Research conducted in low-income countries often yield contradictory results where sensitive issues are concerned even in similar studies and populations [1]. One source of bias is socially desirable responses in face-to-face (FTF) interviewing. Consequently, data from studies on stigma-associated diseases such as HIV/AIDS may be flawed with potentially far-reaching implications on decision-making and public health policies.

The foregoing argument implies that traditional FTF interviewing is unsuitable for collecting potentially sensitive data and calls for alternative methods of interviewing that increase privacy and are less confrontational. ACASI is a method of interviewing conducted in a private setting and the interaction happens only between the respondent and the computer. This person-machine interaction eliminates the need for respondents to reply in a socially desirable manner, as would be the case in FTF interviewing. Studies have shown that privacy provided by ACASI increases reporting of sensitive behaviours [2-7]. Apparently, this increased reporting is not attributable to random errors [4,8]. A plausible explanation given by Cooley and Turner suggests that increased reporting is because respondents are more willing to respond in an ACASI than in FTF interviewing [9]. ACASI is more so apt where education is low since it eases the response task [2,7,8,10]. ACASI program has other advantages such as preventing non-response by guiding respondents through the interview by providing prompts. Skips and branching are automated preventing punching errors occurring in manual data entry. With ACASI, the interview is standardised hence avoiding interviewer variations that would occur in many research settings where FTF interviewing is carried out with the assistance of several interviewers.

ACASI has been shown to be feasible among women in Africa [8]. However, information on its use for interviewing mothers with low education and on sensitive issues in a clinical setting was lacking.

For implementation, a laptop, attached keypad (Targus™ Numeric keypad), and headphones were used (figure 1). All respondents went through two practice questions after which they were left on their own. It was necessary to guide respondents through the two practice questions in order to familiarise them with the use of the input device (external keypad) especially on how much pressure to exert on the keys since too much pressure would have resulted in repeated character entry. At the end of the ACASI, each interviewee's inputs were automatically added to a tab-delimited text record file and the amount of time that each respondent spent answering questions with ACASI was similarly recorded automatically. In the FTF interviewing inputs were directly entered by use of the keypad. At the end of the FTF interview, a differently named text file similar to the one produced in the ACASI was automatically generated. To determine preferences and attitudes for the two interviewing methods, we asked respondents questions about their experiences with both methods in an exit interview.

Figure 1 The audio computer-assisted self-interviewing (ACASI) set-up as it was used in Kenya. The mother being interviewed is sitting in front of a portable computer listening to the questions in her own language in the headset. She provides the answers by the (more ...)

Data were analysed data using SPSS version 10.1 [14] and Epi 6.04 [15]. To compare means for age, marital-time and parity, we used a paired Student's t-test when measuring interview mode effect. We measured time taken by each mother to complete ACASI and used a Student's t-test to compare mean time taken by two groups of mothers – those educated up primary school level compared to mothers educated up to secondary school level. Level of education was determined for each mother from the exit interview. A Wilcoxon Signed Rank test was used for continuous variables for which distributions were skewed. For categorical variables, a Pearson Chi-Square (χ²test) was used to compare proportions for independent variables and when the expected value was less than 5, the Fisher's Exact Test was used. McNemar's χ²test was used to compare differences between paired proportions for categorical variables. For qualitative data, responses were reported literally and numbers of respondents giving similar responses indicated. Since the sample was larger than 30, we assumed normality in all analysis and a p-value of < 0.05 was considered statistically significant.

The Kenya PMTCT pilot project and the questionnaires used in this study had ethical approval from the Kenyatta national hospital ethical review committee and the national council of science and technology. The regional committee for medical research ethics, Norway-West also approved the study. Additional approval was not sought but all respondents orally consented to participate in the study.

Respondents were asked whether they had introduced a selection of 13 food items (table 1). In all cases apart from 'baby formula' and 'fresh packet milk', the number of respondents who indicated that they had introduced the foods ≤3 months of age was fewer in the FTF interview than in the ACASI. These differences were significant (p < 0.05) for 7 food items compared to differences for only 3 food items in proportions of mothers who had introduced food items at 1 year of an infants age (table 1). A graphical presentation of 3 of the 13 food items: milk, protein-rich foods, and herbal tea is presented in (figure 2).

Table 1 Comparison of proportions of mothers who claimed to have introduced different food items at 3 and at 12 months by mode of interviewing: Audio computer-assisted self-interviewing (ACASI) or face-to-face (FTF) interview.

Figure 2 Three examples of difference in reporting infant feeding by method of interviewing: Audio computer-assisted self-interviewing (ACASI) or face-to-face (FTF). Number of mothers, who report giving milk or protein-rich foods (egg, meat or fish), whole milk (more ...)

Irrespective of interview mode, equal proportions of respondents 42 (53%) said they used a cup and spoon to feed their children in both ACASI and FTF interviewing respectively (table 2). However, more respondents reported use of a cup with holes on the snout in the ACASI than in the FTF interview (p = 0.001).

Table 2 Reported utensils used to feed infants by mode of interviewing: Audio computer-assisted self-interviewing (ACASI) or face-to-face (FTF) interview.

Table 3 Comparison of the ability to use audio computer-assisted self-interviewing (ACASI) between mothers with primary versus secondary education.

To assess the importance that respondents assigned to confidentiality, we asked them to tell us how important it was for their responses not to be seen by others. Twenty-six, (33%) indicated that 'it did not matter' while the majority 54 (68%) indicated that 'it was important that others do not see their responses'.

This is the first study comparing ACASI and FTF conducted in an African rural clinical setting among mothers with low formal education. ACASI and FTF interviewing yielded similar results to neutral questions but differed consistently whenever questions of a sensitive nature were asked. Since our aim was not to validate any of the methods of interviewing, any of the two methods could have yielded the more correct answers. However, consistency of data divergence between the methods suggests that ACASI yielded responses that were less influenced by perceptions of what was socially desirable compared to FTF interviewing.

At the antenatal clinic, the nurses stress in their morning health talk sessions the importance of exclusive breastfeeding for the first 6 months of an infant's life and advocate for extended breastfeeding for at least two years. Our study demonstrates early (≤ 3 months) and extensive introduction of food items, irrespective of interview mode. This information concurs with a study done in Malawi showing that exclusive breastfeeding is uncommon and complementary foods are introduced early in rural families [16]. In the ACASI mode of interviewing even higher proportions of mothers reported early introduction of feeds, which suggests that the FTF responses were biased towards socially desirable responses.

Our data also indicates a bias of the FTF data in the direction of socially desirable responses for the intended length of breastfeeding. A larger proportion of mothers responded that they intended to breastfeed their infants for 24 months during FTF interviewing (44 respondents, 55%) as opposed to during ACASI (11 respondents, 14%). A plausible reason would be that respondents would want to confine their responses in FTF interviews to what they had been taught during the health talk sessions. Reported mean age of introducing protein-rich foods such as meat, eggs, or fish by ACASI mode was 117 days (approximately 4 months) and 173 days (approximately 6 months) by FTF interviewing. This discrepancy could mean that respondents knew that they should introduce protein-rich foods when their babies were older (as was taught during health talks) but in practice, they introduced protein-rich foods to babies at a young age.

More respondents (25%) indicated that they used a closed cup with holes on the snout for infant feeding in the ACASI interview compared to 6% in the FTF interview. Since this utensil is hard to clean, its use is discouraged in the morning health talks. It appears that the respondents gave socially desirable responses in the FTF interview hence the lower proportion reporting.

This study had some limitations. The ratio of respondents who undertook each of the interviewing methods in the first and subsequent interview were dissimilar (70:10 – ACASI first to FTF interviewing first). This was not due to design but because of lack of sufficient number of respondents in the second sequencing arm and we could not therefore effectively measure the effect of sequencing on responses. However, changing the order of interviewing only resulted to a decline in time respondents took to complete the ACASI interview among those who started on FTF interview first and ACASI second (t (72) = 2.626 p = 0.011, CI: 2 – 13 minutes) and had no effect on responses. Secondly, assignment of respondents to each of the interviewing sequencing arms was not random since it was done as a group of 70 (ACASI interviewing first) and the second group of 10 (FTF interviewing first) rather than an alternate assignment of respondents during recruitment to each of the interviewing mode.

Studies have shown that ACASI increases reporting of sensitive issues in high-income countries but until recently, few studies had tried this methodology in resource-poor settings. This is the first study on ACASI conducted in an African rural clinical setting among mothers with low formal education. Our study results are in agreement with results of several studies that have shown that ACASI increases reporting of sensitive issues [2-7]. Differences in responses with the two methods of interviewing show that data on infant feeding in the sensitive context of HIV when collected during face-to-face interviews may be influenced by the respondents tendency to give socially desirable responses. Relying on such data may lead to imprecise decisions on public health issues. Use of ACASI in research or program settings involving sensitive issues can improve data, information and conciseness of decision-making.

The author(s) declare that they have no competing interests.

A Waruru participated in all aspects of the study including planning, design of the ACASI program, data collection and data analysis, and in writing the manuscript, R Nduati designed the initial questionnaire, prepared the study site, supervised data collection, participated in analysis and interpretation of data and in the writing of the manuscript. T Tylleskär had the initial idea for the study, supervised the study planning, the design of the ACASI interviews and participated in the analysis and interpretation of data and in the writing of the manuscript.

The pre-publication history for this paper can be accessed here:

http://www.biomedcentral.com/1472-6947/5/24/prepub

Research assistants of the PMTCT project at Karatina District Hospital-Kenya helped with the translation of the English questionnaire to Kikuyu and Kiswahili, did voices for the ACASI and were instrumental in informing the respondents about the study. We also thank the staff at the Karatina District Hospital, antenatal clinic for their cooperation. The University of Bergen, Norway funded the study and had no influence on any decisions taken in the study.