A Pilot Study of a Picture- and Audio-assisted Self-interviewing Method (PIASI) for the Study of Sensitive Questions on HIV in the Field

Abstract

Self-interview methods such as audio computer-assisted self-interviewing (ACASI) are used to improve the accuracy of interview data on sensitive topics in large trials. Small field studies on sensitive topics would benefit from methodological alternatives. In a study on male involvement in antenatal HIV testing in a largely illiterate population in Malawi, we piloted picture- and audio-assisted self-interviewing (PIASI). Out of 388 participants, 96 were randomized to answer six sensitive questions using PIASI after the face-to-face interview (FTF). Participants who responded by PIASI were more likely to express accepting attitudes toward childbearing by HIV-positive women than were FTF respondents. Most considered PIASI easy to use, and one in five preferred it to FTF. PIASI may offer a feasible, low-cost alternative to ACASI for small and simple studies on sensitive topics and in illiterate populations.

Keywords

male involvement self-interview interview HIV sensitive topics

Background

It has been asserted that face-to-face interviews (FTF) on sensitive and stigmatized topics, such as sexual behavior, substance abuse, or HIV, have yielded answers that are socially desirable and reflect perceived social expectations rather than real behavior or opinions (Hewett et al. 2004). Methods that are considered to reduce social desirability bias include both self-interviewing, where the interview takes place in a private and confidential environment, without the direct participation of the interviewer, and in-depth interviews, participant observation, or daily diaries, which engage the person for longer times to build rapport (Lara et al. 2004; Morrison-Beedy et al. 2006; Plummer et al. 2004).

Audio computer-assisted self-interviewing (ACASI) has become the self-interviewing method of choice in much of health research globally (Rathod et al. 2011). In HIV research in low-income countries, ACASI has so far mainly been used to capture sexual behaviors in HIV prevention trials (National Institute of Mental Health [NIMH] 2007; van der Elst et al. 2009) and to determine HIV risk (Gorbach et al. 2012; Mensch et al. 2008). ACASI is well accepted among respondents (Dolezal et al. 2012; van de Wijgert et al. 2000), and the quality of ACASI data is higher than that of self-administered questionnaire data because of built-in reminder functions and direct data entry (Minnis et al. 2007; Morrison-Beedy et al. 2006).

Studies in developing countries have found increased reporting of risky sexual behaviors via ACASI (Gorbach et al. 2012; Hewett et al. 2004; Potdar and Koenig 2005; van der Elst et al. 2009) and a closer association between sexually transmitted infection biomarkers and reports on sexual risk behavior obtained via ACASI than those obtained via FTF (Hewett et al. 2008). Other studies, however, have shown no difference in or less reporting of sensitive behaviors by ACASI compared to FTF or biomarker findings (Mensch et al. 2008; NIMH 2007), and some studies show poorer internal consistency between responses in ACASI than in FTF, probably because in ACASI no interviewer is present to clarify unclear questions (Gorbach et al. 2012; Hewett et al. 2004; Mensch et al. 2008).

How, if at all, ACASI responses differ from FTF responses can have various plausible explanations. Generally, sensitive behaviors are considered to be reported reluctantly to an interviewer, and therefore reporting sensitive behavior more frequently in ACASI may be an indication that the reporting in ACASI is more truthful (Dolezal et al. 2012). Many researchers agree that the more sensitive the behavior, the more the reporting is increased by ACASI (Hewett et al. 2004; Mensch et al. 2008; Minnis et al. 2007; Potdar and Koenig 2005). However, social desirability bias can also work the other way, making the respondents report certain types of behavior, such as adherence to a trial product, more in FTF than in ACASI, to impress the interviewer (Gorbach et al. 2012; Potdar and Koenig 2005). These effects can differ by gender, subgroup, and study setting; therefore, the social context of the study needs to be well understood for the interpretation of ACASI findings (Dolezal et al. 2012; NIMH 2007; Potdar and Koenig 2005).

Self-interviewing methods for low-income countries have to function in field settings, in populations with low literacy and limited experience with computers, and with limited resources. ACASI has been found to be a feasible methodology in low-income country settings (NIMH 2007). With the help of modified applications, ACASI has been used in illiterate populations (Gorbach et al. 2012; Hewett et al. 2004; NIMH 2007; Potdar and Koenig 2005). However, ACASI results are less reliable in less-educated groups, and this diminishes the advantages of ACASI in this group (Lara et al. 2004; Mensch et al. 2008; Potdar and Koenig 2005; van de Wijgert et al. 2000). In addition, the inability to clarify questions may lead to misunderstanding of terms and questions when ACASI is used among illiterate participants (NIMH 2007; Potdar and Koenig 2005). Furthermore, ACASI is a relatively expensive method, mainly due to its high setup costs and the need to develop applications for different subpopulations. While large studies benefit from the savings in data handling offered by ACASI, costs may become a barrier in small studies (van de Wijgert et al. 2000). Apart from costs, technological complexity, unfamiliarity with the method, and the mixed results achieved have been cited to explain the relatively limited use of ACASI in Africa (van der Elst et al. 2009).

In this study, we piloted a modified version of ACASI, picture- and audio-assisted self-interviewing (PIASI), as an attempt to develop methodology suitable for small studies in field settings and illiterate populations. We aimed to investigate the differences in the reporting of sensitive behaviors as well as to roughly determine the acceptability, feasibility, and data quality of the methodology. This article describes a sub-study of a study, which investigated men’s perceptions of HIV in pregnancy and their involvement in antenatal HIV testing and counseling in rural Malawi (Aarnio et al. 2009). The sub-study was carried out to examine the accuracy of the responses to sensitive questions on HIV in FTF by comparing these responses to those given on self-interviewing.

Methods

Study Setting and Population

The study setting in Mangochi district consisted of 11 rural, mainly Muslim fishing villages. The details of the study setting and population are described in Aarnio et al. (2009). A total of 388 men were interviewed by local research assistants using FTF. Eligible participants for the study were married men with spouses of reproductive age (15–49 years). Systematic random sampling was used to select participants. A subsample of 96 out of the 388 participants was randomly assigned to respond to six sensitive questions on HIV by PIASI at the end of the FTF. The rest of the participants (n = 292) responded to all questions, including the sensitive ones, in FTF. The PIASI subsample size of 96 was chosen to detect a difference in reporting between the two interview modes within 10–15% of the true difference depending on the prevalence of the response (Lwanga and Lemeshow 1991).

Tools and Procedure

Six sensitive study questions in the study questionnaire were chosen for PIASI based on their sensitivity in men’s focus group discussions (FGD) in another arm of this study. In the FGD, few knew or were open about their HIV status. Men were expected to be the initiators of new ideas in the family, and for women to suggest HIV testing as recommended at antenatal care was outside the norm. Divorce due to HIV was a sensitive topic that was linked to sensitive behaviors like promiscuity and blame for bringing HIV in the family (Aarnio et al. 2009). The six sensitive questions, two control questions, and two questions on the acceptability of the method, together with the response options to each, were tape recorded in the local language, Chiyao, using the male voice of one of the interviewers. Questions were worded identically in the FTF and on the tape (Figure 1). A tape recorder with C-cassettes (Sony) and earphones was used. Picture booklets containing the response options for each question were created with help of local research assistants, using symbols familiar from antenatal care education materials (Appendix A). Culturally appropriate hand-drawn representations were used since these have shown to be well recognized (Mierzwa et al. 2013). The booklets were coded to allow matching with the FTF questionnaires. Prior to the study, the self-interviewing tools were pretested with voluntary local male participants and were modified accordingly.

Figure 1.

Format of questions in picture booklet and audio tape: sample question on acceptance of childbearing by HIV-positive women.

Before the interview, oral consent for the study was sought from all participants, and the 96 respondents randomly selected for PIASI were also asked to consent to being interviewed using this method. The tape-recorded questions were administered in the village setting at the end of the FTF—a solution that has been recommended to save time and improve the concentration of illiterate participants (Potdar and Koenig 2005). The interviewer first guided the respondent in the use of the tape recorder, earphones, and booklet. The respondent was then left in privacy to listen to the questions from the tape using the earphones and to answer by ticking on the image with their preferred answer in the booklet. After completing the responses, the respondent sealed the booklet in an envelope.

Analysis

PIASI responses were compared with FTF responses using cross-tabulations. The χ² test was used to determine the statistical significance of the identified differences. Missing answers accounted for less than 3% of the responses for each question (n ≤ 11 out of 388) in the total sample and for less than 9.4% (n ≤ 9 out of 96) in the PIASI subsample.

Results

Background of the Participants

The respondents were married men with spouses of reproductive age. Their mean age was 31 years (range 17–54), and 11.9% (n = 46) were polygamous. Most belonged to the Yao tribe (81.2%, n = 315) and were of the Muslim faith (80.7%, n = 313). Only 22.8% (n = 88) had completed primary school, and 41.0% (n = 159) were illiterate by their own definition. Their main income source was fishing (48.5%, n = 188). More detailed background characteristics are presented in Aarnio et al. (2009). Respondents who responded by PIASI did not differ from the rest in terms of background characteristics.

PIASI Feasibility, Acceptability, and Data Quality

Inclusion of the PIASI component in the study required development and pretesting of the picture booklet and recorded questions, interviewer training, separate handling of picture booklet data, and investment in four tape recorders. Against the total setup costs and time of the study, this was a minimal addition. The FTF of 35–45 minutes was lengthened by 10–15 minutes, which we considered feasible for a field study.

PIASI was well accepted by the respondents, that is, 89.1% (n = 82) of the respondents considered it an easy form of interview. Only one participant refused to respond using PIASI after initially consenting to do so. However, only 22% (n = 20) preferred PIASI to FTF, and the rest preferred FTF. There was a statistically insignificant tendency toward illiterate respondents finding PIASI more difficult but also more enjoyable. Respondents’ skills in answering by PIASI were tested by asking them to answer two control questions (“Are you married now?” and “How many wives do you have?”) in both the FTF and the PIASI sections of the interview. No differences in the answers between the two methods were found. Missing responses were on average somewhat more common in the PIASI than in the FTF interviews, accounting for 5.7% and 2.5%, respectively. This level was considered acceptable, bearing in mind that there were no reminder functions in the PIASI system, and the interviewer could not be consulted during the interview.

Responses to Sensitive Questions on HIV via PIASI

Participants who responded to the six sensitive questions by PIASI were more likely to express accepting attitudes toward HIV-positive women having children than were participants in FTF (10.0% vs. 2.8%, p < .05). There was no difference in responses to the other PIASI questions (Table 1). There were no significant differences between the PIASI responses of illiterate and literate respondents.

Table 1.

Comparison of Men’s Responses to Sensitive Questions on HIV Given in FTF and PIASI.

	FTF (n = 292)	PIASI (n = 96)
Knows a man who has divorced his wife due to HIV	15.9% (n = 46)	15.7% (n = 14)
Knows a woman who has divorced her husband due to HIV	13.4% (n = 39)	13.8% (n = 12)
Has (ever) been tested for HIV	35.2% (n = 101)	40.7% (n = 37)
Would attend HIV testing if his wife requested him to do so	93.4% (n = 271)	87.1% (n = 81)
Accepts that HIV-positive women should be allowed to bear children	2.80% (n = 8)*	10.0% (n = 9)*
Would continue family life if his wife was found to have HIV	89.7% (n = 260)	88.0% (n = 81)

Note: Missing responses (<2.9% for each question) are excluded from the analysis. FTF = face-to-face interview; PIASI = picture- and audio-assisted self-interview.

*p < .05.

Discussion

The relevance of self-interviewing methods to research in low-resource settings rests heavily on their applicability to populations with lower levels of literacy (Potdar and Koenig 2005). We found PIASI to be a suitable methodology for a study that involves illiterate respondents. In our study, we found no difference in the number of missing responses by illiterate and literate subjects. This is probably due to the relative simplicity of the PIASI interview, good guidance to using PIASI by the interviewers, the small sample size, and the relatively small variation in education level of respondents. In studies using ACASI, a low level of literacy has been shown to negatively affect data quality (Mensch et al. 2008; Potdar and Koenig 2005; van de Wijgert et al. 2000), and our finding suggests that working with simpler technology may be an advantage in illiterate populations in that a familiar study setting may help the respondents to maintain focus.

In PIASI, the use of audio and images to collect responses limits the complexity of questions and study designs, probably more than is the case with ACASI. However, we argue that with careful design, the needs of most small studies or sub-studies can be met with this methodology. The overall level of missing responses in our study was higher in PIASI than in FTF answers, which we ascribe mainly to the lack of reminder and checkup functions in PIASI—an aspect that needs further development. But, we also concur with Hewett et al. (2004) in their view that fewer missing answers in FTF than in self-interviewing may indicate pressure to respond in FTF.

PIASI was found to be easy by most respondents but was chosen as the preferred method by only 22%. This is much less than the 85% preference for ACASI found among slightly more educated, urban women in Malawi (Gorbach et al. 2012).

In a multicenter study using ACASI in various low-income settings, more than 80% felt very comfortable entering answers on a computer. However, whether respondents felt ACASI or FTF to be more private or to be their preferred interview method varied by setting (NIMH 2007). Some studies have found less preference for or trust in the confidentiality of ACASI among the least educated (NIMH 2007; Rathod et al. 2011), which we suggest also explains part of the comparatively low preference for self-interviewing found in our study. Other factors may include our rural study population with a strong oral culture and a lack of attractive computer equipment. A limitation of our study is that we did not examine how the participants perceived the length and privacy of the PIASI section, and these factors may also affect acceptability (van der Elst et al. 2009).

We reported more accepting attitudes toward childbearing by HIV-positive women in PIASI than by those in FTF. In our opinion, this must be seen as respondents attempting to conform to social norms and expectations in the interview situation by modifying their responses (Hewett et al. 2004). Given the very recent introduction of prevention of mother-to-child transmission programs in our study area and the involvement of men in these (Aarnio et al. 2009), society is likely not to accept childbearing by HIV-positive women. The male respondents’ private angle on the issue is, however, likely to be more liberal, knowing that HIV-positive people’s wishes to have a family life and have children are no different from those of other people.

By contrast, we found no differences between PIASI and FTF in responses to divorce due to HIV, which may be because the questions are of limited sensitivity (Hewett et al. 2004; Mensch et al. 2008; Minnis et al. 2007). The indirect and hypothetical formulation of the questions may have reduced the sensitivity more than we expected. However, lack of difference between self-interview and FTF responses is a common finding in developing country settings (NIMH 2007), and more research is needed to understand the methodological and locally relevant reasons behind this. The lack of differences between FTF and PIASI responses to the questions on HIV testing may also be due to our rather small sample size. In our opinion, self-interviewing continues to be a promising methodology for research on social aspects of HIV and could be added as a component in research based in hospitals or other institutions.

Conclusion

PIASI may provide an alternative to ACASI in small, simple studies that wish to use self-interviewing in field settings and low-income country settings and among illiterate populations. Our pilot study opens the way for more research in this direction. We also recommend more use of self-interviewing in social studies on HIV.

Footnotes

Appendix A

Acknowledgment

We are grateful to the research assistants and staff at the College of Medicine, Mangochi Office, who have been invaluable in carrying out the study. We thank the staff at the University of Tampere for their support and help.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The study was financially supported by the Academy of Finland and by the Finnish Cultural Foundation.

References

Aarnio

Olsson

Chimbiri

Kulmala

2009. Male involvement in antenatal HIV counseling and testing: Exploring men’s perceptions in rural Malawi. AIDS Care 21:1537–46. doi:10.1080/09540120902903719.

Dolezal

Marhefka

S. L.

Santamaria

E. K.

Leu

C.-S.

Brackis-Cott

Mellins

C. A.

2012. A comparison of audio computer-assisted self-interviews to face-to-face interviews of sexual behavior among perinatally HIV-exposed youth. Archives of Sexual Behavior 41:401–10. doi:10.1007/s10508-011-9769-6.

Gorbach

P. M.

Mensch

B. S.

Husnik

Coly

Mâsse

Makanani

Nkhoma

. 2012. Effect of computer-assisted interviewing on self-reported sexual behavior data in a microbicide clinical trial. AIDS Behavior. doi:10.1007/s10461-012-0302-2.

Hewett

P. C.

Mensch

B. S.

Erulkar

A. S.

2004. Consistency in the reporting of sexual behaviour by adolescent girls in Kenya: A comparison of interviewing methods. Sexually Transmitted Infections 80:ii43–ii48. doi:10.1136/sti.2004.013250.

Hewett

P. C.

Mensch

B. S.

Ribeiro

M. C. S. de A.

Jones

H. E.

Lippman

S. A.

Montgomery

M. R.

van de Wijgert

J. H. H. M.

2008. Using sexually transmitted infection biomarkers to validate reporting of sexual behavior within a randomized, experimental evaluation of interviewing methods. American Journal of Epidemiology 168:202–11. doi:10.1093/aje/kwn113.

Lara

Strickler

Díaz Olavarrieta

Ellertson

2004. Measuring induced abortion in Mexico: A comparison of four methodologies. Sociological Methods & Research 32:529–58. doi:10.1177/0049124103262685.

Lwanga

S. K.

Lemeshow

1991. Sample Size Determination in Health Studies. A Practical Manual. Geneva, Switzerland: World Health Organization.

Mensch

B. S.

Hewett

P. C.

Gregory

Helleringer

2008. Sexual behavior and STI/HIV status among adolescents in rural Malawi: An evaluation of the effect of interview mode on reporting. Studies in Family Planning 39:321–34.

Mierzwa

Souidi

Friedland

Littlefield

S. A.

Katzen

L. L.

Savel

Boccio

. 2013. Approaches That Will Yield Greater Success when Implementing Self-administered Electronic Data Capture ICT Systems in the Developing World with an Illiterate or Semi-literate Population. New York: The Population Council.

10.

Minnis

A. M.

Muchini

Shiboski

Mwale

Morrison

Chipato

Padian

N. S

. 2007. Audio computer assisted self-interviewing in reproductive health research: Reliability assessment among women in Harare, Zimbabwe. Contraception 75:59–65.

11.

Morrison-Beedy

Carey

M. P.

2006. Accuracy of audio computer-assisted self-interviewing (ACASI) and self-administered questionnaires for the assessment of sexual behavior. AIDS Behavior 10:541–52. doi: 10.1007/s10461-006-9081-y.

12.

National Institute of Mental Health (NIMH) Collaborative HIV/STD Prevention Trial Group. 2007. The feasibility of audio computer-assisted self-interviewing in international settings. AIDS 21:S49–58.

13.

Plummer

M. L.

Ross

D. A.

Wight

Changalucha

Mshana

Wamoyi

Todd

. 2004. “A bit more truthful”: The validity of adolescent sexual behaviour data collected in rural northern Tanzania using five methods. Sexually Transmitted Infections 80:ii49–56. doi:10.1136/sti.2004.011924.

14.

Potdar

Koenig

M. A.

2005. Does audio-CASI improve reports of risky behavior? Evidence from a randomized field trial among young urban men in India. Studies in Family Planning 36:107–16.

15.

Rathod

Alexandra

Minnis

Subbiah

Krishnan

2011. ACASI and face-to-face interviews yield inconsistent estimates of domestic violence among women in India: The Samata health study 2005–2009. Journal of Interpersonal Violence 26:2437–56. doi:10.1177/0886260510385125.

16.

van de Wijgert

Padian

Shiboski

Turner

2000. Is audio-computer-assisted self-interviewing a feasible method of surveying in Zimbabwe? International Journal of Epidemiology 29:885–90.

17.

van der Elst

Haile Selassie Okuku

Nakamya

Muhaari

Davies

McClelland

R. S.

Price

M. A.

. 2009. Is audio computer-assisted self-interview (ACASI) useful in risk behaviour assessment of female and male sex workers, Mombasa, Kenya? PLoS One 4:e5340. doi:10.1371/journal.pone.0005340.