Introduction
Adolescent girls and young women (AGYW) in sub-Saharan Africa are at high risk of HIV infection. HIV infection rates increase substantially among AGYW after the age of 18. In South Africa HIV prevalence rises from 6% among young women aged 15–19 years to 17% by ages 20–24. This age period marks a time of transition into adulthood and coincides with a number of key life events, such as finishing the mandatory years of schooling, leaving home, entering first sexual relationships and experiencing first pregnancies.(Human Sciences Research Council [HSRC], 2017) Key life events that include first pregnancy, coital debut, leaving school and parental death have all been found independently to be associated with an increased risk of HIV infection in young women. Young women who experience their first vaginal sex before the age of 15 are more likely to be living with
Despite documentation of the importance of life event to HIV risk, little is understood about the timing of key life events and how they shape HIV risk as adolescents transition into adulthood. While coital debut, pregnancy and leaving school are all key life transitions that many young people will experience as part of their expected life course (MacMillan and Copher, 2005), occurrence earlier in adolescence can result in negative outcomes for health and development (Evans et al, 2013). Life course theory highlights that the timing and sequence of life events or transitions result in different meaning and consequences (Elder, 1998). While experience of particular life events earlier may have immediate impacts on health and social factors, their impacts may also extend into the future by affecting subsequent transitions (Elder, 1998). While there is a robust body of literature documenting how adverse child events can impact health in adulthood (Hughes et al, 2017), this evidence is just starting to emerge from low- and middle-income countries (Kidman and Kohler, 2019). Further, negative life events that occur in early adolescence (ages 9–14), when major brain and social transitions are occurring, may have long-lasting consequences (Falconi et al, 2014; UNICEF, 2017). However, changes that occur in early adolescence in the brain may also provide an opportunity for intervention to improve future health and developmental trajectories (Fuhrmann et al, 2015; UNICEF, 2017). Life course research that examines how the timing of events in earlier life can impact later
The purpose of this research was to explore the impact of key life events, specifically sexual debut, pregnancy, death of a parent and leaving school, on HIV incidence among a cohort of AGYW in rural South Africa. This research utilized latent class growth analysis (LCGA) to classify adolescent girls based on growth trajectories according to the age at which they experienced these life events. In addition, LCGA was used to examine the cumulative impact of multiple key life events within the same time period of these young women’s lives, and how these events affected their risk of HIV acquisition as they aged.
Methods
Study design
The HIV Prevention Trials Network (HPTN) 068 was a phase 3 randomized controlled trial to determine whether a monthly cash transfer, conditional on school attendance, would reduce HIV incidence for adolescent girls and young women in South Africa (Pettifor et al, 2016). The study was undertaken by the Medical Research Council/Rural Public Health and Health Transitions Research Unit, which runs the Agincourt Health and Socio-Demographic Surveillance System (HDSS) site in the rural Bushbuckridge subdistrict in Mpumalanga province, South Africa.
Women aged 13–20 years were included in the HPTN 068 study if they were enrolled in school grades 8–11 of the South African government educational system, HIV negative, not married or pregnant, able to read, able to open a bank account, and currently residing in the study area. Potential participants were identified from the Agincourt HDSS sampling frame. Each young woman and her parent or guardian provided written informed consent. Written assent was obtained for those younger than 18 years. Consent and assent forms were available in English and Shangaan.
Institutional Review Board approval for this study was obtained from the University of North Carolina at Chapel Hill and the University of the Witwatersrand Human Research Ethics Committee.
Statistical analysis
Latent class growth models are a method for fully capturing information about between-person differences in within-person patterns of change over time. Individuals are grouped into latent classes based upon similar patterns of data over time. The method assumes that the observed distribution of values may be a combination of two or more subpopulations whose membership is unknown. As such, latent class growth analysis probabilistically assigns individuals to these subpopulations by inferring each individual’s membership to latent classes from the observed growth model data. Life events at different stages in life can be classified as separate events to investigate how the event and when it happens affects the individual’s risks of HIV infection.
The classes of different life event trajectories over time were categorized using LCGA, allowing the individual’s variation to occur around one of multiple trajectories. After classifying life event trajectories, it was determined which trajectories were predictors of the outcome, HIV status. Each class therefore described a course over time of each of the four life events: sexual debut, pregnancy, leaving school and death of a parent. Each trajectory was determined over four time points: whether the event occurred by age 14, by age 17, by age 21 and by age 23. However, for the ‘leaving school’ model all participants had left by age 23 and so the trajectories are determined
Models were fitted over a range of class numbers to identify the ideal number of classes in describing patterns in the timing of all life events. Models were also fitted for each of the life events – sexual debut, pregnancy, leaving school and death of a parent –, over a range of class numbers. Two to four classes were examined for the individual life events and three to six for the combined classes. The model fitting is provided in Table 3.1. Fit was assessed using the Bayesian Information Criterion (BIC), the Bootstrap likelihood ratio tests (BLRT) and entropy values (Jung and Wickrama, 2008). In addition to model fit statistics, the interpretability and practical coherence of the model classes were considered. The models were run with multiple random starting values to avoid settling on local solutions; each model was run with 100 random starts and 10 final optimizations. These values were increased when convergence was not achieved. Having determined the best fitting model, the latent classes were added to a generalized linear model (GLM), regressing HIV status on the latent class trajectories of the life events, using a binomial distributional family and log link, to estimate the risk ratio (RR) of HIV infection for each latent class.
Finally, dominance analysis, a technique which rank-ordered the relative importance or contribution of the life events classes in predicting HIV incidence, was used based on the individual variables’ contribution to the overall model fit. This method is based on the average pseudo R2 explained by each life event across all possible subsets regression models (Azen and Traxel, 2009; Budescu, 1993).
Missing data
Parameter estimates from Mplus were adjusted for missingness using a robust full information maximum likelihood (FIML) estimator, which assumes data is missing at random. Mplus generates the covariance coverage matrix to assess the proportion of observations available for each pair of variables, the minimum recommended coverage in Mplus
Measures
The study looked at events that could occur at only one time and where the timing of the event could be measured (for example, left school at age 15) so that the temporality of exposure and outcome (HIV sero-conversion) could be assessed. In order to assess time since exposure on the outcome, coital debut was measured by asking young women whether they had ever had vaginal sex and if yes, at what age they first had vaginal sex. For pregnancy, young women were asked if they had ever been pregnant and if so the age at first pregnancy. For school leaving, young women were asked if they were still attending secondary school or not and the age they left secondary school. Parental death was assessed by asking, separately, if their mother and father were alive, and if not what age the young person was when they died. Measures were adjusted for the following covariates: age at enrolment, length of study enrolment, trial arm, school attendance, anxiety, depression, interpersonal violence (IPV) and HIV knowledge.
Results
Specific life events and HIV incidence
Figure 3.1 shows the observed and expected trajectories from the LCGA model, for each of the life events. The selected models included two linear trajectory classes for leaving school and death of a parent and three linear trajectory classes for sexual debut and pregnancy.
Latent class growth analysis observed and expected trajectories
Source: Author’s ownLatent class growth analysis observed and expected trajectories
Source: Author’s ownLatent class growth analysis observed and expected trajectories
Source: Author’s ownLatent class growth analysis observed and expected trajectories
Source: Author’s ownLatent class growth analysis observed and expected trajectories
Source: Author’s ownIn Figure 3.1a, the trajectories for leaving school, a two-class model fit the data best, with classes identified as those who leave school early (class 1) and those leaving later (class 2). Table 3.1 shows the unadjusted and adjusted risk ratios (RR) for the log-binomial regression of the two classes on HIV infection; the risk of HIV infection in those who leave school early (class 1) is 2.9 times higher (p-value = 0.001) than those who leave later (class 2) in the unadjusted model. When the model is adjusted for age at enrolment and length of enrolment, the participants’ risk ratio increases to 4.3 (p-value < 0.001) (Table 3.1).
Risk ratio of log-binomial regressions of classes on HIV, unadjusted, adjusted for age at enrolment and length of enrolment and adjusted for these and other covariates including study arm
Life event |
Model |
Variable |
Risk ratio |
P-value |
95% CI |
|
|---|---|---|---|---|---|---|
Left school |
Unadjusted |
left school class 1 |
2.894 |
0.001 |
1.577 |
5.309 |
(N = 2,374) |
||||||
Adjusted: age at entry, |
left school class 1 |
4.275 |
0.000 |
2.313 |
7.900 |
|
time enrolled |
age |
1.272 |
0.000 |
1.187 |
1.364 |
|
enrolled_len |
1.170 |
0.030 |
1.015 |
1.348 |
||
Adjusted: |
left school class 1 |
2.799 |
0.009 |
1.297 |
6.037 |
|
age |
1.208 |
0.000 |
1.114 |
1.309 |
||
enrolled_len |
1.218 |
0.014 |
1.040 |
1.426 |
||
arm |
0.996 |
0.979 |
0.765 |
1.299 |
||
attendance |
0.991 |
0.012 |
0.985 |
0.998 |
||
anxiety |
1.039 |
0.092 |
0.994 |
1.085 |
||
depression |
1.055 |
0.041 |
1.002 |
1.110 |
||
parental monitor |
0.997 |
0.920 |
0.936 |
1.062 |
||
HIV knowledge |
1.054 |
0.738 |
0.776 |
1.432 |
||
Sexual debut |
Unadjusted: |
debut class 1 |
1.141 |
0.720 |
0.554 |
2.349 |
(N = 2,482) |
debut class 2 |
1.498 |
0.305 |
0.693 |
3.238 |
|
Adjusted: age at entry, |
debut class 1 |
2.077 |
0.051 |
0.996 |
4.328 |
|
time enrolled |
debut class 2 |
2.781 |
0.010 |
1.276 |
6.058 |
|
age |
1.277 |
0.000 |
1.192 |
1.369 |
||
enrolled_len |
1.178 |
0.024 |
1.022 |
1.359 |
||
Adjusted: |
debut class 1 |
2.093 |
0.094 |
0.882 |
4.970 |
|
debut class 2 |
2.462 |
0.054 |
0.986 |
6.148 |
||
age |
1.196 |
0.000 |
1.109 |
1.289 |
||
enrolled_len |
1.205 |
0.014 |
1.039 |
1.396 |
||
arm |
1.033 |
0.793 |
0.810 |
1.318 |
||
attendance |
0.991 |
0.001 |
0.985 |
0.996 |
||
anxiety |
1.035 |
0.095 |
0.994 |
1.078 |
||
depression |
1.051 |
0.036 |
1.003 |
1.101 |
||
IPV |
1.058 |
0.710 |
0.785 |
1.428 |
||
HIV knowledge |
1.040 |
0.791 |
0.781 |
1.384 |
||
Pregnancy |
Unadjusted: |
pregnancy class 1 |
0.764 |
0.185 |
0.513 |
1.138 |
(N = 2,447) |
pregnancy class 3 |
1.580 |
0.183 |
0.806 |
3.097 |
|
Adjusted: age at entry, |
pregnancy class 1 |
1.633 |
0.042 |
1.017 |
2.623 |
|
time enrolled |
pregnancy class 3 |
2.832 |
0.007 |
1.336 |
6.006 |
|
age |
1.298 |
0.000 |
1.199 |
1.404 |
||
enrolled_len |
1.202 |
0.019 |
1.031 |
1.403 |
||
Adjusted: |
pregnancy class 1 |
1.525 |
0.062 |
0.979 |
2.377 |
|
pregnancy class 3 |
2.026 |
0.040 |
1.034 |
3.971 |
||
age |
1.238 |
0.000 |
1.148 |
1.335 |
||
enrolled_len |
1.192 |
0.020 |
1.028 |
1.382 |
||
arm |
1.024 |
0.852 |
0.802 |
1.307 |
||
anxiety |
1.040 |
0.062 |
0.998 |
1.083 |
||
depression |
1.057 |
0.018 |
1.010 |
1.107 |
||
IPV |
1.150 |
0.349 |
0.859 |
1.539 |
||
HIV knowledge |
1.017 |
0.910 |
0.766 |
1.350 |
||
Parent death |
Unadjusted: |
parent death class 2 |
2.030 |
0.000 |
1.602 |
2.574 |
(N = 2,071) |
||||||
Adjusted: age at entry, |
parent death class 2 |
1.994 |
0.000 |
1.563 |
2.543 |
|
time enrolled |
age |
1.252 |
0.000 |
1.170 |
1.339 |
|
enrolled_len |
1.145 |
0.055 |
0.997 |
1.316 |
||
Adjusted: |
parent death class 2 |
1.865 |
0.000 |
1.460 |
2.383 |
|
age |
1.201 |
0.000 |
1.119 |
1.290 |
||
enrolled_len |
1.155 |
0.049 |
1.001 |
1.333 |
||
arm |
1.003 |
0.978 |
0.787 |
1.280 |
||
anxiety |
1.036 |
0.080 |
0.996 |
1.078 |
||
depression |
1.061 |
0.012 |
1.013 |
1.110 |
||
Figure 3.1b shows the three-class model trajectories for sexual debut, with those with an ‘early’ debut (class 2), those with no debut or a
Figure 3.1c shows a three-class model for pregnancy, with those with an ‘early’ first pregnancy (class 3), those with no pregnancy or a ‘late’ first pregnancy (class 2) and those with a first pregnancy between these (class 1). Compared to those with a ‘late’ first pregnancy, the risk of HIV infection of a neither late nor early pregnancy (class 1) is 1.6 times higher (p-value = 0.04) and the risk of an early pregnancy (class 3) is 2.8 times higher (p-value = 0.01), when adjusted for age at enrolment and length of enrolment.
Figure 3.1d shows the two-class model trajectories for the death of a parent. The two classes are those who do not experience a death or
Generally, across all of the life events, the effect of adjusting for the study arm was not significant.
Cumulative life events and HIV incidence
Figure 3.2 shows the four-class model that was selected to describe the pattern of cumulative life events. Table 3.2 shows the characteristics of each of the cumulative life event classes; class 1 (‘Blue’) had no one who experienced coital debut, pregnancy or school leaving by age 15 and had a relatively low percentage of participants who experienced the death of a parent by age 23 (14% compared to 80–91%). Class 4 (‘Red’) experienced the most negative life events by age 14. The vast majority (88.6%) had debuted by age 14, 21% had experienced a pregnancy by age 14, 90% had experienced parental death by age 14 and 3.5% had left school. Class 2 (‘Green’) and 3 (‘Pink’) started out with a similar number of events by age 14 (more than class 1 and less than class 4), however the gradient of the trajectory for class 3 is steeper than class 2, resulting in a higher number of negative events experienced by age 23. By age 23 class 3 had caught up to class 4 in
Latent class growth analysis trajectories for the four-class model of cumulative life events, where the events included are sexual debut, pregnancy, leaving school and death of a parent
Source: Author’s ownLatent class growth analysis trajectories for the four-class model of cumulative life events, where the events included are sexual debut, pregnancy, leaving school and death of a parent
Source: Author’s ownLatent class growth analysis trajectories for the four-class model of cumulative life events, where the events included are sexual debut, pregnancy, leaving school and death of a parent
Source: Author’s ownFrequency of attributes by class from the model of cumulative life events
Class 1 |
Class 2 |
Class 3 |
Class 4 |
||||||
|---|---|---|---|---|---|---|---|---|---|
N |
% |
N |
% |
N |
% |
N |
% |
||
HIV |
Negative |
974 |
91.1 |
270 |
89.7 |
408 |
87.9 |
97 |
85.1 |
Positive |
95 |
8.9 |
31 |
10.3 |
56 |
12.1 |
17 |
14.9 |
|
Sexual debut by 14 |
No |
1,069 |
100.0 |
243 |
80.7 |
390 |
84.1 |
13 |
11.4 |
Yes |
0 |
0.0 |
58 |
19.3 |
74 |
15.9 |
101 |
88.6 |
|
Missing |
0 |
0.0 |
0 |
0.0 |
0 |
0.0 |
0 |
0.0 |
|
Sexual debut by 17 |
No |
481 |
45.0 |
234 |
77.7 |
45 |
9.7 |
3 |
2.6 |
Yes |
441 |
41.3 |
67 |
22.3 |
331 |
71.3 |
110 |
96.5 |
|
Missing |
147 |
13.8 |
0 |
0.0 |
88 |
19.0 |
1 |
0.9 |
|
Sexual debut by 21 |
No |
37 |
3.5 |
22 |
7.3 |
2 |
0.4 |
1 |
0.9 |
Yes |
590 |
55.2 |
108 |
35.9 |
350 |
75.4 |
110 |
96.5 |
|
Missing |
442 |
41.3 |
171 |
56.8 |
112 |
24.1 |
3 |
2.6 |
|
Sexual debut by 23 |
No |
3 |
0.3 |
3 |
1.0 |
0 |
0.0 |
0 |
0.0 |
Yes |
594 |
55.6 |
109 |
36.2 |
351 |
75.6 |
110 |
96.5 |
|
Missing |
472 |
44.2 |
189 |
62.8 |
113 |
24.4 |
4 |
3.5 |
|
Pregnant by 14 |
No |
1,069 |
100.0 |
299 |
99.3 |
449 |
96.8 |
90 |
78.9 |
Yes |
0 |
0.0 |
2 |
0.7 |
15 |
3.2 |
24 |
21.1 |
|
Missing |
0 |
0.0 |
0 |
0.0 |
0 |
0.0 |
0 |
0.0 |
|
Pregnant by 17 |
No |
783 |
73.2 |
297 |
98.7 |
225 |
48.5 |
37 |
32.5 |
Yes |
127 |
11.9 |
4 |
1.3 |
124 |
26.7 |
59 |
51.8 |
|
Missing |
159 |
14.9 |
0 |
0.0 |
115 |
24.8 |
18 |
15.8 |
|
Pregnant by 21 |
No |
95 |
8.9 |
39 |
13.0 |
15 |
3.2 |
3 |
2.6 |
Yes |
263 |
24.6 |
33 |
11.0 |
191 |
41.2 |
69 |
60.5 |
|
Missing |
711 |
66.5 |
229 |
76.1 |
258 |
55.6 |
42 |
36.8 |
|
Pregnant by 23 |
No |
15 |
1.4 |
7 |
2.3 |
3 |
0.6 |
2 |
1.8 |
Yes |
272 |
25.4 |
35 |
11.6 |
193 |
41.6 |
70 |
61.4 |
|
Missing |
782 |
73.2 |
259 |
86.0 |
268 |
57.8 |
42 |
36.8 |
|
Left school by 14 |
No |
1,069 |
100.0 |
301 |
100.0 |
464 |
100.0 |
110 |
96.5 |
Yes |
0 |
0.0 |
0 |
0.0 |
0 |
0.0 |
4 |
3.5 |
|
Missing |
0 |
0.0 |
0 |
0.0 |
0 |
0.0 |
0 |
0.0 |
|
Left school by 17 |
No |
870 |
81.4 |
301 |
100.0 |
306 |
65.9 |
75 |
65.8 |
Yes |
26 |
2.4 |
0 |
0.0 |
29 |
6.3 |
15 |
13.2 |
|
Missing |
173 |
16.2 |
0 |
0.0 |
129 |
27.8 |
24 |
21.1 |
|
Left school by 21 |
No |
84 |
7.9 |
32 |
10.6 |
35 |
7.5 |
6 |
5.3 |
Yes |
257 |
24.0 |
50 |
16.6 |
125 |
26.9 |
38 |
33.3 |
|
Missing |
728 |
68.1 |
219 |
72.8 |
304 |
65.5 |
70 |
61.4 |
|
Left school by 23 |
No |
8 |
0.7 |
2 |
0.7 |
2 |
0.4 |
0 |
0.0 |
Yes |
296 |
27.7 |
64 |
21.3 |
140 |
30.2 |
39 |
34.2 |
|
Missing |
765 |
71.6 |
235 |
78.1 |
322 |
69.4 |
75 |
65.8 |
|
Parent death by 14 |
No |
805 |
75.3 |
46 |
15.3 |
66 |
14.2 |
8 |
7.0 |
Yes |
0 |
0.0 |
241 |
80.1 |
375 |
80.8 |
103 |
90.4 |
|
Missing |
264 |
24.7 |
14 |
4.7 |
23 |
5.0 |
3 |
2.6 |
|
Parent death by 17 |
No |
502 |
47.0 |
46 |
15.3 |
31 |
6.7 |
5 |
4.4 |
Yes |
97 |
9.1 |
241 |
80.1 |
383 |
82.5 |
103 |
90.4 |
|
Missing |
470 |
44.0 |
14 |
4.7 |
50 |
10.8 |
6 |
5.3 |
|
Parent death by 21 |
No |
62 |
5.8 |
6 |
2.0 |
3 |
0.6 |
0 |
0.0 |
Yes |
150 |
14.0 |
243 |
80.7 |
384 |
82.8 |
104 |
91.2 |
|
Missing |
857 |
80.2 |
52 |
17.3 |
77 |
16.6 |
10 |
8.8 |
|
Parent death by 23 |
No |
6 |
0.6 |
0 |
0.0 |
0 |
0.0 |
0 |
0.0 |
Yes |
152 |
14.2 |
243 |
80.7 |
384 |
82.8 |
104 |
91.2 |
|
Missing |
911 |
85.2 |
58 |
19.3 |
80 |
17.2 |
10 |
8.8 |
|
We then examined the risk of HIV acquisition by class. Compared to class 1, the class with the fewest events by age 14, all other classes had an increased risk of HIV infection although only class 3 and 4 have a significantly increased risk of infection (Table 3.3). While class 3 and 4 experienced the same number of negative life events by age 23, class 4, the group that experienced the most negative events by age 14, had a greater risk of HIV acquisition (RR 1.81 95% CI 1.31, 2.89) than class 3 (RR 1.44, 95% CI 1.06, 1.96). Even in class 2, where by age 23 the participants have experienced fewer events than those in class 1, the risk of HIV infection is higher, though the difference is not statistically significant. This suggests that there could be an increased impact of these life events when they occur earlier in the participants’ lives as compared to later.
Risk ratios of HIV and the cumulative life event trajectories illustrated in Figure 3.2 both unadjusted and adjusted for age at enrolment and length of enrolment
Risk ratio – unadjusted |
||||||
|---|---|---|---|---|---|---|
Colour |
Class |
Risk ratio |
Std. Err. |
p-value |
95% CI |
|
Blue |
1-Reference |
1.00 |
||||
Green |
2 |
1.16 |
0.227 |
0.45 |
0.79 |
1.70 |
Pink |
3 |
1.36 |
0.216 |
0.05 |
0.99 |
1.86 |
Red |
4 |
1.68 |
0.410 |
0.03 |
1.04 |
2.71 |
Risk ratio – adjusted for age at enrolment and length of enrolment |
||||||
Colour |
Class |
Risk ratio |
Std. Err. |
p-value |
95% CI |
|
Blue |
1-Reference |
1.00 |
||||
Green |
2 |
1.11 |
0.22 |
0.59 |
0.76 |
1.63 |
Pink |
3 |
1.44 |
0.23 |
0.02 |
1.06 |
1.96 |
Red |
4 |
1.81 |
0.44 |
0.01 |
1.13 |
2.90 |
Dominance analysis ranked pregnancy as the most important of the life events in increasing risk of HIV infection, followed by leaving school, while sexual debut was third (Table 3.4). The death of a parent was ranked fourth of these life events.
Dominance analysis of life events on risk of HIV infection
Life event |
Dominance statistic |
Standardized dominance statistic |
Ranking |
|---|---|---|---|
Pregnancy |
0.0197 |
0.3185 |
1 |
Leaving school |
0.0185 |
0.2992 |
2 |
Sexual debut |
0.0173 |
0.2786 |
3 |
Parent death |
0.0064 |
0.1037 |
4 |
It was found that adolescent girls who experienced key life events before the age of 15, specifically pregnancy, coital debut, leaving school and parental death, were at increased risk of HIV infection. Using latent growth class analysis, adolescent girls who experienced more life events before the age of 15 had the highest risk of HIV acquisition by the age of 23. Those who experienced the same number of life events by age 23 were also at increased risk of HIV but not at as great a risk as the group with the most events that occurred in early adolescence. Interestingly, the class who by age 23 had the fewest events but more life events occurring in early adolescence than the other classes also had an increased risk of HIV infection compared to the group with more events later in adolescence, although the difference was not statistically significant.
These results add to the evidence base that timing of events in the adolescent life course, not just the occurrence of the events, impact the risk of HIV acquisition and that these events have not just short-term impacts on increasing risk but long-term effects well into early adulthood. All of the events examined are part of the natural transition from adolescence into adulthood; however, when these events happen early in adolescence they may place adolescents at increased risk. All of the life events examined have been found independently to be associated with HIV infection, but few studies are longitudinal so that the evidence of these events being associated with an increased risk of future infection is limited. Importantly this work demonstrates that it is not just the occurrence of a life event or the number of life events that a young person experiences that shapes HIV risk but when they occur in the life course. This work highlights that earlier life events play
There are many pathways through which early negative life events could lead to increased HIV risk. It is likely there are direct risks for HIV associated with these exposures, such as sexual behaviour, but also other more distal risk factors such as mental health and environmental stressors that likely place young people who experience these events as they transition to adulthood at greater risk. In this analysis, both early coital debut and early pregnancy indicate early initiation of sexual behaviour, meaning adolescent girls have more time when they are at risk for HIV exposure. However, even after adjusting for time, early debut and pregnancy increase risk of HIV acquisition, suggesting that other mechanisms may be increasing risk such as older partners or unprotected sex. Earlier work in this cohort has found that low school attendance and leaving school increase HIV risk and that this risk is explained by older partners and more sexual partners.(Stoner et al, 2017, 2018). Thus school provides a protective environment from HIV infection. Loss of a parent may also lead to increased risk through similar pathways such as lack of parental monitoring and parental care and supervision (Operario et al, 2011).
There is a growing recognition of the importance of childhood and early adolescence on shaping future health. There is a large body of literature demonstrating that adverse childhood events have many negative health outcomes, including increased adult mortality, chronic disease, mental health and sexual behaviour (Hughes et al, 2017; Boullier and Blair, 2018; Balistreri and Alvira-Hammond, 2016). However, much of the work on ACEs is retrospective, asking adults about negative life events that happened before the age of 18 (including, for example, abuse, neglect, death of a parent, living in a negative household/environment). Thus, these measures often do not allow for the examination of how when these events occurred in an adolescent’s life impacted future risk. There is a growing body of literature documenting early adolescence as an important time in life where negative events can have a significant impact on later life (Evans et al, 2013), but it is also a time for opportunity to intervene to counter negative future outcomes (Fuhrmann et al, 2015). While negative experiences during early adolescence can set up negative patterns of behaviour that continue into adulthood, it also provides an opportunity to intervene to prevent these risky trajectories from continuing (UNICEF, 2017). In addition, the life events measured do not occur in isolation but within a larger environment that adolescents and their caregivers operate in; cultural expectations and the social
The implications of these findings are that HIV prevention programmes need to not only screen for current risk factors but take into account earlier life events that may place adolescents on a trajectory of increased risk into adulthood and perhaps beyond that. Successfully meeting SDG goal 3.1, ending the HIV epidemic by 2030, will require a holistic view of prevention over the life course and understanding that events that happen to individuals earlier in their lives impact their future HIV risk. While HIV incidence does not increase rapidly until after age 18, experiences that happen in early adolescence and even childhood put young women at increased risk well into early adulthood. A two-fold approach to HIV prevention programming and policy in adolescents involves both intervening with younger adolescents to prevent negative early life events, working with young adolescents experiencing negative life events to prevent future negative health outcomes and also working with older adolescents who have also experienced negative life events to address trauma, coping and resilience skills to support prevention behaviours. In particular, programmes and policies that support young women staying in secondary school and completing school and those providing comprehensive, adolescent health services to prevent early pregnancies and comprehensive sexuality education in schools are some of the key steps that may help provide safety nets to help adolescents transition safely to adulthood.
References
Azen, R. and Traxel, N. (2009) Using dominance analysis to determine predictor importance in logistic regression. Journal of Educational and Behavioral Statistics 31(3): 293–318.
Balistreri, K.S. and Alvira-Hammond, M. (2016) Adverse childhood experiences, family functioning and adolescent health and emotional well-being. Public Health 132: 72–8.
Boullier, M. and Blair, M. (2018) Adverse childhood experiences. Paediatrics and Child Health 28(3): 132–7.
Budescu, D.V. (1993) Dominance analysis: a new approach to the problem of relative importance of predictors in multiple regression. Psychological Bulletin 114(3): 542–51.
Christofides, N.J., Jewkes, R.K., Dunkle, K.L., Nduna, M., Shai, N.J. and Sterk, C. (2014) Early adolescent pregnancy increases risk of incident HIV infection in the Eastern Cape, South Africa: a longitudinal study. Journal of the International AIDS Society 17(1): 18585.
Elder, G.H. (1998) The life course as developmental theory. Child Development 69(1): 1–12.
Evans, G.W., Li, D. and Whipple, S.S. (2013) Cumulative risk and child development. Psychological Bulletin 139(6): 1342–96.
Falconi, A., Gemmill, A., Dahl, R.E. and Catalano, R. (2014) Adolescent experience predicts longevity: evidence from historical epidemiology. Journal of Developmental Origins of Health and Disease 5(3): 171–7.
Fuhrmann, D., Knoll, L.J. and Blakemore, S.J. (2015) Adolescence as a sensitive period of brain development. Trends in Cognitive Sciences 19(10): 558–66.
Hughes, K., Bellis, M.A., Hardcastle, K.A., Sethi, D., Butchart, A., Mikton, C. et al (2017) The effect of multiple adverse childhood experiences on health: a systematic review and meta-analysis. The Lancet Public Health 2(8): e356–66.
Human Sciences Research Council (HSRC) (2017) The fifth South African national HIV prevalence, incidence, behaviour and communication survey, 2017: HIV impact assessment summary report. Cape Town.
Jung, T. and Wickrama, K.A.S. (2008) An introduction to latent class growth analysis and growth mixture modeling. Social and Personality Psychology Compass 2(1): 302–17.
Kidman, R. and Kohler, H.P. (2019) Adverse childhood experiences, sexual debut and HIV testing among adolescents in a low-income high HIV-prevalence context. AIDS 33(14): 2245–50.
MacMillan, R. and Copher, R. (2005) Families in the life course: interdependency of roles, role configurations, and pathways. Journal of Marriage and Family 67(4): 858–79.
Muthén, L. and Muthén, B. (2005) Mplus: statistical analysis with latent variables: user’s guide.
Operario, D., Underhill, K., Chuong, C. and Cluver, L. (2011) HIV infection and sexual risk behaviour among youth who have experienced orphanhood: systematic review and meta-analysis. Journal of the International AIDS Society 14: 25.
Pettifor, A.E., Van Der Straten, A., Dunbar, M.S., Shiboski, S.C. and Padian, N.S. (2004) Early age of first sex: a risk factor for HIV infection among women in Zimbabwe. AIDS 18(10): 1435–42.
Pettifor, A., O’Brien, K., MacPhail, C., Miller, W.C. and Rees, H. (2009) Early coital debut and associated HIV risk factors among young women and men in South Africa. International Perspectives on Sexual and Reproductive Health 35(2): 74–82.
Pettifor, A., MacPhail, C., Hughes, J.P., Selin, A., Wang, J., Gómez-Olivé, F.X. et al (2016) 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. The Lancet Global Health 4: e978–88.
StataCorp (2017) Stata Statistical Software: Release 15, 2017.
Stoner, M.C.D., Pettifor, A., Edwards, J.K., Aiello, A.E., Halpern, C.T., Julien, A. et al (2017) The effect of school attendance and school dropout on incident HIV and HSV-2 among young women in rural South Africa enrolled in HPTN 068. AIDS 31: 2127–34.
Stoner, M.C.D., Edwards, J.K., Miller, W.C., Aiello, A.E., Halpern, C.T., Julien, A. et al (2018) Does partner selection mediate the relationship between school attendance and HIV/Herpes Simplex Virus-2 among adolescent girls and young women in South Africa. Journal of Acquired Immune Deficiency Syndromes 79(1): 20–27.
United Nations International Children’s Emergency Fund (UNICEF) (2017) The Adolescent Brain: A Second Window of Opportunity. Florence: UNICEF.
Wand, H. and Ramjee, G. (2012) The relationship between age of coital debut and HIV seroprevalence among women in Durban, South Africa: a cohort study. BMJ Open. https://doi.org/10.1136/bmjopen-2011-000285
