General population surveys — an overview of the methods and definitions used
Prevalence and patterns of drug use in the general population (adults or school children) is assessed through surveys based on representative probabilistic samples of the whole population under study. Interview surveys are based on self-report of participants regarding present and past behaviours, personal characteristics, knowledge and attitudes regarding drugs or other health topics.
Surveys allow a direct estimation of prevalence of drug use, as a proportion of the population that declare having used specific drugs. Also, they facilitate estimating patterns of use and other factors considered as potential determinants or consequences of drug use for each individual under study. The factors that are investigated retrospectively have the limitations of self-report (concealment) and memory biases on recall of past events.
The prevalence estimates of surveys are based on standard periods of time. For illegal drugs, the more usual measures are:
- any use during the person’s life (or lifetime prevalence), also called ‘lifetime experience’,
- any use during the previous year (or last 12 months prevalence), also called ‘recent use’,
- any use during the previous month (or last 30 days prevalence), also called ‘current use’.
‘Lifetime experience’ alone will not capture the current drug situation among adults (although it is considered useful among school children) as it also includes people that tried drugs a long time ago. On the other hand, it is a framework measure; it can give a first rough estimation of the extent of drug experience for low prevalence drugs. Lifetime experience measures can help to estimate patterns of use such as incidence, length of drug use, or continuation or discontinuation of use. Last year prevalence (‘recent use’) produces lower figures, but better reflects the more recent situation, although often use could be occasional. Combination of lifetime experience and recent use can give basic information on drug use patterns (e.g. as discontinuation of use — people that have used in the past, have not used in the last year — or continuation of use — those that have used in the past, have also used in the last year — when incidence of use is taken into account, although after a certain age, incidence of drug use is usually very low).
Last month prevalence (‘current use’) gives some indication of more regular use and will include the more intensive users, although in fact many current users will not be intensive users.
Estimates of ‘recent’ or ‘current’ use for the whole population (15–64 years) are usually low in many countries, except in the case of cannabis Epidemiological and policy perspectives focus the analysis on young people (15–24 or 15–34 years) particularly among males, where drug use concentrates. Analysis may also consider geographical location and social or behavioural characteristics.
Samples sizes should be large enough to enable specific analysis with a minimum level of precision. In particular, when policy evaluation requires reliable estimations broken down by gender, age groups or regions, or when is necessary to increase the reliability of estimates for substances with low prevalence. To increase the sample size, certain sampling strategies may help to improve estimations among groups of particular interest; for instance over-sampling of young people, ethnic minorities, or inner city areas.
It should be acknowledged that there are limitations of surveys in estimating prevalence of the more marginalised forms of drug use (e.g. heroin injection, crack use) due to the low prevalence figures, but in particular due to non probabilistic errors (exclusion from the sampling frame, absence in household, non-response). On the other hand, surveys can identify a small but substantial number of people that used heroin in the past, but not at present. The characteristics of these users may provide insights into the reasons for their discontinuation of use, compared to long-term users identified in treatment centers.
Many countries collect information on ‘age of first use’ of drugs, which allows analysis of incidence. Also, intensity of use can be assessed to help identify higher risk groups. Age of first use and frequency of use are included in the EMCDDA guidelines (European Model Questionnaire — EMQ).
The concept of ‘intensive users’ is used most often in relation to cannabis use. It is a broad term meaning use of a drug (e.g. cannabis) that exceeds a certain threshold of frequency. It does not necessarily imply the existence of ‘dependence’ or other problems, but it is considered to increase the possibility of negative consequences, including dependence. In EMCDDA reports, ‘daily use’ is an indicator of intensive use of cannabis (in operative terms it is the use in 20 days or more, in the last 30 days). This frequency has been used in different studies, and can be derived from the European Model Questionnaire (EMQ). For other substances (e.g. cocaine) it may be necessary to define different thresholds of frequency as an indicator of ‘intensive use’.
Although frequency measures are relatively simple and comparable, there is ongoing work in several countries and at the EMCDDA, to further assess the presence of problematic use or dependence through specific brief scales included in surveys. Similar work is being developed in other countries (Australia, Canada, and the USA). The EMCDDA is coordinating collaboration among EU experts working in this area. Two short scales, the Cannabis abuse screening test (CAST) and the Severity of dependence scale (SDS), are now being validated against DSM-IV criteria in seven European countries in a cross-national project seeking European standards. The project is expected to result in European guidelines on data collection and reporting of prevalence of cannabis use disorders among the general population. In 2012, in the PDU indicator revision, a new category was added to the broadened PDU, now called High risk drug use: Frequent and high risk cannabis use. Frequent use is going to be measured as daily or almost daily use (as mentioned in the text, 20+ days of use in the 30 days preceding survey) and high risk cannabis use is going to be estimated using the short screening scales, in particular CAST.
In the case of adults, but even more so in the case of children, the age ranges used to report results might have an influence in the results of prevalence estimates and comparisons should be based on the same age groups. For adults, the EMCDDA recommends range 15–64 years for the whole adult population and 15–34 years for young adults. If wider age groups are used (e.g. 12 to 75 years) prevalence estimates tend to be lower, because illegal drug use is relatively low at higher ages. If narrower groups are used (e.g. 18 to 49), estimates will tend to be higher because drug use concentrates among young adults. For children, see below.
Information provided by surveys is particularly useful when they are repeated at regular intervals, using the same questionnaires and methodology (a survey series). These allow tracking of trends over time that cannot be identified by a single survey or two consecutive surveys without further continuation. A survey series requires a long-term commitment from public institutions and research institutions.
Most Member States have conducted representative national surveys for adults during recent years, although in some cases sample sizes are limited. Several countries have recently conducted their first national drug surveys, in all cases with high compatibility with the EMQ.
Information on trends in prevalence of use, and on the characteristics of users are particularly valuable. Several countries have established series of national surveys, or have started them in recent years. For the time being, only a limited number of countries have long term series with big sample sizes. As these series continue and new ones become consolidated, the possibility of interpreting trends will increase. This will help to plan and evaluate drug policies.
Differences across countries remain in survey context, data collection methods and sampling procedures. In addition to methodological questions, several factors can contribute to differences in overall national figures. The relative proportion of urban and rural population in each country may explain, in part, some overall national figures. Also, national figures may be explained in part by generational factors, including the different rates of convergence between the lifestyles of young males and females. Social context can influence self-reporting of drug use. Comparative analysis across countries should be made with caution, in particular where differences are small, and formulation and evaluation of drugs policy should take into careful consideration age groups, birth cohorts, gender and urbanisation and other relevant factors.
In the case of adult surveys, the EMCDDA has developed guidelines to improve comparability of population surveys in the EU. These guidelines include a set of common core items (‘European model questionnaire — EMQ’) and basic methodological recommendations. The set of items can be used to report data from existing surveys, or can be inserted into broader questionnaires. The set includes basic prevalence measures and use patterns of certain illegal and legal substances, basic socio-demographic characteristics and opinion and risk perception questions. Questions about drug policies are considered optional. The guidelines have been compiled and are available in the EMCDDA Handbook.
Regarding surveys among school children data see the section on ‘Methodology of school surveys’ in the Statistical bulletin.
Finally, in 2002 the EMCDDA developed a prototype of an EU Databank on Population Surveys on Drugs. This databank collated, on a voluntary basis, databases from existing national surveys already analysed and exploited at national level (from Germany, Greece, Spain and the United Kingdom). The aim was to demonstrate the feasibility and added value of further analysis at EU level, after national analysis and use of the data has been completed (a period of about three to five years). The national databases were harmonised following the European Model Questionnaire (ex-post harmonisation), including only the common set of core items. In 2009 a voluntary EMQ module for monitoring perceived availability of drugs was developed and made available. Also in 2009 a project was launched to promote collaborative analysis among groups of countries willing to harmonise a common dataset in order to work with the EMCDDA on common analysis (e.g. data analysis workshops) and this is continuing with a total of fourteen countries in 2013.
Calculating estimates of illicit drug use prevalence levels and numbers of users
Estimates of the prevalence of illicit drug use, overall and by specific drug, and the resulting estimates of the numbers of drug users, by drug, country and the EU as a whole, are of sufficient interest to justify the use of general population surveys to achieve this end.
A simple method of calculating number of users of a drug by country has been adopted. For countries reporting data, the estimate of the prevalence level is applied to population figures obtained from Eurostat. This is broken down by age range: 15–64, 15–34 and 15–24; and by the time window of prevalence: lifetime, last 12 months, and last 30 days. To obtain the European prevalence rates, the numbers are summed and divided by the sum of the populations within the reporting countries. To obtain the numbers of users at the European level, the European prevalence rates are applied to the population within Europe.
Given the nature of the surveys, described above, a number of simplifying assumptions are required, which should be considered when using or quoting these estimates:
- The prevalence levels derived from a country’s general population survey are a reasonable estimate of the prevalence for the country and can be applied to the country population as a whole. Provided the survey samples are representative of the population, this is a reasonable assumption.
- The estimates of country level prevalence are comparable across countries. Given the range of methods used in the surveys, and the differing years when the surveys are completed, this is a weaker assumption. The European level estimates should therefore by treated with some caution, though in the absence of a working alternative they do provide an indication of the scale of use.
- The estimations of prevalence at country level are comparable across time, allowing the construction of trends. Provided the survey methodology remains constant over time this is a reasonable assumption. Where the EMCDDA knows a substantial change in method has occurred, the data is not used in trends.
- The estimated prevalence levels for Europe as a whole are a reasonable estimate for prevalence within the small number of countries with missing data. Missing data results from countries either not having a general population survey or not having questions on specific drugs within their general population survey. The countries with missing data generally do not represent a large proportion of the total European population. This is the case for lifetime and last year prevalence for all drugs other than Amphetamines, where the missing countries account for less than 3% of the total European population. For Amphetamines, the figure rises to just under 6%. For last month prevalence, some of the more populous countries are missing, resulting in figures ranging between 15 and 20% for all drugs other than cannabis, where the figure is less than 1%. Strong country differences may also bring this assumption into question.
Full details of the nature of the surveys are held by the countries, and a list of bibliographic references by country are to be found in the Statistical bulletin. The main characteristics of the surveys are recorded in a table entitled Survey Methods in the Statistical bulletin. A number of reports on the assumption of comparability of different country surveys are available in the General population surveys page of the Key indicator portal.