Sex, lies and survey data: adolescent sexual behaviour in Tanzania

Collecting accurate data on the sexual behaviour of adolescents can be difficult.  Respondents may give inaccurate answers for a variety of reasons. Many young people find it difficult to speak openly about their sexual health and behaviour. In rural Tanzania, where the standard of education is low, poor literacy is also a barrier to completion of questionnaires. Methods of data collection must be responsive to young people's particular circumstances.

Researchers from the London School ofHygiene and Tropical Medicine tested the usefulness of assisted self-completionquestionnaires (ASCQs).  These are considered a compromise betweenface-to-face interviews and self-completion questionnaires (SQAs).In the former, adolescents may be more inclined to give answers that theybelieve are socially acceptable, whether true or not. Whilst SQAs guarantee more privacy, they demand a standard ofliteracy adequate to read and understand the questions. 

Each question and possible answer from anASCQ is read aloud by a researcher. To fill in their answers, the respondents onlyneed to follow the question numbers and the letters for each answer option.

The researchers compared the results of a 1998ASCQ survey of 6,079 primary school students with those of three overlappingsub-groups: 4,958 students in a 1998 face-to-face interview (FFQ); 4,424students in an ASCQ conducted 18 months later; and a relatively small number ofyoung people who participated in qualitative research (i.e. in-depthinterviews, focus group sessions and participant observation. Answers toquestions asked in both the 1998 ASCQ and FFQ were compared assess consistencyduring the same time period.

The 1998 and 2000 ASCQ results werecompared to assess consistency over time, for example if an individual reporteda higher total number of partners in the 1998 survey than in 2000. Qualitative findingswere used to assess the plausibility of the quantitative results. The analysisfound that:

• The 1998 ASCQ socio-demographic characteristics and some of thesexual and reproductive knowledge, attitudes and behaviour results seemedplausible.  However some of thesexual behaviour results were problematic, due to bias related to answer orderor inconsistencies between answers.
• A minority of ASCQ respondents selected only first or only lastanswer options in sections for which this was illogical. This may haverelated to poor comprehension, unease with the sexual topic, and/orfrustration with questionnaire length.
• In 1998 and 2000, the majority of respondents did not appear tohave difficulty completing the ASCQ, suggesting this method haspromise.  Screening of respondentswho could not or would not answer logically and consistently was generallyeffective.
• Reports of sensitive behaviour were sometimes higher in the1998 ASCQ than in the FFQ. Overall the 1998 surveys had similar results,but comparison at an individual level showed some great inconsistencies.
• The high proportion of respondents who unreliably reportedvaginal intercourse in two surveys raises questions about the validity ofsexual behaviour data collected in surveys in general.

Implications for future ASCQ surveysinclude:

• This method in an African setting has most potential amongolder, male and/or more educated adolescents.
• Biases may decrease with shorter questionnaires and random allocationof answer order.  Many questionsshould be repeated to assess internal consistency.
• Careful screening should be done to exclude individuals whocannot or do not complete the questionnaire correctly. This should takeplace both before administering the questionnaire, and during dataanalysis.