Francis, Suzanna C. PhD*,†; Lees, Shelley S. MSc, PhD*,†; Andrew, Bahati BA†; Zalwango, Flavia MA‡; Vandepitte, Judith MD, MSc*,‡; Ao, Trong ScD*,†; Baisley, Kathy MSc*; Kapiga, Saidi MD, ScD*,†; Grosskurth, Heiner PhD*,‡; Hayes, Richard DSc*
Approximately 25 million people are infected with HIV in sub-Saharan Africa—two-thirds of the worldwide total. The greatest burden is shouldered by 13.3 million African women, who make up 60% of the epidemic in this region.1 This inequity may reflect not only greater biologic susceptibility, but also behaviors that stem from gender inequities and social and sexual norms, which will differ between populations. There is evidence that some types of intravaginal practices (IVP), such as intravaginal cleansing or insertion (i.e., placing something inside the vagina, such as herbs), may be risk factors for HIV infection; however, there have been conflicting results from prospective studies.2–4 Results from a systematic review suggested an association between IVP and incident HIV; however, a high level of heterogeneity was found between studies, and this was attributed to different terminology and descriptions of IVP.5 In addition, IVP may interfere with HIV prevention methods such as vaginal microbicides; thus, it is important to obtain accurate measures of IVP in microbicide trials.
Collecting valid behavioral data in populations of women at increased risk of HIV is critical for designing appropriate interventions for HIV prevention. A systematic review on the collection of sexual behavioral data in developing countries concluded that questionnaire delivery mode affects the reporting of sexual behaviors.6 Most studies use structured face-to-face interviews (FTFI) to collect retrospective data on IVP. This method is inexpensive and relatively easy to administer to large cohorts, but it is subject to social desirability and recall bias.7,8 In addition, data from FTFI may oversimplify IVP use by asking women to summarize their practices over time.
Prospective diaries have been used to collect data on sensitive behaviors among women at high risk for HIV in sub-Saharan Africa since the late 1990s.9–12 Diaries are daily self-completed data collection tools that can be in written, pictorial,10,13 electronic,14 or verbal (e.g., daily telephone reports15) format. In addition to decreasing social desirability bias by being self-completed, diaries have the advantage of collecting daily events near real time, thereby decreasing recall bias.9,11,16
This article describes the first vaginal practices diary developed for use in research and its application in a multisite microbicide feasibility study among women at increased risk for HIV in Uganda and Tanzania. Our aims were to compare the diary data with IVP data obtained by an FTFI and to investigate the consistency of reporting between methods and within each method over time.
MATERIALS AND METHODS
Vaginal Practice Diary Development
The purpose of the vaginal practice diary was to collect data on the frequency and types of vaginal practices performed by a woman each day and to capture key associated behaviors. The diary was first developed in Tanzania and then adapted to the local study population in Uganda. Initial diary designs and drawings were developed by a team of investigators, including social scientists, reproductive health experts, HIV epidemiologists, and a local artist. A self-completed, pictorial diary was designed to be administered to low-literacy populations by showing pictures of vaginal practice acts. The diary was revised after 6 focus group discussions and 2 pilots with the study population. A final design was based on the acceptability and understandability of the design as rated by the pilot participants (see Fig. 1A). The diary was adapted by a local artist and piloted in Uganda (Fig. 1B). The diary text, instructions, and interviews were in Swahili (Tanzania) or Luganda (Uganda). Although this diary was adapted from a coital diary that was administered for 4 weeks, the vaginal practices diary was designed to be administered for 6 weeks to capture at least 1 complete menstrual cycle.
Main Cohort Studies
In Tanzania, the main cohort enrolled 970 HIV-negative women at high risk of HIV from 3 towns located close to gold or diamond mines in North-West Tanzania. All participants were employed in bars, guesthouses, or other food and recreational facilities and enrolled between August 2008 and August 2009. In Uganda, the main cohort consisted of 1027 women (645 HIV-negative and 382 HIV-positive) living in the capital city, Kampala. Participants were either self-identified sex workers or employed in entertainment facilities, such as bars, nightclubs, and lodges, and enrolled between April 2008 and April 2009. In both settings, participants were seen every 3 months. At each visit, the participant had an FTFI conducted by a research nurse, which included retrospective questions on sexual behavior and IVP during the preceding 3 months. The IVP questions from the FTFI are presented in Figure 2.
Diary Study: Comparison of Diary With FTFI in the Main Cohort
In Tanzania, 100 participants were randomly selected from participants enrolled in the main cohort at 2 of the 3 sites (N = 663) between the months of September 2008 and June 2009. Participants were invited to participate in the Diary Study at the end of their main cohort enrollment visit. Each consenting participant was asked to fill out the diary every day for 6 weeks (42 days) to ensure that at least 1 menstrual period was captured. A female research assistant (RA) taught participants how to use the diary during the enrollment visit, and then visited the participant 2 days later to support learning for how to use the diary. Each participant was visited every week by the same RA to collect the diary, carry out a quality check, issue a new diary, and answer questions about the study. Participants were encouraged to fill in the diary every day and not just before the RA visit. In the Tanzanian cohort, the enrollment and 3-month follow-up FTFI were compared with the diary, representing the visit before and after Diary Study enrollment (see Fig. 3).
A sample size of 100 or approximately 10% of the main study was expected to be sufficient to describe the vaginal practices of the cohort with reasonable precision while remaining logistically feasible; for example, if a practice was reported by 80% of women, the 95% confidence interval would be 0.71 to 0.87.
In Uganda, between July and September 2009, 100 women were enrolled into the Diary Study by selecting every fourth participant at any follow-up visit for the main cohort study. All consenting participants for the Diary Study were asked to follow the same protocol as described earlier for the Tanzania cohort. In the Ugandan site, the diary data were compared with the 3-monthly FTFI before and after Diary Study enrollment (see Fig. 3).
This study was approved by the ethics committees of the London School of Hygiene and Tropical Medicine, the Tanzanian National Institute for Medical Research, the Science and Ethics Committee of the Uganda Virus Research Institute, and the Uganda National Council for Science and Technology. Informed consent was obtained from all participants before enrollment.
Diary and FTFI data were double entered. All analyses were performed using Stata, version 11 (StataCorp, College Station, TX). Three analyses were carried out: a comparison of the second FTFI with the diary data, as the reporting period of this FTFI would overlap with the 6 weeks of the Diary Study; a comparison of the first FTFI to the second FTFI to examine the consistency over time of practices reported in the FTFI; and a comparison of the first 3 weeks to the last 3 weeks of the diary data to assess the consistency over time of practices reported in the diary. The first 2 analyses were restricted to participants who provided a second FTFI. The third analysis was restricted to participants with complete data for the 6-week study period.
Diary data were summarized on a per-woman basis. On days when participants reported cleansing more than 4 times or insertion more than twice, it was assumed that they cleansed 5 times or inserted 3 times. Cleansing frequency was calculated as the total number of cleansing acts recorded in diaries by each woman, divided by the total number of days with diary data, and rounded to the nearest whole number. Substance used for cleansing was categorized as “water only” or “other” if a substance other than water was used at least once during the 6-week study. Application method for cleansing was categorized as “fingers only” or “cloth” if a cloth was used at least once during the study period, or “other” if another applicator was used at least once during the study period.
Proportions from aggregate data were compared using McNemar's test for matched binary data and a marginal homogeneity test for data with more than 2 categories between the first and second FTFI, the second FTFI and the diary, and the first and last 3 weeks of the Diary Study. A κ statistic was calculated to measure consistency of reporting for each individual participant. We used the Landis and Koch interpretation for the strength of agreement for the κ coefficient.17
Comparison of FTFI and Diary Data
Participants in Tanzania were young (65%, <30 years old), worked in bars or guesthouses (60%), and over half completed primary school education (data not shown). Of the 100 participants who enrolled in the Diary Study in Tanzania, 78 had a second FTFI at the 3-month follow-up visit. Of the 22 women who did not follow-up at 3 months, there was evidence that they were younger, more likely to have worked in a bar, more likely to have taken part in transactional sex, and cleansed with less frequency than attenders. In addition, being a participant in the Diary substudy was not associated with missing the second FTFI in the main study. Table 1 shows that the aggregate data collected in second FTFI and the diary were similar except for reporting cleansing frequency (P < 0.01) and reporting the use of a cloth or other applicator (P < 0.01). The diary data reported a higher frequency of daily cleansing, with more than two-thirds of the cohort cleansing 4 or more times a day, whereas in the FTFI most participants reported usually cleansing 2 to 3 times a day. The diary reported that 34.7% of participants used a cloth at least once; yet, the FTFI reported that only 4.1% of women reported usually using cloth when they cleanse. However, the comparison of applicators was limited, as the diary provided data on ever used, whereas the FTFI summarized data by asking about “usual” practice. Kappa statistics revealed almost perfect agreement for overall cleansing but only fair agreement for cleansing substance and only slight agreement for cleansing by application, cleansing frequency, cleansing before and after sex and insertion.
Participants in Uganda were also young (75%, <30 years old), 43% worked as bar workers and 43% as commercial sex workers, and 42% had not completed primary school whereas 31% had commenced secondary school (data not shown). Of the 100 participants who enrolled in the Diary Study in Uganda, 93 participants had a second FTFI at the next 3-month visit. Table 2 shows that the aggregate data collected from the second FTFI and the diary differed for most variables. The frequency of daily cleansing was higher in the diary data, with more than 95% of the participants cleansing 4 or more times a day, whereas in the FTFI, only 49% of the participants reported usually cleansing 4 or more times a day. The diary reported a higher frequency of using a substance other than water (79.6% vs. 56.5%, P < 0.01). In the diary, 48.4% of participants reported using cloth at least once, whereas only 17.4% of participants reported usually using cloth in the FTFI (P < 0.01). The diary reported high percentages of participants cleansing before (94.6%) or after (95.7%) sex at least once. In contrast, in the FTFI, only 48.9% of the participants reported cleansing to prepare for sex (P < 0.01), and 88.0% reported cleansing after sex (P < 0.01). In the diary, 48.4% of participants reported inserting at least once compared with only 31.2% in the FTFI (P < 0.01). Kappa statistics for these comparisons revealed fair agreement for reporting substance and insertion but only slight consistency for application, frequency of cleansing, and cleansing related to sex.
Consistency of FTFI Data
In Tanzania, there were few statistically significant differences in the aggregate data between the first and second FTFI (Table 1). Kappa statistics for these comparisons revealed fair to moderate consistency of participant reporting for cleansing, substance, and application and insertion but only slight consistency for frequency of cleansing and poor consistency for cleansing related to sex.
In Uganda, there were few differences in the aggregate data reported in the first and second FTFI (Table 2). During the second FTFI, there was an increase in number of participants who reported cleansing to prepare for sex as one of the reasons why they cleansed, 34.1% to 48.9% (P = 0.02). Kappa statistics for the comparison of individual participant consistency between the 2 FTFI revealed substantial agreement for insertion; moderate agreement in participant reporting for cleansing and substance; fair agreement for cleansing application, frequency of cleansing, and cleansing to prepare for sex; and only slight agreement for cleansing after sex.
Consistency of Diary Data
The diary data from the first and last 3 weeks were compared to see how consistent reporting was during the 6-week period (Table 3). In general, the consistency during the 6-week period was better than between the 2 FTFI or between the second FTFI and diary data. The agreement for intravaginal cleansing for Tanzania was perfect (κ = 1.0, P < 0.01), and in Uganda, both methods recorded all the women as engaging in intravaginal cleansing. Both cohorts also had perfect agreement for insertion.
Both cohorts had substantial agreement for substance used to cleanse, and moderate agreement for method of application. However, in both cohorts, there was an increase over time in the proportion of participants that reported use of water only for cleansing (Tanzania: 35%–52%, P < 0.01; Uganda: 24%–32%, P = 0.02) and a corresponding decline of the proportion of participants who reported the use of other substances. In Uganda, there was also an increase in the proportion of participants reporting using fingers only (44%–56%, P = 0.01). There was also high agreement for frequency of cleansing in both cohorts. In Tanzania, there was substantial agreement for cleansing related to sex whereas only moderate to fair agreement for these variables in Uganda.
A daily vaginal practices diary was developed and tested to quantitatively measure IVP, and to our knowledge, this was the first reported use of a vaginal practices diary. Our aims were to compare the consistency of IVP data obtained in the diary with FTFI and to report the consistency of the IVP data over time within each method. Overall, we found that participants reported more IVP in the diary than in the FTFI, and over time, the diary data were more internally consistent than the FTFI. In addition, most of the inconsistencies between the 2 FTFIs and between the FTFI and diary were from reported frequency of IVP or IVP related to sexual intercourse.
In Uganda, the proportions of women reporting IVP were higher in the diary than in the FTFI, and in both sites, a greater proportion of women reported insertion in the diary and a higher frequency of daily cleansing. This latter finding is congruous with results from a coital diary study among female sex workers in South Africa that found higher reported frequency of sex in the diaries than in the FTFI,10 and with a coital diary study among women at increased risk for HIV in Mwanza City, Tanzania, that found that twice as many sex acts were recorded in a diary compared with the FTFI.11 An advantage of the diary was that interval of memory recall was short and closer to direct counts of the behavior, whereas FTFI required the participant to average or summarize a practice over a longer period. However, these results should be interpreted with caution as the FTFI explicitly used summary questions for cleansing application, timing, and frequency (e.g., “About how often do you cleanse inside the vagina?” “When do you usually cleanse inside the vagina?”). In addition, our capacity to compare direct counts of IVP was limited as the Diary Study period was for 6 weeks, whereas the length of recall for the FTFI was 3 months.
There was more consistency of IVP over time in the diary than in the FTFI. Although the comparison between the first and second FTFI showed similar proportions of reported IVP over a 3-month period, there were considerable differences in IVP related to sex. These differences may be related to bias, as it may have been more sensitive to discuss IVP in relation to sex than routine hygiene. Comparison between IVP in the first and last 3 weeks of the diary showed few differences, with the exception of cleansing by substance, in both cohorts, in which there was greater use of other substances (e.g., soaps) in the first 3 weeks. It is possible that weekly visits from the same RA may have encouraged participants to either change their practice or to report in a more socially desirable way; although study staff were trained to respond neutrally to reports of any type of IVP. Investigating IVP use over time within each data collection mode revealed which method was more reliable; nevertheless, it is possible that the diary seemed to be more reliable because of a shorter reporting period of 6 weeks versus 3 months for the FTFI. The apparent reliability of the diary may also suggest that there was little participant fatigue associated with filling out a daily diary. In addition, although this is the first time that a diary data collection tool of this type has been administered for longer than 4 weeks, there was only a small drop-off in retention between weeks 4 and 5, and no change between weeks 5 and 6 in both cohorts, suggesting that longer periods of diary data collection may be feasible.
The comparison of the diary to the FTFI data highlights 2 limitations of FTFI: most inconsistencies between the methods were found in the reports of IVP frequency and IVP related to sex. The former was likely to be an indicator of poor memory recall, and the latter may have demonstrated a degree of social desirability bias. This finding is especially important in the context of future microbicides research where IVP may interfere with a product used before or after sexual intercourse. Reporting from the MDP 301 PRO2000 Trial, Pool et al strongly suggested the use of diaries as part of a mixed methods approach to better understand complex behavior within microbicide trials.13
After analyzing the data from the diary, we noted several limitations with the diary design. The original design was developed in Tanzania, where there was evidence from past studies of cleansing 2 to 3 times a day and rare insertion9; thus, the decision to record up to 4 acts of cleansing and 2 acts of insertion was justified. However, in Uganda, where we had less information on IVP frequencies, we underestimated the higher frequencies of IVP, and this limited our ability to obtain detailed data in women for all IVP acts. This limitation highlights the necessity of documenting specific practices in different populations for future studies.
In conclusion, our study suggests that the reporting of IVP may be improved by a daily, self-administered diary, especially for frequency of cleansing and cleansing in proximity to sexual intercourse. Given the importance of understanding IVP in the context of clinical trials for vaginal products, such as vaginal microbicides, the vaginal practices diary should be used as part of a mixed methods approach to understand these sensitive behaviors in future clinical trials.
2. Myer L, Denny L, de Souza M, et al.. Distinguishing the temporal association between women's intravaginal practices and risk of human immunodeficiency virus infection: A prospective study of South African women. Am J Epidemiol 2006; 163:552–560.
3. McClelland RS, Lavreys L, Hassan WM, et al.. Vaginal washing and increased risk of HIV-1 acquisition among African women: A 10-year prospective study. AIDS 2006; 20:269–273.
4. van de Wijgert JH, Morrison CS, Cornelisse PG, et al.. Bacterial vaginosis and vaginal yeast, but not vaginal cleansing, increase HIV-1 acquisition in African women. J Acquir Immun Defic Syndr 2008; 48:203–210.
5. Hilber AM, Francis SC, Chersich M, et al.. Intravaginal practices, vaginal infections and HIV acquisition: Systematic review and meta-analysis. PLoS One 2010; 5:e9119.
6. Langhaug LF, Sherr L, Cowan FM. How to improve the validity of sexual behaviour reporting: Systematic review of questionnaire delivery modes in developing countries. Trop Med Int Health 2010; 15:362–381.
7. Catania JA, Gibson DR, Chitwood DD, et al.. Methodological problems in AIDS behavioral research: Influences on measurement error and participation bias in studies of sexual behavior. Psychol Bull 1990; 108:339–362.
8. Fenton KA, Johnson AM, McManus S, et al.. Measuring sexual behaviour: Methodological challenges in survey research. Sex Transm Infect 2001; 77:84–92.
9. Allen CF, Lees SS, Desmond NA, et al.. Validity of coital diaries in a feasibility study for the Microbicides Development Programme trial among women at high risk of HIV/AIDS in Mwanza, Tanzania. Sex Transm Infect 2007; 83:490–496; discussion 6–7.
10. Ramjee G, Weber AE, Morar NS. Recording sexual behavior: Comparison of recall questionnaires with a coital diary. Sex Transm Dis 1999; 26:374–380.
11. Lees S, Cook C, Vallely A, et al.. Comparison of sexual behavior data collected using a coital diary and a clinic-based interview during a microbicide pilot study in Mwanza, Tanzania. Sex Transm Dis 2010; 37:497–501.
12. Ferguson AG, Morris CN, Kariuki CW. Using diaries to measure parameters of transactional sex: An example from the Trans-Africa highway in Kenya. Cult Health Sex 2006; 8:175–185.
13. Pool R, Montgomery CM, Morar NS, et al.. A mixed methods and triangulation model for increasing the accuracy of adherence and sexual behaviour data: The Microbicides Development Programme. PLoS One 2010; 5:e11600.
14. Mustanski B. Moderating effects of age on the alcohol and sexual risk taking association: An online daily diary study of men who have sex with men. AIDS Behav 2008; 12:118–126.
15. Kiene SM, Simbayi LC, Abrams A, et al.. High rates of unprotected sex occurring among HIV-positive individuals in a daily diary study in South Africa: The role of alcohol use. J Acquir Immun Defic Syndr 2008; 49:219–226.
16. Graham CA, Catania JA, Brand R, et al.. Recalling sexual behavior: A methodological analysis of memory recall bias via interview using the diary as the gold standard. J Sex Res 2003; 40:325–332.
17. Landis JR, Koch GG. “The measurement of observer agreement for categorical data.” Biometrics 1977; 33:159–174.