anthony’s illustrated ICD-9-CM code book. Reston (VA): St Anthony’s Publishing; 1998.

25. Goodwin FK. Manic depressive illness. New York: Oxford University Press; 1990.

26. van Beeck E, van Roijen L, Mackenbach J. Medical costs and economic production losses due to injuries in the Netherlands. J Trauma 1997;42:1116–23.

R�sum�

Objectif : Comparer les co�ts estim�s productivit� attribuables aux d�c�s pr�matur�s li�s � la schizophr�nie au Canada en 1996, � l’aide de l’approche du capital humain (CH) et de la m�thode des co�ts frictionnels (CF).

M�thodes : Le nombre de d�c�s directement attribuables � la schizophr�nie ont �t� combin�s avec le nombre estim� de d�c�s par suicide attribuables � la schizophr�nie. Ces d�c�s pr�matur�s ont �t� �valu�s � l’aide deux m�thodes : 1) l’approche du CH traditionnelle bas�e sur la production � potentielle � perdue � l’�ge de la retraite, et 2) la m�thode des CF bas�e sur la recherche d’un travailleur suppl�ant.

R�sultats : Au Canada en 1996, il y a eu 342 d�c�s d’hommes et de femmes attribu�s � la schizophr�nie, directement ou indirectement par suicide. La plupart de ces d�c�s se sont produits chez des hommes (78 %) et par suicide (97 %). Le co�t de productivit� de ces d�c�s a �t� estim� � 105 millions de dollars selon l’approche du capital humain, mais seulement � 1,53 million de dollars selon la m�thode des co�ts frictionnels.

Conclusions : Les estimations des co�ts de productivit� obtenues gr�ce � l’approche du CH sont sensiblement plus �lev�es que celles obtenues � partir de la m�thode des CF (69 fois plus �lev�es). En situation de ch�mage, l’approche du CH est une surestimation des pertes de production futures attribuables � un d�c�s pr�matur�.

Appendix 1. The algebraic formulas for estimating the productivity cost asociated with premature mortality using the human-capital (HC) and friction-cost (FC) methods.
Appendix 1 is not available online. Please refer to the print copy.

Manuscript received March 1999.

¹Research Coordinator, Department of Clinical Epidemiology and Biostatistics, McMaster University; Centre for Evaluation of Medicines, St Joseph’s Hospital, Hamilton, Ontario.

²Associate Professor, Department of Clinical Epidemiology and Biostatistics, McMaster University; Centre for Evaluation of Medicines, St Joseph’s Hospital, Hamilton, Ontario.

³Cost Analyst, Department of Clinical Epidemiology and Biostatistics, McMaster University; Centre for Evaluation of Medicines, St Joseph’s Hospital, Hamilton, Ontario.

⁴Research Fellow, Centre for Evaluation of Medicines, St Joseph’s Hospital, Hamilton, Ontario.

⁵Director, Health Systems Research Unit, Centre for Addiction and Mental Health, Clarke Division, Toronto, Ontario.

Address for correspondence: R Goeree, Centre for Evaluation of Medicines, St Joseph’s Hospital, 50 Charlton Avenue East, H-304, Hamilton, ON L8N 4A6

email: goereer@fhs.csu.mcmaster.ca

Can J Psychiatry, Vol 44, June 1999