The relative risks and etiologic fractions of different causes of death and disease attributable to alcohol, tobacco and illicit drug use in Canada

Eric Single, Jürgen Rehm, Lynda Robson, Minh Van Truong

CMAJ 2000;162:1669-75

• Abstract
• Introduction
• Methods
• Results
• Interpretation
• References

Background: In 1996 the number of deaths and admissions to hospital in Canada that could be attributed to the use of alcohol, tobacco and illicit drugs were estimated from 1992 data. In this paper we update these estimates to the year 1995.

Methods: On the basis of pooled estimates of relative risk, etiologic fractions were calculated by age, sex and province for 90 causes of disease or death attributable to alcohol, tobacco or illicit drugs; the etiologic fractions were then applied to national mortality and morbidity data for 1995 to estimate the number of deaths and admissions to hospital attributable to substance abuse.

Results: In 1995, 6507 deaths and 82 014 admissions to hospital were attributed to alcohol, 34 728 deaths and 194 072 admissions to hospital were attributed to tobacco, and 805 deaths and 6940 admissions to hospital were due to illicit drugs.

Interpretation: The use and misuse of alcohol, tobacco and illicit drugs accounted for 20.0% of deaths, 22.2% of years of potential life lost and 9.4% of admissions to hospital in Canada in 1995.

[Contents]

There have been relatively few attempts to estimate mortality and morbidity attributable to substance abuse in Canada.³ Before 1996, estimates were generally based on expert opinion concerning the proportion of different major causes of disease that could be reasonably attributed to the use of alcohol, tobacco or illicit drugs.⁴ Such estimates grouped causes of disease into major categories (e.g., all cancers) and failed to control for age or sex.

In 1996 a major study estimating the economic costs attributable to substance abuse in Canada (1992 data) was reported.^2,5,6 As part of this analysis, the numbers of deaths and admissions to hospital that could be attributed to the use of alcohol, tobacco and illicit drugs were estimated. These estimates, based on the methods described below, considered specific International Classification of Disease (ICD-9) categories⁷ and controlled for age, sex and province. This paper updates the estimates for 1992 to the year 1995 and presents information on the relative risks of alcohol, tobacco and illicit drug use and their etiologic fractions.

Following the lead of Collins and Lapsley,⁸ the term "substance abuse" here encompasses any use that involves a social cost additional to the resource costs of the provision of that drug. Therefore, the consequences of "abuse" are not limited to those associated with heavy use or with dependence or abuse as defined by international classifications such as ICD-9 or the Diagnostic and Statistical Manual of Mental Disorders,⁹ and they include mortality and morbidity associated with moderate use if such use incurs social costs to the community.

[Contents]

Potential causes of death and disease associated with substance abuse have been identified in reviews and meta-analyses of large-scale epidemiological studies for alcohol,^1,10,11 tobacco^1,11,12 and illicit drugs.^1,11,13 Causes include both chronic conditions and acute consequences such as drug overdose and trauma resulting from substance-related accidents. In this study, there were essentially three ways in which the use of alcohol, tobacco or illicit drugs may be implicated as a cause of death or disease. First, we included conditions that are by definition directly related to substance abuse (e.g., alcoholic psychosis or drug dependence). All reported cases of death or admission to hospital from such causes were considered attributable to substance abuse (i.e., etiologic fraction 1.0). Second, for some conditions in which substance abuse is a contributory but not a necessary cause, such as suicide or trauma resulting from impaired driving accidents, etiologic fractions were determined directly from case series (both US^{10,11,12,13,14} and Canadian^1,15,16,17 sources). Finally, for most chronic disease conditions in which substance abuse is a contributory cause, estimates of the relative risk of particular disorders for alcohol, tobacco or illicit drug use were combined with prevalence data on the number of people consuming the substance to derive the etiologic fraction, according to the following formula:

Etiologic fraction = [P₀ + P₁(RR₁) ­ 1]/[P₀ + P₁(RR₁)]

where P₀ and P₁ are the prevalence rates for non-users and users respectively, and RR₁ is the relative risk for users relative to non-users. Where the etiologic fraction takes different levels of use into account, as in the case of alcohol consumption, an appropriate alternative computation was used (see formula 1.9 in English and associates¹).

It is important to note that the etiologic fractions are based on pooled estimates of relative risk rather than on single studies. For some causes of death and disease, where there were sufficient data, separate estimates of the relative risks of mortality and morbidity were calculated. The etiologic fractions represent gross rather than net estimates of the proportion of deaths and admissions to hospital caused by substance abuse. Where it has been established that the use of a psychoactive substance actually prevents rather than causes certain disorders (e.g., the protective effect of low-level alcohol consumption against coronary heart disease), the numbers of deaths and admissions to hospital prevented were calculated; these data have been previously reported.⁵ However, the cases prevented were not subtracted from the numbers of deaths and admissions to hospital caused by the use of these substances, because our primary aim was to estimate total mortality and morbidity caused by substance abuse. Hence the etiologic fractions reported in Table 1, Table 2 and Table 3 refer only to events caused by, and do not adjust for events prevented by, substance abuse.

Prevalence data for the 1992 estimates of etiologic fractions for alcohol were based on a linear interpolation of findings from the 1990 Health Promotion Survey¹⁸ and the 1994 Canada's Alcohol and Other Drugs Survey (CADS),¹⁹ adjusted to correspond to the normal quantity and frequency measures used in epidemiologic studies of relative risk.²⁰ Prevalence data for the 1995 estimates were estimated from the mean of the 1994/95 and the 1996/97 National Population Health Surveys.²¹ Abstinence, low, hazardous and harmful drinking were defined respectively as 2.5 or less, 2.6–40, 41–60, and 61 or more grams of ethanol per day for men and 2.5 or less, 2.6–20, 21–40, and 41 or more grams per day for women.¹ This scale corresponds most closely with the manner in which relative risks are reported in the epidemiologic studies in the meta-analyses. Prevalence estimates for drinking among pregnant women for both the 1992 and 1995 estimates were taken from the 1990 Ontario Health Survey,²² and the proportion of pregnant women who smoke was estimated from the Survey on Smoking in Canada, 1994.²³ For both the 1992 and 1995 estimates, the prevalences of opiate and cocaine use were taken from the 1994 Canada's Alcohol and Other Drugs Survey.¹⁹

In the study estimating mortality and morbidity attributable to substance abuse in 1992,^2,5 estimates of relative risks and etiologic fractions were calculated, and the etiologic fractions were then applied to the reported number of deaths and hospital discharges for each cause of death or disease by age, sex and province to estimate mortality and morbidity attributable to alcohol, tobacco and illicit drugs in Canada. These estimates have been updated to 1995, using the same relative risk estimates but updating the prevalence estimates and applying the resulting etiologic fractions to 1995 data on deaths and admissions to hospital in that year. In addition, the etiologic fraction for motor vehicle accidents was updated on the basis of a more recent study.²⁴

[Contents]

The relative risks associated with alcohol and tobacco use for selected causes of death and disease are presented Tables 1 and 2; the etiologic fractions indicating the proportion of cases attributable to substance misuse are shown in Tables 1–3 for alcohol, tobacco and illicit drug use. The etiologic fractions given in Tables 1–3 are for 1992 but they are very similar to those for 1995 because they are based on the same relative risk estimates and there were no major changes in prevalence over the years. The table does not include causes fully attributable to substance use (i.e., alcohol or drug dependence categories), for which the etiologic fraction is 1.0.

Mortality and morbidity attributable to substance abuse in 1992, based on these relative risks and etiologic fractions, have been reported elsewhere.^2,5 Table 4 presents the number of deaths, the potential years of life lost (the difference between age of death and life expectancy, with age and sex taken into account), the number of hospital separations and the length of hospital stay attributable to alcohol, tobacco and illicit drug use, updated to 1995.

Alcohol

We estimated that 6507 Canadians died in 1995 because of alcohol consumption. The largest number of alcohol-related deaths stemmed from impaired-driving accidents. We estimated that 787 men and 357 women died in motor vehicle accidents caused by alcohol impairment. Alcoholic liver cirrhosis accounted for 1037 deaths, and there were 955 alcohol-related suicides.

The findings regarding years of potential life lost indicate that many of these deaths involved relatively young people. Because of the high number of alcohol-related accidental deaths and suicides, the total potential years of life lost is relatively high, at 172 126. This represents 26.4 years of potential life lost per alcohol-related death. Motor vehicle accidents accounted for 17.6% of all alcohol-related deaths and 27.2% of potential years of life lost, data that indicate the relatively young age of people killed in alcohol-related traffic accidents.

We estimated that 82 014 Canadians were admitted to hospital because of alcohol misuse in 1995. The greatest number of separations attributed to alcohol were for alcoholic psychosis, alcohol dependence syndrome or alcohol abuse (19 744 in total), accidents other than motor vehicle accidents (19 412) and motor vehicle accidents (9591). Although accidents other than motor vehicle accidents accounted for only 12.2% of alcohol-related deaths, they accounted for 23.7% of admissions to hospital attributed to alcohol. These figures include victims who were killed or injured as a result of others' intoxication. In contrast, motor vehicle accidents accounted for 17.6% of alcohol-attributed deaths but only 11.7% of admissions due to alcohol.

These estimates of alcohol-attributable mortality and morbidity represent 3.1% of total mortality, 5.4% of total years of potential life lost and 2.7% of all admissions to hospital for any cause in Canada for 1995.

It should be noted that the 1992 estimates included an estimate of the number of deaths and admissions to hospital prevented by low-level alcohol use.⁵ In 1992 alcohol prevented more deaths (7401) than were caused by alcohol misuse (6701). However, alcohol-related mortality frequently involves relatively young people, whereas the benefits apply mainly to older adults; therefore, the number of potential years of life lost due to alcohol was much greater than the potential years of life saved by alcohol use (186 257 v. 88 656). Furthermore, the number of admissions to hospital averted by alcohol use (45 414) was much lower than the number caused by alcohol (86 076).

Tobacco

The number of tobacco-related deaths in Canada was estimated at 34 728 for 1995. Smoking-related lung cancer accounted for the largest number of deaths (12 151), which represented 35.0% of all deaths attributed to tobacco use. Tobacco-related chronic obstructive pulmonary disease (COPD) accounted for 6671 deaths and ischemic heart disease for 6542 deaths. More than two-thirds of those who died from tobacco-related causes in Canada were men. An estimated 194 072 admissions to hospital were due to tobacco use. The largest number of smoking-related admissions were for ischemic heart disease (52 297), COPD (37 506) and lung cancer (21 212). Tobacco-attributed mortality and morbidity accounted for 16.5% of total mortality, 15.7% of total potential years of life lost and 6.5% of all admissions to hospital for any cause in Canada in 1995.

Illicit drugs

The number of deaths related to illicit drug use in 1995 was estimated at 805, which represents 0.4% of all deaths. Most of these deaths (691 or 85.8%) affected males. Suicide accounted for 329 (40.9%) of deaths related to illicit drug use; opiate and cocaine poisoning accounted for 160 (19.9%) and 78 (9.7%) respectively. Death from AIDS acquired through illicit drug use has been increasing, accounting for 61 (8.3%) of the 762 deaths related to illicit drug use in 1992 and for 83 (10.3%) of the 805 deaths related to illicit drug use in 1995.² Although deaths caused by illicit drug use were less common than deaths attributable to alcohol and tobacco use, the people who died were younger. The 805 deaths resulted in 33 662 potential years of life lost or 41.8 years per death and 1.1% of total years of life lost through any cause in 1995. In addition, 6940 admissions to hospital (0.2% of all admissions) were attributable to illicit drug use in 1995. Drug psychosis (1777), cocaine abuse (980) and assaults (975) were the most common causes of admission related to illicit drug use.

[Contents]

These estimates of mortality and morbidity attributable to substance abuse are generally lower on a per capita basis than estimates for the United States^13,25 and those obtained in prior Canadian studies.^4,26 In particular, the estimate of tobacco-related deaths for 1995 is 15% lower than a Canadian estimate for 1991.²⁶ Although part of the discrepancy in these estimates may be due to different reference years, most of the difference is probably due to differences in the methods used to estimate relative risk. Whereas the 1991 report relied on a single study for estimates of relative risk,²⁷ we used pooled estimates of relative risk from several studies. For example, the prior study estimated that the relative risk of lung cancer for male smokers relative to those who had never smoked was 22, but the corresponding estimate in our study (based on pooled estimates from 11 studies, including the one used for the 1991 estimate) was only 13. Pooled estimates of relative risk for COPD and ischemic heart disease in our study were similarly lower.

Our estimates of alcohol-attributed mortality and morbidity were also lower than prior estimates. The number of deaths attributed to alcohol in 1995 (6507) was much lower than the estimate for 1991 (19 163), which was based on expert opinion.²⁸ This and similar differences for morbidity are largely due to several differences in methodology. For the current estimates, we were able to exploit much more precise estimates of relative risk from a growing body of epidemiologic studies,¹ many of which were not available for earlier estimates. For the estimates presented here, we used specific rather than broad disease categories. For example, prior estimates applied a single etiologic fraction to all forms of cancer, including many types for which there is no evidence of a causal connection to alcohol use, but for this paper the estimates were based on separate calculations of relative risks and etiologic fractions for each type of cancer. Finally, we controlled for age, sex and province, whereas prior estimates did not take these variables into account. For example, the earlier estimates attributed 10% of all cancers to alcohol use. Thus, 1 in 10 deaths from childhood leukemia were attributed to alcohol use, despite the fact that they involved nondrinkers (i.e., children) and the fact that there is insufficient evidence to conclude that any alcohol consumption is causally connected to leukemia. Using more precise methods, we found that only 2% of all cancer deaths are attributable to alcohol use.

Not only are our estimates of alcohol-attributed mortality and morbidity lower than prior estimates, but they also indicate that the relative contribution of accidents to overall alcohol-related mortality and morbidity is much greater than previously thought.²⁹ In the prior estimates, most (62%) alcohol-attributed deaths were accounted for by three broad chronic disease categories — diseases of the respiratory system, diseases of the circulatory system and cancer. The new estimates indicate that less than one-fifth of alcohol-related deaths in 1995 (1207/6507 or 18.5%) were due to these chronic conditions. In contrast, accidents and other acute causes accounted for nearly half (3064/6507 or 47.1%) of all alcohol-attributed deaths and well over half (107 554/172 126 or 62.5%) of potential years of life lost. These findings have important implications for alcohol policy and programming.^30,31

Although the estimates are lower than in prior studies, they nevertheless indicate that alcohol, tobacco and illicit drug use represent a major source of death and illness in Canada. In 1995 substance abuse accounted for 20.0% of total deaths, 22.2% of total potential years of life lost and 9.4% of total admissions to hospital for any cause.

Competing interests: None declared for Drs. Rehm, Robson and Truong. Dr. Single has accepted travel funds to attend meetings of the alcohol industry-funded Alcohol Issues Panel of the Brewers Association of Canada and the International Center for Alcohol Policies.

Send a letter to the editor Envoyez une lettre à la rédaction

[Contents]

Reprint requests to: Dr. Eric Single, Single & Associates Research Consulting Ltd., 6 Mervyn Ave., Etobicoke ON M9B 1M6; e.single@utoronto.ca

1. English D, Holman D, Milne E, Winter M, Hulse C, Codde G, et al. The quantification of drug caused morbidity and mortality in Australia, 1992. Canberra: Commonwealth Department of Human Services and Health; 1995.
2. Single E, Robson L, Xie X, Rehm J. The costs of substance abuse in Canada. Ottawa: Canadian Centre on Substance Abuse; 1996.
3. Robson L, Single E. Literature review of studies on the economic costs of substance abuse. Ottawa: Canadian Centre on Substance Abuse; 1995.
4. Adrian M. Social costs of alcohol. Can J Public Health 1988;79:316-22. [MEDLINE]
5. Single E, Robson L, Rehm J, Xie X. Morbidity and mortality attributable to substance abuse in Canada. Am J Public Health 1999;89:385-90. [MEDLINE]
6. Single E, Robson L, Xie X, Rehm J. The economic costs of alcohol, tobacco and illicit drugs in Canada, 1992. Addiction 1998;93:983-98.
7. International classification of diseases, 9th revision (clinical modification). 3rd ed. Washington: US Department of Health and Human Services; 1998. Publ no.: (PHS) 89-1260-1.
8. Collins D, Lapsley H. Estimating the economic costs of drug abuse in Australia. Canberra: Commonwealth of Australia National Campaign Against Drug Abuse; 1991. Monograph no. 15.
9. American Psychiatric Association. Diagnostic and statistical manual of mental disorders (DSM-IV). 4th rev. Washington: The Association; 1994.
10. Shultz J, Rice D, Parker D, Goodman R, Stroh G, Chalmers N. Quantifying the disease impact of alcohol with ARDI software. Public Health Rep 1991;106:443-50. [MEDLINE]
11. Fox K, Merrill J, Chang H, Califano J. Estimating the costs of substance abuse to the Medicaid hospital care program. Am J Public Health 1995;85:48-54. [MEDLINE]
12. Shultz J, Novotny T, Rice D. Quantifying the disease impact of cigarette smoking with SAMMEC II software. Public Health Rep 1991;106:326-33. [MEDLINE]
13. Rice D, Kelman S, Miller L, Dunmeyer S. The economic cost of alcohol and drug abuse and mental illness 1985. DHEW publ no.: (ADM) 90-1694. Rockville (MD): Department of Health and Human Services; 1990.
14. Blincoe T, Faigin B. Economic impact of motor vehicle crashes — United States 1990. MMWR 1993;42:443-8.
15. Traffic Injury Research Foundation. Motor vehicle accidents and alcohol. Ottawa: The Foundation; 1996.
16. Association of Canadian Fire Marshalls and Fire Commissioners. Annual report: fire losses in Canada, 1992. Ottawa: The Association; 1996.
17. Rehm J, Ialomiteaunu D, Walsh G, Adlaf E, Single E. The quantification of morality caused by illicit drugs in Canada, 1992. Toronto: Addiction Research Foundation; 1996.
18. National public health survey 1990. Ottawa: Statistics Canada; 1992. Cat no 82-567.
19. MacNeil P, Webster I. Canada's alcohol and other drug survey 1994: a discussion of the findings. Ottawa: Ministry of Public Works and Government Services Canada; 1994. Cat no H39-338/1-1994E.
20. Rehm J, Walsh G, Adlaf E, Single E. Assessment methods, prevalence of high risk drinking and harm — a sensitivity analysis. Toronto: Addiction Research Foundation; 1996.
21. National population health survey 1996–97. Ottawa: Statistics Canada; 1998. Cat no 82-567.
22. Adlaf E, Ives F, Smart R. Alcohol and other drug use among Ontario adults in 1994 and changes since 1977. Toronto: Addiction Research Foundation; 1994.
23. Survey on smoking in Canada, cycle 4. Ottawa: Health Canada; 1995.
24. Mayhew DR, Brown SW, Simpson HM. Alcohol use among drivers and pedestrians fatally injured in motor vehicle accidents: Canada, 1996. Ottawa: Traffic Injury Research Foundation; 1996.
25. Public Health Practices Program Office. Cigarette smoking — attributed mortality and years of life lost — United States, 1990. MMWR 1990;42:645-9.
26. Makamaski-Illing E, Kaiserman M. Mortality attributable to tobacco use in Canada and its regions, 1991. Can J Public Health 1995;86:257-65. [MEDLINE]
27. US Department of Health and Human Services. Reducing the health consequences of smoking: 25 years of progress. A report of the Surgeon General. Washington: Public Health Service; 1989. DHHS Publ no.: (CDC) 89-8411.
28. Williams R, Single E, McKenzie D. Canadian profile: alcohol, tobacco and other drugs 1995. Ottawa: Canadian Centre on Substance Abuse; 1995.
29. Robson LS, Single E, Xie X, Rehm J. The cost of alcohol-attributable injuries and poisonings in Canada, 1992. Contemp Drug Probl 1998;25:421-39.
30. Rehm J. Draining the ocean to prevent shark attacks? Nord Stud Alcohol Drugs 1999;16:46-54.
31. Single E. The concept of harm reduction and its application to alcohol: the 6th annual Dorothy Black lecture. Drugs Educ Prevent Pol 1997;4:7-22.

© 2000 Canadian Medical Association or its licensors