Prevention Science

https://doi.org/10.1007/s11121-019-01015-3

Sexual Risk Behavior: a Multi-System Model of Risk and Protective

Factors in South African Adolescents
Kaymarlin Govender 1 & Richard G. Cowden 2 & Kwaku Oppong Asante 3 & Gavin George 1 & Candice Reardon 1

# Society for Prevention Research 2019

Abstract
Adolescent sexual risk behavior has typically been studied within singular, isolated systems. Using a multi-system approach, this
study examined a combination of individual, proximal, and distal factors in relation to sexual risk behavior among adolescents. A
large cross-sectional sample of 2561 adolescent (Mage = 14.92, SDage = 1.70) males (n = 1282) and females in Grades 8 (n =
1225) and 10 completed a range of self-report measures. Hierarchical ordinal logistic regression results supported a multi-system
perspective of adolescent sexual risk behavior. Although individual and peer levels were identified as the primary contributors to
the final model, a range of factors at varying levels of proximity to the individual were associated with sexual risk behavior.
Specifically, being male, black, attaining increased age, greater alcohol use (individual level), parent risk behavior (family/home
level), and peer risk behavior, feeling more pressure from peers to have sex (peer level), and lower social cohesion (community
level) were associated with increased sexual risk behavior. These findings suggest multiple individual, proximal, and distal
factors are salient to understanding sexual risk behavior among adolescents. Implications of the findings for interventions
targeting the prevention of adolescent sexual risk behavior are discussed.

Keywords Adolescents . Ecological . Multi-system . Protective factors . Risk factors . Sexual risk behavior . Youth

Adolescence is a critical developmental period characterized by gender, and self-identities (Schlüter-Müller et al. 2015). They
marked changes in a person’s internal attributes (e.g., physio- also interrogate the assumptions and people (i.e., parents) that
logical, psychological) and external environment (e.g., social, governed their past thoughts and actions. As adolescents devel-
cultural). During this phase, individuals formulate sexual, op their sense of autonomy (Labouvie-Vief 2015), they often
experience a shift in the circle of individuals that influence them
(i.e., away from parents towards peers) and begin to participate
* Kaymarlin Govender in experimental behaviors and activities (Liao et al. 2013; van
Govenderk2@ukzn.ac.za de Bongardt et al. 2014). Often underprepared for the typically
abrupt transformations that occur, adolescents must apply their
Richard G. Cowden
richardgregorycowden@gmail.com developing decision-making abilities to potentially risky situa-
tions (Leijenhorst et al. 2010) in which they may be conflicted
Kwaku Oppong Asante
kwappong@gmail.com
by a combination of individual (e.g., sensation-seeking tenden-
cies) and socio-contextual (e.g., peers) influences (Casey 2011).
Gavin George As reward-seeking processes tend to develop earlier in adoles-
georgeg@ukzn.ac.za
cence than self-regulatory processes (Steinberg 2008), deci-
Candice Reardon sions can favor risk behaviors that may have detrimental health
candiceannereardon@gmail.com
consequences (Reyna et al. 2015).
1 Although a variety of health risk behaviors (e.g., alcohol
Health Economics and HIV and AIDS Research Division, University
of KwaZulu-Natal, Westville Campus, Private Bag X54001, use, physical inactivity) emerge during adolescence (Houck
Durban 4000, South Africa et al. 2016; Sawyer et al. 2012), susceptibility to adverse
2
Department of Psychology, University of the Free State, 205 Nelson health consequences is pronounced for adolescents growing
Mandela Drive, Bloemfontein, Free State 9301, South Africa up in countries that carry the global burden of infectious dis-
3
Department of Psychology, University of Ghana, Legon Boundary, eases (Patton et al. 2016). In South Africa, where HIV preva-
Accra, Ghana lence (up to 7.1%) and incidence (up to 1.5%) rates among
 Prev Sci

individuals between 15 and 24 years of age (Shisana et al. items, scoring ranges, and internal consistency estimates for
2014; Zuma et al. 2016) are among the highest globally all scale measurements, are reported in Table 1.
(Piot et al. 2015), adolescents are especially vulnerable to
experiencing undesirable health consequences of high-risk Sexual Risk Behavior Given the complexities of measuring
sexual behavior (Govender et al. 2013). sexual risk behavior (Wilkinson et al. 2017), we used several
In light of this apparent vulnerability, an abundance of re- items commonly applied in the measurement of sexual risk
search has focused on identifying factors that affect the likeli- behavior (for reviews, see Ssewanyana et al. 2018; Toska et al.
hood of South African adolescents’ engaging in sexually risky 2017). Participants who first indicated that they had experi-
behaviors, but most studies have isolated selected risk or pro- enced sexual intercourse were presented with six additional
tective factors (e.g., Harrison et al. 2012; Kaufman et al. items assessing a variety of sexual risk behaviors, including
2014). A recent review found that across Sub-Saharan age of sexual debut, condom use at last sex, number of sexual
Africa, few studies have examined factors beyond the individ- partners, any partner > 5 years older, prior pregnancy, and
ual or family level, and studies involving a combination of transactional sex. Responses to each item were dichotomized
factors across a range of distal (e.g., community) and proximal into categories representing low and high levels of sexual risk
(e.g., individual) levels have been rare (Mmari and Sabherwal behavior (first category for each item represents sexual prac-
2013). With the success of adolescent sexual risk intervention tices characterized by low risk): sexual debut (≥ 15 years, <
programs depending on the appropriate targeting of specific, 15 years), condom use at last sex (yes, no), multiple sexual
adaptable areas at multiple systemic levels (Patton et al. 2016), partners (no, yes), any partner > 5 years older (no, yes), prior
comprehensive approaches to studying sexual risk are re- pregnancy (no, yes), and transactional sex (no, yes). Our ap-
quired in order for evidence-based, context-specific interven- proach to grading participants’ sexual risk behavior is detailed
tions to be developed. In this study, a multi-system approach is in the Fig. 1.
used to examine individual, proximal, and distal factors asso-
ciated with sexual risk behavior in a sample of South African Substance Use Substance use (not including alcohol use) was
school-going adolescents residing in KwaZulu-Natal, a high measured using an aggregated score of three items that in-
HIV burdened region in South Africa. quired about the frequency with which participants had used
tobacco, marijuana, and other illegal drugs (e.g., heroin, co-
caine, ecstasy) in the last 30 days, respectively. Responses
were provided using a three-point rating scale (1 = none; 3 =
Method more than two times).

Participants Alcohol Use We administered five items that were com-

bined for a total index of alcohol use. Items measured
The sample (N = 2561) characteristics are presented in the frequency of general alcohol use, typical quantity of
Table 1. Participants (Mage = 14.92, SDage = 1.70) were drawn alcoholic beverages when consuming alcohol, frequency
from 12 secondary schools located in the Bergville Education of binge drinking occurrences, involvement in alcohol-
(N = 7) and Central Durban Education (N = 5) circuits within related incidents, and whether others have expressed con-
the province of KwaZulu-Natal, South Africa. For state cern about participants’ drinking patterns. Sample items
funding allocations, the schools are categorized into poverty include Bhow often do you have a drink containing
quintiles,1 with the Bergville circuit comprising rural, lower alcohol?^ (1 = never; 5 = four or more times a week)
income black families (poverty quintiles 1 to 3). The Central and Bhow many drinks containing alcohol do you have
Durban circuit has greater racial diversity and families of on a typical day when drinking?^ (1 = I do not drink
higher socioeconomic status (poverty quintiles 4 and 5). The alcohol; 6 = ten or more drinks).
majority of the participants were male (50.08%), in Grade 10
(52.17%), and black (85.55%) students. Leisure Opportunities A nine-item scale was constructed to
measure adolescents’ perceptions about leisure time, the
leisure time activities they partake in, and the activities
Materials available in the community for them to engage in during
their free time. All items were rated using a five-point
Participants completed the following measures. Response op- response scale (1 = strongly disagree; 5 = strongly agree)
tions for all single-item measures, along with the number of and aggregated for a total score. Sample items include BI
1 know of places in the community where there are lots of
All South African public schools are classified into one of five quintiles
based on socioeconomic indicators of the communities that surround each things to do^ and Bfor me, free time just drags on and
school (Murray 2016). on^ (reverse scored).
 Prev Sci
Table 1 Summary statistics and bivariate associations (ordinal logistic regression) with sexual risk behavior

Systemic level Variable n (%) Items Item/scale range M (SD) ω DV = sexual risk behavior
(none = 0, low = 1,
high = 2)

Individual Age 2559 14.92 (1.70) OR = 1.56***, 95%

CI [1.47, 1.67]
Gender 2560
Female (ref) 1278 (49.92)
Male 1282 (50.08) OR = 3.30***, 95%
CI [2.65, 4.11]
Race 2546
Other (ref) 368 (14.45)
Black 2178 (85.55) OR = 2.65***, 95% CI
[1.83, 3.82]
School grade 2561
Grade 8 (ref) 1225 (47.83)
Grade 10 1336 (52.17) OR = 3.26***, 95%
CI [2.60, 4.08]
Leisure 2261 9 9 to 45 21.79 (5.54) .73 OR = 1.00, 95%
opportunities CI [.98, 1.02]
Substance use 2513 3 3 to 9 3.31 (.89) .71 OR = 1.96***, 95%
CI [1.75, 2.19]
Alcohol use 2212 5 5 to 22 6.18 (2.45) .82 OR = 1.34***, 95%
CI [1.29, 1.40]
Family/home Household resident 2489
with HIV
No (ref) 1532 (61.55)
I do not know 856 (34.39) OR = 1.07, 95%
CI [.86, 1.33]
Yes 101 (4.06) OR = 2.00**, 95%
CI [1.28, 3.12]
Parental status 2448
Both parents living (ref) 1623 (66.30)
Father deceased 492 (20.10) OR = 1.54***, 95%
CI [1.20, 1.98]
Mother deceased 141 (5.76) OR = 1.47, 95%
CI [.96, 2.23]
Both parents 192 (7.84) OR = 1.99***, 95%
deceased CI [1.41, 2.82]
Parental monitoring 2396 6 6 to 30 22.45 (5.84) .85 OR = .94***, 95%
CI [.92, .96]
Communication 2026 16 16 to 80 43.27 (12.42) .89 OR = 1.02**, 95%
about sex CI [1.01, 1.02]
Sibling risk 2410 5 5 to 20 7.78 (4.05) .92 OR = 1.01, 95%
behavior CI [.99, 1.04]
Parental/caregiver 2522 3 3 to 6 5.45 (.80) .51 OR = .71***, 95% CI [.64, .80]
support
Violence at home 1997 12 12 to 60 14.75 (3.55) .74 OR = 1.05**, 95% CI [1.01, 1.08]
Parent risk behavior 2496 4 4 to 8 4.53 (.86) .62 OR = 1.29***, 95% CI [1.16, 1.44]
Table 1 (continued)

Systemic level Variable n (%) Items Item/scale range M (SD) ω DV = sexual risk behavior
(none = 0, low = 1,
high = 2)

Peer Pressure to have sex 2531

No (ref) 1687 (66.65)
I am not sure 293 (11.58) OR = 2.67***, 95% CI [1.96, 3.63]
A little bit 222 (8.77) OR = 4.80***, 95% CI [3.50, 6.57]
Yes 329 (13.00) OR = 7.39***, 95% CI [5.60, 9.75]
Prosocial peers 2496 2 2 to 8 6.21 (1.50) .60 OR = .83***, 95% CI [.78, .89]
Peer support 2465 3 3 to 12 9.01 (2.70) .77 OR = .98, 95% CI [.94, 1.01]
Peer risk behavior 2507 4 4 to 12 5.38 (1.84) .80 OR = 1.56***, 95% CI [1.48, 1.65]
School Frequency of lessons about HIV/AIDS 2540
Not often (ref) 854 (33.62)
Often 1109 (43.66) OR = .84, 95% CI [.67, 1.06]
Very often 577 (22.72) OR = .87, 95% CI [.67, 1.15]
Frequency of lessons about sex 2526
Not often (ref) 824 (32.62)
Often 1136 (44.97) OR = 1.18, 95% CI [.93, 1.49]
Very often 566 (22.41) OR = 1.32, 95% CI [1.00, 1.74]
Information about HIV/AIDS helpful 2537
No (ref) 323 (12.73)
Yes 2214 (87.27) OR = .89, 95% CI [.66, 1.20]
Information about sex helpful 2528
No (ref) 409 (16.18)
Yes 2119 (83.82) OR = .99, 95% CI [.75, 1.31]
School provides information 2515
about organizations to get help
No (ref) 970 (38.57)
Yes 1545 (61.43) OR = .96, 95% CI [.78, 1.18]
School 2561
A (quintile 5) (ref) 419 (16.36)
B (quintile 4) 237 (9.25) OR = 1.15, 95% CI [.77, 1.71]
C (quintile 5) 242 (9.45) OR = .39***, 95% CI [.23, .64]
D (quintile 3) 151 (5.90) OR = 1.15, 95% CI [.73, 1.82]
E (quintile 2) 113 (4.41) OR = 1.26, 95% CI [.78, 2.04]
F and G# (quintile 1) 187 (7.30) OR = .85, 95% CI [.54, 1.33]
H (quintile 3) 113 (4.41) OR = .93, 95% CI [.53, 1.62]
I (quintile 5) 301 (11.75) OR = .51**, 95% CI [.33, .77]
J (quintile 3) 260 (10.15) OR = 1.12, 95% CI [.76, 1.65]
K (quintile 5) 287 (11.21) OR = .78, 95% CI [.53, 1.16]
L (quintile 2) 251 (9.80) OR = 1.31, 95% CI [.90, 1.91]
Violence at school 1986 12 12 to 60 15.54 (3.68) .74 OR = 1.06***, 95% CI [1.03, 1.09]
Ease of learner engagement in risk behaviors at school 2437 4 4 to 16 7.88 (3.17) .81 OR = 1.07***, 95% CI [1.04, 1.11]
School connectedness 1997 18 18 to 90 55.60 (9.65) .83 OR = .99, 95% CI [.98, 1.00]
Community Community support 2534
No (ref) 1052 (41.52)

Prev Sci
Yes 1482 (58.48) OR = .94, 95% CI [.77, 1.15]
Violence in community 2141 12 12 to 60 19.49 (6.00) .82 OR = 1.07***, 95% CI [1.05, 1.09]
Prev Sci

The initial analysis for school could not be estimated, likely due to the comparably smaller sample size in school F (n = 37, 1.44%) relative to the other schools. We combined participants in school F with
those in school G (n = 150, 5.86%), as both schools are located in the Bergville Education district and are classified as Quintile 1 schools. Neither school was associated with a higher odds of sexual risk
Parental Vital Status and HIV Status of Household Residents

OR = .91***, 95% CI [.88, .95]

DV = sexual risk behavior Participants completed two single items to determine if their
biological parents (none, one, or both) were alive and whether
a household resident is living with HIV.
(none = 0, low = 1,

Parent Risk Behaviors Four items were administered to assess

high = 2)

whether participants’ parents/caregivers had smoked ciga-

rettes, been visibly intoxicated in front of them, used illegal
substances, and been in trouble with the police in the last
6 months. These items were rated using a two-point scale
(1 = no; 2 = yes) and combined for a total measure of parent
.55
ω

risk behavior.

Parental Support Three items were developed to measure sup-

13.10 (2.79)

portive parent/caregiver behaviors. Participants were

M (SD)

prompted to respond to each item (e.g., BI talk to my

parents/caregiver about my homework assignments and
projects^) by selecting one of two dichotomous options (1 =
no; 2 = yes). The items were aggregated for an index of family
Item/scale range

support.
4 to 20

Parental Monitoring Six items were selected from the Parental

Monitoring Scale (Small and Kerns 1993) to measure the ex-
tent to which participants were monitored by their parents/
caregivers. A five-point response format (1 = never; 5 = very
Items

often) was used to rate the items, which were summed for a
4

total parental monitoring score.

Parent-Child Communication About Sex Participants

responded to 16 items measuring quality of communication
n (%)

2297

about sexual topics with their parents/primary caregivers

(Jaccard et al. 2000). Items were rated on a five-point scale
(1 = strongly disagree; 5 = strongly agree) and combined for a
total score.

Sibling Risk Behaviors Five items were developed to assess

OR odds ratio, 95% CI 95% confidence intervals for odds ratio

whether participants’ siblings had used tobacco, been intoxi-

cated, used illegal substances, engaged in sexual activity, and
been in trouble with the police during the last 6 months.
Participants rated each item using a four-point response scale
(1 = I do not have siblings; 4 = yes), which were combined for
an index of sibling risk behavior.
Social cohesion

Peer Risk Behavior Four items assessed peer involvement in

problem behaviors and sexual activity (Brook et al. 2006).
p < .05, ** p < .01, *** p < .001
Variable

Items were rated on a three-point response scale (1 = none;

3 = most) and combined for a total score.
Table 1 (continued)

behavior (p = .793)

Prosocial Peers Prosocial peer behavior was assessed using

Systemic level

two items (e.g., Bmy friends try to do what is right^). Both

items were rated on four-point scale (1 = not at all true; 4 =
very much true) and summed for a total measure of peer
prosociality.
#

*
 Prev Sci

No Yes
Ever had sex
(n = 1952) (n = 591)

Yes Condom use No

(n = 374) at last sex (n = 217)
No sexual risk
behavior †
n = (1952)
Age of sexual
No Yes
debut < 15
(n = 261) (n = 246)
years

Multiple
No Yes
lifetime sexual
(n = 300) (n = 263)
Low sexual partners High sexual
risk behavior†† risk behavior‡
(n = 85) (n = 385)
No Any partner Yes
(n = 411) > 5 years older (n = 178)

No Transactional Yes
(n = 500) sex (n = 76)

No Yes
Ever pregnant
(n = 519) (n = 64)

Fig. 1 Gradation of sexual risk behavior. † Participants have not endorsed high sexual risk-taking behavior on at least one subsequent
experienced sexual intercourse; ††Participants have experienced sexual item. Participants with missing responses to one or more items (n =
intercourse and endorsed low sexual risk-taking behavior on all 139) were omitted from sexual risk behavior gradation
subsequent items. ‡Participants have experienced sexual intercourse and

Peer Support Three items inquired about the extent to which Information About HIV, Sex, and Availability of Support We
adolescents received support from peers. Participants used a developed five single-item measures to assess frequency of
four-point response format (1 = not at all true; 4 = very much HIV and sex education at school (along with the perceived
true) to respond to each of the items, which were combined for helpfulness of such information) and whether schools had
an index of perceived peer support. Sample items include BI provided participants with information about organizations
have a friend my own age who really cares about me^ and BI in the community (e.g., health clinics) where they could re-
have a friend my own age who helps me when I’m having a ceive support.
difficult time.^
Violence Scale Adolescents completed 12 items from the
School Connectedness The Psychological Sense of School Screen for Adolescent Violence Exposure scale (Hastings
Membership scale (Goodenow 1993) was used to measure and Kelley 1997) to assess experiences of violence in the
school connectedness. The 18 items assess participants’ sense home, school, and community. Using a five-point response
of inclusion, acceptance, respect, encouragement, and belong- format (1 = never; 5 = always), participants rated the 12 items
ing at school. Items were rated on a five-point response scale with reference to each of the three aforementioned contexts
(1 = strongly disagree; 5 = strongly agree) and aggregated for (36 responses in total). Respective responses were combined
a total school connectedness score. for subscale totals referring to each context.

Ease of Learner Engagement in Risk Behaviors at School Four Social Cohesion Four items from the Collective Efficacy Scale
items were constructed to determine the ease with which (Sampson et al. 1997) were used to measure perceived social
learners are able to engage in problem behaviors and sexual cohesion in the community. Items were completed using a
activity on school premises (e.g., Bhow easy is it for learners to five-point response scale (1 = strongly disagree; 5 = strongly
drink alcohol on the school premises and not get caught?^). agree) and aggregated for an index of social cohesion.
Items were rated using a four-point response scale (1 = you
would get caught for sure; 4 = very easy, no educator would Community Support A single item was used to assess percep-
notice) and aggregated for a composite score. tions of community support.
Prev Sci

Table 2 Summary statistics of the hierarchical ordinal regression model (sequential block entry) predicting sexual risk behavior

Systemic level Determinant DV = sexual risk behavior (none = 0, low = 1, high = 2)

Estimate (SE) AOR [95% CI]

Individual (block 1) Age .29*** (.08) 1.33 [1.13, 1.57]

Gender
Female (ref)
Male .98*** (.26) 2.68 [1.61, 4.46]
Race
Other (ref)
Black 1.46** (.46) 4.32 [1.76, 10.60]
School grade
Grade 8 (ref)
Grade 10 .49 (.29) 1.62 [.92, 2.86]
Substance use .01 (.14) 1.01 [.77, 1.31]
Alcohol use .18*** (.04) 1.19 [1.10, 1.29]
Model χ2 (df) 250.86*** (6)
Nagelkerke R2 .315
Family/home (block 2) Household resident with HIV
No (ref)
I do not know − .14 (.22) .87 [.56, 1.33]
Yes .34 (.53) 1.40 [.50, 3.97]
Parental status
Both parents living (ref)
Father deceased .29 (.27) 1.33 [.79, 2.26]
Mother deceased .45 (.37) 1.57 [.76, 3.22]
Both parents deceased .42 (.35) 1.52 [.76, 3.03]
Parental monitoring − .01 (.02) .99 [.96, 1.03]
Communication about sex .00 (.01) 1.00 [.98, 1.02]
Parental/caregiver support − .12 (.12) .89 [.70, 1.12]
Violence at home .03 (.04) 1.03 [.96, 1.11]
Parent risk behavior .25* (.12) 1.28 [1.01, 1.63]
Model χ2 (df) 269.14*** (16)
Block χ2 (df) 18.28 (10)
Nagelkerke R2 (R2 change) .335 (.020)
Peer (block 3) Pressure to have sex
No (ref)
I am not sure .57* (.29) 1.77 [1.01, 3.12]
A little bit 1.07*** (.31) 2.90 [1.60, 5.28]
Yes 136*** (.31) 3.89 [2.10, 7.20]
Prosocial peers .11 (.07) 1.12 [.97, 1.28]
Peer risk behavior .29*** (.07) 1.33 [1.17, 1.52]
Model χ2 (df) 313.56*** (21)
Block χ2 (df) 44.43*** (5)
Nagelkerke R2 (R2 change) .383 (.048)
School (block 4) School
A (quintile 5) (ref)
B (quintile 4) .29 (.44) 1.33 [.57, 3.13]
C (quintile 5) − .12 (.47) .89 [.36, 2.22]
D (quintile 3) − .37 (.49) .69 [.26, 1.81]
E (quintile 2) − .15 (.51) .86 [.32, 2.33]
 Prev Sci

Table 2 (continued)

Systemic level Determinant DV = sexual risk behavior (none = 0, low = 1, high = 2)

Estimate (SE) AOR [95% CI]

F and G# (quintile 1) .11 (.43) 1.11 [.47, 2.60]

H (quintile 3) − .62 (.73) .54 [.13, 2.24]
I (quintile 5) − .01 (.49) .99 [.38, 2.61]
J (quintile 3) .40 (.40) 1.49 [.68, 3.29]
K (quintile 5) − .27 (.44) .77 [.33, 1.80]
L (quintile 2) − .15 (.42) .86 [.37, 1.96]
Ease of learner engagement in risk behaviors at school .01 (.03) 1.01 [.95, 1.08]
Violence at school .01 (.04) 1.01 [.93, 1.09]
Model χ2 (df) 320.25*** (33)
Block χ2 (df) 6.69 (12)
Nagelkerke R2 (R2 change) .390 (.007)
Community (block 5) Violence in community − .01 (.02) .99 [.94, 1.03]
Social cohesion − .08* (.04) .93 [.86, .99]
Model χ2 (df) 324.66*** (35)
Block χ2 (df) 4.41 (2)
Nagelkerke R2 (R2 change) .394 (.004)

*p < .05, **p < .01, ***p < .001

Procedure ordinal (proportional odds) logistic regression to examine bi-

variate relationships between the study variables and sexual
Permission to conduct the study was granted by the University of risk behavior. Variables that revealed significant (p < .05) bi-
KwaZulu-Natal Human and Social Science Research Ethics variate associations with sexual risk were retained for use in
Committee and the Provincial Department of Basic Education, the primary analysis.
KwaZulu-Natal. Subsequently, the permission of the 12 schools Systemic level predictors of sexual risk behavior were es-
in the Bergville and Central Durban Education circuits was ac- timated using a hierarchical ordinal (proportional odds) logis-
quired in order to obtain access to the students. This process also tic regression model. Variables were entered into the model in
involved informing adolescents’ parents about the study and the a series of sequential blocks (five in total), beginning with
nature of their child’s prospective participation in it. For those those at the individual level. The remaining blocks of vari-
students who volunteered to participate in the study, written con- ables were entered from most proximal (i.e., family) to most
sent (from parents or legal guardians) and assent (from adoles- distal (i.e., community) with regard to the individual.2 For the
cents) were obtained prior to their participation. The question- full model, collinearity diagnostics (all VIF values ≤ 3.28) did
naire was translated from English into isiZulu, which was not reveal any multicollinearity issues (Tabachnick and Fidell
checked for accuracy using backtranslation. Participants from 2013). Analysis of parallel lines (all p values ≥ .065) indicated
schools in the Bergville Education circuit completed the ques- the proportional odds assumption was appropriately met
tionnaire in IsiZulu, whereas those within the Central Durban (Williams 2016).
Education circuit requested to complete the English version. A
multilingual team of fieldworkers was available to provide guid-
ance and support to participants. Questionnaires were completed
Results
in school classrooms during predetermined timeframes allocated
by each school for data collection.
Bivariate Analyses

Data Analyses Results of the bivariate relations between sexual risk and each
of the study variables are displayed in Table 1. Within the
Statistical computations were performed in R (R Core Team
2
2018). Analyses were performed using a pairwise deletion Block 1 (Individual), Block 2 (Individual + Family/home), Block 3
(Individual + Family/home + Peer), Block 4 (Individual + Family/home +
approach. For all measures containing at least two items, in- Peer + School), Block 5 (Individual + Family/home + Peer + School +
ternal consistency was estimated using omega total. We used Community).
Prev Sci

individual level, sexual risk behavior was higher among ado- multi-system perspective of sexual risk behavior among adoles-
lescents who were older, male, black, in Grade 10, and those cents, as a range of factors were associated with sexual risk
who reported higher alcohol and substance use. For the family/ behavior at varying levels of proximity to the individual. This
home level, participants who had someone diagnosed with highlights the importance of considering a multi-level combina-
HIV/AIDS living in their home, were paternal or double or- tion of psychosocial factors to understanding sexual risk behavior
phans, reported less parental monitoring and communication in adolescence (Salazar et al. 2010; Tenkorang and Maticka-
about sex with their parents, experienced violence at home, Tyndale 2014). The findings also suggest that there may be dis-
and endorsed greater parent risk behavior tended to engage in tinctions in the importance of systemic levels depending on
higher sexual risk behavior. Associations with the peer-level where factors fit in within the proximal-distal spectrum, as pre-
variables indicated sexual risk behavior was higher among ad- dominant effects were found for the individual and peer systemic
olescents who experienced more pressure to have sex and re- levels. This notion is consistent with prior research that has found
ported greater peer risk behavior, whereas sexual risk behavior proximal factors tend to be stronger predictors of adolescent
was lower when peers were more prosocial. Within the school sexual risk behavior (James et al. 2013; Lansford et al. 2014).
level, sexual risk behavior was higher when it was easier for Based on the overall model, biological factors (i.e., age, race,
learners to engage in risk behaviors at school and there was and sex) were among the strongest predictors of sexual risk
greater violence at school. Lower sexual risk behavior was behavior at the individual level. These findings are largely con-
found among two schools in quintile 5. For the community sistent with a number of reviews reporting distinctions in sexual
level, greater violence in the community and lower social co- risk behavior propensities based on biographical attributes (e.g.,
hesion were associated with higher sexual risk behavior. Toska et al. 2017). Further, the results coincide with a wealth of
research highlighting the salience of alcohol use as a behavioral
Primary Analysis antecedent of sexual risk behavior (Patrick et al. 2015). A note-
worthy finding was the relative importance of alcohol use, as
Results of the hierarchical ordinal logistic regression are pre- compared to substance use, in predicting sexual risk, advocat-
sented in Table 2. The overall model containing five blocks ing the importance of separating the two behavioral choices
was statistically significant (Nagelkerke R2 = .39, p < .001), when examining sexual risk behavior. Although alcohol and
with systemic-level effects found for the individual substance use are often highly correlated among adolescents
(p < .001) and peer (p < .001) levels. Although the family/ (Capaldi 2014; Kelly et al. 2015), the neurocognitive effects
home (p = .051), school (p = .878), and community of alcohol use (i.e., reduction in behavioral inhibition and im-
(p = .110) level blocks did not contribute significantly to mod- pulse control) on behavior (Winward et al. 2014) may have a
el fit, the results indicated a combination of individual, prox- greater influence on sexually risky behavioral choices.
imal, and distal variables were associated with sexual risk Considering the prominence of alcohol use with regard to sex-
behavior. Specifically, being older, male, black, higher alcohol ual risk behavior found in this study, the finding that leisure
use (individual level), higher parent/caregiver engagement in opportunities was unrelated to sexual risk behavior is of partic-
risk behavior (family/home level), feeling more pressure from ular interest. With adolescents less likely to partake in alcohol
peers to have sexual intercourse, higher peer risk behavior use outside of leisure time periods (Weybright et al. 2016), the
(peer level), and lower social cohesion (community level) types of activities adolescents engage in during their free time
were associated with increased sexual risk behavior. All other may be of greater importance than their perceptions of and
determinants were unrelated to sexual risk behavior (p > .05). attitudes towards leisure opportunities.
Comparable to prior research that has found links between
risky parental behavior and health-risk behaviors among ado-
Discussion lescents (e.g., Donaldson et al. 2016), parent risk behavior was
the single predictor of sexual risk behavior at the family/home
Adolescents growing up in countries where the global burden of level. Adolescents’ sexual risk behavior was linked to their
sexually transmitted infections (including HIV) is highest are engagement in activities that resembled those included in the
especially vulnerable to the health-related consequences of sex- measure of parent risk behavior (e.g., alcohol use), suggesting
ual risk behavior. Developing effective prevention programing that exposure to parents’ maladaptive alcohol use may have a
requires a comprehensive understanding of contextually relevant profound effect on adolescents’ decisions to participate in al-
factors associated with sexual risk behavior among key popula- cohol consumption (van der Zwaluw et al. 2008). That is,
tions living in such countries. In this study, a multi-system ap- parental behaviors may indirectly endorse the kinds of activ-
proach was used to examine relations between a combination of ities that heighten adolescents’ proclivity to engage in such
individual, proximal, and distal factors and sexual risk behavior behaviors themselves. Furthermore, parent risk behavior (e.g.,
in a sample of South African adolescents living in a high alcohol intoxication) impairs parents’ functioning and the
HIV burdened region of the country. The findings supported a ability to effectively fulfill parental responsibilities (e.g.,
 Prev Sci

monitoring), which may lead to parent-child conflict that advantage of capturing a comprehensive range of behav-
drives adolescents towards other extrafamilial influences and iors, caution should be applied in generalizing the findings
affords adolescents greater freedom to partake in health-risk to any singular sexual risk behavior. Fourth, all variables
behaviors (Latendresse et al. 2008). were derived from self-report measures, the accuracy of
Within the peer level, peer risk behavior and feeling greater which may have been affected by task-related demands
pressure to have sex were associated with heightened sexual and the social context in which participation occurred
risk behavior. Peer influences affect adolescents’ sexual risk (Schroder et al. 2003). Although self-report ratings are
behavior directly through their need to belong and feel accepted commonly applied to the study of sexual risk behavior
(Selikow et al. 2009), as well as indirectly through the types of (DiClemente et al. 2013), a broader range of factors (par-
sexual risk-related activities (e.g., alcohol use) that are promot- ticularly at more distal levels) could be captured in future
ed by peers who engage in delinquent behavior (Tomé et al. research by gathering data from other informants. Research
2012). The present findings underscore the increased role of is also needed to identify the relevant processes, contexts,
peer influences during adolescence (Liao et al. 2013), with and interplay between the determinants hitherto identified
negative peer influences exerting a stronger effect on sexual as affecting risk sexual behavior, a precursor to HIV infec-
risk behavior than positive peer influences (e.g., peer support). tion. Additionally, sophisticated research designs are nec-
Similar to previous research that has found facets of social essary to investigate the longitudinal influences of various
capital tend to promote safer sexual practices (Crosby et al. individual, proximal, and distal factors on adolescents’
2003), social cohesion emerged as a community-level factor health risk behaviors, particularly among at-risk and mar-
associated with lower sexual risk behavior. This finding iterates ginalized populations. Policies and programs that attend to
the relevance of distal levels of influence on risky sexual be- the economic and social needs of families and communi-
havior (Salazar et al. 2010), likely due to the impact distal ties, as well as those that seek to build individual compe-
systems (e.g., community) have on more proximal (e.g., indi- tencies, will be critical for adolescents to safely navigate
vidual) systems (Hutchison and Wood 2007). For example, in their development, particularly within AIDS-affected com-
communities characterized by higher levels of social cohesion, munities (Govender et al. 2018).
adolescents have more opportunities to develop bonds with
community members who can supervise, monitor, and positive-
ly shape their values (Sampson et al. 2002). Moreover, inter- Conclusion
connected communities might have a greater influence over the
norms and behavioral choices (e.g., parental monitoring) of This study represents one of the few studies that has
families living within such communities (Valdimarsdóttir and adopted a multi-system approach to examining South
Bernburg 2015). Collective social capital may also produce African adolescents’ sexual risk behavior, a country that
health benefits by diffusing knowledge about health-related is- continues to have some of the highest global high HIV
sues in communities and invoking informal social control over incidence and prevalence rates (Shisana et al. 2014;
health-related behaviors (Boyce et al. 2008). Zuma et al. 2016). The findings of this study support the
understanding that narrowly focusing on individual risk
Limitations and Future Research Directions and protective factors, while ignoring risk and protective
factors across multiple levels, will likely undermine efforts
While our application of a multi-system approach offers targeting maladaptive health-risk behaviors, including the
promising insight into the proximal and distal factors asso- effectiveness of HIV prevention programing.
ciated with sexual risk behavior among adolescents living
in an HIV endemic region of South Africa, selected meth- Compliance with Ethical Standards
odological limitations ought to be considered. First, the
findings of this study are based on cross-sectional data, Conflict of Interest The authors declare that they have no conflict of
thereby preventing determinations of causality. Second, interest.
participants were conveniently sampled from two school
Ethical Approval All procedures involving participants in this study
districts within a single province. Given South Africa’s
were performed in accordance with the ethical standards of the institu-
geographically varied demography and socioeconomic cli- tional and/or national research committee and with the 1964 Helsinki
mates, indiscriminate application of the conclusions drawn Declaration and its later amendments or comparable ethical standards.
in this study may neglect to appreciate contextually specif- Study approval was granted by the University of KwaZulu-Natal
Human and Social Science Research Ethics Committee and the
ic distinctions in adolescents’ sexual risk behavior. Third,
Provincial Department of Basic Education, KwaZulu-Natal.
gradation of sexual risk behavior was based on a combina-
tion of items that captured a relatively heterogenous range Informed Consent Informed consent was obtained from all individual
of risky sexual behaviors. While this approach has the participants included in the study.
Prev Sci

References James, S., Montgomery, S., Leslie, L., & Zhang, J. (2013). Sexual risk
behaviors among youth in the child welfare system. Child and Youth
Services Review, 31, 990–1000. https://doi.org/10.1016/j.
Boyce, W., Davies, D., Gallupe, O., & Shelley, D. (2008). Adolescent risk
childyouth.2009.04.014.
taking, neighborhood social capital, and health. Journal of
Kaufman, Z., Braunschweig, E., Feeney, J., Dringus, S., Weiss, H.,
Adolescent Health, 43, 246–252. https://doi.org/10.1016/j.
Delaney-Moretlwe, S., & Ross, D. (2014). Sexual risk behavior,
jadohealth.2008.01.014.
alcohol use, and social media use among secondary school students
Brook, D. W., Morojele, N. K., Zhang, C., & Brook, J. S. (2006). South
in informal settlements in Cape Town and Port Elizabeth, South
African adolescents: Pathways to risky sexual behavior. AIDS
Africa. AIDS and Behavior, 18, 1661–1674. https://doi.org/10.
Education and Prevention, 18, 259–272. https://doi.org/10.1521/
1007/s10461-014-0816-x.
aeap.2006.18.3.259.
Kelly, A. B., Evans-Whipp, T. J., Smith, R., Chan, G. K., Toumbourou, J.
Capaldi, D. M. (2014). Prevention science supplemental issue commen-
W., Patton, G. C., … Catalano, R. F. (2015). A longitudinal study of
tary promoting healthy sexual practices: What we have learned from
the association of adolescent polydrug use, alcohol use and high
100 years of work. Prevention Science, 15, S78–S80. https://doi.org/
school non-completion. Addiction, 110, 627–635. https://doi.org/
10.1007/s11121-013-0434-2.
10.1111/add.12829.
Casey, S. (2011). Elements of psychological maturity and its influence on Labouvie-Vief, G. (2015). Cognitive–emotional development from ado-
antisocial and criminal behavior. The Open Criminology Journal, 4, lescence to adulthood. In G. Labouvie-Vief (Ed.), Integrating emo-
32–39. https://doi.org/10.2174/1874917801104010032. tions and cognition throughout the lifespan (pp. 89–116). Cham:
Crosby, R. A., Holtgrave, D. R., DiClemente, R. J., Wingood, G. M., & Springer International Publishing.
Gayle, J. A. (2003). Social capital as a predictor of adolescents’ Lansford, J., Dodge, K., Fontaine, R., Bates, J., & Pettit, G. (2014). Peer
sexual risk behavior: A state-level exploratory study. AIDS and rejection, affiliation with deviant peers, delinquency, and risky sex-
Behavior, 7, 245–252. https://doi.org/10.1023/A:1025439618581. ual behavior. Journal of Youth and Adolescence, 43, 1742–1751.
DiClemente, R. J., Swartzendruber, A. L., & Brown, J. L. (2013). https://doi.org/10.1007/s10964-014-0175-y.
Improving the validity of self-reported sexual behavior: No easy Latendresse, S. J., Rose, R. J., Viken, R. J., Pulkkinen, L., Kaprio, J., &
answers. Sexually Transmitted Diseases, 40, 111–112. https://doi. Dick, D. M. (2008). Parenting mechanisms in links between parents’
org/10.1097/OLQ.0b013e3182838474. and adolescents’ alcohol use behaviors. Alcoholism, Clinical and
Donaldson, C. D., Handren, L. M., & Crano, W. D. (2016). The enduring Experimental Research, 32, 322–330. https://doi.org/10.1111/j.
impact of parents’ monitoring, warmth, expectancies, and alcohol 1530-0277.2007.00583.x.
use on their children’s future binge drinking and arrests: A longitu- Leijenhorst, L., Moor, B., Op de Macks, Z., Rombouts, S., Westenberg,
dinal analysis. Prevention Science, 17, 606–614. https://doi.org/10. P., & Crone, E. (2010). Adolescent risky decision-making:
1007/s11121-016-0656-1. Neurocognitive development of reward and control regions.
Goodenow, C. (1993). The psychological sense of school membership NeuroImage, 51, 345–355. https://doi.org/10.1016/j.neuroimage.
among adolescents: Scale development and educational correlates. 2010.02.038.
Psychology in the Schools, 30, 79–90. https://doi.org/10.1002/1520- Liao, Y., Huang, Z., Huh, J., Pentz, M., & Chou, C. (2013). Changes in
6807(199301)30:1<79::AID-PITS2310300113>3.0.CO;2-X. friends’ and parental influences on cigarette smoking from early
Govender, K., Naicker, S., Meyer-Weitz, A., Fanner, J., Naidoo, A., & through late adolescence. Journal of Adolescent Health, 53, 132–
Penfold, W. (2013). Associations between perceptions of school 138. https://doi.org/10.1016/j.jadohealth.2013.01.020.
connectedness and adolescent health risk behaviors in south Mmari, K., & Sabherwal, S. (2013). A review of risk and protective
African high school learners. Journal of School Health, 83, 614– factors for adolescent sexual and reproductive health in developing
622. https://doi.org/10.1111/josh.12073. countries: An update. Journal of Adolescent Health, 53, 562–572.
Govender, K., Masebo, W., Nyamaruze, P., Cowden, R. G., Schunter, B. https://doi.org/10.1016/j.jadohealth.2013.07.018.
T., & Bains, A. (2018). HIV prevention in adolescents and young Murray, M. (2016). Does poor quality schooling and/or teacher quality
people in the Eastern and Southern African region: A review of key hurt black South African students enrolling for a degree at the
challenges impeding actions for an effective response. The Open University of KwaZulu-Natal? PLoS One, 11, e0153091. https://
AIDS Journal, 12, 53–67. https://doi.org/10.2174/ doi.org/10.1371/journal.pone.0153091.
1874613601812010053. Patrick, M. E., Maggs, J. L., & Lefkowitz, E. S. (2015). Daily associa-
Harrison, A., Smit, J., Hoffman, S., Nzama, T., Leu, S., Mantell, J., … tions between drinking and sex among college students: A longitu-
Exner, T. (2012). Gender, peer and partner influences on adolescent dinal measurement burst design. Journal of Research on
HIV risk in rural South Africa. Sexual Health, 9, 178–186. https:// Adolescence, 25, 377–386. https://doi.org/10.1111/jora.12135.
doi.org/10.1071/SH10150. Patton, G. C., Sawyer, S. M., Santelli, J. S., Ross, D. A., Afifi, R., Allen,
Hastings, T. L., & Kelley, M. L. (1997). Development and validation of N. B., …Viner, R. M. (2016). Our future: A Lancet commission on
the Screen for Adolescent Violence Exposure (SAVE). Journal of adolescent health and wellbeing. Lancet, 387, 2423–2478. https://
Abnormal Child Psychology, 25, 511–520. https://doi.org/10.1023/ doi.org/10.1016/S0140-6736(16)00579-1.
A:1022641916705. Piot, P., Abdool Karim, S., Hecht, R., Legido-Quigley, H., Buse, K.,
Houck, C. D., Hadley, W., Barker, D., Brown, L. K., Hancock, E., & Stover, J., … UNAIDS–Lancet Commission. (2015). Defeating
Almy, B. (2016). An emotion regulation intervention to reduce risk AIDS—Advancing global health. The Lancet, 386, 171–218.
behaviors among at-risk early adolescents. Prevention Science, 17, https://doi.org/10.1016/S0140-6736(15)60658-4.
71–82. https://doi.org/10.1007/s11121-015-0597-0. R Core Team. (2018). R: A language and environment for statistical
Hutchison, M. K., & Wood, E. B. (2007). Reconceptualizing adolescent computing. R Foundation for Statistical Computing, Vienna.
sexual risk in a parent-based expansion of the theory of planned Retrieved from https://www.R-project.org/. Accessed 16 Oct 2018
behavior. Journal of Nursing Scholarship, 39, 141–146. https:// Reyna, V. F., Wilhelms, E. A., McCormick, M. J., & Weldon, R. B.
doi.org/10.1111/j.1547-5069.2007.00159.x. (2015). Development of risky decision making: Fuzzy-trace theory
Jaccard, J., Dittus, P. J., & Gordon, V. V. (2000). Parent-teen communi- and neurobiological perspectives. Child Development Perspectives,
cation about premarital sex: Factors associated with the extent of 9, 122–127. https://doi.org/10.1111/cdep.12117.
communication. Journal of Adolescent Research, 15, 187–208. Salazar, L. F., Bradley, E. L. P., Younge, S. N., Daluga, N. A., Crosby, R.
https://doi.org/10.1177/0743558400152001. A., Lang, D. L., & DiClemente, R. J. (2010). Applying ecological
 Prev Sci

perspectives to adolescent sexual health in the United States: Explanatory model. Global Journal of Health Science, 4, 26–35.
Rhetoric or reality? Health Education Research, 25, 552–562. https://doi.org/10.5539/gjhs.v4n2p26.
https://doi.org/10.1093/her/cyp065. Toska, E., Pantelic, M., Meinck, F., Keck, K., Haghighat, R., & Cluver, L.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods (2017). Sex in the shadow of HIV: A systematic review of preva-
and violent crime: A multilevel study of collective efficacy. Science, lence, risk factors, and interventions to reduce sexual risk-taking
277, 918–924. https://doi.org/10.1126/science.277.5328.918. among HIV-positive adolescents and youth in sub-Saharan Africa.
Sampson, R. J., Morenoff, J. D., & Gannon-Rowley, T. (2002). Assessing PLoS One, 12, e0178106. https://doi.org/10.1371/journal.pone.
Bneighborhood effects^: Social processes and new directions in re- 0178106.
search. Annual Review of Sociology, 28, 443–478. https://doi.org/10. Valdimarsdóttir, M., & Bernburg, J. G. (2015). Community disadvantage,
1146/annurev.soc.28.110601.141114. parental network, and commitment to social norms: Multilevel study
Sawyer, S. M., Afifi, R. A., Bearinger, L. H., Blakemore, S.-J., Dick, B., of self-reported delinquency in Iceland. Journal of Research in
Ezeh, A. C., & Patton, G. C. (2012). Adolescence: A foundation for Crime and Delinquency, 52, 213–244. https://doi.org/10.1177/
future health. Lancet, 379, 1630–1640. https://doi.org/10.1016/ 0022427814548685.
S0140-6736(12)60072-5. van de Bongardt, D., Reitz, E., Sandfort, T., & Deković, M. (2014). A
Schlüter-Müller, S., Goth, K., Jung, E., & Schmeck, K. (2015). meta-analysis of the relations between three types of peer norms and
Assessment and treatment of identity pathology during adolescence. adolescent sexual behavior. Personality and Social Psychology
Scandinavian Journal of Child and Adolescent Psychiatry and Review, 19, 203–234. https://doi.org/10.1177/1088868314544223.
Psychology, 3, 63–70. https://doi.org/10.21307/sjcapp-2015-006. van der Zwaluw, C., Scholte, R., Vermulst, A., Buitelaar, J., Verkes, R., &
Schroder, K. E., Carey, M. P., & Vanable, P. A. (2003). Methodological Engels, R. (2008). Parental problem drinking, parenting, and ado-
challenges in research on sexual risk behavior: II. Accuracy of self- lescent alcohol use. Journal of Behavioral Medicine, 31, 189–200.
reports. Annals of Behavioral Medicine, 26, 104–123. https://doi. https://doi.org/10.1007/s10865-007-9146-z.
org/10.1207/S15324796ABM2602_03.
Weybright, E. H., Caldwell, L. L., Ram, N., Smith, E. A., & Wegner, L.
Selikow, T., Ahmed, N., Flisher, A. J., Mathews, C., & Mukoma, W.
(2016). Trajectories of adolescent substance use development and
(2009). I am not Bumqwayito^: A qualitative study of peer pressure
the influence of healthy leisure: A growth mixture modelling ap-
and sexual risk behaviour among young adolescents in Cape Town,
proach. Journal of Adolescence, 49, 158–169. https://doi.org/10.
South Africa. Scandinavian Journal of Public Health, 37, 107–112.
1016/j.adolescence.2016.03.012.
https://doi.org/10.1177/1403494809103903.
Shisana, O., Rehle, T., Simbayi, L., Zuma, K., Jooste, S., Zungu, N., … Wilkinson, A. L., El-Hayek, C., Fairley, C. K., Roth, N., Tee, B. K.,
Onoya, D. (2014). South African National HIV Prevalence, McBryde, E., … Stoové, M. (2017). Measuring transitions in sexual
Incidence and Behaviour Survey, 2012. Cape Town: Human risk among men who have sex with men: The novel use of latent
Sciences Research Council Press. class and latent transition analysis in HIV sentinel surveillance.
Small, S. A., & Kerns, D. (1993). Unwanted sexual activity among peers American Journal of Epidemiology, 185, 627–635. https://doi.org/
during early and middle adolescence: Incidence and risk factors. 10.1093/aje/kww239.
Journal of Marriage and the Family, 55, 941–952. https://doi.org/ Williams, R. (2016). Understanding and interpreting generalized ordered
10.2307/352774. logit models. The Journal of Mathematical Sociology, 40, 7–20.
Ssewanyana, D., Mwangala, P. N., van Baar, A., Newton, C. R., & https://doi.org/10.1080/0022250X.2015.1112384.
Abubakar, A. (2018). Health risk behaviour among adolescents liv- Winward, J. L., Hanson, K. L., Tapert, S. F., & Brown, S. A. (2014).
ing with HIV in sub-Saharan Africa: A systematic review and meta- Heavy alcohol use, marijuana use, and concomitant use by adoles-
analysis. BioMed Research International, 2018, 7375831. https:// cents are associated with unique and shared cognitive decrements.
doi.org/10.1155/2018/7375831. Journal of the International Neuropsychological Society, 20, 784–
Steinberg, L. (2008). A social neuroscience perspective on adolescent 795. https://doi.org/10.1017/S1355617714000666.
risk-taking. Developmental Review, 28, 78–106. https://doi.org/10. Zuma, K., Shisana, O., Rehle, T. M., Simbayi, L. C., Jooste, S., Zungu,
1016/j.dr.2007.08.002. N., … Abdullah, F. (2016). New insights into HIV epidemic in
Tabachnick, B. G., & Fidell, L. S. (2013). Using multivariate statistics South Africa: Key findings from the National HIV Prevalence,
(6th ed.). New York: Pearson. Incidence and Behaviour Survey, 2012. African Journal of AIDS
Tenkorang, E. Y., & Maticka-Tyndale, E. (2014). Assessing young peo- Research, 15, 67–75. https://doi.org/10.2989/16085906.2016.
ple’s perceptions of HIV risks in Nyanza, Kenya: Are school and 1153491.
community level factors relevant? Social Science and Medicine,
116, 93–101. https://doi.org/10.1016/j.socscimed.2014.06.041. Publisher’s Note Springer Nature remains neutral with regard to juris-
Tomé, G., Matos, M., Simões, C., Diniz, J. A., & Camacho, I. (2012). dictional claims in published maps and institutional affiliations.
How can peer group influence the behavior of adolescents: