Assessment of non-marital sexual behaviours of men in Bangladesh: a methodological experiment using a modified confidential ballot-box method

Abstract

This study assessed the effectiveness of a modified ballot-box method (MBBM) in eliciting non-marital sexual behaviours compared with face-to-face interview (FTFI). A cross-sectional survey collected data from men aged 18–49 years in Bangladesh using a multistage cluster sampling method. In total, 3499 and 3623 respondents were interviewed by MBBM and FTFI, respectively. In the MBBM, pre-recorded questions were administered using a portable audio-cassette player with two pairs of headphones used concurrently by the respondent and the interviewer. Overall, 18% of the respondents had non-marital sexual exposure in the past year. The MBBM elicited higher responses of non-marital sex (adjusted odds ratio (aOR) 1.3, 95% confidence interval [CI]: 1.1, 1.5) compared with FTFI. The interview methods did not, however, revealed significant differences in response to condom-use rates and the number of non-marital sexual partners. The MBBM is more effective than the FTFI method in eliciting higher responses rates of non-marital sexual contacts and may be recommended for reliable estimates of sexual behaviours.

Keywords

men sexual behaviour survey ballet box interview method STI/HIV Bangladesh social desirability bias

INTRODUCTION

In many societies, including Bangladesh, sexual behaviours are traditionally considered shameful and not to be discussed in public or within the family.^1–3 Surveys collecting data on sexual behaviours could thus be subject to under-reporting due to social desirability bias.^4–7 This under-reporting may be due to embarrassment and fear of revealing information on sexual behaviours to others in society. Available limited data indicate that sexual behaviours that might put people at risk of sexually transmitted infections (STIs), including HIV, are not uncommon among the general male population in Bangladesh.^8–10 High-risk sexual behaviours of men put their wives or other non-marital sexual partners at risk of HIV infection because of their concurrent exposure to paid sex.^11,12 Further understanding of these behaviours and correlates will help prevent and control STIs/HIV in Bangladesh.

A number of methods have been explored to address the challenges of collecting accurate and reliable data on sexual behaviours.^7,13–17 Methods such as self-administered questionnaire, computer-assisted self-interview (audio-CASI) and telephone survey techniques allow less interaction between respondents and interviewers to help reduce social desirability bias.^7,15,16 However, each of these methods has its own challenges in administration in a developing country context. Although self-administered questionnaires provide an alternative approach for conducting interviews in-person by reducing fear of embarrassment or disclosure, they are limited by high rates of non-response.^7,13 Moreover, this method is not very suitable for populations where the literacy level is low. In audio-CASI using portable lap-top computers, respondents listen to questions through headphones and enter answers by pressing a labelled key that provides a private mode of collecting data.^14,17 This method is, however, not suitable in communities where the study population has limited literacy and is not familiar with computers.¹⁷ Although telephone survey techniques are effective in reducing cost,¹⁸ telephone accessibility is not homogeneous for the population, particularly in many developing countries.

The simple ballot-box method (SBBM) is effective in reducing social desirability bias in developing country settings.¹⁹ In this method, the interviewer reads out the question from the questionnaire, and the respondent provides the answer by means of a secret voting procedure using ballot-box and ballot-slips. Although the SBBM is a step towards increasing confidentiality in answering sensitive questions in a community survey, particularly in urban settings, it is often difficult to find an isolated place to ask the sensitive questions face-to-face. This problem is critical while conducting interviews in household settings. Therefore, we modified the SBBM by incorporating an audio-cassette player and headphones for administration of sensitive questions on sexual behaviours (Figure 1). The key difference between this modified ballot-box method (MBBM) and the SBBM is that the sensitive questions are administered through the audio system instead of asking them face-to-face. We hypothesized that increased confidentiality of the MBBM would elicit higher reporting of non-marital sexual exposures, condom-use and numbers of non-marital sexual partners among men compared with conventional face-to-face interviews. This study aimed to assess the differences between using MBBM and FTFI to estimate non-marital sexual contacts, numbers of non-marital sexual partners and condom-use reported by men from the general population of Bangladesh.

Figure 1

Logistics used in the modified-ballot-box method

METHODS

Study design and settings

The study was conducted during February–August 2005 among men aged 18–49 years in three selected divisions (Dhaka, Chittagong and Rajshahi) in Bangladesh. Three urban sites (Dhaka metropolitan, Chittagong metropolitan and Bogra town) and three corresponding rural sites (Faridpur, Rajshahi and Cox's Bazar) were purposively selected from the three divisions to accommodate urban and rural variations in sexual behaviours of men. The Research Review Committee and the Ethical Review Committee of the International Centre for Diarrhoeal Diseases Research, Bangladesh (ICDDR,B) approved the study.

Sampling technique

A three-stage cluster sampling technique was used in each study site involving selection of clusters in the first stage, segmentation of the selected clusters at the second stage and systematic selection of subjects in the third stage (FHI/Impact, 2000). In urban areas – mahallas (the smallest identifiable areas within city corporation/municipality) – and in rural areas – mouzas (a revenue village with a jurisdiction list number) – were considered as primary sampling units (PSUs).

In each district, 30 PSUs were selected by the probability-proportional-to-size method using the census data of 2001. Based on our formal request, the Bangladesh Bureau of Statistics helped draw the PSUs using the census data of 2001. In each study area, the selected clusters were divided into a number of segments such that each segment consisted of an equal number of households (varying from 150 to 300 households among the study areas). One segment was randomly selected by a simple-random sampling method, and all the households in the selected segment were visited to identify eligible respondents. Men aged 18–49 years, who were residing in the area during the last one year or visited home at least once in the last one year of interview, were listed. All the identified eligible respondents were listed in a logbook, and 50 were systematically selected. Half of the selected subjects were randomly allocated to FTFI, and the remaining to MBBM for interview. Thus, in the six study areas, a total of 4500 respondents were selected for interviewing in FTFI and another 4500 in MBBM.

The interviewers visited the households of the selected respondents and identified them by their age, profession and relationship with the household head, as name was not recorded in the listing form to maintain anonymity. After explaining the objectives of the study by the interviewer, the respondent was asked for his verbal consent to participate in the study. Only subjects who willingly provided their consent were finally interviewed. For anonymity, no names were recorded in the questionnaire.

Selection and training of interviewers and supervisors

Twenty-four interviewers and six supervisors who were male and non-public health personnel were recruited in this study. All the interviewers were graduates in social science and experienced in data collection on sexual behaviour. They received two weeks of intensive training on data-collection tools, including administration of prerecorded questions in MBBM, development of sampling frame, maintenance of logbook, etc. Training was also given on rapport-building with respondents for asking sensitive questions on sexual behaviours, fixing a suitable time with respondents to conduct the interview, isolation of subjects for interviews and handling of unforeseen situations that might arise while interviewing. They also received one day of field training for hands-on-experience of working with the data-collection tools. All the interviewers were trained and employed in data collection for both the MBBM and FTFI.

Interview in modified-ballot-box method

For interview in the MBBM, although sensitive questions were administered using audio-cassette players and headphones, the interviewers still emphasized the identification of a suitable isolated place to ensure better privacy. Before starting the interview, the MBBM and its usefulness were explained to the respondents to maintain confidentiality.

Each interview comprised three consecutive steps. First, the interviewers collected sociodemographic information through a face-to-face interview. Second, the interviewers acquainted the respondents with the MBBM. They played a simple story by audio-cassette and then asked four dummy questions of different types and explained to the subject how to answer each question using ballot-slips. During this exercise, the interviewers made adequate interactions with the respondents to familiarize them with the system. Finally, once the interviewers were convinced that the respondents understood the procedure, nine selected sensitive questions (three for each on female sex workers [FSWs], casual female partners and males/transgenders) were administered. Examples of sensitive questions asking about FSWs are shown in Figure 2. During this step, the interviewers provided one ballot-slip to the respondent at a time. Then, the interviewer administered the questions in audio with headphones. After listening to each question, the respondents provided an answer on a ballot-slip by putting a circle around a tick or cross mark (to indicate ‘yes’ or ‘no’), or crossing lines to indicate numbers. When a question was not applicable, the respondent was asked not to mark anything on the ballot-slip and simply drop it in the ballot-box. In this way, the nine sensitive questions were administered one after another. During the interview, the respondent was asked to raise his hand if he did not understand any question. The interviewer who operated the cassette player then replayed the same question. The field supervisor opened the ballot-box at the end of the day to match the voting responses to the appropriate questions using the questionnaire serial number and question numbers that were written beforehand on the back of each ballot-slip.

Figure 2

Example of questions asked and ballot-slips used in the modified-ballot-box method for sex with female sex workers. ¹This question was asked expecting a yes/no response. ²This question was asked expecting a numeric response. *Respondents were asked to circle for ‘yes’ and for ‘no’ answers. **Respondents were asked to cross the bars to indicate a number they wanted to answer

Interview in face-to-face interview

For FTFI, the interviewers identified a suitable place to maintain confidentiality during the interview. The interview was preferably conducted in a private room in the household if available. In the absence of such a room, the interview was conducted in an isolated place, such as a corner of the courtyard of the household, where other members were advised not to interrupt. Respondents sometimes requested the interviewers to interview them in their working place. In rural areas, in some cases, the interview was conducted in isolation under a tree, in a field, in a garden or in other locations.

Operational definition of variables

‘Non-marital sex’ was defined as vaginal, anal or oral penetrative sex with a person the interviewee was not married to. ‘Female sex worker’ was defined as a female who participated in vaginal, anal or oral penetrative sex in exchange for money. ‘Casual female partner’ was defined as female neighbours, friends or relatives with whom the respondent participated in vaginal, anal or oral penetrative sex without exchange of money. ‘Male or transgender partners’ was defined as a man either actively or passively participating in anal sex with another man or transgender.

Data entry and analysis

All data were double-entered and validated for any inconsistency. Analysis of data was done using the survey data module of Stata (version 8.0, Stata Corp, College Station, TX, USA) by employing weighted and cluster analysis techniques. Descriptive analysis was done to compare sociodemographic characteristics of the study participants by two different interview methods using the chi-square test. Logistic regression models were fitted on non-marital sexual exposure separately for any partners, sex worker partners, casual female partners, and male or transgender partners to assess the role of the MBBM. Partially adjusted odds ratios (aORs) are reported with adjustment for clustering and weight effect while aORs are reported with adjustment for clustering, weight and sociodemographic factors, including age, education, marital status, religion, income and occupation.

RESULTS

In the FTFI and MBBM, 3623 (80.6%) and 3499 (77.8%) respondents, respectively, were successfully interviewed. Overall refusal rates in the FTFI and MBBM were 0.3% and 0.5%, respectively; absenteeism rates were 19% in FTFI and 22% in MBBM. The major causes of absenteeism were: in-country migration (FTFI = 11.3%; MBBM = 13.0%) and unable to contact after five visits (FTFI = 7.8%; MBBM = 8.7%).

Background characteristics of respondents

In each interview method, nearly three-quarters of the respondents were aged 20–39 years, and less than 10% were under 20 years of age (Table 1). Overall, 22% of the respondents never attended any school, and only about 30% had completed 10 or more years of schooling. About one-third were never-married. The study population was predominantly Muslim. More than half (59.0%) of the respondents had monthly family expenditure of Tk. 5000 (US$75) or less. More than three-quarters were from four professional categories such as general services, business, farming and transport labourers.

Table 1

Sociodemographic characteristics of the respondents by two different interview methods

	% of respondents
Characteristics	FTFI	MBBM	Total	P value
Age (years)	n = 3623	n = 3499	n = 7122
≤19	7.4	9.4	8.4
20–29	42.2	40.4	41.3
30–39	30.8	30.2	30.5
≥40	19.6	19.9	19.8	0.134
Schooling (years)	n = 3623	n = 3499	n = 7122
0	22.2	21.1	21.7
1–4	14.6	15.7	15.1
5–9	33.4	33.2	33.3
≥10	29.8	30	29.9	0.526
Marital status	n = 3623	n = 3499	n = 7122
Never-married	32.8	34.6	33.7
Ever-married	67.2	65.4	66.3	0.15
Religion	n = 3623	n = 3499	n = 7122
Islam	91.8	92.4	92.1
Others*	8.2	7.6	7.9	0.373
Monthly family expenditure	n = 3620	n = 3497	n = 7117^†
≤3000 Tk	27.2	26.6	26.9
3001–5000 Tk	33.6	30.3	32.0
5001–10,000 Tk	28.8	31.2	30.0
>10,000 Tk	10.3	11.9	11.1	0.062
Occupation	n = 3623	n = 3499	n = 7122
General service-holder	31.6	29	30.3
Businessman	23	22.1	22.6
Farmer	12.1	12.4	12.3
Transport labourer	10.6	10.6	10.6
Day labourer	5.9	5.5	5.7
Student	6.8	7.6	7.6
Unemployed	2.3	2.4	2.4
Fisherman	0.8	0.9	0.8
Service in defence/police	0.5	0.3	0.4
Others^‡	6.3	8.3	7.3	0.016

FTFI = face-to-face interview; MBBM = modified ballot-box method

*Hinduism/Christianity/Buddhism

^†Non-response = 5

^‡Mechanics/builders/carpenters /painter/village doctors/missionary

Non-marital sex by modes of interview

Overall, about 18% of the respondents reported having had non-marital sexual exposure in the last one year. For sex with FSWs, casual female partners and male/transgender the corresponding figures were 10%, 9% and 2%, respectively. Interviews using the MBBM for having non-marital sex elicited higher responses than the FTFI for sex with FSWs and males/transgenders (Figure 3). However, responses to sex with casual female partners were similar in both the methods.

Figure 3

Percentages of men (aged 18–49 years) who had non-marital sex in the last one year by types of partner, and by modes of interview. FTFI = face-to-face interview; MBBM = modified ballot-box method

After adjusting for potential confounding factors, the MBBM was found to be effective in eliciting a 30% higher response for having sex with any type of non-marital sexual partners compared with FTFI (aOR: 1.3, 95% CI: 1.1, 1.5) (Table 2). When separated by type of sexual partner, in the MBBM, the respondents were 1.5 and 2.0 times more likely to report having sexual relationship with FSWs (aOR: 1.5, 95% CI: 1.2, 1.9) and males or transgenders (aOR: 2.2, 95% CI: 1.2, 3.7) compared with FTFI, respectively.

Table 2

ORs for having non-marital sex with different types of partners in the last one year by two different interview methods

Types of partner	Partially adjusted OR* (95% CI)	Adjusted OR^† (95% CI)
Sex with any type of partner
Face-to-face	1.0	1.0
Modified ballot-box	1.3 (1.1–1.5)	1.3 (1.1–1.5)
Sex with female sex workers
Face-to-face	1.0	1.0
Modified ballot-box	1.5 (1.2–1.8)	1.5 (1.2–1.9)
Sex with casual female partners
Face-to-face	1.0	1.0
Modified-ballot-box	1.0 (0.8–1.3)	1.1 (0.9–1.3)
Sex with males/transgenders
Face-to-face	1.0	1.0
Modified ballot-box	2.0 (1.2–3.6)	2.16 (1.2–3.7)

95% CI = 95% confidence interval

*Adjusted for clustering and weight

^†Adjusted for clustering, weight and sociodemographic factors (age, education, marital status, religion, income and occupation)

Number of non-marital sexual partners by modes of interview

There was no substantial variation in reporting of number of non-marital sexual partners by interview method. Of the respondents who reported non-marital sex in the last year, the majority (56.0%) had more than one non-marital sexual partner. One-third (36.0%) of the respondents had two to three and another one-fifth (20.0%) had >3 non-marital sexual partners. When segregated by types of non-marital sexual partner, of the respondents who had sex with FSWs, about 62% had sex with more than one FSW. The corresponding figures were 35% and 49% for sex with casual female partner and male/transgender, respectively.

Condom-use by modes of interview

The overall condom-use rate during last penetrative sex in the past year with a FSW or a casual female partner or a male/transgender was 40%, 30% and 9%, respectively. There was no substantial variation in response to condom-use between the interview methods for each of the three types of non-marital sexual partners.

DISCUSSION

The study has demonstrated that the MBBM produced significantly higher levels of responses on non-marital sexual exposure compared with FTFI (19% versus 16%; P < 0.05) among men from the general population of Bangladesh. Surveys using this new method elicited significantly higher response rates separately for sexual relationships with FSWs and male/transgenders compared with traditional FTFI. The interview method did not, however, reveal any significant difference in condom-use rate during last sex and the number of non-marital sexual partners in the last year. It is assumed that once the respondents have answered positively for non-marital sexual exposure in the past year in either of these two interview methods, the interview method makes almost no difference to the likelihood of answering further questions on condom-use and numbers of partners. It appears that the MBBM ‘opens the door’ for respondents to discuss this issue but once that door is open, the choice of methodology no longer plays a vital role.

A major strength of this new interviewing method is that it is equally administrable and feasible among different sociodemographic groups, which is indicated by the low refusal rates to participate in the study. Although about one-fifth of the respondents could not been reached at home, even after five visits, once contacted only a few of them refused to participate (0.3% in FTFI and 0.5% in MBBM). Unlike highly technical methods such as audio-CASI which has operational difficulties and suffers from producing inconsistent results among respondents who are unfamiliar with the computer technology¹⁷, our simple technology-based method succeeded in generating consistent results across different sociodemographic groups. Low literacy rate is considered a major problem in administering the SBBM in developing countries.¹⁹ We were able to overcome this problem by using simple ballot-slips and incorporating a demonstration session on the MBBM. A short demonstration before the real interview also allowed us to ensure that the respondents understood the complicated questions, which is one of the limitations of the audio-CASI method.²⁰

Although our new interviewing method is technology dependent, it is still simple and easy to use in community settings in developing countries. Our interviewers rarely faced any technical difficulty in using the battery-operated audio system. However, additional training is needed for interviewers for properly administering the method. The additional cost involved in this study was not high; the cost was Tk 25 (US$ 0.37) per interview where we used one unit of audio-system for each of the 24 interviewers. Cost could, however, be further reduced by decreasing the number of audio-systems and allocating those to different interviewers during the study period. Depending on the study population, there might be a need for recording questions in more than one dialect, which would further increase cost. On average, the MBBM needed 12 minutes more time to complete an interview compared with FTFI. The longer time required, mainly due to demonstrating the audio-system to the respondents and how to use ballot-slips for different types of dummy questions, was needed for familiarizing the respondents with the new system.

This study was conducted only among male respondents and is thus unable to determine the role of MBBM method in elicitation of sexual behaviours among female participants. Previous studies have indicated that under-reporting of sexual behaviours could be more common among women compared with men because of social desirability bias against non-marital sex and perceived lack of confidentiality in the data-collection process.^10,11,21 It could be assumed that the MBBM would be more effective in the elicitation of non-marital sexual behaviours among women. One-fifth of the study subjects were not available for interview (19% in FTFI and 22% in MBBM) despite repeated visits during the study period, but further analysis of the reasons of absenteeism was not found to be significant by their assigned interview methods.

The findings of this study provide evidence that the MBBM may be useful in improving response rates to sensitive questions on sexual behaviours among men. In an effort to reduce social desirability bias in the elicitation of sexual behaviours, various adaptations of FTFI methods were suggested but only a few formal evaluations of such methods in developing-country settings have been published.^17,19,22 The MBBM is well worth the small additional effort and cost in achieving the important goal of making respondents feel more secure to answer sensitive survey questions openly and honestly. We recommend this method for assessing sexual behaviours of men in the community for reliable estimates. We do, however, suggest undertaking more evaluation of the method in women and also in culturally diverse population groups along with assessing cost-effectiveness. We also recommend conducting qualitative studies to document the perceptions of the study participants about the new interview method.

Footnotes

ACKNOWLEDGEMENTS

This study was funded by the United States Agency for International Development (USAID) through Family Health International-Bangladesh Country Office and Asia Regional Programme-Bangladesh. ICDDR,B acknowledges with gratitude the commitment of USAID to the Centre's research efforts. We are also thankful to the Bangladesh Bureau of Statistics for their assistance in selecting samples (clusters) from the population census of 2001.

References

Khan

, Hudson-Rodd

, Saggers

, Bhuiya

. Men who have sex with men's sexual relations with women in Bangladesh. Cult Health Sexualit 2005;7:159–69

Sivaram

, Johnson

, Bentley

, Sexual health promotion in Chennai, India: key role of communication among social networks. Health Promot Int 2005;20:327–33

Islam

, Conigrave

. HIV and sexual risk behaviours among recognized high-risk groups in Bangladesh: need for a comprehensive prevention programme. Int J Infect Dis 2008;12:363–70

Fay

, Turner

, Klassen

, Gagnon

. Prevalence and patterns of same-gender sexual contact among men. Science 1989;243:338–48

Catania

, Gibson

, Chitwood

, Coates

. Methodological problems in AIDS behavioural research: influences on measurement error and participation bias in studies of sexual behaviour. Psychol Bull 1990;108:339–62

Konings

, Bantebya

, Carael

, Bagenda

, Mertens

. Validating population surveys for the measurement of HIV/STD prevention indicators. AIDS 1995;9:375–82

Anderson

, Stall

. Increased reporting of male-to-male sexual activity in a national survey. Sex Transm Dis 2002;29:643–46

Caldwell

, Pieris

, Barkat-e-Khuda , Caldwell

. Sexual regimes and sexual networking: the risk of an HIV/AIDS epidemic in Bangladesh. Soc Sci Med 1999;48:1103–16

Hawkes

, Morison

, Chakraborty

, Reproductive tract infections: prevalence and risk factors in rural Bangladesh. Bull World Health Organ 2002;80:180–88

10.

Sabin

, Rahman

, Hawkes

, Sexually transmitted infections prevalence rates in slum communities of Dhaka, Bangladesh. Int J STD AIDS 2003;14:614–21

11.

Gibney

, Saquib

, Metzger

. Behavioural risk factors for STD/HIV transmission in Bangladesh's trucking industry. Soc Sci Med 2003;56:1411–24

12.

Commission on AIDS in Asia. Redefining AIDS in Asia; Crafting an Effective Response. New Delhi: Oxford University Press, 2008

13.

Boekeloo

, Schiavo

, Rabin

, Conlon

, Jordan

, Mundt

. Self-reports of HIV risk factors by patients at a sexually transmitted disease clinic: audio vs. written questionnaires. Am J Public Health 1994;84:754–60

14.

Turner

, Ku

, Rogers

, Lindberg

, Pleck

, Sonenstein

. Adolescent sexual behaviour, drug use, and violence: increased reporting with computer survey technology. Science 1998;280:867–73

15.

Gribble

, Miller

, Roger

, Turner

. Interview mode and measurement of sexual behaviours: methodological issues. J Sex Res 1999;36:16–24

16.

Lau

, Tsui

, Wang

. Effects of two telephone survey methods on the level of reported risk behaviours. Sex Transm Infect 2003;79:325–31

17.

Potdar

, Koenig

. Does audio-CASI improve reports of risky behaviour? Evidence from a randomized field trial among young urban men in India. Stud Fam Plann 2005;36:107–16

18.

Midanik

, Greenfield

, Rogers

. Reports of alcohol-related harm: telephone versus face-to-face interviews. J Stud Alcohol 2001;62:74–78

19.

Gregson

, Zhuwau

, Ndlovu

, Nyamukapa

. Methods to reduce social desirability bias in sex surveys in low-development settings: experience in Zimbabwe. Sex Transm Dis 2002;29:568–75

20.

Kissinger

, Rice

, Farley

, Application of computer-assisted interviews to sexual behaviour research. Am J Epidemiol 1999;149:950–54

21.

Alam

, Rahman

, Gausia

, Sexually transmitted infections and risk factors among truck stand workers in Dhaka, Bangladesh. Sex Transm Dis 2007;34:99–103

22.

Van de Wijgert

, Padian

, Shiboski

, Turner

. Is audio computer-assisted self-interviewing a feasible method of surveying in Zimbabwe? Int J Epidemiol 2000;29:885–90