A Cross-Sectional Analysis of Randomized Survey Modality to Assess Impact of Survey Modality on Demographic,Behavioral,and Psychosocial Measures among Men Who Have Sex with Men in Kenya

Abstract

Objectives

We assessed whether survey mode influenced reporting of sexual behaviors and psychosocial factors among men who have sex with men (MSM) in Kenya.

Methods

In cross-sectional analysis of baseline data from 493 MSM in Kisumu and Nairobi enrolled in a prospective cohort study, participants were randomized 1:1 to Computer-Assisted Personal Interview (CAPI) or Audio Computer-Assisted Self-Interview (ACASI). We compared responses across survey modes using Poisson regression with robust variance, adjusting for socio-demographics.

Results

In both sites, CAPI users more frequently reported sex with a female partner. In Kisumu, CAPI users were less likely to report receptive anal intercourse. In Nairobi, CAPI users were less likely to report food insecurity, transactional sex, and STI symptoms.

Conclusion

While most responses were similar across modes, ACASI prompted higher reporting of sensitive behaviors, highlighting added value for capturing stigmatized and sensitive information. Offering both methods may enhance data quality and respect participant preferences.

Plain Language Summary Title

Different ways of asking survey questions affect the answers given for sensitive and stigmatized behaviors

Plain Language Summary

We wanted to know if the way surveys are given changes how men who have sex with men answer sensitive questions about their lives and behaviors. We surveyed 493 men, who were randomly assigned to either answer questions face-to-face with an interviewer (CAPI) or privately on a computer with audio support (ACASI). We found that men answering face-to-face were more likely to report having sex with women, but less likely to report certain behaviors and challenges, such as receptive anal sex, food insecurity, transactional sex, and STI symptoms. Overall, most answers were similar, but the computer based method encouraged more reporting of sensitive and stigmatized behaviors. We recommend offering both methods to improve data quality and give participants a choice in how they share information.

Keywords

Audio computer assisted personal interview computer personal interview sensitive behaviour MSM sexual gender minority antibiotic use cis gender RAI

Introduction

Despite declining HIV incidence and prevalence globally, men who have sex with men (MSM) and other sexual and gender minorities continue to bear the greatest burden of HIV due to biological, social, and structural barriers.¹

In behavioural research, the accuracy of self-reported data is crucial. Computer assisted personal interviewing (CAPI) involves an interviewer administering the survey using a computer, which can facilitate complex questionnaire designs and improve data accuracy. However, the presence of an interviewer may lead to social desirability bias, either due to familiarity with the interviewer or fear of disclosure which may lead to underreporting of stigmatized behaviours. Audio computer assisted self-interviewing (ACASI), allows respondents to listen to recorded questions through headphones and read the same on computer screen and enter their responses directly into a computer, providing a greater sense of privacy. A systematic literature review including results from 20 different studies published 2000–2021 comparing ACASI to other survey modes in Asia and Sub-Saharan Africa found that ACASI was more likely to elicit sensitive behaviours, such as forced sex, multiple partners, and transactional sex, compared to face-to-face interviews (FTFI).²

However, ACASI is not always more effective than CAPI in collecting sensitive information. A randomized study conducted in rural South Africa in 2015 with 504 adults compared different electronic survey methods for sexual behaviour. The study found that self-interview methods like ACASI and computer-assisted self-interviewing (CASI) were feasible and acceptable, but they required more effort and reading ability from participants. Additionally, item non-response rates were higher in self-interview arms. As a result, authors recommended the choice of method should be based on context-specific criteria.³ In addition, a United States study of 22,862 ACASI interviews conducted between 2006 and 2010, found that interviewer behaviour mattered. For example, being close enough to see the computer screen or helping participants during ACASI influenced how participants answered sensitive questions.⁴

Given these mixed findings, it is essential to compare ACASI and CAPI within the specific cultural and social context of MSM populations in Kenya, where same sex behaviours are highly stigmatized and criminalized.⁵ The objective of this analysis was to assess how survey modality may impact responses among a cohort study of MSM in Kisumu and Nairobi, Kenya, especially regarding behaviours that are stigmatized. We did not pre-specify a hypothesis that one modality would lead to more frequent reporting than the other regarding sensitive sociodemographics, behavioral practices, or psychosocial measures.

Methods

Ethical Approval and Informed Consent

This study was approved by the institutional review boards of Jaramogi Odinga Oginga Teaching and Referral Hospital in Kenya (JOOTRH, ISERC/JOOTRH/754/23), Rush University Medical Center in United States (RUMC, #23060502-IRB1), and University of Manitoba in Canada (UM, HS26665). Written informed consent was obtained for all participants in their preferred language (English, DhoLuo, Kiswahili) for the collection of data and samples and processing and analysis of those data and samples.

Study Setting, Design, and Participants

The reporting of this study conforms to STROBE guidelines for reporting observational studies⁶ (Supplemental File 1). Mbili Pamoja (“Two Together” in Swahili), is a prospective study conducted in Nairobi (the capital of Kenya) and Kisumu (Kenya's third largest city, and located in the west adjacent to Lake Victoria). The 2022 Kenyan Demographic Health Survey demonstrates the differences between Nairobi and Kisumu counties. In Kisumu county, 84.3% of men aged 15–49 report ever being tested for HIV, and this is 89.2% for men in Nairobi county.⁷ Socioeconomic indicators also differ, with 70.5% of Nairobi City residents being in the highest wealth quintile compared to just 13.5% of Kisumu residents. While 26.6% and 30.7% of Nairobi City men have completed secondary and post-secondary education, respectively, these figures lie at 17.0% and 11.2% for Kisumu men.

Recruitment was done between June 7, 2024, and November 14, 2024. Participants were recruited from existing HIV prevention programs in both study sites. In the two sites, peer educators were informed about the study and were tasked with contacting their peers; contact was either physical, online, or through phone calls. Potential participants were also informed during their quarterly visits. Both sites have WhatsApp groups where the information about the study was posted. In Kisumu, information was also passed during weekly activities. The participants then self-selected and came for screening and enrolment. Some participants booked for enrolment appointment and others came as walk-ins. Rescheduling was done appropriately where the numbers exceeded the daily target.

To be eligible, participants had to be aged 18–39 years, assigned male sex at birth, reporting to have had sex with another man in the past 3 months, agreeing to study procedures, and signing an informed consent. Because Mbili Pamoja study is measuring the microbiome and rectal samples are taken, having had diarrhoea or vomiting in the past 14 weeks was considered as temporarily ineligibility. After symptoms resolved, potential participants could come for rescreening while enrolment remained open. Participants would be followed for one year with an enrolment visit, 6 months visit and 12 months visit.

Sample Size Calculation

Sample size for the prospective Mbili Pamoja cohort study was calculated to estimate the change in bacterial composition and features from time of infection to post-treatment. Expecting to test up to 300 features, assuming 90% proportion of nulls (ie, 10% true associations), with false discovery rate (FDR) < 0.05, and effect sizes 0.20–0.25, we estimated needing 75 incident STIs to achieve 80% power. Based on the STI incidence of 18 per 100 person years observed in our prior study,⁸ we would need to enroll 417 men. Estimating 85% follow-up based on the prior cohort, we sought to enrol 491 men and rounded this to 500. One extra participant was enrolled in Nairobi as he had already come for his appointment.

Data Collection

Data was collected through audio computer assisted self-interview (ACASI) and computer assisted personal-interview (CAPI). Participants were randomized in a ratio of 1:1 to either ACASI or CAPI and could complete the survey in their preferred language (English, DhoLuo, and Kiswahili). The randomization was done in Excel (subscription-based Microsoft 365 with automatic updates) stratified by study site using the random number generator function, with values > 0.5 assigned to ACASI and values ≤ 0.5 assigned to CAPI. The randomization assignment was assigned according to the pre-generated study ID numbers, and shared to study sites via university hosted shared folders. There was no allocation concealment as the survey mode was assigned along with the study identification number – ie, as participants were enrolled. The receptionist checked the randomization assignment spreadsheet and placed a coloured sticker corresponding to assigned mode on the participant file. To verify that participants completed the survey according to their randomization assignment, the survey mode completed was compared to the randomization assignment. We identified eight observations in which the randomization assignment did not match the mode of survey completed (Supplemental File 4). Two occurred in Nairobi and six in Kisumu, and arose from an error at reception whereby the wrong colour sticker was placed on the participant file; these eight observations were excluded from analysis.

For CAPI, a trained member of the research staff administered the survey (ie, read the questions and responses to the participant) and recorded participant responses directly into the computer. Staff was trained over a 2-week period. The first week was familiarizing with study procedures and practicing the consents, eligibility screening, and surveys among each other. The second week of training involved practicing eligibility screening, consent, and surveys with program clients and peers who would not otherwise be eligible for the study. One of the study investigators (JK, FO, LM, SM) observed the staff members during these practice sessions, providing feedback and ensuring survey script was followed.

Data was collected on sociodemographic factors, sexual practices, alcohol and substance use, depressive symptoms, and social support (Supplemental File 2 for complete survey in English, Swahili, DhoLuo; Supplemental File 3 for survey administration protocol). Many of the sociodemographic and behavioral practice questions have been used in our prior studies, having known understandability and relation to sexual health outcomes.^8,9 The nine question Patient Health Questionnaire (PHQ-9) assessed depressive symptoms, dichotomized at ≥5 for moderate or greater symptoms.¹⁰ Alcohol dependency was assessed using the Alcohol Use Disorders Identification Test (AUDIT) tool dichotomized at ≥8 for harmful alcohol use.¹¹ The 10-item drug abuse screening test (DAST) was used to measure potential drug dependency, with responses dichotomized at a threshold of ≥3 for harmful drug use.¹² Stigma related to same sex behaviour was assessed with a series of 11 questions, summed and normalized to a 0 to 100 scale, and dichotomized at the median for analysis.¹³ Prior to initiating the study, surveys were piloted with staff and peers at both study sites to ensure accuracy of skip patterns and understandability.

Statistical Analysis

In this cross-sectional analysis of baseline data, the main objective was to estimate the effect of survey modality (ACASI vs CAPI) on responding to demographics, psychosocial measures, and sensitive behavioural questions. Frequency distributions were compared with Chi-square test, and Fishers exact test was adopted for categorical variables with n < 5 in any cells. Poisson regression with robust variance estimate was used to model the association between survey mode and the independent variables and prevalence ratios (PR) with 95% confidence intervals (CI) are reported. Multivariable analysis was performed for variables differing by survey mode at the p < 0.10 level, adjusting for age, educational attainment, employment status, and HIV status. We adjusted for these variables because we hypothesized based on the literature that they would be associated with how participants may respond to sensitive sexual practice questions,¹⁴ and might also differ by survey mode.²

Results

In Kisumu, 256 individuals presented for screening, of whom 250 were eligible, and none refused participation. In Nairobi, 270 participants were screened, 262 were eligible, and 251 were enrolled. Reasons for ineligibility included (not mutually exclusive): not agreeing to specimen collection (n = 13), diarrhea or vomiting in past 2 weeks (n = 5), or not reporting sex with another man in the past 3 months (n = 3). Among the 11 individuals who were eligible and refused, three provided reasons: uncertain and wants to think about it (n = 1), does not want to take HIV test (n = 1), does not want to have penile specimen taken (n = 1). As noted above, eight observations in which the completed survey mode did not match the randomization assignment were excluded from these analyses (Supplemental File 4).

Characteristics of Study Sample

Participant characteristics differed substantially by study site (Table 1). Participants in Kisumu were older than participants in Nairobi (median age of 27 years vs 23 years, p = 0.001). Having secondary education or more was more common among Nairobi participants (89.6% vs 73.4% Kisumu). Being non cis-male was more common among Nairobi participants (PR = 1.14; 95% CI: 1.05–1.24). Reporting antibiotic use in the past 30 days was more commonly reported in Nairobi than Kisumu (30.1% vs 20.5%, PR = 1.47; 95% CI: 1.08–2.01). Regarding sexual practices, ever having receptive anal intercourse was more commonly reported among Nairobi than Kisumu participants (PR = 1.32; 95% CI: 1.19–1.45), while having sex with a female was less common among Nairobi participants (PR = 0.57; 95% CI: 0.48–0.69). Reporting recent sex with a female partner was more common among Kisumu participants, and having a current boyfriend/main partner was less common in Nairobi than in Kisumu. Compared to Kisumu participants, Nairobi participants were more likely to report urethral discharge (PR = 2.04; 95% CI: 1.31–3.18), dysuria (PR = 1.52; 95% CI: 1.07–2.14), or anal discharge (PR = 3.63; 95% CI: 1.84–7.13). HIV prevalence was over four times higher in Nairobi than in Kisumu (PR = 4.19; 95% CI: 2.59–6.79).

Table 1.

Distribution of Participant Characteristics by Study Site.

	Kisumu, N = 244 N (%)	Nairobi, N = 249 N (%)	p-Value	Prevalence Ratio (95% CI)
Age in years (median, IQR)	27 [23–30]	23 [21–27]	<0.001	0.91 (0.88-0.94)
Educational Attainment
Less than Secondary school	65 (26.6)	26 (10.4)
Secondary school or more	179 (73.4)	223 (89.6)	<0.001	1.22 (1.12-1.33)
Worried about whether food would run out in the past 30 days	167 (68.4)	163 (65.5)	0.482	0.96 (0.84-1.08)
Employed versus Unemployed	182 (74.6)	140 (56.2)	<0.001	0.75 (0.66-0.86)
Used antibiotics in the past 30 days	50 (20.5)	75 (30.1)	0.014	1.47 (1.08-2.01)
Gender Identity	227 (93.0)	207 (83.1)
Cis-male	17 (7.0)	42 (16.9)	0.003	1.14 (1.05-1.24)
Transgender Woman, gender queer, other identity
Receptive anal intercourse ever	163 (66.8)	219 (88.0)	<0.001	1.32 (1.19-1.45)
Vaginal or anal sex with a cisgender female ever (missing, n = 1)	159 (65.2)	93 (37.4)	<0.001	0.57 (0.48-0.69)
Currently have a boyfriend/main partner/husband (missing, n = 1)	178 (72.9)	153 (61.7)	0.008	0.85 (0.75-0.96)
Any female sex partner past one week (missing, n = 1)	22 (9.0)	4 (1.6)	<0.001	0.18 (0.06-0.51)
Received money or gifts from a partner at last sex (missing, n = 1)	97 (39.8)	113 (45.4)	0.283	1.05 (0.83-1.34)
Self-reported penile discharge	25 (10.3)	52 (20.9)	0.001	2.04 (1.31-3.18)
Self-reported anal discharge	10 (4.1)	37 (14.9)	<0.001	3.63 (1.84-7.13)
Self-reported pain or burning on urination	42 (17.2)	65 (26.1)	0.017	1.52 (1.07-2.14)
Known HIV positive	18 (7.4)	77 (30.9)	<0.001	4.19 (2.59-6.79)
Among HIV negative participants
Never taken PrEP	48 (21.2)	59 (34.3)	0.008	Ref
Taken PrEP in past	89 (39.4)	64 (37.0)		0.91 (0.84-0.99)
Currently taking PrEP	89 (39.4)	49 (28.5)		0.87 (0.80-0.95)
Stigma scale median or higher versus below median (≥21 vs <21)	106 (43.4)	144 (57.8)	0.001	1.33 (1.11-1.59)
Patient Health Questionnaire elevated depressive symptoms (≥ 5 vs < 5)	98 (40.2)	147 (59.0)	<0.001	1.47 (1.22-1.77)
AUDIT harmful or hazardous alcohol use (≥8 vs <8)	97 (39.8)	73 (29.3)	0.015	0.74 (0.58-0.94)
DAST drug dependency (≥ 3 vs < 3)	82 (33.6)	62 (24.9)	0.034	0.74 (0.56-0.98)

Distribution of Participant Characteristics by Survey Mode in Kisumu and Nairobi

ACASI surveys were completed in English (88.5%), Swahili (8.2%), or Luo (3.3%), while CAPI surveys were completed in English (92.9%), Swahili (5.6%), or Luo (1.6%) (p = 0.230). The median duration for survey completion was 32.9 min for ACASI (IQR 26.2-39.0 min) and 26.8 min for CAPI (IQR 22.7–34.3 min) (p < 0.001).

In Kisumu, employment was more commonly reported on CAPI than ACASI (78.9% vs 70.3%, p = 0.122), while gender identity and educational attainment were similarly distributed (Table 2). Recent antibiotic use was 46% less likely to be reported in CAPI than in ACASI (PR = 0.46; 95% CI: 0.27–0.79). Reporting to have ever had vaginal or anal sex with a female partner was more common in CAPI than ACASI (78.1% vs 52.5%, p < 0.001). Psychosocial measures (depressive symptoms, alcohol use, drug use, stigma) varied somewhat by survey mode, but differences were attenuated and non-significant in adjusted analyses. Adjusted for age, educational attainment, employment status, and HIV status, the differences by survey mode in reported employment, recent antibiotic use, receptive anal intercourse, and ever having had sex with a female partner remained statistically significant (Table 2 and Figure 1).

Figure 1.

Results of multivariable regression: prevalence ratio of responses on CAPI versus ACASI by study site. Legend: The figure shows the adjusted prevalence ratio (center node) and 95% confidence intervals (bar width) for variables that differed p < 0.05 by survey mode (CAPI vs ACASI), adjusted for age, educational attainment, employment status, and HIV status.

Table 2.

Distribution of Survey Responses by Survey Modality in Kisumu.

	ACASI, N = 121 N (%)	CAPI, N = 123 N (%)	P-Value¹	PR (95% CI)	Adjusted PR² (95%CI)
Median age in years (IQR)	27 (23-30)	26 (22-29)	0.142	0.99 (0.98-1.01)
Educational Attainment
< Secondary school	35 (28.9)	30 (24.4)		Ref
Secondary school or more	86 (71.1)	93 (75.6)	0.423	1.06 (0.91-1.24)
Employed	85 (70.3)	97 (78.9)	0.122	1.12 (0.97-1.30)	1.16 (1.01-1.34)
Worried about whether food would run out in the next 30 days	87 (71.9)	80 (65.0)	0.249	0.90 (0.76-1.07)
Antibiotics use past 30 days	34 (28.1)	16 (13.0)	0.004	0.46 (0.27-0.79)	0.52 (0.30-0.91)
Gender Identity
Cis-male	110 (90.9)	117 (95.1)		Ref
Other Identity	11 (9.1)	6 (4.9)	0.196	0.89 (0.75-1.06)
Receptive anal intercourse ever	90 (74.4)	73 (59.4)	0.013	0.80 (0.67-0.96)	0.80 (0.67-0.96)
Vaginal or anal sex with a cisgender female ever (missing, n = 1)	63 (52.5)	96 (78.1)	<0.001	1.49 (1.22-1.81)	1.51 (1.24-1.84)
Currently have a boyfriend/main partner/husband	85 (70.3)	93 (75.6)	0.346	1.08 (0.92-1.26)
Any female sex partner past one week	9 (7.4)	13 (10.6)	0.393	1.42 (0.63-3.21)
Given money or gifts by partner at last sex (missing, n = 1)	53 (44.2)	44 (35.8)	0.182	0.81 (0.59-1.11)
Self-reported penile discharge	11 (9.1)	14 (11.4)	0.555	1.25 (0.59-2.65)
Self-reported anal discharge	6 (5.0)	4 (3.3)	0.538	0.66 (0.19-2.27)
Self-reported pain or burning on urination	22 (18.2)	20 (16.3)	0.691	0.89 (0.51-1.55)
Among HIV negative participants			0.997
Never taken PrEP	19 (17.8)	20 (18.2)		Ref
Taken PrEP in past	44 (41.1)	45 (40.9)		1.00 (0.88-1.23)
Currently taking PrEP	44 (41.1)	45 (40.9)		1.00 (0.88-1.23)
Known HIV positive	9 (7.4)	9 (7.3)	0.971	0.98 (0.40-2.40)
Test positive for urethral CT and/or NG	10 (8.3)	13 (10.6)	0.538	1.28 (0.58-2.81)
Test positive for rectal CT and/or NG	5 (4.2)	4 (3.3)	0.747	0.78 (0.21-2.84)
Depressive symptoms (PHQ-9 score)
< 5	66 (54.5)	80 (65.0)		Ref
>=5	55 (45.5)	43 (35.0)	0.095	0.77 (0.56-1.05)	0.80 (0.58-1.09)
Hazardous or harmful alcohol use (AUDIT score)
< 8	66 (54.6)	81 (65.9)		Ref
>=8	55 (45.5)	42 (34.2)	0.074	0.76 (0.55-1.03)	0.76 (0.55-1.03)
Hazardous or harmful drug use (DAST score)
< 3	86 (71.1)	76 (61.8)		Ref
>=3	35 (28.9)	47 (38.2)	0.125	1.32 (0.92-1.89)
Stigma scale
Below Median (<21)	63 (52.1)	75 (61.0)		Ref
At or Above Median (>=21)	58 (47.9)	48 (39.0)	0.160	0.81 (0.61-1.09)

PR = Prevalence Ratio; aPR = Adjusted Prevalence Ratio; 95% CI = 95% Confidence Interval.

Chi-square p-value or Fisher's exact where cell size n < 5; Wilcoxon rank sum p-value for non-normally distributed continuous variables.

Each model is adjusted for: age, educational attainment, employment status, and HIV status.

In Nairobi, in adjusted analyses, worrying about food running out (aPR = 0.77; 95% CI: 0.64–0.92) and recent antibiotic use (aPR = 0.56; 95% CI: 0.38–0.83) were less likely to be reported in CAPI than in ACASI (Table 3 and Figure 1). Vaginal or anal sex with a cisgender female was 55% more likely to be reported in CAPI than in ACASI (aPR = 1.55; 95% CI:1.11–2.16), whereas practising transactional sex was 26% less likely to be reported in CAPI than ACASI (aPR = 0.74; 95% CI: 0.56–0.97). Urethral discharge (aPR = 0.46; 95% CI: 0.27–0.78) and anal discharge (aPR = 0.38; 95% CI: 0.19–0.76) were less likely to be reported in CAPI than ACASI.

Table 3.

Nairobi: Distribution of Survey Responses by Survey Modality.

	ACASI, N = 122 N (%)	CAPI, N = 127 N (%)	p-Value	PR (95% CI)	APR¹ (95%CI)
Age in years (median, IQR)	23 [21-27]	23 [21-27]	0.920	1.00 (0.95- 1.05)
Educational attainment
< Secondary school	14 (11.5)	12 (9.5)		Ref
Secondary school or more	108 (88.5)	115 (90.6)	0.601	1.02 (0.94-1.11)
Employed	72 (59.0)	68 (53.5)	0.384	0.91 (0.73-1.13)
Worried about whether food would run out in the next 30 days	90 (73.8)	73 (57.5)	0.007	0.78 (0.65-0.94)	0.77 (0.64-0.92)
Antibiotics use past 30 days	48 (39.3)	27 (21.3)	0.002	0.54 (0.36-0.81)	0.56 (0.38-0.83)
Gender identity
MSM	104 (85.3)	103 (81.1)		Ref
Other identity	18 (14.8)	24 (18.9)	0.383	1.04 (0.94-1.17)
Receptive anal intercourse ever	105 (86.1)	114 (89.8)	0.370	1.04 (0.95-1.14)
Vaginal or anal sex with a cisgender female	36 (29.5)	57 (44.9)	0.012	1.52 (1.09-2.13)	1.55 (1.11-2.16)
Current boyfriend/main partner/husband	75 (62.0)	78 (61.4)	0.927	0.99 (0.81-1.21)
Any female sex partner past one week	2 (1.6)	2 (1.6)	0.974	0.97 (0.14-6.79)
Given money or gifts by partner at last sex	63 (51.6)	50 (39.4)	0.052	0.76 (0.58-1.01)	0.74 (0.56-0.97)
Self-reported penile discharge	35 (28.7)	17 (13.4)	0.003	0.47 (0.28-0.79)	0.46 (0.27-0.78)
Self-reported anal discharge	26 (21.3)	11 (8.7)	0.005	0.41 (0.21-0.79)	0.38 (0.19-0.76)
Self-reported pain or burning on urination	33 (27.1)	32 (25.2)	0.739	0.93 (0.61-1.42)
Among HIV negative participants
Never taken PrEP	21 (27.6)	23 (28.4)	0.993	ref
Taken PrEP in past	31 (40.8)	33 (40.7)		1.00 (0.88-1.13)
Currently taking PrEP	24 (31.6)	25 (30.9)		1.11 (0.87-1.13)
Known HIV positive	36 (29.5)	41 (32.3)	0.636	1.09 (0.75-1.59)
Test positive for urethral CT and/or NG	11 (9.0)	15 (11.9)	0.458	1.32 (0.63-2.76)
Test positive for rectal CT and/or NG	38 (31.2)	34 (27.0)	0.470	0.87 (0.59-1.28)
Depressive symptoms (PHQ-9 score)
< 5	48 (39.3)	54 (42.5)		Ref
>=5	74 (60.7)	73 (57.5)	0.611	0.95 (0.77-1.17)
Hazardous or harmful alcohol use (AUDIT score)
< 8	84 (68.9)	92 (72.4)		Ref
>=8	38 (31.2)	35 (27.6)	0.534	0.88 (0.60-1.30)
Hazardous or harmful drug use (DAST score)
< 3	92 (75.4)	95 (74.8)		Ref
>=3	30 (24.6)	32 (25.2)	0.912	1.02 (0.66-1.58)
Stigma scale
Below median (<21)	50 (41.0)	55 (43.3)		Ref
At or above median (>=21)	72 (59.0)	72 (56.7)	0.711	0.96 (0.78-1.19)

Discussion

Our study examined the sociodemographic and behavioural characteristics of MSM and gender-diverse individuals in Kisumu and Nairobi, Kenya, and evaluated the impact of two data collection methods, CAPI and ACASI, on the distribution of responses. We found differences in participant characteristics by study site, and some differences in reporting by survey modality.

Participant characteristics differed between the two sites. Participants in Nairobi were generally younger, more educated, and showed higher frequency of gender diversity than those in Kisumu, which could be a reflection of urbanization characteristics.¹⁵ The Kenya Demographic and Health Survey report of 2022 affirms these differences, as urban areas are more likely to provide access to better educational resources.⁷ Our study findings are consistent with previous studies in Kenya and other sub-Saharan African settings, which have highlighted urban-rural disparities in sexual behaviours, HIV prevalence, and STI burden among MSM.^16,17 The higher HIV prevalence in Nairobi aligns with prior evidence of elevated risk in urban centres due to more extensive sexual networks and varying access to HIV prevention services.¹⁸ There was a higher prevalence of receptive anal intercourse and lower likelihood of female partners among Nairobi participants as compared to Kisumu participants, which could reflect differences in social norms, exposure to diverse sexual networks, and urban anonymity, which may increase the likelihood of exploration of diverse sexual identities and practices.¹⁹

Most response frequencies did not differ by survey modality, but ACASI elicited more frequent reporting of certain sensitive behaviours compared to CAPI: antibiotic use, receptive anal intercourse (RAI), and transactional sex. This is consistent with other studies in which ACASI has the potential of reducing social desirability bias in comparison to face-to-face interviews.²⁰ Greater likelihood of reporting RAI through ACASI is an important self-disclosure of sensitive HIV risk behaviour, as previous research has shown that unprotected RAI increases the risk of HIV acquisition by 17 times compared to unprotected vaginal sex.²¹ The increased frequency of reporting sex with cis-female partners in CAPI than ACASI may be due to cultural or community-level influences where bisexual behaviours are less stigmatized.²² Thus, sex with females may be overreported via CAPI to meet these expectations. However, in a prior study of 158 MSM in Kisumu initiating HIV pre-exposure prophylaxis (PrEP), having a regular female sex partner was a risk for STI acquisition.²³ Therefore, it is unclear whether sex with cis-females is being over-reported on CAPI, under-reported on ACASI, or both. “Sensitive” topics remain sensitive due to persistent stigma. Stigma and discrimination among MSM, especially in Sub-Saharan Africa, may lead them to marry women due to pressure to fulfil societal and cultural obligations while continuing to have sexual relationships with their male partners.²⁴ Addressing these transmission pathways requires targeted interventions, including promoting safe sex practices, encouraging regular HIV testing, reducing stigma associated with same-sex behaviour, and providing culturally sensitive education to both MSM and the general population.²⁵ Policies that may support sustained, population level changes to reduce stigma include protections against discrimination based on sexual orientation and decriminalization of same sex behaviour.^26,27

Self-reported recent antibiotic use was common, reported by 20% and 30% of Kisumu and Nairobi participants, respectively. In our study, the most common reasons for antibiotic use (results not shown) were a cold (n = 49), fever (n = 22), discharge from penis or burning on urination (n = 10), an STI (n = 18), skin infection (n = 10), diarrhoea (n = 8), and a variety of conditions (n = 20) that included few specific diagnoses that supported provider prescribed treatment (eg, tonsillitis, typhoid). Among participants at both sites, recent antibiotic use was less likely to be reported on CAPI than ACASI, suggesting there may be some stigma involved in the reporting of this. We hypothesize that this is more likely an indication of self-medication rather than provider-prescribed treatment. Meta-analyses find that self-medication with antibiotics is common in low- and middle- income countries in which antibiotics are not regulated.²⁸ This practice can be a response to difficulty affording healthcare.²⁹ Solutions to self-treatment with antibiotics are outside the scope of the current study, but future work can assess common reasons for use and develop and strengthen referrals.

Participants in Nairobi were more likely to report STI symptoms. Analyses of factors associated with STI is outside the scope of the current paper, but notably, the prevalence of urethral chlamydia and/or gonorrhoea did not differ by study site (9.2% Kisumu vs 10.4% Nairobi, p = 0.662), though rectal infection was more common in Nairobi (29.5% vs 3.7% Kisumu, p < 0.001). Increased reporting of urethral or anal discharge on ACASI among Nairobi participants may reflect confidentiality concerns. Procedures were the same in Kisumu, where this discrepancy in symptom reporting by survey mode was not observed, and at both study sites the CAPI was conducted by a non-clinical member of staff. Although most gonorrhoea and chlamydia infections detected by nucleic acid amplification test are asymptomatic,³⁰ the disclosure of symptoms is important to support syndromic treatment.

Other differences in reporting by modality between Kisumu and Nairobi related to being employed, worrying about food running out, and transactional sex. Employment status and worrying about food running out are both related to socioeconomic status, which people may be reluctant to report having lower income.³¹ Future studies may potentially mitigate this by using standardized indices that capture multiple dimensions of socioeconomic status. The frequency of transactional sex reported differentially by survey mode for Nairobi and not Kisumu. This may stem from greater perceived stigma associated with transactional sex among Nairobi participants, though we cannot determine if this is due to perceptions about study staff, underlying differences in participants, or different experiences between cities, highlighting the importance of multisite studies.

Reporting of stigma did not differ by CAPI versus ACASI for either study site. The effectiveness of ACASI in eliciting sensitive behavioural data aligns with global research showing its usefulness in reducing social desirability bias.³² Our results suggest ACASI can be more suitable for capturing nuanced behaviours among sexual- and gender-diverse individuals, while CAPI may suffice for less sensitive questions. Using both methods could enrich data collection while enhancing participant comfort and data accuracy. At future study rounds, we will offer participants their choice of ACASI or CAPI, and will be able to examine how reporting varies by self-selected modality. We have held dissemination meetings with the research participants and summarized the results of this analysis to them in lay terms. Overall, they were not surprised by the findings and agreed with our recommendation to offer both options so that participants may choose. They gave feedback for future studies that rely on ACASI to break it into shorter portions or build in breaks to reduce participant fatigue. Given the reporting differences by survey mode, future analyses of certain variables in relation to STI or HIV risk will need to account for mode of data collection.

Strengths and Limitations

Our study enhances the generalizability of findings due to the multi-site design by capturing diverse sociocultural contexts, though our findings are specific to MSM and gender-diverse individuals in Kisumu and Nairobi and may not be generalizable to other populations or regions. Our study provides insights into the influence of data collection modes on behavioural reporting and the importance of methodological considerations when conducting sensitive research. The self-reported data are subject to recall bias and may still be influenced by cultural norms despite the advantages of ACASI. Although we used validated scales where available, questions we had used in previous studies, and piloted our surveys for accuracy and understandability prior to study implementation, we did not record any meta-data on the pilot testing, nor any interviewer performance metrics for CAPI prior to implementation. We did not measure digital literacy, which could affect how participants responded on ACASI versus CAPI. Measure of this in relation to response patterns and how they differ by survey modality could help inform construction of future ACASI and CAPI surveys. We did not record the number of instances in which assistance was required for ACASI, but study coordinators and staff who administered surveys report this was uncommon and primarily during internet outages; in “rare” instances, a participant asked to change the language. As reported, survey time was approximately 5 min faster for CAPI versus ACASI, but we did not measure time per section, which could have informed which sections may have been more cumbersome for ACASI takers. We cannot infer whether the 5-min difference, on average, was related to digital fluency, understanding of questions, or some other factor. Whether participants were randomized to ACASI or CAPI was not concealed from staff, but we did not find time trends in survey mode.

Conclusions

It also supports the utility of ACASI in eliciting sensitive behavioural data among MSM and gender diverse individuals in Kisumu and Nairobi. Based on our findings, studies of similar nature should consider: (1) the acceptability, feasibility, and impact of offering participants their choice of ACASI or CAPI, and (2) whether using both modalities may reduce participant burden and facilitate data collection. Given the reporting differences by survey mode, statistical analyses of certain variables in relation to STI or HIV risk will need to account for mode of data collection.

Supplemental Material

sj-docx-1-jia-10.1177_23259582261434295 - Supplemental material for A Cross-Sectional Analysis of Randomized Survey Modality to Assess Impact of Survey Modality on Demographic, Behavioral, and Psychosocial Measures among Men Who Have Sex with Men in Kenya

Supplemental material, sj-docx-1-jia-10.1177_23259582261434295 for A Cross-Sectional Analysis of Randomized Survey Modality to Assess Impact of Survey Modality on Demographic, Behavioral, and Psychosocial Measures among Men Who Have Sex with Men in Kenya by Joseph Nzioka, Patriciah Wambua, Fredrick Otieno, Joshua Kimani, Rhoda Kabuti, Felix Ochieng, Richard Gichuki, Mary Wanjiru, Monica Okumu, Lyle McKinnon and Supriya D Mehta in Journal of the International Association of Providers of AIDS Care (JIAPAC)

Supplemental Material

sj-docx-2-jia-10.1177_23259582261434295 - Supplemental material for A Cross-Sectional Analysis of Randomized Survey Modality to Assess Impact of Survey Modality on Demographic, Behavioral, and Psychosocial Measures among Men Who Have Sex with Men in Kenya

Supplemental material, sj-docx-2-jia-10.1177_23259582261434295 for A Cross-Sectional Analysis of Randomized Survey Modality to Assess Impact of Survey Modality on Demographic, Behavioral, and Psychosocial Measures among Men Who Have Sex with Men in Kenya by Joseph Nzioka, Patriciah Wambua, Fredrick Otieno, Joshua Kimani, Rhoda Kabuti, Felix Ochieng, Richard Gichuki, Mary Wanjiru, Monica Okumu, Lyle McKinnon and Supriya D Mehta in Journal of the International Association of Providers of AIDS Care (JIAPAC)

Supplemental Material

sj-docx-3-jia-10.1177_23259582261434295 - Supplemental material for A Cross-Sectional Analysis of Randomized Survey Modality to Assess Impact of Survey Modality on Demographic, Behavioral, and Psychosocial Measures among Men Who Have Sex with Men in Kenya

Supplemental material, sj-docx-3-jia-10.1177_23259582261434295 for A Cross-Sectional Analysis of Randomized Survey Modality to Assess Impact of Survey Modality on Demographic, Behavioral, and Psychosocial Measures among Men Who Have Sex with Men in Kenya by Joseph Nzioka, Patriciah Wambua, Fredrick Otieno, Joshua Kimani, Rhoda Kabuti, Felix Ochieng, Richard Gichuki, Mary Wanjiru, Monica Okumu, Lyle McKinnon and Supriya D Mehta in Journal of the International Association of Providers of AIDS Care (JIAPAC)

Supplemental Material

sj-docx-4-jia-10.1177_23259582261434295 - Supplemental material for A Cross-Sectional Analysis of Randomized Survey Modality to Assess Impact of Survey Modality on Demographic, Behavioral, and Psychosocial Measures among Men Who Have Sex with Men in Kenya

Supplemental material, sj-docx-4-jia-10.1177_23259582261434295 for A Cross-Sectional Analysis of Randomized Survey Modality to Assess Impact of Survey Modality on Demographic, Behavioral, and Psychosocial Measures among Men Who Have Sex with Men in Kenya by Joseph Nzioka, Patriciah Wambua, Fredrick Otieno, Joshua Kimani, Rhoda Kabuti, Felix Ochieng, Richard Gichuki, Mary Wanjiru, Monica Okumu, Lyle McKinnon and Supriya D Mehta in Journal of the International Association of Providers of AIDS Care (JIAPAC)

Footnotes

ORCID iDs

Fredrick Otieno

Joshua Kimani

Felix Ochieng

Richard Gichuki

Mary Wanjiru

Monica Okumu

Supriya D Mehta

Ethical Approval and Informed Consent Statements

Author Contributions

JN – Drafting of manuscript, Investigation; PW – Data Analysis, Investigation, Drafting of manuscript; FO – Supervision, Methodology, Resources, Review and Revision of Manuscript; JK – Supervision, Resources, Review and Revision of Manuscript; RK – Investigation, Supervision, Review and Revision of Manuscript; FO – Investigation, Review and Revision of Manuscript; RG – Investigation, Review and Revision of Manuscript; MW – Investigation, Supervision, Review and Revision of Manuscript; MO – Investigation, Review and Revision of Manuscript; LM – Methodology, Resources, Funding Acquisition, Review and Revision of Manuscript; SDM – Data Analysis, Methodology, Resources, Funding Acquisition, Review and Revision of Manuscript.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by grant number R01-AI177005 (Contact MPI: Mehta) from the National Institutes of Health, National Institutes of Allergy and Infectious Diseases.

Declaration of Conflicting Interests

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

Data Availability Statement

The data used in this analysis may be available upon reasonable request and with appropriate ethical approvals.

Supplemental Material

Supplemental material for this article is available online.

References

Beyrer

Baral

Van Griensven

, et al. Global epidemiology of HIV infection in men who have sex with men. Lancet. 2012;380(9839):367-377. doi:10.1016/S0140-6736(12)60821-6

Phoo

NNN

Lobo

Vujcich

Reid

. Comparison of the ACASI mode to other survey modes in sexual behavior surveys in Asia and Sub-Saharan Africa: Systematic literature review. J Med Internet Res. 2022;24(5):e37356. doi:10.2196/37356

Harling

Gumede

Mutevedzi

, et al. The impact of self-interviews on response patterns for sensitive topics: A randomized trial of electronic delivery methods for a sexual behaviour questionnaire in rural South Africa. BMC Med Res Methodol. 2017;17(1):1–14. doi:10.1186/s12874-017-0403-8

West

Peytcheva

. Can interviewer behaviors during ACASI affect data quality? Surv Pract. 2014;7(5). doi:10.29115/SP-2014-0023

Shaw

. From disgust to dignity: Criminalisation of same-sex conduct as a dignity taking and the human rights pathways to achieve dignity restoration. Afr Human Rights Law J. 2018;18(2):684-705. doi:10.17159/1996-2096/2018/v18n2a12

Von Elm

Altman

Egger

Pocock

Gøtzsche

Vandenbroucke

. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies. Lancet. 2007;370(9596):1453-1457. doi:10.1016/S0140-6736(07)61602-X

Stastics KNBo. 2022 KDHS Key Indicators Report. Kenya National Bureau of Stastics. 2023.

Otieno

Ng'ety

Okall

, et al. Incident gonorrhoea and chlamydia among a prospective cohort of men who have sex with men in Kisumu, Kenya. Sex Transm Infect. 2020;96(7):521-527. doi:10.1136/sextrans-2019-054166

Mehta

Otieno

Kimani

, et al. Transgender women in Kenya experience greater stigma, depressive symptoms, alcohol and drug use and risky sexual practices than cis-gendered men who have sex with men. BMC Public Health. 2023;23(1):1493. doi:10.1186/s12889-023-16348-6

10.

Kroenke

Spitzer

. The PHQ-9: a new depression diagnostic and severity measure. Slack Incorporated Thorofare, NJ; 2002:509-515.

11.

Saunders

Aasland

Babor

De la Fuente

Grant

. Development of the alcohol use disorders identification test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption-II. Addiction. 1993;88(6):791-804. doi:10.1111/j.1360-0443.1993.tb02093.x

12.

Skinner

. The drug abuse screening test. Addict Behav. 1982;7(4):363-371. doi:10.1016/0306-4603(82)90005-3

13.

Secor

Wahome

Micheni

, et al. Depression, substance abuse and stigma among men who have sex with men in coastal Kenya. Aids. 2015;29(3):S251–S2S9. doi:10.1097/QAD.0000000000000846

14.

Catania

Turner

Pierce

, et al. Response bias in surveys of AIDS-related sexual behavior. Methodological issues in AIDS behavioral research. 1993:133-162.

15.

Van Maarseveen

. Urbanization and educational attainment: Evidence from Africa. J Dev Econ. 2025;179:103654.

16.

Mwaniki. Integrated biological and behavioural assessment of human immunodeficiency virus and sexually transmitted infections among tertially student men who have sex with men, Nairobi, Kenya; PROJECT BESPOKE. University of the Witwatersrand Johannessburg; 2023.

17.

Jin

Restar

Beyrer

. Overview of the epidemiological conditions of HIV among key populations in Africa. J Int AIDS Soc. 2021;24(Suppl 3):e25716. doi:10.1002/jia2.25716

18.

Ziraba

Orindi

Muuo

, et al. Understanding HIV risks among adolescent girls and young women in informal settlements of Nairobi, Kenya: Lessons for DREAMS. PloS one. 2018;13(5):e0197479. doi:10.1371/journal.pone.0197479

19.

NASCOP

. 2011 Integrated biological and behavioral surveillance survey among key populations in Nairobi and Kisumu, Kenya. Accessed April. 2010;20:2020.

20.

Ghanem

Hutton

Zenilman

Zimba

Erbelding

. Audio computer assisted self interview and face to face interview modes in assessing response bias among STD clinic patients. Sex Transm Infect. 2005;81(5):421-425. doi:10.1136/sti.2004.013193

21.

Lavoie

Fisher

. Receptive anal intercourse and HIV infection. World J AIDS. 2017;7(4):269–278. doi:10.4236/wja.2017.74023

22.

Rotheram-Borus

Batterham

, et al. Measuring sexual risk using audio computer-assisted self-interviewing (ACASI) versus computer-assisted personal interview (CAPI) in China. Int J Sex Health. 2007;19(1):25-30. doi:10.1300/J514v19n01_04

23.

Mehta

Okall

Graham

N’gety

Bailey

Otieno

. Behavior change and sexually transmitted incidence in relation to PREP use among men who have sex with men in Kenya. AIDS Behav. 2021;25(7):2219–2229. doi:10.1007/s10461-020-03150-3

24.

Fiorentino

Gravier-Dumonceau Mazelier

Yanwou

, et al. Marriage and steady relationships with women in men who have sex with men in Sub-Saharan Africa: A mixed-method systematic review and meta-analyses. AIDS Behav. 2025;29(4):1–34.

25.

Phillips

McCuskey

Ruprecht

Curry

Felt

. Structural interventions for HIV prevention and care among US men who have sex with men: A systematic review of evidence, gaps, and future priorities. AIDS Behav. 2021;25(9):1–13.

26.

Mpabe Bodjongo

. Can the decriminalization of homosexuality counter religious and traditional homophobia in Africa? Rev Law Econ. 2024;20(3):357-401. doi:10.1515/rle-2024-0079

27.

Mogotsi

Otchere

Botchway

Muthoni

Gariseb

Ramalepe

. Psychosocial correlates of LGBTIQ+ experiences in selected African countries: Reimagining LGBTIQ+ research. J Soc Issues. 2024;80(3):1079-1111. doi:10.1111/josi.12640

28.

Torres

Chibi

Middleton

Solomon

Mashamba-Thompson

. Evidence of factors influencing self-medication with antibiotics in low and middle-income countries: A systematic scoping review. Public Health. 2019;168:92-101. doi:10.1016/j.puhe.2018.11.018

29.

Green

Keenan

Fredricks

, et al. The role of multidimensional poverty in antibiotic misuse: A mixed-methods study of self-medication and non-adherence in Kenya, Tanzania, and Uganda. Lancet Glob Health. 2023;11(1):e59-e68. doi:10.1016/S2214-109X(22)00423-5

30.

Kenyon

Herrmann

Hughes

de Vries

. Management of asymptomatic sexually transmitted infections in Europe: Towards a differentiated, evidence-based approach. Lancet Regional Health–Europe. 2023;34:100743. doi:10.1016/j.lanepe.2023.100743

31.

Tadesse

Abate

Zewdie

. Biases in self-reported food insecurity measurement: A list experiment approach. Food Policy. 2020;92:101862. doi:10.1016/j.foodpol.2020.101862

32.

Langhaug

Sherr

Cowan

. How to improve the validity of sexual behaviour reporting: Systematic review of questionnaire delivery modes in developing countries. Trop Med Int Health. 2010;15(3):362-381. doi:10.1111/j.1365-3156.2009.02464.x

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.05 MB

0.02 MB

0.16 MB

0.03 MB

0.00 MB