Economic Burden of Illness in Canada, 1993

Discussion

Methods for estimating the cost of illness have been refined extensively from our original report, Economic Burden of Illness in Canada, 1986. This section provides an overview of how methods have been improved and what limitations remain. Further details about each specific cost component can be found in the section entitled "Expenditures by Cost Component". We also provide a more general context for our methods, comparing our results with other sources of information, in Appendix 7.

Drug Expenditures

Our total drug expenditure ($9.9 billion) now includes estimates of prescription drugs, non-prescription drugs and personal health supplies bought in retail stores and prescription drugs purchased for hospital use. Although costs of personal health supplies and most non-prescription drugs are still not available by diagnostic category, we are able to categorize more than 90% of all prescription drugs sold in pharmacies by diagnostic category, age and sex. We also estimate hospital drug expenditures by diagnostic category, age and sex.

We have compared our total hospital drug expenditure with estimates provided by Statistics Canada.¹ After adjusting Statistics Canada's estimates for underreporting, our estimates were almost identical to theirs. Thus, we are confident that we have represented the majority of hospital drug expenditures.

Cost estimates for some diagnostic categories (e.g. cancer) may be underestimated since the data sources we used to categorize drugs by diagnostic category excluded the activities of specialized health centres or clinics outside of a hospital. As much as 15% of the cancer drug market could be missing.² However, the cost of drugs for these patients is represented in the total drug expenditure.

Our cost estimates exclude drugs prescribed in other institutions that are included in the component called "Expenditures for Care in Other Institutions."

Physician Care Expenditures

To categorize the national physician care expenditure by diagnostic category, age and sex, we have compared the distribution of medical service costs in seven provinces by diagnostic category and sex. At the time of this comparison, most provinces could only provide 1991/92 data. We assumed that the distribution by diagnostic category for each province remained fairly constant between 1991/92 and 1993/94. The National Physician Database (NPDB)³ provides a distribution of fee-for-service physician expenditures for all provinces except Quebec. Although the NPDB could not provide data to verify the distribution by diagnostic category, it verified that the distribution by sex remained fairly constant between 1990/91 and 1992/93.

The seven provincial distributions are comparable for all diagnostic categories except ill-defined conditions and well-patient care (see Appendix 3). Provincial officials have suggested that disparities in these two categories can be attributed to the use of a limited set of diagnostic codes and over-classification of many diagnoses to these categories that capture poorly-defined diagnoses.⁴ As in our 1986 cost-of-illness estimates, we apply Manitoba's 1993/94 distribution of physician expenditures by diagnostic category, age and sex to the total national physician expenditure; Manitoba's data have been validated, whereas officials in other provinces lacked confidence in the accuracy of their diagnostic coding.

Unfortunately, we were unable to obtain physician care expenditures by diagnostic category from Ontario, Quebec, New Brunswick and the territories. These regions represent 67.5% of the total national physician care expenditure;⁵ distribution of physician care costs by diagnostic category in these areas has an important impact on the overall distribution of physician care costs in Canada. We assume physician care cost distributions for the missing provinces and territories to be similar to those of the other seven provinces.

Although we were unable to verify that the distribution of physician expenditure by diagnostic category for Ontario, Quebec and New Brunswick was similar to that of the other provinces, the NPDB did confirm that the sex distribution for Ontario and New Brunswick was similar to that of the other provinces for 1990/91 and 1992/93.³

Hospital Care Expenditures

The national hospital expenditure by diagnostic category, age and sex is estimated separately for acute care, long-term care and psychiatric hospitals to account for the impact of differences in patient composition, duration of stay and intensity of care on costs within these types of hospitals.

Acute care hospitals represent 86.6% of the total hospital expenditure.¹ We were able to account for the duration of stay and intensity of care for the inpatient "acute care" portion of the acute care hospitals using total Resource Intensity Weights (RIWs). General and other short-stay hospitals (i.e. acute care hospitals) also include long-term care bed, outpatient and emergency ward costs that are not considered in the calculation of total RIWs. These costs are assumed to have diagnostic category and age/sex distributions similar to those for acute care.

In the current report, we estimate the average cost of hospitalization per patient day for extended care and rehabilitation hospitals to represent patients in long-term care hospitals. Although we are unable to identify the intensity of care in these hospitals, we know that long-term care hospitals make up only 6.9% of the total hospital expenditure,¹ and the intensity of care among patients in these hospitals is likely to be more homogeneous than for acute care patients.

The total psychiatric hospital expenditure is assigned to mental disorders.

Expenditures for Care in Other Institutions

We use the term "other institutions" to refer to residential care facilities providing Type II care or higher (see definitions in Appendix 2). Although we are unable to estimate costs by diagnostic category, age and sex, we have grouped these institutions by principal characteristic of the predominant group of residents. The majority of the expenditure for care in other institutions is for homes for the aged. However, attributing costs to the category "aged" obscures the fact that, for a very high proportion of residents of these institutions, it is a health problem (e.g. cognitive impairment, musculoskeletal diseases), and not age alone, that requires professional health care. Approximately 12% of the expenditure for care in other institutions was spent on treating mental disorders.

Health Science Research Expenditures

Our health science research expenditure represents the vast majority of health science research conducted in Canada ($752 million). We have categorized approximately 70% of this expenditure. Research categories have been expanded to include basic research (e.g. metabolism, immunology) and other non-disease areas of health science research (e.g. medical history, equipment grants), providing a more accurate estimate of the distribution of research by diagnostic category. In our previous publication, basic research was assigned to endocrine and related disorders and to blood diseases, overestimating the cost of research in these diagnostic categories.

The total research expenditure includes a combination of 1992/93, 1993, and 1993/94 grants (see Health Science Research Expenditures); however, the majority of grants were for 1993/94. We assumed the type and value of research to be similar for the two years and reported all grants and awards as if for 1993/94.

Additional Direct Health Expenditures

Unfortunately, we are unable to categorize care by other professionals, capital and some other expenditures by diagnostic category. Many of these costs are either system costs that should not be attributed to specific diagnostic groups (i.e. public health, prepayment administration) or areas that could be allocated at aggregate levels (i.e. dentistry and denturists, $4.7 billion, to digestive diseases; eye care and hearing aids, $1.4 billion, to nervous system and sense organ diseases).

We did not include these additional disease costs in our Summary of Results because information about them is not detailed enough to justify allocating the total figure to a diagnostic category. For example, a large portion of dentistry might be considered well-patient care. To allocate the total cost to a diagnostic category could misrepresent the proportion of total costs for that diagnostic category; for instance, digestive diseases would rank seventh overall largely because of dental costs.

Nevertheless, it is important to focus more attention on these additional health expenditures since they represent a considerable portion of the total direct costs.

Mortality Costs

Our calculation of mortality costs (i.e. the present value of lost productivity due to premature mortality) now considers the loss of labour force and unpaid work resulting from premature mortality.

In our 1986 estimates, we calculated productivity loss using the average annual income of individuals. Income overestimated productivity losses because it included many sources of revenue (e.g. interest from stocks and bonds, pensions and benefits) in addition to earnings from employment. Loss of productivity would not affect these additional income sources: only employment income would be affected.

We add the replacement value of unpaid work ⁶ to earnings from employment because employment income alone underestimates the loss resulting from illness, especially for the women, children and elderly persons who are not in the work force.^7,8 Although we have reduced disparities between some groups (i.e. men and women, middle-aged and elderly), mortality costs for children are still underestimated since earnings from employment and the value of unpaid work for children under age 15 are not available. Attention to measuring the value of lost productivity for this age group is necessary in subsequent work.

The arithmetic sum of lifetime earnings and unpaid work overestimates the present value of an individual since the future value of capital is less than its present worth (i.e. capital can be invested and increases in value over time).⁹ We express the value of future earnings from employment and the value of unpaid work in terms of equivalent present-day dollars. This is accomplished by discounting by an assumed rate of return on investments for the time period (i.e. 6% discount rate). Because investment rates are difficult to predict, we provide a sensitivity analysis comparing our estimates at a 6% discount rate to a range of estimates using discount rates between 2% and 10%.¹⁰

Morbidity Costs Due to Long-term and Short-term Disability

Our calculations of lost productivity due to long-term and short-term disability are significantly higher than our 1986 estimates. This reflects the use of the 1994 National Population Health Survey (NPHS), which provides a more recent prevalence of disability than the 1978/79 Canada Health Survey used previously. The NPHS also provides information about long-term disability among the institutionalized population; only psychiatric hospitals were included in the cost of long-term disability in the 1986 estimates.

However, surveys that provide estimates of the number of "self-reported" people affected by illnesses are hard to validate, and it is even more difficult to determine the degree of disability involved. For example, survey data tend to underestimate the extent of disability due to mental illness and its severity. This could result in a significant underestimation of the indirect costs for some diagnostic categories.

The NPHS uses V-codes developed by Statistics Canada to code musculoskeletal diseases, which include injuries. Although we have attempted to solve this problem, long-term disability costs for injuries may be underestimated.

NPHS data on short-term disability are not available by diagnostic category; the distribution of the Quebec Health and Social Survey is used to classify this NPHS data by diagnostic category.

As in estimating premature mortality, we use earnings (adjusted for wage supplements) and the replacement value of unpaid work to measure lost productivity. We have developed different weights to adjust for "loss of productivity" at different levels of disability. This is an improvement from our first report, where the use of "loss of health" indexes may not reflect "loss of productivity."^11-14 We now provide a sensitivity analysis, comparing morbidity costs by using a range of weights for future productivity losses at different levels of disability.

Transfer Payments

In our 1986 estimates, we considered pensions and benefits as direct costs and included them in the total cost of illness to society. This led to double counting of costs by almost $7 million since we counted both the individual's lost productivity and reallocation of resources used to compensate the loss. In this report, we exclude transfer payments from our total cost estimate. The cost of administration for these transfer payments is a legitimate cost;^15-20 unfortunately, we are unable to obtain this information at present.

Other Costs

Inevitably there are other costs we have not included in our cost-of-illness estimates. Direct costs should include, among other considerations, patient out-of-pocket expenses (i.e. transportation costs to health providers; the cost of relocating or altering property to adapt to a patient's needs; special diets, clothing or equipment for rehabilitation or comfort; and education and counselling of patients and caregivers),⁷ social services (i.e. community programs/services, corrections/legal system costs) and service provider training. Indirect costs of illness might well include foregone employment opportunities for family members and the cost of reduced working hours or leisure time used in providing for a sick or disabled family member.¹⁴ Although we have been unable to account for such costs, they should be considered in subsequent work.

Comorbidity

We have attributed direct and indirect costs to a single diagnostic category even though they may represent individuals with multiple diseases or conditions.²¹ Costs have been apportioned to the primary cause of illness based on the following list.

Cost estimates may thus present a one-dimensional picture of illness that fails to address the costs of secondary causes.

Health care databases (i.e. mortality, hospitalization) generally do not provide information about multiple contributing or associated secondary causes. Suicide, for example, is coded as an injury even though physical or mental illness may be a contributing factor.

Similarly, our morbidity costs do not reflect short-term limitations for people with long-term disabilities. Although this avoids overestimation from double counting, our cost values may underestimate certain diagnostic categories that are often secondary causes of illness. For example, diabetes, arthritis and mental disorders are often secondary causes of illness, whereas cancer is more likely to be recorded as a primary cause of illness or death.

[Previous] [Next][Table of Contents]