SEARCH

SEARCH BY CITATION

Keywords:

  • adolescent;
  • illiteracy;
  • methodology;
  • reliability and validity;
  • sex behaviour;
  • Africa

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Objectives  To develop and test a sexual behaviour survey method for semi-literate populations, combining the privacy of a self-completion questionnaire (SCQ) with the clarity of a face-to-face questionnaire (FFQ).

Methods  In 1998, 6079 Tanzanian primary school students (mean age 15.1 years) were surveyed using an innovative assisted self-completion questionnaire (ASCQ). The format of the questionnaire was simple, all responses were closed, and conceptually complex questions such as those involving ranking or multiple answers were avoided. The ASCQ was administered to groups of 20 by a research assistant who read questions and answers aloud in two languages so pupils could tick or write responses independently. A total of 4958 of respondents from the 1998 ASCQ Cohort also participated in a 1998 FFQ interview and, in 2000, 4424 again completed an ASCQ.

Results  In the 1998 ASCQ survey, 55.0% of males and 21.1% of females reported they had had vaginal intercourse, of whom 71.5% and 66.0%, respectively reported their first sexual relationship lasted for a week or less, and 49.5% and 59.6%, respectively reported they had had sex in the last 4 weeks. After adjustment for age, reported sex was associated with alcohol use in both males (OR = 1.57) and females (OR = 1.69), earning money for males (OR = 1.32) and not living with a mother for females (OR = 0.77). The vast majority of respondents did not appear to have difficulty completing the ASCQ, but 7.4% of 1998 respondents and 2.9% of 2000 respondents selected all first or all last answers in a section for which this was inconsistent. This bias was associated with female, less educated and more geographically remote respondents. Of those respondents who reported sex in the 1998 ASCQ survey, 32.1% reported fewer total partners in the 2000 ASCQ survey, 25.2% reported having had sex fewer times than originally reported, and 61.9% of those who reported having used a condom in 1998 reported never having used one in 2000. While the proportions reporting sex were very similar in the 1998 ASCQ and FFQ surveys, 37.9% of males and 59.2% of females reporting sex only did so on one of the two questionnaires. Higher proportions of respondents reported sensitive information in the ASCQ than the FFQ, although in some cases this may have related to answer order bias.

Conclusion  The results suggest that an ASCQ may be useful in assessing sexual behaviour in African adolescents, particularly for older, male and/or educated respondents. However, triangulation with data from other surveys raises questions about the validity of self-reported sexual behaviour in general.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

In many parts of sub-Saharan Africa, sexual activity commences at a young age (Meekers 1994; Blanc & Way 1998; Singh 1998), and HIV incidence is often highest among young people, particularly among girls and young women (Konde-Lule et al. 1997; Zabin & Kiragu 1998; UNAIDS 2000). Many African countries have implemented HIV interventions for adolescents (Kuhn et al. 1994; Klepp et al. 1997; MacLachlan et al. 1997; Schapink et al. 1997; Kinsman et al. 1999), but such efforts have rarely been rigorously evaluated. Collection of valid data on sexual behaviour is challenging because of problems with recall, ambiguous terminology, and the sensitive nature of sexual information (Wight & West 1999). Self-reported data may be invalid if informants tell researchers what they believe is socially desirable, whether or not it is true (Nachman 1984; Bleek 1987; West 1990), and such bias may be particularly great among adolescents concerned about peer stigma or adult punishment. For example, sexual norms in the Sukuma ethnic group of northern Tanzania include that school pupils should be abstinent, women should maintain their sexual respectability, and sex should not be discussed in public, especially not in groups of mixed generation or sex (Wight et al. 2001). Each of these norms may make it difficult for adolescents to honestly report their sexual behaviour in a survey. In attempting to assess self-reported data validity, researchers have often assumed that illegal or socially stigmatized behaviours are more likely to be under-reported than over-reported. However, something that is socially unacceptable in one group (e.g. multiple sexual partners) may be considered desirable in another.

One important means of exploring the validity of sexual behaviour results is through assessment of data reliability, i.e. whether participants provide consistent responses to the same question asked more than once within an interview or between interviews (Smith & Morrow 1996). A reliable series of responses does not ensure data validity, as a person can consistently report false information, but an unreliable series of responses clearly suggests that one or both reports are invalid. Comparison of data from contemporaneous surveys reduces inaccuracy that may result from changing recall over time. When different methods are used, such comparisons may also provide insights into the strengths and weaknesses of each method (James et al. 1991; Davoli et al. 1992; Rosen et al. 2000; Liau et al. 2002). Comparison of young people's longitudinal reports can instead provide insight into how information and biases may change with different age, education or period of recall (Siegel et al. 1998; Couper & Stinson 1999; McFarlane & St. Lawrence 1999).

Some researchers have proposed that self-completion questionnaire (SCQ) data are likely to be more valid than face-to-face questionnaire (FFQ) data, because a greater sense of confidentiality for the respondent may reduce social desirability bias (Catania et al. 1986; Catania et al. 1990; Wellings et al. 1990; Wolff et al. 2000). However, SCQ surveys can have lower response rates and more missing data than FFQ surveys (Hingson & Strunin 1993; Copas et al. 2002). In face-to-face interviews participant misperceptions can be readily clarified, illiterate or blind respondents can participate easily, and complex question series can be used, such as those involving repeated skips, multiple choices or open-ended answers (Denzin & Lincoln 2000). The question of illiteracy is particularly important in developing countries (Ankrah 1989; Pillai & Benefo 1995; Matasha et al. 1998; Kuleana 1999).

One method which may combine the best qualities of a traditional SCQ (e.g. privacy and confidentiality) with those of a traditional FFQ (e.g. clarity and simplicity) is an assisted self-completion questionnaire (ASCQ), in which an interviewer reads question and answer options aloud, while the respondent complete questionnaires independently. This paper describes the design of this method and the results of a 1998 ASCQ survey of Tanzanian primary school pupils. Data reliability and validity are assessed by comparing them with contemporaneous FFQ, longitudinal ASCQ, and qualitative data collected from subsets of the same cohort. The research took place as part of the MEMA kwa Vijana Project, a community-randomized adolescent sexual and reproductive health intervention trial in rural Mwanza, Tanzania (Cleophas-Frisch et al. 1999; Plummer et al. 1999; Ross et al. 1999).

Mwanza Region is located on the southern shore of Lake Victoria in northern Tanzania, and has an estimated population of 3 million people. Most of the region is rural, and the main occupations are subsistence farming, cattle rearing and fishing. In Tanzania, children cannot start primary school until the age of 7 years, but in fact most only begin school between the ages of 8 and 11 years (Kuleana 1999). Many students repeat years, so it is not uncommon for those who reach the final year, Standard Seven, to vary in age from 14 to 20 years. Most Tanzanian children attend at least part of primary school, but only 5–6% go on to attend secondary school (UNICEF 1998). The quality of education in primary schools is very low, particularly in rural areas (Kuleana 1999). As a result, many children who graduate from primary school in rural Mwanza are illiterate or semi-literate.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Prior to beginning the research, the study proposal was reviewed and approved by the London School of Hygiene and Tropical Medicine Ethics Committee, the Tanzanian Medical Research Coordinating Committee, the Tanzanian Ministry of Health, and the Tanzanian Ministry of Education and Culture.

MEMA kwa Vijana Project participant recruitment took place from September to December 1998 in 20 communities in rural Mwanza. It involved both a clinical/FFQ survey and an ASCQ survey; these surveys shared enrolment, registration and informed consent procedures. Prior to trial enrolment, information meetings were held with school committees, teachers, parents and pupils, and all pupils received information sheets, which they were asked to discuss with their parents. School committee chairs and head teachers provided written consent for their schools to participate in the trial, while parents were given the opportunity to refuse to allow their children to participate, and pupils were asked for written consent after the information sheet had been read aloud to them on the day of registration.

1998 assisted self-completion questionnaire (ASCQ) survey

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Pupils were pre-selected for the ASCQ survey according to the following criteria: (i) in each school up to 20 girls and 20 boys (and alternates) were selected from both Standard 5 and Standard 6; (ii) in each standard equal numbers of pupils were selected from each stream; (iii) in each stream those pupils born in 1984 or earlier were prioritized, followed by those born in 1985, then those born in 1986, etc. and (iv) if more than the required number of pupils were available, they were selected in the order of enrolment.

The ASCQ focused on socio-demographic, future expectation, sexual behaviour, knowledge and attitude questions, in that order. The format of the questionnaire was simple, all responses were closed, and conceptually complex questions such as those involving ranking or multiple answers were avoided.

The draft ASCQ underwent a series of translations and back-translations from English to Swahili (the national language) and Sukuma (the most common local language in the study area). Participant questionnaires were only written in Swahili, as Swahili is the official language of instruction in Tanzanian primary schools, and Sukuma is a tonal language that is rarely used in a written form (Gleason 1961). However, preliminary, small-scale research suggested that many rural Mwanza primary school pupils were not fluent in Swahili, so research assistants were trained to read aloud both standardized Sukuma and Swahili versions of the questionnaire. Three successive drafts of the questionnaire were pilot tested in one rural school each, resulting in substantial modifications of questionnaire content and administration.

ASCQ participants completed questionnaires in same-sex groups of up to 20 individuals, who were spread throughout a classroom to maximize privacy. Prior to beginning the questionnaire, a 20–24 year old, same-sex research assistant gave a standardized introduction in both Swahili and Sukuma. In the introduction, the research assistant explained the research, emphasized confidentiality, showed how to answer questions with the use of a large demonstration board and asked if any pupil would prefer the questions and answers to be read aloud in Sukuma and Swahili.

The research assistant then slowly read aloud each of the 87 questions and each answer option in Swahili (and Sukuma, if a pupil requested it, or the research assistant deemed it necessary), allowing time for pupils to tick or write their responses independently. In this way, respondents with a low literacy level could follow the number of each question and the letter of each answer option to select an appropriate answer, without necessarily being able to read the questions and answers themselves. Pupils were asked to listen to the full question and all the answer options prior to answering. For 80 questions respondents were required to answer with one tick, and for seven questions they were required to write a number.

In the questionnaire, the terms used for ‘to have vaginal intercourse’ were euphemisms, ‘to make love’ (Swahili –kufanya mapenzi) and ‘to sleep with a person’ (Sukuma –kulala na munhu), as pilot studies found that more explicit or technically accurate terms were often either offensive or unfamiliar to the general public. To ensure understanding of these euphemisms, prior to asking the sexual behaviour questions the research assistant provided a standardized definition: ‘when a penis is inside of a vagina (or a man or boy's private parts are inside a woman or girl's private parts)’. At that time, research assistants also provided standardized synonyms, such as ‘sexual intercourse’ (Swahili –kujamiiana; Sukuma –kwilala), ‘meeting of bodies’ (Swahili –kukutana kimwili), ‘sleeping with a person’ (Swahili –kulala na mtu) and ‘act of love’ (Sukuma –shitwa sha witogwa). Respondents were asked about their experience of seduction (Swahili –kutongoza; Sukuma –kukanya), but they were not specifically asked about rape. Pilot studies had found that the term ‘rape’ in Swahili (kubaka) was generally only used for extreme examples of violent, forced sex followed by quasi-judicial proceedings. In an attempt to assess unwilling sex in other forms, the term ‘force’ (Swahili –kulazimisha; Sukuma –kuhadikija) was used instead.

Respondents who had not had vaginal intercourse were asked to continue answering questions during the sexual behaviour section, selecting the last answer ‘I have never had vaginal intercourse’ for each of those 33 questions, so that no one's sexual experience would be obvious to the other respondents in the room. Because of the length of the questionnaire, five breaks were scheduled between sections, when the research assistant briefly led pupils in games or songs, which pupils seemed to enter into enthusiastically. Including these breaks, bi-lingual questionnaire administration took approximately 2 h; when no Sukuma was used, questionnaire administration time was reduced by 30–45 min.

1998 face-to-face questionnaire (FFQ) survey

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Pupils who were eligible and provided written consent to participate in the FFQ interview received clinical screening and treatment and provided urine specimens for pregnancy and STD testing. Approximately one-third of all FFQ survey participants were in Standard 4 and did not participate in the ASCQ survey, as a preliminary, small-scale assessment suggested that many may not have been sufficiently literate. In contrast, pupils born in 1984 or later (i.e. about 14 years old or younger) were not eligible for the FFQ survey. In 14 communities, the ASCQ survey took place 5 weeks before the FFQ survey, while in six communities the reverse was the case.

The FFQ was drafted in English, translated into Swahili, and back-translated into English for modification prior to pre-tests and pilot tests in four primary schools. Same-sex interviewers aged 20–24 years administered the FFQ in Swahili, and when requested also read aloud written Sukuma translations of specific questions. However, unlike ASCQ research assistants, FFQ interviewers varied in their Sukuma-speaking ability, as only a few were fluent, and some did not speak it at all. With simpler procedures and 50 rather than the 87 questions used in the ASCQ, the FFQ usually only took 15–20 min to administer. The narrative introduction to the sexual behaviour section and 17 of the questions were very similar or exactly the same as in the ASCQ. All but four of the FFQ questions required a single answer and did not involve prompting or probing by the interviewer.

2000 ASCQ survey

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

An ASCQ survey was again conducted 18 months after the 1998 ASCQ survey, from February to June 2000. Prior to ASCQ administration in 2000, participant identities were confirmed by comparing faces to the project identification card photographs, and by comparing their answers to 5–10 identification questions with data collected at enrolment. In 2000, pupils were again asked for written consent after an information sheet had been read aloud to them. A new procedure was also introduced in which research assistants asked all respondents if they could read and write letters and numbers, and excluded them from ASCQ participation if they said no.

The 2000 ASCQ followed the same general format as in 1998. However, a few questions that were found to have been poorly designed or too complex were dropped or re-designed. For example, responses requiring written numbers were sometimes illegible, so all such answer options were changed to require only the tick of a box. In addition, in 1998 many participants appeared to have been frustrated by the length of the questionnaire, particularly those who did not report vaginal intercourse, but researchers were unable to identify a way to shorten the questionnaire for those individuals without potentially stigmatizing pupils. In 2000, therefore, efforts were made to reduce the questionnaire for all participants, and 32 of the 87 questions asked in 1998 were dropped. However, some new questions were added because of the participants’ increasing age (e.g. marriage) and experience (e.g. intervention participation). These questions, combined with the division of complex questions into multiple, simple ones, meant that the total number of questions was only reduced from 87 to 73 or 77, depending on intervention status. Bi-lingual questionnaire administration took approximately 1.5 h in 2000; when only Swahili was used, questionnaire administration and completion time was reduced by 30–45 min.

Data processing, analysis and interpretation

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

All data were double entered, verified and cleaned using Dbase Version 4 (Borland Corporation, Scotts Valley, CA, USA). Data were analysed using STATA Version 6 (STATA Corporation, College Station, TX, USA). The Mantel Haenszel method was used to calculate odds ratios (OR) and 95% confidence intervals (CI). Pearson's chi-squared test and McNemar's test were used to compare proportions for unpaired and paired data, respectively (Siegel & Castellan 1998). PPos and Pneg statistics were used to assess the level of agreement between two questions (Landis & Koch 1977). The proportion of positive agreements, Ppos, is defined as 2p++/(2p++ + p+− + p−+), where p++, p+− and p−+ are the proportions of subjects with positive results on both questions, or on one or the other question, with an analogous definition for the proportion of negative agreements, Pneg. For continuous variables, the per cent agreement was defined as the proportion of respondents providing exactly the same response in both questionnaires.

Some 1998 ASCQ data were excluded at two different points during analysis. According to the initial exclusion criteria, an individual's responses to specific sections or questions were excluded if 50% or more of the questions within a section had missing or multiple answers, or if responses to similar questions about experience of vaginal intercourse, number of sexual partners, relationships, or condom use were inconsistent.

Preliminary analysis revealed that a small proportion of respondents appeared to have answered questions based on first or last answer order rather than answer content. The proportion of respondents who entirely selected the first or last answer in one or more sections for which this was illogical or inconsistent (i.e. who demonstrated ‘answer order bias’) was thus used as an indicator of ASCQ data validity, and sociodemographic variables associated with it were examined. Additional exclusion criteria were then developed for the final analysis, in which part or all of questionnaire responses were excluded for those individuals who demonstrated answer order bias.

Comparison of 1998 ASCQ and FFQ data involved assessing the degree and nature of response agreement for the 17 questions that were asked in the same or very similar form in both surveys. In 1998 and 2000 ASCQ data comparison, the longitudinal reliability of sexual behaviour reports was assessed, e.g. if an individual reported experience of sex or condom use in 1998 but not in 2000, and for those who had reported sex in 1998, whether key variables changed (e.g. age at first sex), or decreased illogically (e.g. total number of partners, total number of sexual encounters). In addition, the extent of answer order bias in the two surveys was compared, as were socio-demographic variables associated with it. Where data on reported vaginal intercourse were compared between the three surveys, reported vaginal intercourse in the 1998 ASCQ was arbitrarily chosen as the standard of comparison.

Finally, qualitative findings from the Health and Lifestyles Research (HALIRA) Program, the social science research component of the MEMA kwa Vijana Project, were used to assess and interpret the plausibility of some of the quantitative results presented here. Specifically, data were drawn from HALIRA research conducted between July 1999 and August 2002 that included in-depth interviews with 93 primary school pupils who were participating in the MEMA kwa Vijana Project, a series of 3–4 focus group discussions with young people in each of three project villages, and a total of 158 person-weeks of participant observation with rural young people, primarily conducted in four villages.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

In 1998, 4797 pupils were registered in Standard 5 and 5203 in Standard 6. Of these 10 000 individuals, 8575 (85.8%) were present for trial enrolment, of whom 8564 (99.9%) consented to participate in the research and were enrolled in the trial. Of these, 6857 (80.1%) were pre-selected for the ASCQ, of whom 750 (10.9%) were absent or refused on the day of the survey. Thus, a total of 6107 respondents (61.1% of registered pupils and 89.1% of pre-selected pupils) participated in the 1998 ASCQ survey. However, during questionnaire administration, research assistants identified 28 individuals (0.5%) who were clearly non-literate and unable to complete the questionnaire correctly, so their results were excluded. Usable questionnaires were thus provided by 6079 of the 1998 ASCQ participants (the 1998 ASCQ Cohort), 3059 (50.3%) of whom were male, and 3020 (49.7%) female.

In 1998, of 17 084 registered Standard 4, 5 and 6 pupils of all ages, 14 580 (85.3%) were present for trial enrolment. Of these, 9988 (68.5%) were born in 1984 or earlier and thus were eligible to participate in the FFQ survey, 9976 (99.9%) of whom consented to participate in the research. A total of 9283 (93.1%) of eligible and enrolled pupils were present on the FFQ survey day and participated in an interview. Of the 6079 pupils in the 1998 ASCQ Cohort, 4958 (81.6%) also participated in the FFQ survey (the 1998 ASCQ-FFQ Cohort) (Figure 1).

image

Figure 1. Relationship of 1998 ASCQ Cohort with sub-cohorts. Italicized, negative figures are number of questionnaires partially or entirely excluded because of answer order bias (ASCQ, assisted self completion questionnaire; FFQ, face-to-face questionnaire).

Download figure to PowerPoint

A total of 7255 young people attended and were registered for the 2000 ASCQ survey, of whom 10 (0.1%) refused to participate and 80 (1.1%) did not participate because of self-reported illiteracy, resulting in 7165 (98.8%) usable questionnaires. A total of 4424 (72.8%) of 1998 ASCQ Cohort members participated in this later survey (the 1998–2000 ASCQ Cohort) (Figure 1).

In total, 3590 young people participated in all three surveys (Figure 1). In the majority of both the 1998 and the 2000 ASCQ sessions, research assistants used Sukuma and Swahili. Sukuma was used far less often in FFQ interviews.

1998 ASCQ reporting patterns

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

On average, 1.0% of question responses were either missing or multiple answers, and thus could not be included in analysis. By section, missing or multiple answers made up the following proportions of all responses: 0.0–1.4% (socio-demographic), 0.9–1.3% (future expectations), 0.8–1.9% (sexual behaviour), 0.8–1.1% (knowledge) and 0.5–1.1% (attitudes).

Twelve sexual behaviour questions had an additional 1.6–6.8% (mean 4.7%) of responses excluded from analysis because of internal inconsistencies, as defined by the initial exclusion criteria. Inconsistent responses were greatest for the three pairs of questions focusing on condom use ever (6.1%), at first sex (6.3%) and at last sex (6.8%). Inconsistencies were also evident for sets of questions about the number of times the respondent had had sex and the number of sexual partners (1.6–4.8%) and whether the respondent had a current sexual relationship (1.7–2.4%).

In the final sections of the questionnaire, a high proportion of respondents answered ‘I do not know’: 37.5–55.0% (mean 42.8%) of respondents selected this option for knowledge questions (a last answer option), compared with 8.9–27.4% (mean 18.2%) for attitude questions (a middle answer option). In these sections, males were approximately half as likely as females to select the answer ‘I do not know’.

Both first and last answer reporting biases were evident in the 1998 ASCQ data. Specifically, 210 (3.5%) respondents selected all first answers and 240 (3.9%) respondents selected all last answers in one or more sections for which this was illogical or inconsistent; in this paper such consistent answer order bias will be referred to as ‘first answer bias’ and ‘last answer bias’, respectively. First answer bias was found in both early and late sections of the questionnaire, but last answer bias was only found in late sections. Last answer bias in the final sections probably resulted from response patterns established in the middle section of the questionnaire, on sexual behaviour; 92.4% of those demonstrating last answer bias had earlier reported they had never had vaginal intercourse, and thus for each of the 33 questions in the middle section had appropriately chosen the last answer, ‘I have never had vaginal intercourse’.

All results for the 210 individuals who demonstrated first answer bias, and responses in the last two sections for the 240 individuals demonstrating last answer bias, were excluded from 1998 ASCQ analysis. Exclusion was not associated with research assistant or age, but was associated with sex (2.7% of males, 12.2% of females: P < 0.001) and less education for males (3.3% of Standard 5 males, 2.1% of Standard 6 males: P = 0.005). In addition, exclusion was highly associated with communities that were distant from urban areas or major roads, and/or had an HIV prevalence of <1.5% among 15–19 year olds in a survey conducted in 1997–1998 (P < 0.001).

1998 ASCQ socio-demographic characteristics and future expectations

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Participant ages ranged from 10 to 24 years old, with a median age of 15 years old for both sexes. The mean age of males was 15.6 years, compared with 14.6 years for females. Sukuma ethnicity was reported by 77.6% of respondents, and 79.9% reported that they were Christian, while 12.8% said they had no religion, 6.6% reported they were Muslim and 0.7% said they followed a traditional religion.

Males were significantly more likely than females to report having their own livestock, their own farm plot, and sometimes or often earning money from work (Table 1). The vast majority of respondents (86.2%) thought it likely that they would be farming in 3 years time, while 51.8% thought they would be earning money through other work.

Table 1.  Significant differences in socio-demographic characteristics and future expectations in the 1998 Assisted Self-Completion Questionnaire Cohort, by the respondent's sex
VariablePer cent respondents (no.)P-value
MalesFemales
Either Christian or Muslim82.2 (2455/2987)91.0 (2571/2825)<0.001
Attends church or mosque daily/weekly62.6 (1868/2984)73.9 (2085/2823)<0.001
Has own farm plot65.1 (1945/2986)59.2 (1671/2825)<0.001
Has own livestock30.9 (922/2985)22.3 (630/2819)<0.001
Sometimes/often earns money from work61.8 (1845/2984)38.2 (1078/2820)<0.001
Sometimes/often drinks alcohol4.4 (132/2984)6.4 (181/2825)0.001
Expects will be in primary school in 3 years58.8 (1551/2637)69.9 (1747/2501)<0.001
Expects will be married in 3 years43.0 (1027/2387)34.0 (727/2137)0.001
Expects will have baby in 3 years26.9 (625/2321)32.3 (652/2019)0.003

1998 ASCQ sexual behaviour

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Male (55.0%) and female (21.1%) respondents reported they had had vaginal intercourse. Reported sex was significantly associated with reporting that one has sometimes or often drunk alcohol (both sexes), earned money from work (males only), and/or does not live with her mother (females only) (Table 2). It was also significantly associated with the expectation that in 3 years one will be living in a town (females only), married (both sexes) and/or have a baby (both sexes) (Table 2). All of these associations remained after adjustment for age.

Table 2.  Socio-demographic characteristics and reported vaginal intercourse in the 1998 Assisted Self-Completion Questionnaire Cohort, by the respondent's sex
VariablePer cent respondents had sex (no.)
TotalMaleFemale
Current age
 11 years30.0 (6/20)50.0 (3/6)21.4 (3/14)
 12 years24.7 (41/166)48.1 (25/52)14.0 (16/114)
 13 years25.1 (137/546)48.9 (86/176)13.8 (51/370)
 14 years30.8 (451/1466)52.3 (315/602)15.7 (136/864)
 15 years37.1 (543/1463)54.1 (373/690)22.0 (170/773)
 16 years45.3 (533/1177)56.6 (388/686)29.5 (145/491)
 17 years50.4 (247/490)52.9 (198/374)42.2 (49/116)
 ≥18 years59.2 (273/461)63.7 (249/391)34.3 (24/70)
 Any age38.5 (2231/5790)55.0 (1637/2977)21.1 (594/2813)
 Variable by sex (per cent)Per cent respondents had sex (no.)Crude ORAdjusted OR*95% CI
  1. OR, odds ratio; CI, confidence interval.

  2. * Adjusted for age.

Male
 Sometimes/often drinks alcoholYes (4.3)64.8 (83/128)1.541.571.08–2.28
No (95.7)54.5 (1548/2842)   
 Sometimes/often earns money from workYes (61.8)57.7 (1057/1833)1.331.321.14–1.53
No (38.2)50.5 (573/1135)   
 Expects will be married in 3 yearsYes (42.9)61.2 (624/1019)1.561.541.30–1.82
No (57.1)50.3 (682/1355)   
 Expects will have baby in 3 yearsYes (26.9)61.1 (379/620)1.401.371.13–1.65
No (73.1)53.0 (895/1689)   
Female
 Lives with motherYes (80.1)19.9 (446/2241)0.700.770.62–0.96
No (19.9)26.1 (145/556)   
 Sometimes/often drinks alcoholYes (6.4)30.0 (54/180)1.661.691.21–2.38
No (93.6)20.5 (537/2621)   
 Expects will be married in 3 yearsYes (34.1)31.6 (228/721)2.392.221.79–2.76
No (65.9)16.2 (226/1395)   
 Expects will have baby in 3 yearsYes (32.3)28.4 (183/645)1.981.851.47–2.32
No (67.7)16.6 (225/1352)   
 Expects will live in a town in 3 yearsYes (50.8)23.4 (239/1020)1.381.381.11–1.72
No (49.2)18.2 (180/989)   

The proportion of participants who reported having had sex increased with age (Table 2). Of those respondents who reported sex, 35.4% of males and 12.2% of females reported having first had vaginal intercourse at the age of 11 years or younger (Table 3). Most males and females consistently reported that their first sexual partner was about their own age in response to two separate questions about it. For respondents who reported sex, Table 3 summarizes characteristics of their first vaginal intercourse, and Table 4 summarizes their general sexual and reproductive experience.

Table 3.  Characteristics of first vaginal intercourse for those who reported it in the 1998 Assisted Self-Completion Questionnaire Cohort, by the respondent's sex
Variable*Per cent respondents (no.)
MaleFemaleTotal
  1. * Questions and answers are listed in the order used in the questionnaire; the last answer option, ‘I have never had sex’ is not included.

Age at first sexn = 1603n = 508n = 2111
 ≤11 years old35.4 (567)12.2 (62)29.8 (629)
 12 years old10.2 (163)10.6 (54)10.3 (217)
 13 years old10.9 (174)12.8 (65)11.3 (239)
 14 years old15.0 (241)22.0 (112)16.7 (353)
 15 years old12.3 (197)22.6 (115)14.8 (312)
 16 years old8.4 (134)12.8 (65)9.4 (199)
 17 years old4.4 (70)4.5 (23)4.4 (93)
 ≥18 years old3.6 (57)2.4 (12)3.3 (69)
Partner typen = 1674n = 630n = 2304
 Relative11.3 (189)14.1 (89)12.1 (278)
 Pupil in my school28.8 (482)36.4 (229)30.9 (711)
 Youth my age but not in my school51.3 (858)34.9 (220)46.8 (1078)
 Teacher0.8 (13)2.5 (16)1.3 (29)
 Someone older than me (not teacher)3.5 (58)5.2 (33)4.0 (91)
 Someone I did not know4.4 (74)6.8 (43)5.1 (117)
Seduction or forcen = 1643n = 611n = 2254
 I forced him/her4.5 (74)4.1 (25)4.4 (99)
 I seduced him/her55.6 (914)8.0 (49)42.7 (963)
 No force or seduction23.1 (380)15.6 (95)21.1 (475)
 I was seduced7.2 (118)46.5 (284)17.8 (402)
 I was forced9.6 (157)25.9 (158)14.0 (315)
Received giftn = 1644n = 620n = 2264
 He/she gave me a gift16.7 (274)74.7 (463)32.6 (737)
Gave giftn = 1656n = 616n = 2272
 I gave her her/him a gift42.8 (708)20.3 (125)36.7 (833)
Duration of first sexual relationshipn = 1648n = 603n = 2251
 1 day47.0 (774)50.9 (307)48.0 (1081)
 2–7 days24.5 (404)15.1 (91)22.0 (495)
 8–30 days8.5 (140)8.0 (47)8.3 (187)
 1–6 months9.8 (162)12.8 (77)10.6 (239)
 7 months or more6.3 (103)9.3 (56)7.1 (159)
Used condomn = 1498n = 518n = 2016
 Yes6.5 (98)6.8 (35)6.6 (133)
Table 4.  Sexual and reproductive experience for those who reported vaginal intercourse in the 1998 Assisted Self-Completion Questionnaire, by the respondent's sex
VariablePer cent respondents (no.)
MaleFemaleTotal
  1. * Questions and answers are listed in the order used in the questionnaire; the last answer option, ‘I have never had sex’ is not included.

Number of times had intercourse evern = 1464n = 584n = 2048
 Once29.6 (434)48.0 (280)34.9 (714)
 Twice20.7 (303)22.1 (129)21.1 (432)
 Three to four times26.8 (393)18.7 (109)24.5 (502)
 Five to nine times10.0 (147)6.2 (36)8.9 (183)
 Ten or more times12.8 (187)5.1 (30)10.6 (217)
Number of sex partners in last 4 weeksn = 1485n = 554n = 2039
 One30.7 (456)49.1 (272)35.7 (728)
 Two11.5 (171)7.6 (42)10.5 (213)
 Three3.2 (48)1.8 (10)2.8 (58)
 Four2.8 (41)0.5 (3)2.2 (44)
 Five or more1.3 (19)0.5 (3)1.1 (22)
 None50.5 (750)40.4 (224)47.8 (974)
Number of sex partners evern = 1432n = 583n = 2015
 One41.6 (596)69.8 (407)49.8 (1003)
 Two24.9 (356)17.8 (104)22.8 (460)
 Three11.7 (168)6.2 (36)10.1 (204)
 Four13.5 (193)4.3 (25)10.8 (218)
 Five to nine4.5 (64)1.2 (7)3.5 (71)
 Ten or more3.8 (55)0.7 (4)2.9 (59)
Number of current sex partnersn = 1515n = 553n = 2068
 One38.2 (578)44.1 (244)39.8 (822)
 Two15.1 (229)4.5 (25)12.3 (254)
 Three or more5.1 (77)2.0 (11)4.3 (88)
 None41.7 (631)49.4 (273)43.7 (904)
Ever used a condomn = 1480n = 506n = 1986
12.4 (184)13.8 (70)12.8 (254)
Ever caused pregnancy/been pregnantn = 1637n = 593n = 2230
1.9 (31)4.4 (26)2.6 (57)
Ever forced/was forced to have sexn = 1635n = 580n = 2215
9.2 (151)34.3 (199)15.8 (350)

Finally, 12.4% of males and 13.8% of females who reported sex said that they had used a condom in the past, while 6.5% and 6.8%, respectively, reported using one at first sex. Of respondents who reported condom use, 54.4% of males and 58.6% of females said they had only used a condom once. In addition, 9.2% of males who reported sex reported having forced a girl or woman to have sex, while 34.3% of females who reported sex reported they had been forced to have sex at some time in the past.

1998 ASCQ sexual and reproductive health knowledge and attitudes

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Overall, respondents knew most about condoms and STDs (30.0–48.4% correct answers), less about the appearance of an HIV-positive person (18.5–32.5% correct answers), and only 14.2% knew that a girl could become pregnant the first time that she has vaginal intercourse (Table 5). In the attitude section, the majority of respondents stated that forced sex is unacceptable (76.7%), while one-third to one-half of respondents gave desirable answers about sexual negotiation and refusal preparedness (43.1–53.6%), and ability to access, negotiate, and use condoms (33.6–52.0%). The least desirable reported attitudes had to do with gift-giving and sex, as only one-quarter of respondents (25.1–27.4%) believed a girl is not obliged to have sex if she has received a gift from a boy.

Table 5.  Sexual and reproductive health knowledge in the 1998 Assisted Self-Completion Questionnaire, by the respondent's sex
VariablePer cent correct responses (no.)P-value
MaleFemaleTotal
  1. * Question for which the correct response was a negative, middle answer option, rather than an affirmative, first answer option.

  2. † Figures in brackets, calculations for which ‘I do not know’ responses (a last answer option) were excluded.

  3. ‡ Question for which no significant difference was found when ‘I do not know’ responses were excluded.

  4. § Question for which significant difference was reversed when ‘I do not know’ responses were excluded, i.e., females knew correct answers significantly more often than males.

Do condoms have HIV in them before they are used?*65.8 [92.4]† (1945/2957)28.8 [89.0] (757/2625)48.4 [91.4] (2702/5582)<0.001 [0.003]
Can condoms protect you from HIV infection?65.0 [85.5] (1923/2960)26.1 [68.8] (682/2615)46.7 [80.4] (2605/5575)<0.001 [<0.001]
Are there diseases caused by sex that cause infertility?‡50.0 [77.3] (1480/2960)29.8 [78.4] (781/2620)40.5 [77.7] (2261/5580)<0.001 [0.488]
Can using condoms prevent HIV infection?57.3 [78.3] (1695/2959)20.5 [61.4] (535/2615)40.0 [73.5] (2230/5574)<0.001 [<0.001]
Is a condom a good way to prevent pregnancy?51.4 [73.2] (1519/2957)18.8 [54.8] (492/2616)36.1 [67.6] (2011/5573)<0.001 [<0.001]
If you look carefully, can you know if someone has HIV?*§40.9 [59.8] (1208/2956)23.0 [67.8] (601/2615)32.5 [62.2] (1809/5571)<0.001 [<0.001]
Is abnormal genital discharge a symptom of STDs?43.1 [72.1] (1270/2949)15.5 [53.1] (406/2614)30.0 [66.4] (1676/5563)<0.001 [<0.001]
Can a person who looks strong and healthy have HIV?§22.4 [32.5] (662/2960)14.1 [36.8] (368/2609)18.5 [33.9] (1030/5569)<0.001 [0.018]
Can a girl become pregnant the first time she has sex?§16.3 [23.9] (482/2955)11.9 [31.1] (311/2617)14.2 [26.3] (793/5572)<0.001 [<0.001]

Males were significantly more likely than females to answer each knowledge question correctly, in part because females were twice as likely as males to select the last answer option, ‘I do not know’ (Table 5). Similarly, for almost every knowledge question males and females who reported sex were significantly more likely to select the first answer, ‘Yes’, than those who did not report sex, and for almost all attitude statements there was evidence of answer order bias related to both the respondent's sex and/or reported vaginal intercourse.

Analysis of agreement between similar knowledge and attitude questions supports the possibility that answer order often contributed to significant differences. For the two pairs of questions that required the same order of answer to be consistent, agreement was high (82.5% and 71.6%, respectively). However, for the one pair of questions that required a different answer order to be consistent, agreement was relatively low (48.8%).

Comparison of 1998 ASCQ and FFQ results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

When 1998 ASCQ answer order exclusion criteria were applied to the 1998 ASCQ-FFQ Cohort, 154 individuals (3.1%) had results partially or entirely excluded.

Table 6 shows the degree of response agreement for the 17 questions that were asked in the same or very similar form in both the 1998 ASCQ and the 1998 FFQ. Notably, ASCQ respondents heard and/or read the list of possible answers for each of these questions before answering, whereas FFQ respondents should have instead coded the answer that was provided spontaneously by the respondent. For almost all of these questions, the proportion of affirmative agreements (Ppos) was lower than the proportion of negative agreements (Pneg), particularly in response to questions about sexual behaviour. In addition, for almost all sexual behaviour questions ASCQ respondents reported significantly more sexual experiences than FFQ respondents. In most cases this coincided with an affirmative, first answer, but in a few cases it did not.

Table 6.  Agreement between 1998 Assisted Self-Completion Questionnaire – Face-to-face Questionnaire Cohort responses to questions asked in both questionnaires
VariableNo.Per cent ‘Yes’PposPnegPer cent agreement
ASCQFFQ
  1. ASCQ, assisted self-completion questionnaire; FFQ, face-to-face questionnaire; Ppos, proportion of affirmative agreements; Pneg, proportion of negative agreements; n.a., not applicable.

  2. * Figures in brackets, calculations for which ‘I do not know’ responses were excluded.

  3. † Statement for which the desirable response was a negative, last answer option, rather than an affirmative, first answer option.

  4. ‡ Results for 656 respondents were not analysed because they did not report having had sex, and FFQ interviewers mis-coded their responses as ‘not applicable’.

  5. § More than one answer may have been valid, and questions were not exactly the same.

  6. ¶ In the ASCQ, condom use, forced sex and pregnancy questions were asked of all respondents, not only those who reported sex, with the answer options being ‘Yes’ and ‘No’ only.

All respondents
 Ethnic group4753n.a.n.a.n.a.n.a.94.6
 Religion4761n.a.n.a.n.a.n.a.91.0
 Ever had sex473940.338.371.881.777.8
 Can a girl get pregnant the first time she has sex?4566 [2470]*14.9 [26.6]31.4 [31.3]32.4 [47.8]79.7 [78.7]68.7 [69.8]
 Can a healthy and strong looking person have HIV?4567 [2461]19.4 [34.8]39.7 [41.4]41.2 [59.3]75.3 [74.9]65.3 [69.0]
 If a girl accepts a gift she must have sex†4568 [2954]39.7 [60.9]66.6 [71.4]38.6 [74.6]75.4 [50.4]64.9 [66.4]
 In the future I will use condoms if I have sex‡3907 [3179]45.3 [46.5]62.2 [64.4]67.1 [66.6]61.8 [58.4]64.6 [62.9]
Those who had sex only
 Age at first sex1280n.a.n.a.n.a.n.a.64.4
 No. partners ever1220n.a.n.a.n.a.n.a.47.3
 First partner type§1289n.a.n.a.n.a.n.a.55.9
 Last partner type§1267n.a.n.a.n.a.n.a.54.5
 Sex in last 4 weeks122951.127.448.766.851.8
 Ever used condom¶4451 5.4 1.830.797.495.0
 Used condom first sex¶4719 2.5 1.123.098.697.2
 Used condom last sex¶4726 2.5 1.310.398.597.0
 Ever forced sex (M)¶2572 5.8<0.1 0.097.094.1
 Ever forced to have sex (F)¶216512.3 0.2<0.193.387.5
 Ever been pregnant (F)¶2177 1.4 0.2 5.699.298.4

Agreement was high for demographic information such as ethnic group (94.6%) and religion (91.0%). There was only 77.8% agreement between the surveys for the question that asked respondents whether they had ever had vaginal intercourse. In the ASCQ survey, 40.3% of respondents (55.6% of males and 22.2% of females) reported vaginal intercourse, while 38.3% reported it in the FFQ (51.6% of males and 22.6% of females). However, while overall proportions reporting sexual experience were similar in the two surveys, 37.9% of males and 59.2% of females who reported sex only did so in one of the two questionnaires (Table 7). Males were significantly more likely to report sex in the ASCQ but not the FFQ, than reporting sex in the FFQ but not the ASCQ (P < 0.001). No such significant patterns were found for females.

Table 7.  Reliability of reported vaginal intercourse and number of partners in last 4 weeks, by sub-cohort and the respondent's sex
Have you ever had vaginal intercourse?Per cent respondents (no.)
1998 ASCQ2000 ASCQMale n = 2137Female n = 1821Total n = 3958
YesYes35.8 (766)14.4 (262)26.0 (1028)
YesNo20.2 (432)5.2 (94)13.3 (526)
NoYes13.9 (296)20.0 (364)16.7 (660)
NoNo30.1 (643)60.5 (1101)44.1 (1744)
1998 ASCQ1998 FFQMale (n = 2572)Female (n = 2166)Total (n = 4739)
YesYes41.0 (1056)13.0 (281)28.2 (1337)
YesNo14.5 (374)9.2 (199)12.1 (573)
NoYes10.5 (271)9.6 (208)10.1 (479)
NoNo33.9 (872)68.2 (1478)49.6 (2350)
Number of partners in last 4 weeksMaleFemaleTotal
1998 ASCQ (n = 1295)1998 FFQ (n = 1343)1998 ASCQ (n = 449)1998 FFQ (n = 502)1998 ASCQ (n = 1744)1998 FFQ (n = 1845)
  1. ASCQ, assisted self-completion questionnaire; FFQ, face-to-face questionnaire.

None50.8 (658)76.6 (1029)41.0 (184)61.0 (306)48.3 (842)72.4 (1335)
One30.7 (397)15.6 (209)48.3 (217)32.7 (164)35.2 (614)20.2 (373)
Two11.8 (153)6.2 (83)7.8 (35)5.4 (27)10.8 (188)6.0 (110)
Three2.6 (34)1.1 (15)2.0 (9)0.6 (3)2.5 (43)1.0 (18)
Four2.7 (35)0.2 (3)0.5 (2)0.4 (2)2.1 (37)0.3 (5)
Five or more1.4 (18)0.3 (4)0.5 (2)0 (0)1.2 (20)0.2 (4)

Between the 1998 ASCQ and FFQ surveys, overall agreement for reported age at first sex and total number of partners ever was 64.4% and 47.3%, respectively. While there was high overall agreement in the answers for most questions about pregnancy, forced sex, and condom use, the proportion of positive agreements was very low for those questions. For example, only one (<0.1%) of all female respondents reported that she had been pregnant in both questionnaires, but an additional 1.4% reported pregnancy in the ASCQ only, and 0.2% reported it in the FFQ only. Similarly, no males reported that they had ever forced a girl or woman to have sex in both questionnaires, but 5.8% did in the ASCQ only, and <0.1% did in the FFQ only. Finally, only one (<0.1%) female respondent reported that she had been forced to have sex in both questionnaires, but 12.3% reported it in the ASCQ only, and 0.2% reported it in the FFQ only.

Agreement for responses to questions about the characteristics of the first sex partner was also only fair. However, within each questionnaire response options for this question were not mutually exclusive, and between questionnaires not all of the terms were exactly the same. In the ASCQ survey, 4.4% of respondents reported that their first sex partner was a stranger, compared with only 0.2% of FFQ respondents (P < 0.001), and 0.7% of ASCQ respondents reported their first partner was a teacher, while none reported this in the FFQ. Neither the stranger nor the teacher option was a first or a last answer option.

Of those respondents who reported sex in both questionnaires, 51.7% reported that they had had sex in the last 4 weeks in the ASCQ, compared with only 27.6% in the FFQ (P < 0.001) (Table 7). This discrepancy does not simply relate to answer order bias, as the higher reporting in the ASCQ involved other answer options as well, e.g. 16.5% of respondents reported more than one sex partner in the last 4 weeks in the ASCQ (neither the first nor the last answer option), compared with only 7.4% in the FFQ (P < 0.001).

Finally, ‘I do not know’ responses were much more common in the 1998 ASCQ than in the 1998 FFQ. For example, for four knowledge or attitude statements that were included in both questionnaires, 22.1–44.8% of respondents reported ‘I do not know’ in the ASCQ survey, compared with only 1.6–5.2% of FFQ respondents. However, both before and after exclusion of ‘I do not know’ responses, agreement ranged from 62.9 to 69.8% for responses to these questions (Table 6). For all four questions, ASCQ respondents were significantly more likely to respond with a negative response than FFQ respondents, and thus provided a significantly more desirable answer each time a negative response was more desirable, and vice versa (P < 0.001).

Comparison of 1998 and 2000 ASCQ results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Of the 80 individuals who identified themselves as illiterate during 2000 ASCQ survey registration, only eight (10.0%) had had results included in the 1998 ASCQ analysis, as the other 72 individuals either did not attend in 1998, or their results were excluded during one of the screening procedures prior to analysis.

A total of 410 (9.3%) of the 4424, 1998–2000 ASCQ Cohort members had results entirely or partially excluded from one or both questionnaires because they entirely selected a first or a last answer in sections for which this was inconsistent. 10.7% of those who were excluded had results excluded from both questionnaires, while 68.8% had results excluded from the 1998 ASCQ only, and 20.5% had results excluded from the 2000 ASCQ only. In both questionnaires exactly 81.8% of those excluded were female.

In the post-exclusion 1998–2000 ASCQ data, there was substantial inconsistency in sexual behaviour reports between the two surveys. For example, 13.1% of respondents reported that they had had sex in the 1998 survey, but denied it in the 2000 survey (Table 7). This inconsistency was greater among males (20.2%) than females (5.2%).

The reliability of other responses to sexual behaviour questions asked in both ASCQ surveys was affected by these individuals’ responses, but inconsistencies remained even when their results were excluded from analysis. For example, of those respondents who reported sex in both the 1998 and the 2000 surveys, 42.1% reported an older age at first sex in the 2000 survey, while 17.8% reported a younger age. In addition, 32.1% reported having had a lower number of total partners, 25.2% reported having had sex fewer times than originally reported, and 61.9% of those who reported having used a condom in 1998 reported never having used one in 2000. Males were generally twice as likely as females to be inconsistent in response to most of these questions, with the exception of reported age at first sex.

Plausibility of 1998 ASCQ results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Reported socio-demographic characteristics and future expectations seem plausible. For example, reported religion and ethnicity were similar to the findings of prior face-to-face questionnaire surveys in Mwanza Region (Munguti et al. 1997; Boerma et al. 2002), and expectations of circumstances in 3 years associated with vaginal intercourse were plausible, such as marrying, having children, working for money, and living in a town. For females, such associations were found at both younger and older ages, and may relate to qualitative research findings that, during teenage years, ‘adult’ status may reflect social role and responsibility more than chronological age itself. Alcohol use was rarely reported, but was nonetheless associated with reported sex for both males and females, which also seems plausible.

Some of the sexual and reproductive knowledge, attitudes and behaviour results also seem plausible, such as attitudes that may be desirable (e.g. forced sex as unacceptable) or undesirable (e.g. rarely intending to use a condom, or gift-giving and sex). In addition, the common exchange of a gift or money for sex, the high proportion of one-day sexual relationships, and the small proportion of first partners who were reported to be teachers, are all sensitive findings that are well supported by HALIRA qualitative research results (Mshana et al. 2002). Three-quarters of the females who had had vaginal intercourse (74.7%) reported receiving a gift or money from their partner at their first vaginal intercourse, but less than half of males (42.8%) reported giving a gift or money. This may reflect differences in sexual partner type and age, if some school girls have older sexual partners who can afford to give them gifts, while some school boys cannot afford to provide gifts. The norm of money or a gift for sex could also explain why male respondents who earned money from work were significantly more likely to report having had sex, regardless of age (Mshana et al. 2002).

However, many of the sexual behaviour results were problematic, and it is difficult to interpret the knowledge and attitude data by the sex of the respondent or reported vaginal intercourse because of answer order bias, even after exclusion of the most extreme examples. When the prevalence of reported vaginal intercourse in the ASCQ is compared with HALIRA qualitative results, other SCQ survey results (Matasha et al. 1998) and other FFQ survey results (Konings et al. 1994; Munguti et al. 1997; Boerma et al. 2002), ASCQ reports for adolescent males are similar or relatively high, while those for adolescent females are similar or relatively low. Authors of some of these studies acknowledge possible problems with data validity, such as poor question comprehension among primary school pupils, or low accuracy of recall among adults reporting age at first sex. However, a strong possibility remains that males may have over-reported sexual activity in the ASCQ, while females under-reported it.

A large proportion of the reported ages at first sex also seem implausibly low, particularly for boys. Researchers have found that respondents sometimes have difficulty accurately averaging numbers or estimating age in cultures where people have low basic math skills and rarely think about specific ages (Gil & Omaboe 2000; Stycos 2000; Zarkovich 2000). To recall an earlier age thus may be particularly difficult, and this may be further confounded by illiteracy and problems with writing numbers. In addition, social desirability bias for males (e.g. wanting to claim greater sexual experience), or misunderstanding of the terminology (e.g. believing the euphemism ‘to make love’ included non-penetrative sexual play, imitation, or experimentation) may have contributed to inaccurately low reported ages at first vaginal intercourse. However, it is possible that a proportion of the reports of first sex having occurred at very low ages were accurate, e.g. representing childhood sexual abuse, but if so it is difficult to interpret its extent from the results. Finally, it is possible that difficulty summing up and averaging experiences contributed to the 1.6–4.8% of respondents who inconsistently reported a higher total number of sexual partners than sexual encounters.

More than half of respondents who reported sex in the ASCQ also reported having had it within the last 4 weeks, which seems remarkably high considering that one-third of males and one-half of females reported that they had only had sex once ever. This suggests many respondents had either become sexually active recently, had difficulty estimating time accurately, and/or had under-reported the number of times they had ever had vaginal intercourse earlier in the questionnaire. The latter possibility seems particularly likely given school pupils are forbidden to have sex, and could be severely punished for it if caught by authorities or parents (Wight et al., submitted).

However, while sexual behaviour may have been under-reported in general, particularly for girls, there is reason to believe that condom use may have been over-reported. HALIRA qualitative research has found that negative beliefs about condoms are widespread in rural areas, and condoms are also very difficult to access. Thus the likelihood of 6.6% of sexually active adolescents truly having used condoms at first sexual intercourse – or 12.8% having ever used them – seems remarkably high. Such possible over-reporting may partially result from social desirability bias, e.g. boys trying to exaggerate their sexual experience, or young people trying to report behaviours they believe would please the researcher. However, affirmative response to the ‘condom use’ question was a first answer option, and ‘condom’ may additionally have been an unfamiliar term for some respondents, so non-excluded first answer bias and misunderstanding may also have contributed to implausibly high reports. These issues may have also contributed to the relatively high inconsistency (6.1–6.8%) seen in responses to condom use questions that were repeated in the questionnaire.

Finally, 34.3% of females who reported sex reported that they had been forced to have sex at some time in the past, while 9.2% of males who reported sex also reported having forced someone to have sex. This raises the concern that non-consensual or undesired sex is common for adolescents in rural Mwanza. These results are supported by HALIRA qualitative research findings that approximately one-quarter of young women have been pressured into sex after accepting a gift, and one-half have been pressured into it by intermediaries (Plummer et al. 2002). However, 9.6% of males and 4.1% of females reported that the female forced the male to have sex during their first sexual encounter (Table 3), suggesting that the term ‘force’ was also ambiguous, and this ambiguity is also supported by qualitative research results. For example, HALIRA findings suggest it is not unusual for a girl to affect pretence of refusal at the onset of sexual negotiation or during a sexual encounter itself, in order to appear chaste and/or to have more power in negotiating a gift.

Survey process and reporting patterns

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

In both 1998 and 2000, the vast majority of respondents did not appear to have difficulty completing the ASCQ and often provided plausible answers. In addition, the registration and analysis screening mechanisms used to exclude respondents who either could not or would not answer the questions in a logical and consistent way seem to have been partially effective, increasingly so in the second survey. Although missing answers are a common difficulty in self-completion questionnaire surveys (Hingson & Strunin 1993), they were not a substantial problem in these surveys as only a very small proportion of respondents provided no answer or multiple answers. A somewhat larger portion of respondents provided inconsistent responses for sexual behaviour questions, possibly because of limited ability to conceptualize, recall or honestly report those activities.

The concern remains that only the most extreme examples of answer order bias were identified and excluded from the final analysis of the ASCQ data. Other respondents may have had a similar though less consistent pattern, e.g. selecting most but not all of the first or last answers within one or more sections. In both 1998 and 2000, the small percentage of respondents who demonstrated an extreme answer order bias were disproportionately female, of a lower education level, and/or more geographically remote. Each of these factors is likely to be associated with relatively low literacy in Mwanza, so answer order bias may also have stemmed from poor comprehension and/or ability to complete the questionnaire correctly. Notably, this problem was not as great in 2000, suggesting that answer order bias may have been reduced by greater education or maturity.

In addition, substantially increased last answer reporting in the final sections of the questionnaire suggests that the questionnaire's length and/or prior content contributed to bias. Process evaluation during the ASCQ surveys found that some respondents were tired, bored or frustrated by the time they reached the knowledge section 1–1.5 h into the questionnaire. This may have been a greater problem for those respondents who were fluent in Swahili and did not feel a need to hear a Sukuma translation read aloud, and/or for those respondents who did not report sex and therefore had to select the last response (‘I have never made love’) every time in the preceding section. HALIRA qualitative research has confirmed the 1998 ASCQ pilot study findings that many primary school pupils in rural Mwanza are fluent in Sukuma but have a poor command of Swahili. Nonetheless, it is difficult to know whether improved data validity among such individuals is outweighed by poorer data validity among the fluent Swahili speakers, who were required to listen to each question in Sukuma and Swahili. In addition, while some respondents who did not report sex may have disengaged because of frustration or boredom, some may have been less confident of their opinions when they reached the sexual and reproductive health knowledge and attitude sections, because they were sexually inexperienced.

The high rate of ‘I do not know’ responses reflects a well-known weakness of SCQs, i.e. in the absence of a face-to-face interviewer probing or prompting for responses, participants may feel unmotivated to reflect on a question. However, some of the ‘I do not know’ responses reported only in the ASCQ may have been valid (i.e. the respondent truly did not know or was ambivalent), but respondents may have provided a different answer in the FFQ interview for fear of appearing ignorant, uncooperative or discourteous to the interviewer (Stycos 2000). In addition, the disproportionately high ‘I do not know’ responses among female respondents may represent limited education or confidence in expressing their opinions relative to male respondents.

Finally, for four knowledge and attitude questions, 1998 ASCQ respondents were significantly more likely than 1998 FFQ respondents to select negative answers that were placed either in the middle or at the end of the answer series. While this may to an extent have related to last answer bias, it is also possible that some respondents had affirmative and/or acquiescent biases in face-to-face interviews. For example, respondents may provide affirmative or acquiescent responses for motivational or cognitive reasons, such as a desire to be agreeable (Couch & Keniston 1961; Hayes & Baker 1998) or limited flexibility in searching for contraindicating information (Knowles & Nathan 1997). Either of these factors may have been exacerbated in this face-to-face questionnaire survey, given the direct personal contact, formal school setting, limited time for reflection and very low education level of participants (Heaven 1983).

Data reliability and sensitivity

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

Reliable results do not guarantee that data are valid, as respondents can reliably and consistently lie or misreport for other reasons. However, unreliable cross-sectional or longitudinal results clearly indicate that at least one of two reports was invalid, as was the case for those individuals who reported an experience in one of the 1998 surveys but not the other, or who reported less cumulative experience of an activity in the 2000 ASCQ than in the 1998 ASCQ. For example, the high proportions of respondents who unreliably reported vaginal intercourse raises questions about the degree of validity of sexual behaviour data collected in any of the quantitative surveys. The 1998 surveys were 5 weeks apart, so some inconsistencies can be expected, but not at the scale observed.

HALIRA qualitative research has found that the vast majority of 15-year-old adolescents in rural Mwanza have already experienced vaginal intercourse, and that some adolescents do not feel a need to hide this because they have already left school, are supporting themselves, are living independently, have already married, and/or have had children. In contrast, if an adolescent or young adult is still in school, social norms dictate that they should have no sexual activity regardless of age, and punishment for school pupils who are found to have had sex can include beatings, fines, or expulsion, particularly if a girl is pregnant. As a result, adolescents take great pains to conceal their sexual behaviour from parents and teachers, and to a lesser extent, from peers. Both the ASCQ and the FFQ surveys were conducted in school settings, and despite assurances of confidentiality, it is quite possible that these restrictive norms and actual practices made many pupils dishonestly report their behaviour in either or both of the surveys (Cowan et al. 2002). School-related inhibitions may have been reduced in the 2000 ASCQ survey, however, as approximately half of the participants had already left school by that time. In addition, greater education and experience in 2000 may have contributed to more valid reporting for people who had problems with low literacy in 1998, a possibility supported by findings from other SCQ surveys, in which young people in their late teens generally provided more reliable and accurate sexual behaviour reports than those in their early to mid-teens (Matasha et al. 1998; Siegel et al. 1998).

Males were twice as likely as females to be inconsistent in reporting sexual behaviour in the two ASCQ surveys, and this may relate to the fact that relatively few female respondents reported sex in 1998, and thus relatively few had the potential to be inconsistent in 2000. However, high rates of inconsistency for males may also relate to HALIRA findings that male adolescents sometimes exaggerate their sexual experience in trying to impress one another, and young men who over-reported sexual experience in 1998 may have chosen to report it more honestly with greater maturity in 2000. Nonetheless, while the above interpretation suggests that the 2000 ASCQ data are more valid than the 1998 ASCQ data, it is also possible that some later reports were less accurate, as studies have found that error may increase with the period of recall (Ross & Vaughan 1986; Wellings et al. 1990; Kauth et al. 1991; Morris 1993).

Data on age at first sex were also unreliable in the longitudinal studies, and two aspects of this are noteworthy: (i) respondents tended to increase their estimated age at first sex rather than decrease it and (ii) it is the only example of longitudinal sexual behaviour data in which female reports were even less consistent than male reports. It is possible that a combination of factors contributed to inconsistencies, such as social desirability bias, poor recall and difficulty with basic maths and age estimation.

Finally, there was evidence to suggest that the ASCQ survey may have been more effective than the FFQ survey in collecting data on sensitive behaviours. Significantly higher reporting of more than one sex partner in the last 4 weeks is one example, as are reports of first partners being a teacher or a stranger; none of these were first or last answer options. Qualitative research in the region supports these ASCQ findings (Schapink et al. 1997). For example, in HALIRA in-depth interviews with 93 youths, several adolescents reported that they had first had sex with someone they met on a school sports trip, or when propositioned by a stranger while on an errand. Similarly, three annual HALIRA participant observation visits to each of four villages in the study area found that sexual relationships between male teachers and female school pupils were widely reported to have occurred in three of the villages.

Conclusion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

The vast majority of this adolescent, rural African study population seemed to use the assisted self-completion questionnaire survey method without difficulty, despite low literacy levels. Most of the results of this study were plausible, and this method was more effective in collecting particularly sensitive information than the FFQ survey method. However, the ASCQ method was not without problems, including substantial answer order bias, high proportions of ‘I do not know’ responses, inappropriateness for individuals with very low literacy levels, and complicated analysis. In addition, careful scrutiny and comparison of 1998 ASCQ sexual behaviour data with both contemporaneous FFQ data and longitudinal ASCQ data revealed problems with reliability and validity. Notably, rather than the problems these detailed comparisons revealed being unique to this study, similar problems may have existed but have been undiscovered or ignored in many other SCQ and FFQ survey reports.

The ASCQ method in this rural African setting has the greatest potential with older, male and/or relatively educated adolescents. Recommendations for future ASCQ surveys in this context include using shorter questionnaires, random allocation of answer order, and including as many internal consistency check questions as is feasible. Ideally, a modification would be developed that does not require individuals to participate in questionnaire sections that are not necessary for them, i.e. Sukuma translations for Swahili speakers, or questions about sexual behaviour for those who do not report sex. Finally, any use of this method should involve careful screening and exclusion of individuals who cannot or do not complete the questionnaire correctly, both before administration and during analysis of data.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References

This was a collaborative study involving the Tanzanian National Institute for Medical Research (NIMR), the African Medical and Research Foundation (AMREF), the London School of Hygiene and Tropical Medicine (LSHTM), the UK Medical Research Council's Social and Public Health Sciences Unit, and the Tanzanian Ministries of Health and of Education and Culture. We are grateful to Mwanza regional, ward, village and sub-village government, school and health authorities for their assistance. Special thanks are also due to the MEMA kwa Vijana Project field research, data entry and administrative staff, who worked very hard to ensure the surveys were completed successfully, and also to the HALIRA Programme researchers, whose qualitative results were useful in survey data interpretation. Finally, we thank the young people of Mwanza Region, whose helpful participation made this study possible. The research reported here was primarily funded by the UK Medical Research Council and the European Commission, with additional support provided by the UK Department for International Development.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. 1998 assisted self-completion questionnaire (ASCQ) survey
  6. 1998 face-to-face questionnaire (FFQ) survey
  7. 2000 ASCQ survey
  8. Data processing, analysis and interpretation
  9. Results
  10. 1998 ASCQ reporting patterns
  11. 1998 ASCQ socio-demographic characteristics and future expectations
  12. 1998 ASCQ sexual behaviour
  13. 1998 ASCQ sexual and reproductive health knowledge and attitudes
  14. Comparison of 1998 ASCQ and FFQ results
  15. Comparison of 1998 and 2000 ASCQ results
  16. Discussion
  17. Plausibility of 1998 ASCQ results
  18. Survey process and reporting patterns
  19. Data reliability and sensitivity
  20. Conclusion
  21. Acknowledgements
  22. References
  • Ankrah EM (1989) AIDS: methodological problems in studying its prevention and spread. Social Science and Medicine 29, 265276.
  • Blanc AK & Way AA (1998) Sexual behaviour and contraceptive knowledge and use among adolescents in developing countries. Studies in Family Planning 29, 106116.
  • Bleek W (1987) Lying informants: a fieldwork experience from Ghana. Population and Development Review 13, 314322.
  • Boerma JT, Urassa M, Nnko S et al. (2002) Sociodemographic context of the AIDS epidemic in a rural area in Tanzania with a focus on people's mobility and marriage. Sexually Transmitted Infections 78(Suppl), i97i105.
  • Catania JA, McDermott LJ & Pollack LM (1986) Questionnaire response bias and face-to-face interview sample bias in sexuality research. Journal of Sex Research 22, 5272.
  • Catania JA, Gibson DR, Chitwood DD & Coates TJ (1990) Methodological problems in AIDS behavioral research: influences on measurement error and participation bias in studies of sexual behaviour. Psychological Bulletin 108, 339362.
  • Cleophas-Frisch B, Obasi A, Ross D & Gavyole A (1999) Design and evaluation of an intervention to improve adolescent access to reproductive health services in rural Tanzania. 13th Meeting of the International Society for Sexually Transmitted Diseases Research, Denver, USA, 11–14 July 1999. Abstract no. 139.
  • Copas AJ, Wellings K, Erens B et al. (2002) The accuracy of reported sensitive sexual behaviour in Britain: exploring the extent of change 1990–2000. Sexually Transmitted Infections 78, 2630.
  • Couch A & Keniston K (1961) Agreeing response set and social desirability. Journal of Abnormal and Social Psychology 62, 175179.
  • Couper MP & Stinson LL (1999) Completion of self-administered questionnaires in a sex survey. Journal of Sex Research 36, 321330.
  • Cowan FM, Langhaug LF, Mashungupa GP et al. (2002) School based HIV prevention in Zimbabwe: feasibility and acceptability of evaluation trials using biological outcomes. AIDS 16, 16731678.
  • Davoli M, Perucci CA, Sangalli M et al. (1992) Reliability of sexual behavior data among high school students in Rome. Epidemiology 3, 531535.
  • Denzin NK & Lincoln YS (2000) Handbook of Qualitative Research. Sage, Thousand Oaks.
  • Gil B & Omaboe EN (2000) Population censuses and national sample surveys in developing countries. In: Social Research in Developing Countries: Surveys and Censuses in the Third World (eds MBulmer & DPWarwick) UCL Press, London, pp. 4151.
  • Gleason HA (1961) The role of tone in the structure of S?k?´ma [Book Review]. Language 37, 294308.
  • Hayes RP & Baker DW (1998) Methodological problems in comparing English-speaking and Spanish-speaking patients’ satisfaction with interpersonal aspects of care. Medical Care 36, 230236.
  • Heaven PC (1983) Authoritarianism or acquiescence? South African findings. Journal of Social Psychology 119, 1115.
  • Hingson R & Strunin L (1993) Validity, reliability, and generalizability in studies of AIDS knowledge, attitudes and behavioral risks based on subject self-report (Commentary). American Journal of Preventive Medicine 9, 6264.
  • James NJ, Bignell CJ & Gillies PA. (1991) The reliability of self-reported sexual behavior. AIDS 5, 333336.
  • Kauth MR, St Lawrence JS & Kelly JA (1991) Reliability of retrospective assessments of sexual HIV risk behavior: a comparison of biweekly, three-month, and twelve-month self-reports. AIDS Education and Prevention 3, 207214.
  • Kinsman J, Harrison S, Kengeya-Kayondo J, Kanyesigye E, Musoke S & Whitworth J (1999) Implementation of a comprehensive AIDS education programme for schools in Masaka District, Uganda. AIDS Care 11, 591601.
  • Klepp K, Ndeki SS, Leshabari MT, Hannan PJ & Lyimo BA (1997) AIDS education in Tanzania: promoting risk reduction among primary school children. American Journal of Public Health 87, 19311936.
  • Knowles ES & Nathan KT (1997) Acquiescent responding in self-reports: cognitive style or social concern? Journal of Research in Personality 31, 293301.
  • Konde-Lule JK, Sewenkambo N & Morris M (1997) Adolescent sexual networking and HIV transmission in rural Uganda. Health Transition Review 7(Suppl.), 89100.
  • Konings E, Blattner WA, Levin A et al. (1994) Sexual behaviour survey in a rural area of northwest Tanzania. AIDS 8, 987993.
  • Kuhn L, Steinberg M & Mathews C (1994) Participation of the school community in AIDS education: an evaluation of a high school programme in South Africa. AIDS Care 6, 161171.
  • Kuleana (1999) The State of Education in Tanzania: Crisis and Opportunity. Kuleana, Mwanza, Tanzania.
  • Landis JR & Koch GG (1977) The measurement of observer agreement for categorical data. Biometrics 33, 159174.
  • Liau A, DiClemente RJ, Wingood GM et al. (2002) Associations between biologically confirmed marijuana use and laboratory-confirmed sexually transmitted diseases among African American adolescent females. Sexually Transmitted Diseases 29, 387390.
  • MacLachlan M, Chimombo M & Mpemba N (1997) AIDS education for youth through active learning: a school-based approach from Malawi. International Journal of Educational Development 17, 4150.
  • Matasha E, Ntembelea T, Mayaud P et al. (1998) Sexual and reproductive health among primary and secondary school pupils in Mwanza, Tanzania: need for intervention. AIDS Care 10, 571582.
  • McFarlane M & St Lawrence JS (1999) Adolescents’ recall of sexual behavior: consistency of self-report and effect of variations in recall duration. Journal of Adolescent Health 25, 199206.
  • Meekers D (1994) Sexual initiation and premarital childbearing in sub-Saharan Africa. Population Studies 48, 4764.
  • Morris M (1993) Telling tails explain the discrepancy in sexual partner reports. Nature 365, 437440.
  • Mshana G, Wamoyi J, Plummer M, Wight D, Ross D & Salamba Z. (2002) Sexual negotiation and decision-making among young people in rural Mwanza, Tanzania. 14th International AIDS Conference, Barcelona, Spain, 8–11 July 2002. Abstract no. E12174.
  • Munguti K, Grosskurth H, Newell J et al. (1997) Patterns of sexual behaviour in a rural population in north-western Tanzania. Social Science and Medicine 44, 15531561.
  • Nachman SR (1984) Lies my informant told me. Journal of Anthropological Research 40, 536555.
  • Pillai VK & Benefo K (1995) Teenagers in Zambia: sexual activity among school-going females. International Journal of Contemporary Sociology 32, 125132.
  • Plummer M, Wight D, Ross D et al. (1999) Measurement of the impact of a sexual behaviour intervention: an attempt to separate fact from fantasy. 13th Meeting of the International Society for Sexually Transmitted Diseases Research, Denver, USA. 11–14 July 1999. Abstract no. 244.
  • Plummer M, Wight D, Wamoyi J, Mshana G, Ross D & Salamba Z (2002) Sexual violence, pressure and HIV in rural Mwanza, Tanzania. 14th International AIDS Conference, Barcelona, Spain. 8–11 July 2002. Abstract no. WeOrE1282.
  • Rosen CS, Henson BR, Finney JW & Moos RH (2000) Consistency of self-administered and interview-based Addiction Severity Index composite scores. Addiction 95, 419425.
  • Ross DA & Vaughan JP (1986) Health interview surveys in developing countries: a methodological review. Studies in Family Planning 17, 7894.
  • Ross D, Obasi A, Changalucha J & Moos RH (1999) Community-randomised trial of an intervention to prevent HIV infection and enhance reproductive health among adolescents in rural Tanzania. 11th International Conference on AIDS and STDs in Africa, Lusaka, Zambia. 12–16 September 1999. Abstract no. 15PT38–14: 129.
  • Schapink D, Hema J & Mujaya B (1997) Youth and HIV/AIDS programmes. In: HIV Prevention and AIDS Care in Africa: A District Level Approach (eds JNg'weshemi, JTBoerma, JBennett & DSchapink) Royal Tropical Institute, Amsterdam, pp. 163184.
  • Siegel S & Castellan NJ (1998) Nonparametric Statistics for the Behavioural Sciences, 2nd Edn. McGraw-Hill, New York.
  • Siegel DM, Aten MJ & Roghmann KJ (1998) Self-reported honesty among middle and high school students responding to a sexual behaviour questionnaire. Journal of Adolescent Health 23, 2028.
  • Singh S (1998) Adolescent childbearing in developing countries: a global review. Studies in Family Planning 29, 117136.
  • Smith PG & Morrow RH (1996) Field Trials of Health Interventions in Developing Countries: a Toolbox. Macmillan, Oxford.
  • Stycos JM (2000) Sample surveys for social science in underdeveloped areas. In: Social Research in Developing Countries: Surveys and Censuses in the Third World (eds MBulmer & DPWarwick) UCL Press, London, pp. 5364.
  • UNAIDS (2000) Report on the Global HIV/AIDS Epidemic. UNAIDS, Geneva.
  • UNICEF (1998) The State of the World's Children. Oxford University Press, Oxford.
  • Wellings K, Field J, Wadsworth J, Johnson AM, Anderson RM & Bradshaw SA (1990) Sexual lifestyles under scrutiny (Commentary). Nature 348, 276278.
  • West P (1990) The status and validity of accounts obtained at interview: a contrast between two studies of families with a disabled child. Social Science & Medicine 30, 12291239.
  • Wight D & West P (1999) Poor recall, misunderstandings and embarrassment: interpreting discrepancies in young men's reported heterosexual behaviour. Culture, Health and Sexuality 1, 5578.
  • Wight D, Plummer M, Mshana G, Wamoyi J & Ross D. Divergent sexual norms for young people in northern Tanzania. 12th International Conference on AIDS/STD in Africa, Ouagadougou, Burkina Faso, 9–13.12.01. Abstract Book p. 301, no. 12PT6–526.
  • Wolff B, Tollman S & Collinson M (2000) Improved self reporting of HIV risk behaviours through self-administered forms for rural, low literacy populations. 13th International AIDS Conference, Durban, South Africa. 9–14 September 2000. Abstract no. ThPeD5666: 492.
  • Zabin LS & Kiragu K (1998) The health consequences of adolescent fertility behaviour in sub-Saharan Africa. Studies in Family Planning 29, 210232.
  • Zarkovich SS (2000) Some problems of sampling work in underdeveloped countries. In: Social Research in Developing Countries: Surveys and Censuses in the Third World (eds MBulmer & DPWarwick) UCL Press, London, pp. 101108.