Method
Results
Discussion
Author: G. C. Reginald SMART and Paul WHITEHEAD
Pages: 49 to 55
Creation Date: 1973/01/01
Several previous studies indicated that the use of drugs among high school students described a unimodal distribution with log normal characteristics ( [ 1] , [ 2] ). This is a distribution which contains many infrequent users, fewer moderate users and very few heavy users from which persons with problems are drawn. It is a unimodal distribution in which there is no clear difference between light, moderate and heavy users. It has been shown that a similar distribution obtains for alcohol consumption in a variety of countries ( [ 3] , [ 4] ) and a later study demonstrated that high per capita consumption is associated with heavy use in eight countries ( [ 5] ). Studies of high school students in Canada reveal that total drug use in a variety of cities and at several points in time describe log normal distributions (1). A similar distribution describes the use of drugs such as marihuana, LSD, speed and tranquillizers in high school students in several cities, with very few exceptions ( [ 2] ). It has been argued from these data that programmes of prevention may have to be focused on reducing per capita consumption so that in the next generations the whole curve could be shifted towards a lower profile and hence contain fewer heavy users.
If one could find in the consumption data a clear bimodality dividing users into "normal users" and "abusers" or heavy users prevention could perhaps focus on preventing heavy use. However, the available data for drug use in Canada suggest that this distribution may rarely be found. Nevertheless, the available data concerns only Canadian high school students, albeit in five different areas and at several points in time (1968 to 1970). One purpose of this paper is to expand the data base for these theories by describing data from (i) a study of adults in Toronto in 1971 and (ii) a study of British university students.* These data provide a test of the generality of the distribution of consumption data on populations different in age and nationality from those originally employed. Another purpose is to indicate the correlation between per capita drug consumption and heavy use within high school districts in Toronto. Data from this latter study were included in the original report ( [ 6] ) but only for the school population as a whole. The Toronto area is composed of seventeen districts, some with very high and some with very low rates of drug use. The aim here is to determine whether high per capita use in a district is associated with a high frequency of heavy users or whether per capita use and heavy use are unrelated.
Toronto adults
The details of the sampling procedures employed in this study are described in the complete report (7). Only a few essentials are recorded here. The sample included
The authors are indebted to Dr. D.V. Hawks and Mrs. Adele Kosviner for making these data available.
1,200 persons 18 years and over interviewed in Toronto during 1971. It was a stratified proportionate sample planned to represent all census tracts and to contain the correct proportions according to age, sex and family size. The sample turned out to contain slightly too many females and young people according to the (1961) census data although the discrepancies were small. Participants were asked to co-operate in a study of "health problems and how people cope with them". Interviews about recent illnesses, medical treatment and the use of drugs such as alcohol, tobacco, marihuana, LSD, tranquillizers, stimulants and depressants were held in the home.
British university students
The data from British universities were obtained from 570 students at 2 colleges. The response rate at these colleges was 61 per cent and 67 per cent and sampling involved the total enrolment who were mailed questionnaires. These questionnaires inquired about the extent of use of cannabis, amphetamine, methamphetamine, barbiturates, opiates, methadone, hallucinogens, cocaine and other stimulants. The more commonly used drugs (e.g., alcohol and tobacco) were not included in this present report because the frequency of use categories were so different from the above drugs, all of which had the same frequency categories. The two colleges were both small institutions whose students may not be representative of the total population of British university students.
Toronto high school students
The sampling methods employed in the 1970 Toronto high school study have been discussed in detail in the previous report ( [ 6] ). Basically, the sample involved 120 students from each of grades 7, 9, 11 and 13 in each of 17 high school districts. Intact classes were chosen until 120 students or more were obtained; 399 classes in 94 schools were utilized. The districts were chosen at random from the total number of districts to form a one sixth sample. A high school district is an educational and geographical unit consisting of a high school and the primary schools which send students to it.
The questionnaire inquired about the use of the following drugs in the past six months: alcohol, tobacco, marihuana, glue, other solvents, barbiturates, opiates, methamphetamine, stimulants other than methamphetamine, tranquillizers, LSD and other hallucinogens.
Data from the three studies involved were not collected in a strictly comparable manner. All three studies involved large segments of large populations although all used somewhat different sampling techniques. The three studies used different questionnaires enquiring about different drugs. More important, the response categories were somewhat different. This has meant that the scoring criteria for each study are slightly different. What is being compared in the Toronto adult and British university studies is the general character of the distribution, not the rates of use of the same set of drugs. However, students in all Toronto high school districts did receive the same questionnaire and these data are directly comparable.
Toronto adults
For each user of any of the drugs a score was computed taking into account the frequency of use and the number of drugs used in the past year. The drugs were alcohol, tobacco, tranquillizers, stimulants, depressants, cannabis (marihuana or hashish) and LSD. For tobacco (cigarettes) the categories with their scores were:
None
|
0 |
Smoked occasionally
|
1 |
1-10 a day
|
2 |
11 to about 20 a day
|
3 |
More than 20 a day
|
4 |
For alcohol, tranquillizers, depressants and stimulants the categories and scores were:
None
|
0 |
Less than once a month
|
1 |
Once every 2-3 weeks or less
|
2 |
1-5 times per week
|
3 |
Nearly every day
|
4 |
For marihuana and LSD the categories referred to the past 12 months:
Never
|
0 |
1-2 times
|
1 |
3-4 times
|
2 |
5-6 times
|
3 |
7 or more times
|
4 |
The total scores could vary between 1 and 28 for 7 drugs. The frequency of persons with each score is shown in figure 1 and these data were fitted to the log normal distribution in the manner suggested by Croxton and Cowden ( [ 8] ). It can be seen from figure 1 that the distribution is unimodal but only a fair approximation to the log normal expectancy. Tests of skewness and kurtosis indicate that the curve does not fit the log normal expectancy sufficiently (p>.01). The distribution includes too few moderate users and too many light users although it can be seen that at many points it is close to the expectancy.
British university students
For British students certain drug categories were included: cannabis, amphetamines/methamphetamine, barbiturates, opium, LSD and other hallucinogens, methadone, cocaine, heroin and other opiates. For each of the categories frequency of use data were obtained as follows: once, 2-10 times, and more than 10 times. The questions in this study referred to use at any time, not just in the past year. These frequency categories were arbitrarily given the scores of 1, 2 and 3 respectively so that scores could vary between 1 and 24. Figure 2 shows the distribution to be unimodal and a fairly good fit to the log normal expectancy. Tests of skewness and kurtosis indicate a good fit to the expected log normal distribution (p.05).
Per capita consumption and heavy use in Toronto high school districts
The scores involved alcohol, tobacco, glue, marihuana, LSD, other hallucinogens, opiates, tranquillizers, stimulants and barbiturates. For each drug there were four frequency of use categories: 1-2 times, 3-4 times, 5-6 times, and 7 plus times in the past 6 months. Responses in each category were assigned to the mid-point so that the scores were: 1.5, 3.5, 5.5, and 7.5. Each user's score could vary from 1.5 to 7.5 for each drug and from 1.5 to 75.0 for the 10 drugs. The 17 districts varied enormously in the extent of drug use some being low in all types of drug use, some high, and some varying with the type of drug.
The average score for each district was determined and the proportion of students achieving scores of 30 or more. Earlier studies had shown that all heavy multiple drug users would obtain scores of 30 or more; this was arbitrarily defined as heavy use. It involved only 0 per cent to 7.6 per cent of the drug using population of each district.
It is important to note that the mean or per capita scores were computed with the 30 or more scores removed, so that there is no forced correlation between the two sets of information.
A Spearman rank order correlation coefficient was computed to show the relationship between average scores and amount of heavy use (table 1). The correlation is 0.75 which could occur by chance or random fluctuation fewer than 1 in 1,000 times,
District |
Mean score |
Rank |
Percentage scores 30 + |
Rank |
---|---|---|---|---|
1 | 7.90 | 5 | 2.99 | 10 |
2 | 7.06 | 8 | 3.27 | 8 |
3 | 7.99 | 4 | 7.6 | 11 |
4 | 9.08 | 1 | 7.4 | 12 |
5 | 6.72 | 13 | 3.6 | 17 |
6 | 6.89 | 12 | 2.67 | 11 |
7 | 6.93 | 10 | 1.62 | 13 |
8 | 6.39 | 14 | 2.11 | 12 |
9 | 6.90 | 11 | 1.57 | 14 |
10 | 6.34 | 15 | 3.06 | 9 |
11 | 6.33 | 16 | 1.40 | 16 |
12 | 3.65 | 17 | 0.00 | 17 |
13 | 7.15 | 7 | 1.46 | 15 |
14 | 8.23 | 3 | 6.78 | 3 |
15 | 8.92 | 2 | 5.49 | 4 |
16 | 6.94 | 9 | 3.69 | 6 |
17 | 7.26 | 6 | 3.96 | 5 |
Spearman Rank Order Correlation Coefficient = 0.7451, t = 4.3268, p0.001.
The data presented here show that the unimodal and log normal character of drug use distributions has some generality. The original studies ( [ 1] ) were confined to Canadian high school students surveyed in 1968, 1969 and 1970. The present study shows the same sort of unimodal distribution with selected samples of Canadian adults and British university students. However, the log normal character of the distribution does not hold for the Canadian adults. It may well be that the unimodal, continuous character of the distribution is more important for prevention than the presence or lack of log normality, since families of log normal curves can be drawn through the same mean with different ranges. It is further shown that, within a given population, where average drug consumption is high heavy drug use is also high, even where heavy users have been excluded from the computation of the average. These data suggest that in many populations there may be a correlation between heavy use and average consumption. This, together with the unimodal character of the distributions suggests that efforts to reduce heavy consumption may fail unless average drug use is also reduced.
It would be possible to overstate the generality of the propositions adduced. So far, data from a few samples from only two countries are involved and further crossnational data on drug use are certainly required. Further samples from Canada, Britain and elsewhere should also be studied. However, De Lint and Schmidt have shown that the character of the alcohol distribution is similar in Canada, Finland and the United States and France ( [ 3] ). They have also shown that a correlation between hazardous alcohol consumption and per capita consumption obtains across a wide variety of countries ( [ 5] ). Whether drug use in general will be shown to have the same generality as alcohol consumption is uncertain but becoming more probable. It would be ideal to have comparable data from a variety of countries other than Canada and Britain and from a variety of user groups in each.
The correlation between mean scores and proportions of heavy users is of interest from several viewpoints. It, too, should be extended to cross-national studies and to other types of samples as has been done for alcohol consumption. At present, it strongly suggests that where average drug consumption is high, heavy use will also be high. Since drug abusers are typically heavy multi-drug users it is strongly suggested that heavy average drug use co-relates closely with drug abuse. Naturally, further studies would be necessary to establish to the points on the distribution above which various drug use hazards are likely to occur. The equivalent of liver cirrhosis risk for those regularly consuming 15 cl. of absolute alcohol a day (5) has yet to be established for multi-drug use.
Many of us have thought for some time that some of the more well developed perspectives that stem from the research on alcohol use and alcoholism should have applicability to the study of the use of other drugs and drug abuse. Thus far, success has been limited but the convergence in the area of epidemiological research dealing with the distribution of consumption offers new hope. Parallel findings in these areas are also crucial for another reason. Thus far we have no satisfactory explanation or theory about why log normal distributions exist ( [ 9] ). In the absence of such a theory serious question is legitimately raised about the validity of suggestions made on the basis of such findings. Given the unlikelihood of such a theory being developed in the near future the next best approach is to continue working at the empirical level testing the generalizability of the model in different places, under different circumstances, and over varying periods of time. This paper has been another such attempt.
Another parallel between the alcohol and drug using research has come to our attention and deserves to be noted here. We have previously noted that between 1968 and 1970 per capita use of drugs among adolescent students in Toronto increased from an average of 6.5 to 11.2 ( [ 1] ). In spite of this sizeable increase over a relatively short period of time both distributions were log normal. Ekholm (9) recently reported a similar finding for alcohol consumption in Finland where consumption was 47 per cent higher in 1969 than it had been in 1968. Per capita consumption increased from 2.9 litres of absolute alcohol in 1968 to 4.0 litres in 1969 but in both years the distribution of consumption was log normal in character. Hence, these parallel findings further suggest the generalizability of the consumption model and the stability of the character of the distribution over time even when consumption patterns - in terms of the amount consumed - change considerably over even short periods of time. What is lacking, so far, are data showing that the distribution remains similar when average consumption falls over time.
This study together with the earlier data on distributions on drug use suggests the following:
There appears to be no clear dividing line for normal, moderate and heavy drug users in terms of the distribution of consumption;
The character of many drug use distributions appear to be log normal although there are exceptions to this finding;
Average drug use in a population is positively related to heavy drug use, even where those heavy users are excluded from the computation of the average;
It maybe impossible to reduce heavy drug use unless reductions are also made in average use. That is, many people in the population may have to use fewer drugs in order to have fewer drug abusers or heavy users in the next generation. A set of experiments are required in which average drug use in a population is decreased and corresponding changes in the proportions of heavy users are examined. It may be possible to create these reductions by a variety of legal, educational or preventive approaches although few have been reliably described as yet.
Smart, R. G., Whitehead, P., and Laforest, L. The prevention of drug abuse by young people: an argument based on the distribution of drug use. Bulletin on Narcotics , XXIII, 2, pp. 11-15.
002Smart, R.G., and Whitehead, P.C. The consumption patterns of illicit drugs and their implications for prevention of abuse. Bulletin on Narcotics , XXIV, 1, pp. 39-47.
003DeLint, J. E., and Schmidt, W. The distribution of alcohol consumption in Ontario. Quarterly Journal of Studies on Alcohol , 1968, 29, 968-973.
004Ledermann, S. Alcool, alcoolisme, alcoolisation . Données scientifiques de caractere physiologique, économique et social, lnstitut national d'études démographiques, Travaux et Documents, Cahier (No. 29) Paris, Presses universitaires de France, 1956.
005DeLint, J.E., and Schmidt, W. Consumption averages and alcoholism prevalence: a brief review of epidemiological investigations. British Journal of Addictions , 1971, 66, 97-107.
006Smart, R. G., Fejer, D., and White, J. The extent of drug use in Metropolitan Toronto schools: a study of changes from 1968 to 1970. Addiction Research Foundation, Toronto, 1970.
007Smart, R. G., and Fejer, D. Marihuana use among adults in Toronto, Addiction Research Foundation, Substudy 6-7 and JO-71, 1971.
008Croxton, F.R., and Cowden, D.J. Applied General Statistics, Englewood Cliffs, Prentice-Hall, 1955.
009Ekholm, Anfers, "The log normal distribution of blood alcohol concentrations in drivers". Quarterly J. of Studies on Alcohol , 33: 508-512, 1972.