What can surveys tell us about the nature and extent of drug use?
John Corkery, BA Hons, MSc, MPhil
Honorary Senior Research Fellow
Department of Mental Health - Addictive Behaviour
St George's Hospital Medical School
University of London
Overview
Survey data form one of the EMCDDA's key indicators. In this presentation I will outline the range of surveys that can be used at local (e.g. City), regional or national level. This will include general household surveys, school surveys (including ESPAD), surveys of arrestees, recreational drug users, driving and substance use.
Methodological issues to be covered will include data collection methods and their limitations e.g. paper questionnaires, telephone surveys, self-completed computer questionnaires; sample sizes, stratification, sampling methods, the need for booster samples (e.g. to look at minority ethnic groups). Generalisability of findings to other populations e.g. problem users, marginalised groups such as school excludees/truants, recreational users, other age groups.
Problems associated with using different survey instruments, age groups, user profiles, etc. in different regions and combining results to get a national picture.
I will present examples using data from various parts of the UK and other parts of Europe to illustrate the various points being made. Analysis will cover demographics such as age, gender and ethnicity, as well as substances used, frequency of use, association with crime, dependence, changes over time, geographical variations. I will relate these to the reporting needs outlined in existing EMCDDA guidance, and the use of check lists.
Introduction
A number of topics and associated issues have to be borne in mind or considered when either planning a new survey or examining the results of one that has already been conducted.
Population and geographical area to be surveyed
Sample frame, sample size, sampling methodology
Data collection methods and instruments
Information needed
Analysis
Time-scale(s)
Periods between survey sweeps vs. continuous data collection
Costs
Limitations
Because there is no workshop as such for this topic, real examples of some of the points made will be given to enhance comprehension. These be will be mostly drawn from my own experience in the UK over the past 10 years or more.
Sources of information
Under this heading one needs to consider
- What is the purpose(s) of the survey?
- What group(s) are we interested in and why?
- Do we want to make comparisons with other groups (nationally/internationally)?
- Is this a 'one-off' or will it be repeated?
- What sort of analysis do we want to undertake, i.e. what types of variable?
- Will it provide the information we need to comply with our EMCDDA commitments?
The answers to these questions will help to determine what information is sought, in what form and by what methods
Geographical area
Surveys can be conducted at a number of different levels:
- In unique locales e.g. pop festival or night clubs
- Local (e.g. City)
- Regional (e.g. County)
- National (e.g. Community, Country)
- International (e.g. ESPAD)
The level can be determined by the type of population surveyed e.g. excludees/homeless in a city; general population nationwide
The larger the area the more complex and costly the logistics
Sampling
Sampling frame
- Will depend on population to be surveyed
- There may be ready-made lists from which to select, e.g. administrative records - electoral roll, school roll, telephone directories, patients attending treatment centres
- Where lists don't exist use may have to be made of proxy lists, e.g. clients of shelters for the homeless
Sampling method and sample size
- Where lists are available, can decide on stratification method to suit the aims of the survey, e.g. ranging from 100% sample (census), to be representative of age, gender, geographical area.
- Where some sub-groups are small in number (e.g. minority/hard to reach groups) may have to consider the need for booster samples so that sample sizes are of sufficient size to yield meaningful statistics.
- Where there are no lists, then can select every nth person in a given location, or use the snowball approach
Data collection methods and instruments
These need to be appropriate to the population being surveyed
Postal and telephone surveys
- Cannot be used for homeless/excludees
- But can be used for some recreational users, e.g. Mixmag postal survey of clubbers who read monthly music magazine. Self-selected respondents but can throw light onto a particular type of drug use
- How many mail-outs and what constitutes a reasonable response rate?
Self-completion pen and paper questionnaires cannot be used for some groups (e.g. illiterate, ethnic minorities) if they cannot read or understand the language the questionnaire is written in
Face to face interviews - respondents may want to say what they think the interviewer wants to hear or may hide the truth.
Self-completion questionnaires, especially if anonymised and completed using a laptop computer, may get round these issues.
Internal validity can be enhanced by the use of urine or saliva testing of respondents to see if what they say matches with the physical evidence. This more practical with smaller numbers e.g. those arrested or in prison, drivers randomly stopped. There are ethical issues about the imposition of collecting samples from people. It is likely to be more acceptable for those coming into contact with the criminal justice system, and in fact may be lawful/mandatory e.g. prisons in the UK. It is less likely to be acceptable in schools, although such suggestions made recently in the UK have been broadly welcomed.
Analysis
Although published reports on surveys may not present information as is required for the completion of Standard Tables 01 and 02, it may be possible to get the researchers who carried out the analyses to either re-analyse the data using the age groups used in the Standard Tables or to negotiate access to the data and carry new analyses specifically for the National Report. In some cases this will not be possible for all age groups, e.g. the UK general household surveys look at 16-59 year olds whereas the Standard Tables ask for data on 15-64 year olds.
The basic analysis for general population surveys must include (for Standard Table 01) percentages for use ever, last year and last 30 days for specific age groups, broken down by gender and type of substance used. For school surveys (Standard Table 02), the same type of information is requested, although for different age groups from 11 to 18.
Additional analysis (including cross-tabulations using counts) could cover:
- Change over time
- Ethnicity
- 'Availability', access to substances, offers of substances
- Attitudes towards drugs, including health issues, e.g. smoking
- Regional differences, type of area
- Age of first use, drug careers
- Employment; illegal income, connections with offending,
- Dependence status, risk behaviour, e.g. injecting, sharing equipment, unprotected sex.
Logistic regressions and other correlational methods could be used to identify statistically significant relationships between variables.
Estimates of the numbers of people using substances at a national level can be extrapolated from sufficiently large (in terms of sample size and spread) national general population and school surveys. For example, the British Crime Survey.
Methodological issues
- Surveys, unless they are 100% censuses, cannot tell us about everyone in a population.
- Generalisability/external validity will be affected by the type of area surveyed, i.e. the more specialised the sample the more the findings are unique to that setting, population or situation.
For example, the results from a school survey of 11-18 year olds may not be applicable to all in that age group
- some in upper end of that age group may be in employment (and thus have a disposable income with which to buy drugs)
- they can be excluded from School,
- they can be absent from school, e.g. truanting, or
- they may be homeless and thus exposed to situations that may make them more at risk of certain types of drug taking.
- Definitions should be clearly stated. Where there are differences between surveys in definitions these should be explained, and attempts should be made to use one definition across a country, and across different sweeps of the same survey.
- External validity can only be increased by using sampling methods that mean that a range of locations and/or situations is sampled.
- Is there any evidence of interviewer effect? For example, was the interviewer a government official or someone else 'in authority' whose status could have affected the answers given by respondents? If so, was this effect reduced by the methodology employed e.g. self-completion on a lap-top computer?
- Population frame errors.
- Non response
- Any design effects (clustering, non-proportional designs)
Data quality issues
Data sources - should be briefly described in the National Report, together with issues regarding definitions, completeness, timeliness of data submission, and any limitations.
If there are thought to be any problems with the quality of the data, these should be stated in simple terms, and contacts/references given for further information. Alternatively, technical notes can be included.
If it is thought that the data are not robust then this should be stated rather than using misleading information.
EMCDDA reporting requirements
National Report
Chapter 2 relates to drug use in the population
It is supposed to be an overview or summary on drug use and attitudes to drugs. It covers:
- prevalence (extent) and incidence (new cases) of use
- patterns of use
- characteristics of users (gender, social characteristics, age at first use)
Key topics
Drug use in the general population
Drug use in the school and youth population
Drug use among specific groups - conscripts, minorities, sex workers, etc.
Attitudes to drugs and drug users
Check list
The following need to be included for each of the key topics
New developments and trends
How these results relate, if at all, to other indicators
Placing trends within the wider context of - youth culture; economic, social, cultural & demographic changes, supply, etc.
Main results of new research, studies & surveys.
[Surveys can also ask about access to/availability of drugs - this is useful for Chapter 10 on Drug markets]
There are two Standard Tables required by the EMCDDA:
ST 01 - Basic results and methodology of population surveys on drug use
ST 02 - Methodology and results of school surveys on drug use
Organisation of information
For the National Report to the EMCDDA, this should be done in a logical sequence within each of the sections already set out in the guidelines. These are sufficiently general to allow flexibility within them. Other reports for internal national consumption can follow the same layout or be adjusted to take account of national requirements.
For each topic the following should be given:
- Source of information
- What the main findings are for the latest period available,
- How they compare with previous years (if such information is available),
- Any limitations regarding sample size, quality of data, completeness, etc,
- Any other caveats regarding interpretation,
- How the findings agree/disagree with other (available) indicators
- References for further sources of information/detail
A possible structure for Chapter 2 might be as follows:
- What surveys have been undertaken and by whom, sample sizes and age range; any other studies
- Overview from different sources
- Drug use in the general population, including young people; school, children; vulnerable groups
- National and local estimates of problem drug use and prevalence rates (per 1,000 population)
- Trends in prevalence and incidence - age, gender, main drugs, ethnic groups,
- Attitudes to drugs and other substances
- Availability of/ access to/ offers of drugs
- Overview of figures for year being reported on and comparison with previous year for main drugs of interest
- More detailed analysis for each main drug
Errors made in presentation & terminology
Make sure that the unit of analysis is always given for tables and charts, and the level of counting, e.g., offender vs. offence.
State clearly what is meant by terms such as 'access' to drugs and 'availability' of drugs. These are often taken to mean the same thing but can have very different connotations.
'Problem' drug use can relate to different substances according to local patterns of use. For example, benzodiazepines are more of an issue/problem in Scotland and Northern Ireland than in England or Wales.
'Problematic drug users' may just be heroin users in one survey but 'opiate+' in another, or even 'opioids and benzodiazepines' in yet another.
Make sure that drug names are correctly spelled, especially unusual ones.
Recall periods
- Use ever = 'lifetime use'
- Use in last 12 months = 'recent use'
- Use in last 30 days = 'current use'
Further resources
EMCDDA
http://academy.emcdda.eu.int/reference-documents/key-epi-indicators
/gps/self-reported-surveys-report.pdf
http://academy.emcdda.eu.int/reitox-academy/madrid-seminar
/madrid-gen-pop-survey/madrid-gen-pop-survey-jrosuna.doc
http://academy.emcdda.eu.int/reitox-academy/madrid-seminar
/madrid-gen-pop-survey/madrid-gen-pop-survey-emq-rbless.ppt
http://www.emcdda.eu.int/multimedia/project_reports/situation
/population_survey_handbook.pdf
http://www.emcdda.eu.int/multimedia/project_reports/situation
/druguse_survey_CT00EP14.pdf
Further resources and examples
UNODC -
http://www.unodc.org/unodc/en/drug_demand_gap_m-toolkit_module.html
Standards for school surveys have been developed by the Pompidou Group of the Council of Europe and the Swedish Council for Information on Alcohol and other Drugs (CAN).
