Abstract
Background
Attending school may have a strong preventative association with sexually transmitted infections among young women, but the mechanism for this relationship is unknown. One hypothesis is that students who attend school practice safer sex with fewer partners, establishing safer sexual networks that make them less exposed to infection.
Setting
We used longitudinal data from a randomized controlled trial of young women age 13-20 in the Bushbuckridge district, South Africa, to determine if the percentage of school days attended, school dropout and grade repetition are associated with having a partner five or more years older (age-disparate) and with the number of sexual partners in the previous 12 months.
Methods
Risks of having an age-disparate relationship and counts of sex partners were compared using inverse probability of exposure weighted Poisson regression models. Generalized estimating equations were used to account for repeated measures.
Results
Young women who attended fewer school days (<80%) and who dropped out of school were more likely to have an age-disparate relationship (risk difference (RD) 9.9%, 95% CI 3.9%,16.0%; RD (%) dropout 17.2%, 95% CI: 5.4%,29.0%) and those who dropped out reported having fewer partners (count difference (CD) dropout 0.343, 95% CI: 0.192, 0.495). Grade repetition was not associated with either behavior.
Conclusion
Young women who less frequently attend school or who drop out are more likely to have an age-disparate relationship. Young women who drop out have more partners overall. These behaviors may increase risk of exposure to HIV infection in young women out of school.
Key phrases: HIV, Sexual risk behaviors, Schooling, Young women, South Africa
Introduction
An estimated 5.4 million young people aged 15-24 are living with HIV, accounting for 15% of the total burden of persons living with HIV worldwide.1 In South Africa, the prevalence of HIV among persons of this age group is 7.1 % and is 3 to 4 times higher in young women compared to young men.1–3 Attending school may protect against sexually transmitted infections (STIs) including HIV,4–8 but the mechanism for this relationship is unclear. One hypothesis is that students who attend school are engaged in safer sexual networks with lower exposure to STIs.5,9 Compared to young women who do not attend school, girls who are in school may have fewer lifetime partners5,10,11, and fewer partners older than themselves.5,12,13 Evidence is lacking about young women who repeat grades but we would expect them to have younger partners and safer networks if they remain in school. But to date, a temporal relationship between school attendance, grade repetition, and school dropout with sexual behaviors has not been established.5,10–12,14 Given that attending school has a strong preventative association with HIV acquisition in young women, a better understanding of the sexual behaviors affected by attending school would provide clarity on how to reduce transmission in this population.
This analysis uses longitudinal data from a randomized trial to determine if the percentage of school days attended over the past year, school dropout and grade repetition affect the probability of having an age-disparate relationship and number of partners among high school girls in rural South Africa. We hypothesize that young women who do not drop out of school and who attend school more frequently will have fewer partners and partners closer to their own age compared to girls who attend less school, and that young women who repeat grades will have a similar risk to those who no do repeat.
Methods
Study sample
This study analyzes data from the HIV Prevention Trials Network (HPTN) 068 study, a phase III randomized controlled trial to establish whether providing cash transfers, conditional on school attendance, reduced young women's risk of HIV acquisition.15 The study included 2,533 young women aged 13-20 years who were registered in high school in grades 8-11 at enrollment within the rural Bushbuckridge sub-district of Mpumalanga province, South Africa. Potential participants were excluded if they were pregnant or married at enrollment, or if they did not have a parent or guardian in the household. However, young women who became pregnant, married or who dropped out of school during the study were not excluded. We further excluded participants who did not have at least one follow up visit after baseline.
Young women were seen annually from baseline until study completion or graduation from high school, whichever came first. Each annual study visit included an Audio Computer-Assisted Self-Interview (ACASI), as well as HIV and herpes simplex virus type 2 (HSV-2) testing for those who were not positive at the previous visit. Up to four assessments of the young women and their parent/legal guardians were conducted from 2010 to 2015; the first one at baseline and then every 12 months thereafter. Participant attrition was minimal.16 In addition, school attendance records were collected from all schools where the young women were registered during the study period. For this study we used data from February, May and August when all participants (intervention and control) had school attendance data available and because these months were representative of normal attendance (no holidays or exams)
Measures/Variables
We estimated the effect of school attendance and dropout on the probability of having an age-disparate relationship and number of sexual partners. Age-disparate was defined as a young woman having at least one sexual or non-sexual partner five or more years older than herself at each follow up visit. Partners with whom there is no reported sexual relationship were included to account for potential misreporting about sexual behaviors. Five years was selected as the cutoff to capture partners who are out of school and conceptually in a higher HIV prevalence pool. Young women who did not have a partner were coded as not having an age-disparate relationship. Number of sexual partners was defined as a time varying count variable indicating the self-reported number of sex partners in the last 12 months at each follow-up visit. Young women who had not had sex were defined as having 0 sex partners in the last 12 months. All sexual behaviors were self-reported.
School attendance and school dropout were constructed using high school attendance registers. We expect school registers to be more accurate in our study because they were closely monitored during the trial. School attendance was defined as the average percentage of days attended out of the total number of days in February, May and August between yearly follow-up visits. Attendance was dichotomized as high (≥80% of school days) versus low (<80% of school days) attendance with those who dropped out coded as having 0% attendance. The 80% cutoff was chosen as the original cash transfer trial provided the intervention based on this cut off and to set a reasonable target to which to increase attendance. The original intervention did not have an impact on incident infection or school attendance.16 For further comparison, school attendance was also categorized as fewer than 50% of school days, 50% to 80% and 80% or more school days. School dropout was defined as having dropped out of school in any month between surveys (even if they returned later). Reasons for dropout and low attendance were examined using self-reported ACASI data instead of attendance records, as more detailed information was available in the ACASI survey.
We used the outcome variables from the follow-up visit after the covariates and at the end of the exposure period to determine how attendance affects behavior at the next visit. However, attendance data were averaged over the year between follow-up visits and could have overlapped with sexual behavior questions that were asked retrospectively about that same year. An additional sensitivity analysis was done to explore how assumptions about temporality would affect findings. We used attendance between surveys to predict the outcomes two visits later, controlling for confounders prior to attendance. For example, attendance from baseline to follow-up visit 1 was the independent variable while number of partners at visit 2 was the dependent variable, controlling for confounders at baseline.
Statistical Analysis
Although young women could return after dropout, incidence of second dropout after returning to school was low in the study. Therefore, we used Kaplan Meier curves to descriptively examine cumulative incidence of first dropout by time in years since age 13, and time in years since enrollment in the study. Young women who were older than 13 at study enrollment were treated as late entries. In our descriptive analysis of time to first dropout, young women were removed from the risk set if they dropped out and censored at the visit before moving, graduation, study completion or loss to follow up.
Risks of having an age-disparate relationship and counts of sex partners were compared using Poisson regression models. To compare the risks of having an age-disparate relationship between exposure groups, we used Poisson models with a log or identity link function to estimate risk ratios or risk differences, respectively.17 The Poisson model was used to approximate the binomial model.18,19 To compare the counts of partners between exposure groups, we used Poisson models with log or identity link functions to estimate count ratios and count differences.20 Generalized estimating equations (GEE) with an exchangeable correlation matrix and robust variances were used to create 95% confidence intervals that accounted for repeated measures over the study period and for clustering within schools.21
We identified a minimally sufficient set of confounders using directed acyclic graphs (DAG) for each exposure-outcome relationship. Confounders included age, intervention assignment, orphan status, alcohol use, depression, anxiety, pregnancy, and socioeconomic status (SES). Prior age-disparate relationship was included in the weights for age-disparate relationship and prior partner number was included for number of partners in the past 12 months. Potential confounders that were examined but not included were school, grade repetition, and parental monitoring. Inverse probability of exposure weights (IPW) were used to adjust for time-varying confounding.22 For each binary exposure (dropout (yes/no), grade repetition (yes/no) and attendance (high/low)), the denominator of the weights was estimated using a logistic regression model for the exposure of interest conditional on all confounders. To estimate weights for attendance as three categories, we used a pooled logistic regression model conditional on all confounders. To improve efficiency of our estimates, weights were stabilized by the marginal probability of exposure.
We also examined effect measure modification of the relationship between school attendance and partner age difference by time-varying age of the young woman. Young women aged over the study period to age 23. Stratified estimates were examined to determine if the magnitude of the association varied across strata of young women's ages (13-14 years, 15-16 years, 17-18 years and 19-23 years).
Results
Of 2,533 young women included in the parent study, 163 were excluded because they did not have at least one additional ACASI visit following baseline. A total of 2,360 young women were included in our analysis cohort of which 6.1% (N=144) ever dropped out of school during the study period. About 14% (N=20) of those who dropped out returned to school and 10.0% (N=2) of those who returned dropped out a second time. Roughly 4% (N=97) attended less than 80% of school days from baseline to the first follow up visit. Young women who attended more school (≥80% of school days) from baseline to first follow-up were less likely at baseline to have repeated a grade (33.4% vs. 53.6%), be in the intervention arm (52.0% vs. 67.0%), use any alcohol (8.4% vs. 16.5%), ever been pregnant (8.0% vs. 16.7%) or ever had sex (25.5% vs. 46.4%) (Table 1). Young women with low attendance had a similar level of wealth, mother's educational level, parental monitoring, orphanhood and unprotected sex compared to young women with high attendance.
Table 1. Baseline characteristics of young women aged 13 to 21 by average school attendance from baseline to round 1 in Agincourt, South Africa from March 2011 to December 2012 (N=2360)*.
| Average school attendance from baseline to round 1 | ||||||
|---|---|---|---|---|---|---|
| Low (<80%) (N=97) | High (≥80%) (N=2263) | Total (N=2360) | ||||
| N (%) | Median (IQR) | N (%) | Median (IQR) | N (%) | Median (IQR) | |
| Young women's age at baseline (years) | 16 (15,17) | 15 (14,17) | 15 (14,17) | |||
| Age 13-14 | 20(20.6) | 730(32.3) | 750(31.8) | |||
| Age 15-16 | 38(39.2) | 965(42.6) | 1003(42.5) | |||
| Age 17-18 | 26(26.8) | 477(21.1) | 503(21.3) | |||
| Age 19-21 | 13(13.4) | 91(4.0) | 104(4.4) | |||
| Household wealth (assets) | ||||||
| Low | 28(28.9) | 578(25.6) | 606(25.7) | |||
| Middle to Low | 26(26.8) | 605(26.8) | 631(26.8) | |||
| Middle | 19(19.6) | 547(24.2) | 566(24.0) | |||
| High | 24(24.7) | 529(23.4) | 553(23.5) | |||
| Interventionarm | 65(67.0) | 1177(52.0) | 1242(52.6) | |||
| Ever repeated a grade | 52(53.6) | 756(33.4) | 808(34.2) | |||
| Any alcohol use | 16(16.5) | 189(8.4) | 205(8.7) | |||
| Children's depression index score | 4(1,8) | 2(1,5) | 2(1,5) | |||
| Children's manifest anxiety score | 5 (2,8) | 4(1,7) | 4(1,7) | |||
| Ever pregnant | 16(16.7) | 180(8.0) | 196(8.4) | |||
| Age-disparate relationship | 16 (16.5) | 123(5.4) | 139(5.9) | |||
| Ever had sex | 45(46.4) | 577(25.5) | 622(26.4) | |||
| Any unprotected sex in the last three months | 11(12.2) | 161(7.6) | 172(7.7) | |||
| Single or double orphan | 33(35.1) | 603(28.0) | 636(28.3) | |||
| Parent/guardian monitoring score | 10(8,12) | 10(7,12) | 10(8,12) | |||
| Mother's educational level | ||||||
| No school | 16 (17.8) | 378 (18.2) | 294 (18.2) | |||
| Some primary | 20(22.2) | 416 (20.0) | 436 (20.1) | |||
| Completed primary | 7(7.8) | 93 (4.5) | 100 (4.6) | |||
| Some high school | 26 (28.9) | 617 (29.7) | 643 (29.7) | |||
| Completed high school | 16 (17.8) | 496 (23.9) | 512 (23.6) | |||
| University or Technical school | 5 (5.6) | 77 (3.7) | 82 (3.8) | |||
Missing at baseline: school attendance N=10; wealth N=4; alcohol use N=3; depression N= 107; anxiety N=33; ever pregnant N= 27; ever had sex N=3; unprotected sex N=139; Single or double orphan N=109; parental monitoring N=40; mother's educational level N=205
The average percentage of days not attended each month increased over the study period from 2.1% in May 2011 to 11.1% in August 2014 and was similar by study arm. Cumulative incidence of first dropout increased over the study period from baseline, when all young women were enrolled, to 12.0% at the end of the study period (Figure 1). When looking at cumulative incidence over time since age 13, the risk of dropout increased from 0% in those age 13 to 27.8% in girls 8 years older at the age of 21 but was similar by intervention arm. The most common reason that young women reported for dropping out of school was that they were pregnant or had a child (N=66; 43.7%). Other common reasons were that they were sick or disabled (N=19; 12.6%) or not doing well in school (N=18; 11.9%). The most common reasons reported for not attending school were being sick or disabled (N=1,664; 77.0%), other (N=158; 7.3%), having to help at home (93,4.3%) or being pregnant or having a child (N=89; 4.1%). The main reason for attending school was to get a job in the future (N=1814; 76.5%).
Figure 1. Cumulative incidence of school dropout by 1) time since age 13 and 2) time since study enrollment between 2011 and 2014, by intervention arm.
Over the study period, out of 4,993 women-visits, 8.0% (N=397) of women reported having an age-disparate relationship, and the median number of partners in the last 12 months was 0 (interquartile range= 0,1). The weighted risk of having an age-disparate relationship was 17.8% in young women with low attendance compared to 7.8% in young women with high attendance (Table 2). Young women who attended fewer than 80% of school days had a 9.9% higher one-year risk (95% Confidence interval (CI): 3.9%, 16.0%) of having an age-disparate relationship compared to young women who attended 80% or more school days, accounting for confounding. The weighted effect was similar for those attending fewer than 50% but not as strong in those attending 50-80% of school days. The weighted risk of having an age-disparate relationship for young women who dropped out was 25.2% compared to 8.0% in young women who did not drop out. Young women who dropped out of school had a higher one-year risk of having an age-disparate relationship (Risk difference (RD): 17.2%; 95% CI: 5.4%, 29.0%) but young women who repeated a grade had a similar one-year risk compared to those who did not drop out (RD: -0.2%; 95% CI: -2.0%, 1.6%), accounting for confounding.
Table 2. Unweighted and weightedrisks, risk ratios (RR), risk differences (RD) and 95% confidence intervals (CI) for the effect of school attendance, school dropoutand grade repetition on the probability of having an age-disparate relationshipin HPTN 068 from 2011 to 2015***.
| Unweighted | Weighted | |||||
|---|---|---|---|---|---|---|
| Risk (%) | RR (95% CI) | RD (%; 95% CI) | Risk (%) | RR (95% CI) | RD (%; 95% CI) | |
| School Attendance- Binary* | ||||||
|
| ||||||
| Low | 22.7 | 3.04 (2.34,3.94) | 15.2(9.8,20.7) | 17.8 | 2.28 (1.59,3.23) | 9.9(3.9,16.0) |
| High | 7.5 | 1 | 0 | 7.8 | 1 | 0 |
|
| ||||||
| School Attendance-categorical* | ||||||
|
| ||||||
| <50 percent | 24.9 | 3.34 (2.38,4.68) | 17.4 (9.5,25.5) | 19.6 | 2.32(1.16,4.63) | 11.8 (2.7,20.8) |
| 50-80% | 20.9 | 2.80 (1.97,3.98) | 13.4 (6.4,20.5) | 18.3 | 2.49(1.55,3.98) | 10.4(-2.2,22.9) |
| ≥80% | 7.5 | 1 | 0 | 7.9 | 1 | 1 |
|
| ||||||
| School Dropout* | ||||||
|
| ||||||
| Yes | 27.3 | 3.62(2.75,4.77) | 19.8(12.8,26.8) | 25.2 | 3.17 (1.96,5.10) | 17.2 (5.4,29.0) |
| No | 7.5 | 1 | 0 | 8.0 | 1 | 0 |
|
| ||||||
| Grade Repetition** | ||||||
|
| ||||||
| Yes | 10.6 | 1.52 (1.24,1.87) | 3.6 (1.8,5.5) | 6.8 | 0.97 (0.74,1.26) | -0.2 (-2.0,1.6) |
| No | 7.0 | 1 | 0 | 7.0 | 1 | 0 |
Controlled for confounding using inverse probability of treatments weights including Age, intervention assignment, Orphan Status, SES, Alcohol Use, Depression, anxiety, prior age-disparate relationship
Grade repetition- weighted for age, Alcohol use, Depression, Anxiety, Orphan Status, Pregnancy, School, intervention assignment, prior age-disparate relationship.
No. cases/exposed of 4942 visits: Attendance low N=61/240, high N=333/4702; <50% N=30/107, 50-80% N=31/133, >=80% N=333/4702; dropout yes N=50/158, no N=342/4776; grade repetition yes N=189/1811, no N=208/3184.
Results for the weighted effect of school attendance, school dropout and grade repetition on number of sexual partners were similar to patterns observed for age disparity (Table 3). The weighted mean number of partners in the past 12 months was 0.44 among women with high attendance compared to 0.60 among those with low attendance. Young women who attended fewer than 80% of school days had 0.079 more partners (95% CI: -0.24, 0.182) compared to young women who attended 80% or more school days over a one-year period, accounting for confounding. No difference in outcomes was observed when comparing attendance categories of less than 50% and 50-80% of school days. Young women who dropped out of school had 0.343 more partners (95% CI: 0.192, 0.495) compared to those who did not drop out, while young women who repeated a grade had a similar one-year count of partners compared to those who did not (Count difference (CD): -0.006; 95% CI: -0.060, 0.049).
Table 3. Unweighted and weightedcount ratios (CR), count differences (CD) and 95% confidence intervals (CI) for the effect of school attendance, school dropoutand grade repetition on number of sexual partners in HPTN 068 from 2011 to 2015***.
| Unweighted | Weighted | |||||
|---|---|---|---|---|---|---|
| Mean | CR (95% CI) | CD (95% CI) | Mean | CR (95% CI) | CD (95% CI) | |
| School Attendance- Binary* | ||||||
|
| ||||||
| Low | 0.661 | 1.69 (1.43,2.01) | 0.271 (0.161,0.381) | 0.597 | 1.94 (0.96,1.48) | 0.079 (-024,0.182) |
| High | 0.390 | 1 | 0 | 0.436 | 1 | 0 |
|
| ||||||
| School Attendance-Categorical* | ||||||
|
| ||||||
| <50 % | 0.755 | 1.93 (1.49,2.50) | 0.365 (0.174,0.556) | 0.430 | 1.06(0.75,1.49) | 0.024(-0.122,0.170) |
| 50-80% | 0.585 | 1.50 (1,23,1.83) | 0.195 (0.081,0.310) | 0.506 | 1.24(0.95,1.62) | 0.099(-0.034,0.232) |
| >=80% | 0.390 | 1 | 0 | 0.407 | 1 | 0 |
|
| ||||||
| School Dropout* | ||||||
|
| ||||||
| Yes | 0.821 | 2.11 (1.74,2.55) | 0.432 (0.281,0.582) | 0.748 | 1.85 (1.50,2.28) | 0.343 (0.192,0.495) |
| No | 0.389 | 1 | 0 | 0.405 | 1 | 0 |
|
| ||||||
| Grade Repetition** | ||||||
|
| ||||||
| Yes | 0.553 | 1.72 (1.54,1.92) | 0.232 (0.181, 0.282) | 0.366 | 0.98 (0.85,1.14) | -0,006 (-0.060,0.049) |
| No | 0.321 | 1 | 0 | 0.371 | 1 | 0 |
Weighted for confounding usinginverse probability of treatments weights including Age, intervention assignment, Orphan Status, SES, Alcohol Use, Depression, anxiety, prior partner number
Grade repetition- weighted for age, Alcohol use, Depression, Anxiety, Orphan Status, Pregnancy, School, intervention assignment, prior partner number.
Number of 4942 visits: Attendance low N=240, high N=4702; <50% N=107, 50-80% N=133, >=80% N=4702; dropout yes N=158, no N=4776; grade repetition yes N=1811, no N=3184.
When stratified by age, the percentage with an age-disparate relationship was higher for young women who were aged 19-23 (N=44, 16.4%) and 17-18 (N=166, 12.0%) than those aged 15-16 (N=151, 6.4%) and 13-14 (N=36, 3.7%). Young women who had low attendance in school (<80%) had a higher risk of having an age-disparate relationship among the 13-14 (RD 22.7%; 95% CI: 2.4%,43.0%) and 17-18 year olds (RD 16.7%; 95% CI: 5.9%,27.5%), accounting for confounding (Table 4). The weighted effect of attending school on age disparity was slightly reduced among the 15 to 16 year old age group (RD 2.3%; 95% CI: -4.7%,9.3%) and the 19 to 23 year old age group (RD 13.2%; 95% CI: -4.1%,30.5%). However, few young women were in the oldest age group, which may have resulted in a lack of precision for this estimate.
Table 4. Weighted risk ratio (RR) and risk difference (RD) with 95% confidence intervals (CI) for the association between low attendance in school versus high attendance and having an age-disparate relationship stratified by time-varying age of the young woman***.
| Risk (%) | RR (95% CI)* | RD (%; 95% CI)** | |
|---|---|---|---|
| Unstratified association | 2.69 (1.92,3.77) | 13.0 (6.3,19.7) | |
|
| |||
| Stratified by Age | |||
| Age 13-14 | |||
| Low | 25.9 | 8.07 (3.40,19.19) | 22.7 (2.4,43.0) |
| High | 3.2 | 1 | 0 |
| Age 15-16 | |||
| Low | 9.2 | 1.33 (0.61,2.87) | 2.3 (-4.7,9.3) |
| High | 6.9 | 1 | 0 |
| Age 17-18 | |||
| Low | 27.7 | 2.52 (1.66,3.82) | 16.7 (5.9,27.5) |
| High | 11.0 | 1 | 0 |
| Age 19-23 | |||
| Low | 26.7 | 1.97 (0.95,4.09) | 13.2 (-4.1,30.5) |
| High | 13.6 | 1 | 0 |
Likelihood Ratio Test (LRT) for modification by age: chi2=10.67, DF=3, p-value=0.014;
Likelihood Ratio Test (LRT) for modification by age: chi2=8.38, DF=3, p-value=0.039;
Weighted for confounding using inverse probability of treatments weights including Age, intervention assignment, Orphan Status, SES, Alcohol Use, Depression, anxiety, prior age-disparate relationship
Weighted number of cases/ total exposed: Age 13- 14 N=34/881; Age 15-16 N=141/2156; Age 17-18 N=156/1296; Age 19-23 N=38/250.
In a sensitivity analysis, the weighted effect of attending fewer than 80% of school days versus 80% or more was similar for age disparity (RD: 13.2%; 95% CI: 3.3%, 23.1%) and number of partners (CD: 0.256; 95% CI: 0.032, 0.481) two follow-up visits later, accounting for prior confounders (Appendix 1). The weighted effect of school dropout was stronger on age disparity (RD: 28.0%; 95% CI: 8.2%, 47.8%) and number of partners (CD: 0.584; 95% CI: 0.188, 0.908) two visits later, accounting for prior confounders.
Discussion
In our study the percentage of young women attending fewer than 80% of school days on average in the past year was relatively low (4.1%) but increased with age and time since enrollment. Low attendance in school was associated with the probability of having an age-disparate relationship. School dropout (6.1%) was associated with both having an age-disparate relationship and with having more sexual partners. Conversely, ever repeating a grade was not associated with either outcome.
Our results support the hypothesis that young women who stay in school and who attend school more frequently have partners closer to their own age and fewer partners than young women who attend less school or drop out. These findings based on longitudinal data are consistent with those from prior studies indicating that young women who are in school have fewer lifetime partners, 5,10,11 and fewer partners older than themselves compared to young people who do not attend school.5,12–14,23 Most notably, our results are similar to a cross-sectional study in South Africa that found that young men who were in school were less likely to be HIV infected, but both young men and women in school had fewer lifetime partners, and young women had fewer partners more than 3 years older than themselves.5 Another recent study in Zimbabwe found that age-disparate relationships were associated with incident HIV infection and that completion of secondary school was inversely associated with age-disparate relationships. Students might have a safer sexual network structure, thereby putting them at lower risk of HIV infection.
In the HTPN 068 study, school attendance was associated with incident HIV-infection; partner number and partner age difference were also associated with HIV-infection.16,24 Age disparity has been associated with prevalent HIV infection in several cross-sectional studies from sub-Saharan Africa3,25–27and with incident HIV infection in Zimbabwe and Uganda.13,28 Conversely, age disparity did not appear to be associated with incident HIV infection in other studies from South Africa and Uganda.29–32 While age disparity and number of partners appear to be factors in the relationship between time spent in school and risk of HIV infection, how these behaviors further increase risk of HIV warrants further investigation.
Attending fewer school days was associated with age-disparate relationships and school dropout was associated with both having an age- disparate relationship and having more partners. However, grade repetition was not associated with either behavior. This pattern suggests that the effect of school attendance on partner age and partner number is more strongly related to amount of time spent in a school environment than with educational success. In addition, young women who repeat grades may be exposed to younger men and be less likely to have an age-disparate relationship. It is important to note that there is a cyclical relationship between grade repetition, school attendance and school dropout. Our results indicate that young women who have repeated a grade are more likely to have low attendance in school and young women who have low attendance are more likely to later repeat a grade. Grade repetition and low attendance were also markers for later school dropout. While grade repetition may not be directly associated with age disparity or number of sexual partners, it appears to be a warning sign for later low attendance and school dropout, which are associated with these partnering behaviors.
Our study uses longitudinal data to test the hypothesis that young women in school have younger and fewer partners. However, we use data from an RCT in which all young women were in school at study enrollment. A prior study using data from the trial found evidence of selection bias: young women who participated in the HPTN 068 were already more likely to be enrolled in school than in the underlying population.33 Participation in an RCT may have also resulted in a Hawthorne effect where young women may have been less likely to drop out than they would have otherwise simply because of study participation.33 Young women in the study were also more likely to attend school than has been seen in other studies in sub-Saharan Africa, but we would expect the association between schooling and behaviors to be similar.16 Second, information on sexual behaviors and partner characteristics was self-reported in the study and may be misreported.
Lastly, despite the use of longitudinal data, the time period for the exposure and outcome variables could overlap because school attendance and enrollment information were measured between surveys and sexual behaviors were reported retrospectively. Although reverse causality is plausible (i.e., older partners and more partners may result in low school attendance), our sensitivity analysis shows that these partnering behaviors occur more often after school dropout or after low school attendance. In fact, when restricting our time period to that following dropout, the effect estimates for the associations are even stronger.
In our study, young women who attended fewer days of school were more likely to have an age-disparate relationship and those who dropped out were more likely to both have an age-disparate relationship and more sexual partners. Spending time in school imposes network and time constraints that make frequent attendees more likely to both select other (same-age) students as partners and to have fewer partners overall, thereby reducing their risk of being exposed to partners with HIV. Initiatives aimed at keeping girls in school such as DREAMS are critical to promoting safe sexual behaviors and preventing sexually transmitted infections.34 However, effectively preventing infections in young women should also involve the development of interventions for young women out of school or who have completed school that encourage them to be part of safer networks.
Supplementary Material
Acknowledgments
We thank all of the participants in HPTN 068 and the study staff.
Source of Funding: This work was supported by grant T32 5T32AI007001 from National Institutes of Health. Funding support for the HPTN was provided by the National Institute of Allergy and Infectious Diseases (NIAID), the National Institute of Mental Health (NIMH), and the National Institute on Drug Abuse (NIDA) of the National Institutes of Health (NIH; award numbers UM1AI068619 [HPTN Leadership and Operations Center], UM1AI068617 [HPTN Statistical and Data Management Center], and UM1AI068613 [HPTN Laboratory Center]. The study was also funded under R01MH087118 and R24 HD050924 to the Carolina Population Center. Additional funding was provided by the Division of Intramural Research, NIAID, and NIH. The Agincourt Health and Socio-Demographic Surveillance System is supported by the School of Public Health University of the Witwatersrand and Medical Research Council, South Africa, and the UK Wellcome Trust (grants 058893/Z/99/A; 069683/Z/02/Z; 085477/Z/08/Z; and 085477/B/08/Z). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Footnotes
Meetings where data were presented: Presented at International AIDS Society (IAS) conference 2017. Paris, France: July 23-26, 2017
Conflicts of Interest: None declared
References
- 1.UNAIDS. UNAIDS report on the global AIDS epidemic 2013. Vol. 198. UNAIDS; 2013. doi:JC2502/1/E. [Google Scholar]
- 2.Shisana O, Rehle T, Simbayi LC, Zuma K, Jooste S, Zungu N, Labadarios D, Onoya DEA. South African National HIV Prevalence. Incidence and Behaviour Survey. 2012;2012:194. doi: 10.2989/16085906.2016.1153491. [DOI] [PubMed] [Google Scholar]
- 3.Pettifor AE, Rees HV, Kleinschmidt I, et al. Young people's sexual health in South Africa: HIV prevalence and sexual behaviors from a nationally representative household survey. AIDS. 2005;19(14):1525. doi: 10.1097/01.aids.0000183129.16830.06. doi:00002030-200509230-00012[pii] [DOI] [PubMed] [Google Scholar]
- 4.Gavin L, Galavotti C, Dube H, et al. Factors associated with HIV infection in adolescent females in Zimbabwe. J Adolesc Health. 2006;39(4):596, e11–e18. doi: 10.1016/j.jadohealth.2006.03.002. [DOI] [PubMed] [Google Scholar]
- 5.Hargreaves JR, Morison La, Kim JC, et al. The association between school attendance, HIV infection and sexual behaviour among young people in rural South Africa. J Epidemiol Community Health. 2008;62(2):113–119. doi: 10.1136/jech.2006.053827. [DOI] [PubMed] [Google Scholar]
- 6.Pettifor AE, Levandowski BA, Macphail C, Padian NS, Cohen MS, Rees HV. Keep them in school: The importance of education as a protective factor against HIV infection among young South African women. Int J Epidemiol. 2008;37(6):1266–1273. doi: 10.1093/ije/dyn131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Baird S, Chirwa E, McIntosh C, Ozler B. The short-term impacts of a schooling conditional cash transfer program on the sexual behavior of young women. Health Econ. 2010;(19 Suppl):55–68. doi: 10.1002/hec.1569. [DOI] [PubMed] [Google Scholar]
- 8.Birdthistle I, Floyd S, Nyagadza A, Mudziwapasi N, Gregson S, Glynn JR. Is education the link between orphanhood and HIV/HSV-2 risk among female adolescents in urban Zimbabwe? Soc Sci Med. 2009;68(10):1810–1818. doi: 10.1016/j.socscimed.2009.02.035. [DOI] [PubMed] [Google Scholar]
- 9.Jukes M, Simmons S, Bundy D. Education and vulnerability: the role of schools in protecting young women and girls from HIV in southern Africa. AIDS. 2008;22(4):S41–S56. doi: 10.1097/01.aids.0000341776.71253.04. [DOI] [PubMed] [Google Scholar]
- 10.Zambuko O, Mturi AJ. Sexual risk behaviour among the youth in the era of HIV/AIDS in South Africa. J Biosoc Sci. 2005;37(5):569–584. doi: 10.1017/S0021932004007084. [DOI] [PubMed] [Google Scholar]
- 11.Zuilkowski SS, Jukes MCH. The impact of education on sexual behavior in sub-Saharan Africa: a review of the evidence. AIDS Care. 2012;24(5):562–576. doi: 10.1080/09540121.2011.630351. [DOI] [PubMed] [Google Scholar]
- 12.Harrison A, Cleland J, Frohlich J. Young People's Sexual Partnerships in KwaZulu-Natal, South Africa: Patterns, Contextual Influences, and HIV Risk. Stud Fam Plann. 2008;39(4):295–308. doi: 10.2307/20454477. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Schaefer R, Gregson S, Eaton JW, et al. Age-disparate relationships and HIV incidence in adolescent girls and young women: evidence from a general-population cohort in Zimbabwe. Aids. 2017;0(April) doi: 10.1097/QAD.0000000000001506. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Stroeken K, Remes P, De Koker P, Michielsen K, Van Vossole A, Temmerman M. HIV among out-of-school youth in Eastern and Southern Africa: a review. AIDS Care. 2011;0(0):1–9. doi: 10.1080/09540121.2011.596519. [DOI] [PubMed] [Google Scholar]
- 15.Pettifor A, MacPhail C, Selin A, et al. HPTN 068: A Randomized Control Trial of a Conditional Cash Transfer to Reduce HIV Infection in Young Women in South Africa???Study Design and Baseline Results. AIDS Behav. 2016;20(9):1863–1882. doi: 10.1007/s10461-015-1270-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Pettifor A, MacPhail C, Hughes JP, et al. The effect of a conditional cash transfer on HIV incidence in young women in rural South Africa (HPTN 068): a phase 3, randomised controlled trial. Lancet Glob Heal. 2016;(Hptn 068) doi: 10.1016/S2214-109X(16)30253-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Wacholder S. Binomial regression in GLIM: estimating risk ratios and risk differences. Am J Epidemiol. 1986;123(1):174–184. doi: 10.1093/oxfordjournals.aje.a114212. [DOI] [PubMed] [Google Scholar]
- 18.Williamson T, Eliasziw M, Fick G. Log-binomial models: exploring failed convergence. Emerg Themes Epidemiol. 2013;10(1):14. doi: 10.1186/1742-7622-10-14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Cheung YB. A modified least-squares regression approach to the estimation of risk difference. Am J Epidemiol. 2007;166(11):1337–1344. doi: 10.1093/aje/kwm223. doi:kwm223[pii]\r10.1093/aje/kwm223. [DOI] [PubMed] [Google Scholar]
- 20.Hutchinson MK, Holtman MC. Analysis of count data using poisson regression. Res Nurs Health. 2005;28(5):408–418. doi: 10.1002/nur.20093. [DOI] [PubMed] [Google Scholar]
- 21.Hubbard AE, Ahern J, Fleischer NL, et al. To GEE or not to GEE: comparing population average and mixed models for estimating the associations between neighborhood risk factors and health. Epidemiology. 2010;21(4):467–474. doi: 10.1097/EDE.0b013e3181caeb90. [DOI] [PubMed] [Google Scholar]
- 22.Cole SR, Hernán MA. Constructing inverse probability weights for marginal structural models. Am J Epidemiol. 2008;168(6):656–664. doi: 10.1093/aje/kwn164. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Mathews C, Aarø LE, Flisher AJ, Mukoma W, Wubs AG, Schaalma H. Predictors of early first sexual intercourse among adolescents in Cape Town, South Africa. Health Educ Res. 2009;24(1):1–10. doi: 10.1093/her/cym079. [DOI] [PubMed] [Google Scholar]
- 24.Pettifor Audrey, Jing W, Amanda S, et al. Impact of Male Partners and Schooling on HIV Risk Among South African Girls: HPTN 068e. CROI; Vol Boston, Massachusetts: 2017. [Google Scholar]
- 25.Gregson S, Nyamukapa CA, Garnett GP, et al. Sexual mixing patterns and sex-differentials in teenage exposure to HIV infection in rural Zimbabwe. Lancet. 2002;359(9321):1896–1903. doi: 10.1016/S0140-6736(02)08780-9. [DOI] [PubMed] [Google Scholar]
- 26.Kaiser R, Bunnell R, Hightower A, et al. Factors associated with HIV infection in married or cohabitating couples in Kenya: Results from a nationally representative study. PLoS One. 2011;6(3) doi: 10.1371/journal.pone.0017842. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Businge CB, Longo-Mbenza B, Mathews V. Risk factors for incident HIV infection among antenatal mothers in rural Eastern Cape, South Africa. Glob Health Action. 2016;9(1) doi: 10.3402/gha.v9.29060. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Biraro S, Mayaud P, Morrow RA, Grosskurth H, Weiss Ha. Performance of commercial herpes simplex virus type-2 antibody tests using serum samples from Sub-Saharan Africa: a systematic review and meta-analysis. Sex Transm Dis. 2011;38(2):140–147. doi: 10.1097/OLQ.0b013e3181f0bafb. [DOI] [PubMed] [Google Scholar]
- 29.Kelly RJ, Gray RH, Sewankambo NK, et al. Age differences in sexual partners and risk of HIV-1 infection in rural Uganda. J Acquir Immune Defic Syndr. 2003;32(4):446–451. doi: 10.1097/00126334-200304010-00016. [DOI] [PubMed] [Google Scholar]
- 30.Balkus JE, Nair G, Montgomery ET, et al. Age-disparate partnerships and risk of HIV-1 acquisition among South African women participating in the VOICE trial. J Acquir Immune Defic Syndr. 2015;70(2):212–217. doi: 10.1097/QAI.0000000000000715. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Harling G, Newell M, Tanser F, Kawachi I, Subramanian S, Bärnighausen T. Do Age-Disparate Relationships Drive HIV Incidence in Young Women ? Evidence from a Population Cohort in Rural KwaZulu-Natal, South Africa. J Acquir Immune Defic Syndr. 2014;66(4):443–451. doi: 10.1097/QAI.0000000000000198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Street RA, Reddy T, Ramjee G. The generational effect on age disparate partnerships and the risk for human immunodeficiency virus and sexually transmitted infections acquisition. Int J STD AIDS. 2015;24(5):517–528. doi: 10.1177/0956462415592325. [DOI] [PubMed] [Google Scholar]
- 33.Rosenberg M, Pettifor A, Twine R, et al. Selection and Hawthorne effects in an HIV prevention trial among young South African women. International Aids Society (IAS); Vol Durban, South Africa: 2016. [Google Scholar]
- 34.The United States President's Emergency Plan for AIDS Relief. DREAMS: Working Together for an AIDS-free Future for Girls and Women. [Accessed February 2, 2017];2017 http://www.dreamspartnership.org.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.