Strasbourg, 5 July 2004       

P-PG/Med (2004) 2 E

Validity and reliability of school surveys based on the European ESPAD methodology

in Algeria, Libya and Morocco

(MEDSPAD pilot school survey project)

RUUD BLESS and RICHARD MUSCAT



PREFACE

1.       SAMPLE AND RESPONSE ANALYSIS. 4

2.       MISSING VALUES ANALYSIS. 6

2.1.... Introduction. 6

2.2.... Item non-response. 7

2.3.... Declared missing values. 10

3.       IMPUTATION OF MISSING VALUES. 10

4.       VALIDITY.. 10

5.       RELIABILITY.. 12

6.       CONCLUSIONS. 15

ANNEX 1: ITEM NON-RESPONSE AND MISSING VALUES in % PER (SUB)QUESTION.. 16

ANNEX 2: PATTERNS OF ITEM NON-RESPONSE IN TABLE FORMAT QUESTIONS. 24


PREFACE

The development of a questionnaire that addresses the issue of drug use among 15-16 year olds throughout Europe has as its early foundations in a PG working group that goes back to 1989. Subsequently, the 1994 report on the pilot survey conducted in six European countries paved the way for the first full survey in 1995 in which  26 European countries participated.  Following this first wave, a number of issues arose that were tackled successfully in  a 1998 survey among eight countries in which issues of drunkenness and validity were tested and presented at the 25th annual Alcohol Epidemiology Symposium in Montreal, Canada in 1999. Moreover, the necessary changes were then adopted by all countries for the next survey that was conducted in 1999 among 30 European countries. In 2003 the third survey conducted in 35 European countries was carried out and the report is due at the end of 2004.

Moreover, the success of the European School Survey, a Pompidou Group initiative, demonstrated that it is possible to co-ordinate, collect, collate, compare and publish data in relation to the prevalence of alcohol and other drugs amongst youth. In much the same regard, Health Behaviour in School-Aged Children study (WHO) has been able to collect specific data on children’s health in most European countries and those of North America.

These two projects per se demonstrate that it is viable to collect reliable information on the behaviour of youth that in turn may be used by policy makers to address specific issues. The only apparent caveat with either of these projects in regard to the Mediterranean is that the HBSC does not make any reference to drug use while the ESPAD is mainly based on Western European society norms in the targeted population. Thus it was not considered appropriate to completely adapt either to the Mediterranean context, however in light of the fact that the ESPAD does consider drug use in youth the working group supported the concept of adapting the ESPAD for piloting in the Mediterranean.

This report presents the result of an analysis on validity and reliability of pilot school surveys in Algeria, Libya and Morocco based on the methodology of the European ESPAD survey.
1.         SAMPLE AND RESPONSE ANALYSIS

The pilot surveys are based on convenient samples that more or less cover the variations in school types and socio-economic environments of the areas chosen (Boumerdes and surrounding area in Algeria, metropolitan Tripoli in Libya and Rabat and surrounding area in Morocco). The intention was to cover in all countries the mandatory school-going age group of 15-16 year olds, which would imply more or less equal numbers of boys and girls in the sample.

The results however show that the assumptions underlying the convenience sample are not consistent with the target group actually reached. The age group covered is much more varied, in particular in Morocco (Figure 1). Median as well as modal age is 15 in both Algeria and Libya and 16 in Morocco. In Algeria and Libya females are over-represented in the response (
Figure 2
).

As we don’t have information about the age and gender distributions of the school classes surveyed, we cannot assess if this deviation from the intended survey population is caused by the selection of schools and classes, incorrect assumptions about the expected age distributions in the selected classes or by non-response of the class populations addressed.

Figure 1: Age distributions in the response of the pilot samples


Figure 2: Gender distribution in the response of the pilot samples

2.                  MISSING VALUES ANALYSIS

2.1              Introduction

We distinguish two types of missing values. First, missing values resulting from item non-response, and second, values that are declared as missing because of data entry errors (entering a code that doesn’t correspond to the pre-coded answer categories) or because the answer category itself implies that the respondent cannot (“don’t know”) or doesn’t want to reply to a question (refusal).

Item non-response, i.e. survey questions (items) that deliberately or inadvertently have been skipped by the respondents, affects the accuracy of the population estimates that statistically can be inferred from the survey data, as the net response for the items concerned will be lower than the overall response rate, which results in larger margins of error. When item non-response is not randomly distributed, it can also imply bias in the survey results.

In computer assisted surveys item non-response normally doesn’t occur because the software prevents incorrect skipping of questions by requiring an answer to a question before one can move to the next one. Also in interviewer assisted surveys item non-response is usually rare when the interviewers are well trained and experienced. But in pen-and-paper self-completion surveys, like MEDSPAD, item non-response can be a serious problem. Respondents may skip questions by mistake, but large numbers of skipped questions more likely indicate that they do not understand the questions or answer categories or that they feel uncomfortable with the content of the survey or do not want to answer particular questions, which in turn might indicate poor questionnaire design, failing completion instructions, inadequate survey introduction or might signal that the survey addresses issues that are beyond the interests or experiences of the target group.

Declared missing values can have a similar effect on population estimates when they are caused by data entry errors, but their main problem is that they limit the options for analysis. Many “don’t know” answers can also indicate that a question is not appropriate for the target group. 

The purpose of the analysis below on the pilot school surveys of Algeria, Libya and Morocco is to assess the extent and patterns of missing values and to identify problems in the design and content of the MEDSPAD questionnaire.

2.2              Item non-response

The questionnaire of the MEDSPAD pilot survey contains 46 (Algeria, Morocco) or 48 questions (Libya). Many questions however are split into sub-questions or have a table format in which the rows actually represent separate sub-questions. If we take these as separate questions the total number of questions varies from 190 (Morocco) to 191 (Algeria) and 196 (Libya). The questionnaire is designed in such a way that every respondent should answer each question; there are no instructions to skip questions on the basis of answers to preceding questions. Non-response is left blank in the survey data files (Algeria, Libya) or coded with a value that is labelled as “missing” (Algeria).

The average percentage of questions skipped by the respondents varies from 10.5% (Algeria) to 14.9% (Libya) and 17.2% (Morocco). In Morocco and Libya almost 50% and in Algeria almost 30% of the respondents skipped more than 10% of the questions; in Morocco 10% (Libya 8%, Algeria 5%) of the respondents skipped even more than 50% of the questions (Figure 3).

This level of item non-response should be considered very high as most drug use school surveys show less than 5% item non-response[1]. When non-response is very high, analysis of the items concerned will not give reliable results and researchers should reconsider or abandon the questions.

Item non-response can also imply bias in the survey results if the non-response is associated with respondent attributes or other characteristics of the sampling. We have tested this for gender and age of the pupils and for the schools participating in the survey.

The results, based on analysis of variances within and between groups, show that in Morocco and Libya boys skip more questions than girls; in Algeria it seems the opposite but the difference is not statistically significant (Table 1). In all countries older pupils skip more questions than younger ones, although in Libya pupils of over 17 skip less than 16 year olds (Table 2).

Figure 3.Distribution of % skipped questions

Table 1: % skipped (sub) questions by gender

Gender

Mean

N

Std. Dev.

Sig.

ALGERIA

Male

9.1

154

13.4

n.s.

Female

11.2

276

15.4

LIBYA

Male

17.2

387

21.2

0.002

Female

13.4

592

16.7

MOROCCO

Male

21.5

159

22.6

0.000

Female

11.9

145

16.2

Table 2: % skipped (sub) question by age

Age

Mean

N

Std. Dev.

Sig.

ALGERIA

< 15

7.5

109

11.0

0.000

15

7.9

125

11.6

16

12.6

107

15.8

17 +

15.3

87

19.5

LIBYA

< 15

11.9

345

14.0

0.000

15

15.0

386

19.5

16

22.3

100

25.3

17 +

17.3

82

20.8

MOROCCO

< 15

7.2

38

10.1

0.000

15

11.2

62

15.3

16

15.6

68

19.3

17 +

22.0

129

22.6

At the level of individual schools item non-response varies from 3% to 26% in Algeria and from 9% to 27% in Libya; the differences are statistically significant (Table 3). This implies that the setting (school) of the survey is a key factor for the non-response rates. For Morocco this relation could not be tested because the schools in which the survey took place have not been recorded in the data file.

Table 3: % skipped (sub) questions per school

School

Mean

N

Std. Dev.

Sig.

ALGERIA

Bordj Menaiel

10.5

72

10.9

0.000

Boumerdes

10.7

67

14.1

Cap Djinet

6.5

68

11.4

Corso

7.4

77

12.6

Ouled Moussa

3.0

77

3.4

Sidi Daoued

25.9

69

20.1

LIBYA

School 1

17.6

121

20.1

0.000

School 2

26.6

85

29.3

School 3

13.0

138

16.7

School 4

11.8

112

15.5

School 5

14.7

157

19.8

School 6

9.2

116

9.6

School 7

13.7

126

14.9

School 8

16.0

138

18.1

We have also tested if item non-response relates to the type of question format. The results show that in general table formats with many sub-questions have higher non-response scores than single questions, in particular on the second and consecutive sub-questions (Table 4). This applies to all pilot surveys, although the differences are less prominent in the Algerian survey.

A complete overview of the non-response per question is presented in Annex 1, which also shows that the Algerian pilot survey on almost all questions has less item non-response than the Libyan survey, whereas the Moroccan survey in most cases has the highest non-response rates. At the same time the overview reveals that in general in all pilot surveys high non-responses are found for the same questions, which confirms the suggestion that non-response is related to the type of questions.

Table 4: % skipped questions by type of question format

ALGERIA

LIBYA

MOROCCO

Single questions

6.2

7.6

6.6

Table format questions (average)

11.2

16.2

19.1

   First sub-question

8.4

8.9

10.0

   Consecutive sub-questions

11.6

17.1

20.2

Analysis of the response patterns in table format questions shows that if skipping occurs, the common patterns are to complete only the first sub-question, to halt somewhere halfway down the table or to skip everything, which suggests that the structure of these table formats is not always properly understood or explained in the completion instructions.

In the case of table format questions that ask for life-time, last year and last month prevalence of substance use, skipping might be caused by the fact that a respondent considers asking for last year and last month prevalence obsolete when he already has already stated that he has never used the substance. This assumption can then be used to impute the missing data (see below). The response patterns of table format questions are presented in Annex 2.

2.3              Declared missing values

The data files that we used for the analysis had already been corrected for possible data entry errors. About one-third of all sub-questions have a pre-coded answer category that corresponds to “don’t know” and in most analyses this category would be treated as a missing value.

Combining item non-response and “don’t know” answers substantially increases the percentages of missing values for Q9 (drinking alcohol at 25) and the sub-questions of Q32 (disapprovals), Q33 (risk perceptions) and Q34 (perceived availability of drugs). These combined percentages are specified in Annex 1.

3.      IMPUTATION OF MISSING VALUES

As mentioned before we observed many missing values in sub-questions that ask for last year and last month prevalence when the respondent has already denied lifetime prevalence in the preceding sub-question. In these cases we can assume that the respondent has skipped the last year and last month questions because he thought that these didn’t apply to him. There are several other questions about substance use, in particular related to alcohol use, that in a similar way seem obsolete to the respondent when he has already stated that he didn’t use the substance.

This type of item non-response can be corrected afterwards by imputing the missing values on the basis of the logical argument that once the use of a substance has been denied, skipping of any consecutive question, which phrasing refers to actual use of that substance, should be interpreted as confirming the previous denial of use.

Implementing these imputations on (sub)questions, which implicitly or explicitly require reconfirmation of previous answers, indeed results in a substantial reduction of missing values, in particular in the Libyan and Moroccan pilot surveys (Table 5). This suggests that either the instructions to the respondents should be improved or that the design of these questions should be reconsidered. The effects of the imputations on item non-response of individual (sub)questions are included in the overview of Annex 1.

Table 5: Item non-response before and after imputation of missing values

Number of imputed questions

Average % of item non-response

before imputation

after

imputation

ALGERIA

38

9.7

5.6

LIBYA

48

11.6

4.0

MOROCCO

39

15.1

5.7

4.      VALIDITY

Validity refers to the extent to which the answers to the questions of a survey could be true. Large numbers of item non-response might indicate validity problems and the results of the missing values analysis above suggest that such problems do exist in the pilot surveys.

The MEDSPAD questionnaire contains some questions that directly attempt to assess validity. Two questions ask for the respondent’s honesty with regard to self-reported cannabis (Q44) and heroin (Q45) use. As the answer patterns on Q44 and Q45 are very similar[2], we present only results on Q44 (honesty cannabis). Four table format questions, Q21 (having heard of), Q26 (lifetime prevalence), Q27 (age of onset) and Q28 (first drug), include a sub-question about a non-existent drug, which may indicate exaggeration of drug use.

Honesty

The results with regard to honesty are not very positive. In Algeria and Morocco 32% and in Libya 46% of all respondents state that they would not have reported – probably not or definitively not – cannabis use if they actually would have used it. Considering also the relative high non-response rate on Q44 (see Annex 1) we can hardly expect that the pilot surveys have produced valid cannabis prevalence rates and the same applies to heroin prevalence.

In most countries girls are more honest than boys (Table 6) and younger pupils are more honest than older ones (Error! Reference source not found.) and these differences are statistically significant. Reported dishonesty of course does not mean that respondents have concealed actual drug use, but indicates that the questionnaire was not adequate to measure such use. Extending the survey population to older age groups, which is advocated by the research teams in all countries, might increase the number of respondents that actually have experienced some drug use, but the pilot results suggest that this at the same time would further decrease the validity of survey outcomes.

Table 6: Self-reported honesty regarding cannabis use in % of the response per gender

Already admitted

Definitive YES

Probably YES

Probably

NO

Definitive NO

Total

ALGERIA

Male

14.7

46.0

8.7

8.7

22.0

100

Female

8.5

50.8

8.5

1.7

30.5

100

Total

10.9

49.0

8.5

4.4

27.2

100

LIBYA

Male

6.7

35.9

6.3

4.1

47.0

100

Female

2.2

46.2

8.1

3.4

40.0

100

Total

4.0

42.2

7.4

3.7

42.7

100

MOROCCO

Male

19.4

47.8

3.0

6.7

23.1

100

Female

7.1

51.6

6.3

5.6

29.4

100

Total

13.5

49.6

4.6

6.2

26.2

100


References to Relevin

Despite the fact that 16% of the respondents in Algeria and Morocco and 8% in Libya claim to have heard of the non-existing drug Relevin listed in Q21, only one or two in each pilot survey report actual lifetime use. This dummy test drug therefore does not reveal any further validity problems.

5.      RELIABILITY

Reliability is a necessary, though not sufficient condition for validity and usually refers to the extent to which repeated measurements under the same conditions yield the same results. To assess the reliability of the results of a single survey a more practical way is to check for internal consistency of responses to different questions within the same questionnaire.

The MEDSPAD questionnaire has some build-in options for such consistency checks. For the purpose of this report the following have been explored:

-          Life-time use of substances and age of first use of those substances (Algeria, Morocco; in Libya age of first use has not been recorded);

-          Life-time, last year and last month prevalence of alcohol, cannabis and inhalants (Algeria, Morocco) or all substances (Libya);

-          Honesty of responses on cannabis use and actual reported use of cannabis.

For most substances the Algerian and Moroccan pilot surveys (in Libya age of first use has not been recorded) the rates of inconsistency between reported life-time use (Q6 smoking, Q8a alcohol, Q23a cannabis, Q24a inhalants and Q26 for other drugs) and age of first use (Q27 for all substances) are very high, both for boys and girls (Table 7). In several cases inconsistent answers, i.e. admitting use in one question but denying it in the other, outnumber the consistent answers. The total numbers of users may be small, but given the observed inconsistencies the reported prevalences can hardly be considered reliable.

To some extent these inconsistencies may be related to the phrasing of the questions concerned, as there are subtle differences –at least in the original English or French versions- between the prevalence and the age of first use questions in wording and semantic meaning or interpretation. These differences may have been accentuated in the Arab version of the questionnaire. 

Table 7: Inconsistencies between life-time prevalence of substance use and reported age of first use

Boys

Girls

Total

Country / substance

Valid N

Use reported

% in- consis-tent

Valid N

Use reported

% in- consis-tent

Valid N

Use reported

% in- consis-tent

ALGERIA

Tobacco

152

72

27.8

258

4

50.0

410

76

28.9

Alcohol

151

8

0.0

263

2

50.0

414

10

10.0

Cannabis

148

17

35.3

260

4

25.0

408

21

33.3

Inhalants

147

16

87.5

259

6

66.7

406

22

81.8

Tranquillisers

147

9

55.6

263

8

50.0

410

17

52.9

Amphetamines

146

10

80.0

248

5

60.0

394

15

73.3

LSD

144

3

33.3

248

2

0.0

392

5

20.0

Crack

145

2

50.0

248

2

0.0

393

4

25.0

Cocaine

145

1

100.0

248

2

0.0

393

3

33.3

Relevin

145

0

248

2

0.0

393

2

0.0

Heroin

143

1

0.0

247

2

0.0

390

3

0.0

Ecstasy

145

0

248

2

0.0

393

2

0.0

MOROCCO

Tobacco

157

66

30.3

143

5

40.0

300

71

31.0

Alcohol

154

38

34.2

144

3

33.3

298

41

34.1

Cannabis

155

30

33.3

144

4

50.0

299

34

35.3

Inhalants

152

19

57.9

144

4

100.0

296

23

65.2

Tranquillisers

151

15

66.7

141

15

86.7

292

30

76.7

Amphetamines

136

8

37.5

138

6

66.7

274

14

50.0

LSD

132

4

75.0

137

0

269

4

75.0

Crack

132

3

100.0

137

0

269

3

100.0

Cocaine

131

2

50.0

137

1

100.0

268

3

66.7

Relevin

132

3

66.7

137

0

269

3

66.7

Heroin

131

1

100.0

137

0

268

1

100.0

Ecstasy

132

2

50.0

137

0

269

2

50.0

Rates of inconsistent answers on self-reported life-time, last year and last month prevalences for alcohol, cannabis and inhalants (Algeria, Morocco) or all substances covered in the questionnaire (Libya) are also very high (Table 8). Here inconsistencies can occur either by reporting last month or last year use after having denied last year or life-time use, or by reporting more frequent use in last month or last year than has been reported for last year or life-time use. Again total numbers of users are small, but the prevalences reported are far from consistent and therefore not reliable. In this case inconsistencies cannot be attributed to the phrasing of the questions

Table 8: Inconsistencies between reported life-time, last year and last month prevalences

Boys

Girls

Total

Country / substance

Valid N

Use reported

% in- consis-tent

Valid N

Use reported

% in- consis-tent

Valid N

Use reported

% in- consis-tent

ALGERIA

Alcohol

149

9

44.4

255

1

100.0

404

10

50.0

Cannabis

148

15

26.7

257

3

33.3

405

18

27.8

Inhalants

146

16

6.3

258

6

16.7

404

22

9.1

LIBYA

Alcohol

371

13

23.1

568

3

66.7

939

16

31.3

Cannabis

380

12

16.7

585

9

11.1

965

21

14.3

Inhalants

377

9

22.2

584

1

0.0

961

10

20.0

Tranquillisers

380

10

50.0

584

9

44.4

964

19

47.4

amphetamines

353

11

45.5

559

6

66.7

912

17

52.9

LSD

353

5

40.0

556

0

909

5

40.0

Crack

352

3

33.3

556

0

908

3

33.3

Cocaine

352

3

100.0

556

0

908

3

100.0

Relevin

352

4

75.0

555

0

907

4

75.0

Heroin

353

3

66.7

555

0

908

3

66.7

ecstasy

352

4

75.0

555

0

907

4

75.0

Drug injecting

353

5

80.0

555

1

0.0

908

6

66.7

Alcohol+pills

380

5

40.0

590

0

970

5

40.0

Alcoh.cannabis

382

5

40.0

591

2

50.0

973

7

42.9

Hasj+marihuana

353

6

50.0

555

1

100.0

908

7

57.1

MOROCCO

Alcohol

156

35

34.3

143

3

33.3

299

38

34.2

Cannabis

156

31

25.8

143

3

0.0

299

34

23.5

Inhalants

153

21

47.6

144

4

0.0

297

25

40.0

Finally, comparing the responses on the honesty question Q44 about cannabis use shows that most respondents who declare that they “already said to have used cannabis” in fact previously had denied the use of cannabis in the prevalence question Q23a (Table 9). The reverse could be observed in Libya, where 6 out of 14 self-reported users declare that they “definitively would not have said so if they had used cannabis”.

Table 9: Inconsistencies between self-reported cannabis use (Q23) and honesty with regard to cannabis use *Q44).

Boys

Girls

Total

Country / substance

Valid N

Use reported

% in- consis-tent

Valid N

Use reported

% in- consis-tent

Valid N

Use reported

% in- consis-tent

ALGERIA

144

22

220

20

364

42

85.0

LIBYA

313

21

488

11

801

32

90.6

MOROCCO

131

26

125

9

256

35

46.9

6.      CONCLUSIONS

·         The analyses show that the results of the MEDSPAD pilot surveys in Algeria, Libya and Morocco cannot be considered valid or reliable. Without substantial changes in methods and instruments a survey based on the European ESPAD model will not produce valid and reliable prevalence estimates for these countries. Considering the similarities in the problems encountered in all pilot countries, this might apply to all Arab countries.

·         Some of the validity and reliability problems might be solved by improving the design of the questionnaire, for example by reducing the number of table format questions, or by providing better instructions on how to complete the questions.

·         It is likely that validity and reliability problems relate to the content of the questionnaire itself. Pupils are not familiar with the situation of being subjects of a survey by means of a questionnaire with pre-coded answer categories and pupils are not used to the idea of reporting honestly about issues that are considered taboo or forbidden. This might be addressed by better preparation and instruction prior to administering the questionnaire, but it could also imply that the instrument is not appropriate to assess drug use prevalence in the countries involved.

·         The over-representation of females in the response in Algeria and Libya might indicate that the reality of school participation at the age of 15-16 years differs from the expected situation based on the mandatory age until pupils have to attend school. This affects the basic assumption of the project that the target group chosen will more or less cover the general population of 15-16 year olds.


ANNEX 1: ITEM NON-RESPONSE AND MISSING VALUES in % PER (SUB)QUESTION

The column “item non-response + declared missing values” in the table below only presents figures if declared missing values exist as pre-coded categories. The column “item non-response after imputation” presents only figures if imputations have been made.

< 5%

   

5-15%

15-30%

> 30%

(Sub)

question

Label

ALGERIA

LIBYA

MOROCCO

Item non-response

Item non-response

+ declared missing values

Item non-response after imputation

Item non-response

Item non-response

+ declared missing values

Item non-response after imputation

Item non-response

Item non-response

+ declared missing values

Item non-response after imputation

%

%

%

%

%

%

%

%

%

Q1

Sex

0.0

1.5

4.7

Q2

Age

0.5

8.1

6.9

Q3

Doing things

Q3a

Doing ride

4.2

5.4

2.8

Q3b

Doing games

4.4

7.7

7.5

Q3c

Doing sport

6.5

9.0

6.3

Q3d

Doing read

3.7

8.9

6.3

Q3e

Doing party

6.0

8.6

7.8

Q3f

Doing other

72.6

30.8

6.6

Q4

Missing school

Q4a

Absent illness

9.3

14.8

12.9

Q4b

Absent skipped

41.4

43.0

42.0

Q4c

Absent other

71.2

35.1

29.8

Q5

Grade

1.2

9.2

4.7

Q6

LTF smoke

6.0

3.1

1.6

Q7

LMF smoke

3.3

2.1

4.9

2.0

3.1

1.9

Q8

Prevalence alcohol

Q8a

LTF alcohol

6.7

4.3

4.1

Q8b

LYF alcohol

16.5

6.7

32.3

5.0

29.2

6.0

Q8c

LMF alcohol

17.4

7.7

32.4

5.0

28.5

6.0

Q9

Drink 25

5.8

17.7

3.6

13.9

6.0

21.9

Q10

Last month prevalence alcoholic drinks

Q10a

LMF beer

10.0

6.0

6.7

2.1

5.3

2.8

Q10b

LMF wine

12.8

8.8

15.3

10.7

16.6

14.1

Q10c

LMF spirits

12.8

8.8

14.5

9.9

16.3

13.8

Q11

Last=beer

4.4

3.7

3.7

0.9

4.1

1.9

Q13

Last=wine

4.7

4.0

4.2

1.1

3.1

0.9

Q14

Last=spirits

4.7

4.0

4.0

1.4

4.1

1.3

Q15

Where drink

4.7

3.7

2.3

0.7

2.8

1.3

Q16

LMF 5 drinks

4.7

0.0

2.4

0.3

3.1

0.3

Q17

Perceived effects of alcohol

Q17a

Relaxed

25.6

42.3

37.6

Q17b

Police

21.6

44.2

37.9

Q17c

Health

19.1

37.1

29.5

Q17d

Happy

26.7

46.2

40.1

Q17e

Forget

28.8

46.5

40.8

Q17f

No stop

29.1

47.0

40.8

Q17g

Hangover

26.0

44.7

38.6

Q17h

Friendly

28.6

47.7

42.3

Q17i

Regret

25.6

45.9

39.2

Q17j

Fun

27.9

40.7

35.1

Q17k

Sick

24.2

47.0

41.4

Q17l

Guilty

19.8

44.0

37.6

Q18

Prevalence of drunkenness

Q18a

LTF drunk

8.1

4.9

5.0

0.7

7.5

3.8

Q18b

LYF drunk

15.6

4.9

29.1

0.9

32.0

5.0

Q18c

LMF drunk

15.3

4.4

29.1

0.8

32.0

4.7

Q19

How drunk

5.6

9.3 [3]

24.3

6.9

Q20

Amount drunk

5.6

3.5

3.0

0.2

5.6

1.9

Q21

Having heard of drugs

Q21a

Heard tranq.

35.3

16.6

35.7

Q21b

Heard cannabis

6.5

12.0

12.9

Q21c

Heard LSD

18.6

20.0

35.7

Q21d

Heard amphet.

21.9

21.0

38.9

Q21e

Heard crack

20.5

20.5

36.7

Q21f

Heard cocaine

7.9

17.1

19.4

Q21g

Heard relevin

20.5

19.9

37.9

Q21h

Heard heroin

9.8

13.8

26.3

Q21i

Heard ecstasy

20.7

20.9

37.3

Q21j

Heard methad.

20.9

20.1

37.6

Q22

Want try drug

9.1

4.6

4.7

Q23

Prevalence cannabis

Q23a

LTF cannabis

6.7

1.6

3.8

Q23b

LYF cannabis

12.3

6.0

24.4

1.8

31.0

6.6

Q23c

LMF cannabis

13.0

6.7

24.6

1.9

30.4

5.6

Q24

Prevalence sniffing

Q24a

LTF sniff

6.0

1.8

3.4

Q24b

LYF sniff

13.7

7.4

25.9

2.2

30.7

4.7

Q24c

LMF sniff

13.7

7.4

25.7

2.2

30.4

4.1

Q25

Prescr. tranq.

2.8

3.2

1.9

Q26

Lifetime prevalence of drugs

Q26a

LTF tranq.

5.8

3.5

6.0

Q26b

LTF amphet.

9.1

9.0

12.5

Q26c

LTF LSD

9.5

9.0

13.8

Q26d

LTF crack

9.3

9.0

13.5

Q26e

LTF cocaine

9.3

9.0

13.8

Q26f

LTF relevin

9.3

9.1

13.5

Q26g

LTF heroin

9.3

9.3

13.5

Q26i

LTF ecstasy

9.3

9.6

13.5

Q26j

LTF injecting

9.5

9.4

13.8

Q26k

LTF alc.+pills

9.5

5.3

9.5

1.2

13.8

4.1

Q26l

LTF alc.+cann.

9.5

3.5

9.1

0.7

11.9

1.6

Q26m

LTF steroids

9.5

Q26n

LTF hasj+mari.

9.1

Q26_2

Last month prevalence of drugs (LIBYA only)

Q26_2a

LMF tranq.

4.9

1.8

Q26_2b

LMF amphet.

10.4

6.9

Q26_2c

LMF LSD

10.5

7.1

Q26_2d

LMF crack

10.9

7.2

Q26_2e

LMF cocaine

10.6

7.2

Q26_2f

LMF relevin

10.6

7.3

Q26_2g

LMF heroin

10.4

7.3

Q26_2i

LMF ecstasy

10.5

7.4

Q26_2j

LMF injecting

10.7

7.3

Q26_2k

LMF alc.+pills

10.5

1.0

Q26_2l

LMF alc.+cann.

10.4

0.7

Q26_2n

LMF hasj+mari.

10.4

7.4

Q26_3

Last year prevalence of drugs (LIBYA only)

Q26_3a

LYF tranq.

5.0

2.0

Q26_3b

LYF amphet.

10.1

7.2

Q26_3c

Lyf lsd

10.2

7.2

Q26_3d

LYF crack

10.3

7.3

Q26_3e

LYF cocaine

10.4

7.3

Q26_3f

LYF relevin

10.4

7.4

Q26_3g

LYF heroin

10.6

7.5

Q26_3i

LYF ecstasy

10.3

7.5

Q26_3j

LYF injecting

10.3

7.4

Q26_3k

LYF alc.+pills

10.7

1.0

Q26_3l

LYF alc.+cann.

10.8

0.7

Q26_3n

LYF hasj+mari.

10.7

7.6

Q27

Age of first use substances

Q27a

Age beer

5.1

4.0

6.9

3.4

Q27b

Age wine

10.2

5.8

14.1

4.7

Q27c

Age spirits

10.5

5.8

15.0

4.4

Q27d

Age drunk

10.2

3.7

14.1

3.8

Q27e

Age first cig

9.8

5.1

13.2

6.0

Q27f

Age day smoke

10.7

16.0

Q27g

Age amphet.

10.9

8.8

15.4

11.0

Q27h

Age tranq.

10.5

4.7

15.4

5.3

Q27i

Age cannabis

10.7

5.6

16.9

5.0

Q27j

Age LSD

10.5

17.6

Q27k

Age crack

10.2

8.6

16.9

11.3

Q27l

Age cocaine

10.2

8.6

17.2

11.6

Q27m

Age relevin

10.5

8.6

17.2

11.6

Q27n

Age ecstasy

10.7

8.6

17.2

11.6

Q27o

Age heroin

17.2

11.6

Q27p

Age sniff

10.9

5.8

16.9

4.4

Q27q

Age steroids

10.5

8.6

20.7

20.7

Q28

First drug

4.4

4.7

3.0

2.1

2.3

0.4

5.3

6.6

2.2

Q29

How obtained

5.3

3.0

3.2

0.4

5.3

1.3

Q30

Reason taking

5.1

3.0

2.5

0.5

3.4

1.3

Q31

Easy buy can.

5.3

3.7

6.9

Q32

Disapproval substance use

Q32a

Smoke occas.

2.6

8.6

6.0

16.0

4.7

11.9

Q32b

Smoke 10

3.7

6.7

11.8

19.1

13.8

16.9

Q32c

Drink few year

2.8

8.1

11.7

19.3

12.9

16.6

Q32d

Drink 1-2 week

12.0

19.3

16.6

20.1

Q32e

Drunk once wk

2.8

6.7

12.1

20.0

16.0

18.5

Q32f

Cannabis try

3.5

7.9

12.0

19.6

16.0

19.4

Q32g

Cannabis occ.

2.8

7.9

12.5

19.5

16.9

19.7

Q32h

Cannabis reg.

2.8

8.8

12.1

19.5

16.0

19.1

Q32i

LSD try

3.5

10.2

12.8

20.2

17.6

20.4

Q32j

Heroin try

4.2

10.9

12.2

19.6

17.6

21.0

Q32k

Tranquill. try

3.3

12.1

12.4

20.7

18.2

21.3

Q32l

Amphet. Try

3.7

10.0

12.2

20.0

16.6

19.4

Q32m

Crack try

3.7

10.0

12.4

19.7

17.6

21.0

Q32n

Cocaine try

3.7

8.6

12.3

19.4

17.6

20.4

Q32o

Ecstasy try

4.4

10.9

12.3

19.7

17.2

20.4

Q32p

Sniff try

4.0

10.7

12.3

19.7

16.9

19.7

Q33

Risk perception substance use

Q33a

Smoke occas.

1.4

12.3

7.9

35.5

4.7

31.3

Q33b

Smoke heavy

2.3

10.0

13.0

36.6

10.7

25.7

Q33c

Drink 1-2 day

2.6

14.4

13.7

39.5

11.6

26.0

Q33d

Drink 4-5 day

3.3

15.1

13.9

39.1

12.5

27.9

Q33e

Drink 5 wk’end

3.3

15.6

14.6

40.0

11.0

25.1

Q33f

Cannabis try

3.3

17.9

13.6

40.5

11.6

30.7

Q33g

Cannabis occ.

4.4

19.8

14.4

40.9

14.4

35.1

Q33h

Cannabis reg.

4.0

18.6

13.9

39.5

14.4

31.0

Q33i

LSD try

4.0

21.6

14.5

44.1

15.0

34.5

Q33j

LSD regular

3.7

19.8

14.3

43.4

15.7

34.8

Q33k

Amphet. Try

5.1

23.0

14.3

44.2

15.0

36.