Skip to main content
PMC home page
HHS Author Manuscripts logoLink to HHS Author Manuscripts
. Author manuscript; available in PMC: 2022 Dec 26.
Published in final edited form as: Lancet Glob Health. 2021 Nov 22;10(1):e124–e133. doi: 10.1016/S2214-109X(21)00457-5

Changes in prevalence of violence and risk factors for violence and HIV among children and young people in Kenya: a comparison of the 2010 and 2019 Kenya Violence Against Children and Youth Surveys

Francis B Annor 1, Laura F Chiang 2, Patricia R Oluoch 3, Vivienne Mang’oli 4, Marygorret Mogaka 5, Mary Mwangi 6, Anne Ngunjiri 7, Francis Obare 8, Thomas Achia 9, Pragna Patel 10, Greta M Massetti 11, Linda L Dahlberg 12, Thomas R Simon 13, James A Mercy 14
PMCID: PMC9791664  NIHMSID: NIHMS1854030  PMID: 34822755

Summary

Background

Previous research has shown a high prevalence of violence among young people in Kenya. Violence is a known risk factor for HIV acquisition and these two public health issues could be viewed as a syndemic. In 2010, Kenya became the third country to implement the Violence Against Children and Youth Survey (VACS). The study found a high prevalence of violence in the country. Led by the Government of Kenya, stakeholders implemented several prevention and response strategies to reduce violence. In 2019, Kenya implemented a second VACS. This study examines the changes in violence and risk factors for violence and HIV between 2010 and 2019.

Methods

The 2010 and 2019 VACS used a similar sampling approach and measures. Both VACS were cross-sectional national household surveys of young people aged 13–24 years, designed to produce national estimates of physical, sexual, and emotional violence. Prevalence and changes in lifetime experiences of violence and risk factors for violence and HIV were estimated. The VACS uses a three-stage cluster sampling approach with random selection of enumeration areas as the first stage, households as the second stage, and an eligible participant from the selected household as the third stage. The VACS questionnaire contains sections on demographics, risk and protective factors, violence victimisation, violence perpetration, sexual behaviour, HIV testing and services, violence service knowledge and uptake, and health outcomes. For this study, the main outcome variables were violence victimisation, context of violence, and risk factors for violence. All analyses were done with the entire sample of 13–24-year-olds stratified by sex and survey year.

Findings

The prevalence of lifetime sexual, physical, and emotional violence significantly declined in 2019 compared with 2010, including unwanted sexual touching, for both females and males. Experience of pressured and forced sex among females also decreased between the surveys. Additionally, significantly more females sought and received services for sexual violence and significantly more males knew of a place to seek help in 2019 than in 2010. The prevalence of several risk factors for violence and HIV also declined, including infrequent condom use, endorsement of inequitable gender norms, endorsement of norms justifying wife beating, and never testing for HIV.

Interpretation

Kenya observed significant declines in the prevalence of lifetime violence and some risk factors for violence and HIV, and improvements in some service seeking indicators between 2010 and 2019. Continued prioritisation of preventing and responding to violence in Kenya could contribute to further reductions in violence and its negative outcomes. Other countries in the region that have made substantial investments and implemented similar violence prevention programmes could use repeat VACS data to monitor violence and related outcomes over time.

Introduction

Violence against children and young people is a global crisis—1 billion children experience one or more forms of violence annually.1 Research has shown that violence can lead to short-term and long-term health and social consequences with lasting effects across the life course, including on mental health, sexual and reproductive health, chronic disease, premature death, and violence perpetration and revictimisation.2–4 The combined epidemics of violence and HIV/AIDS contribute to a disproportionate burden of disease in sub-Saharan Africa. Violence can impact HIV transmission and acquisition directly (through forced sex by an infected person) and indirectly (through compromised negotiations for safer sex and increased sexual risk-taking behaviours).5,6 The US President’s Emergency Plan for AIDS Relief (PEPFAR) has included violence prevention in its HIV prevention portfolio, particularly for adolescent girls and young women through the Determined, Resilient, Empowered, AIDS-free, Mentored and Safe (DREAMS) initiative. In 2010, Kenya was the third country to conduct a nationally representative Violence Against Children and Youth Survey (VACS). That survey showed high rates of childhood violence: 76% of young women and nearly 80% of young men aged 18–24 years had experienced at least one type of violence (sexual, physical, or emotional) before the age of 18 years.7 Although few data were available before 2010, research supported these findings.8,9 A study of 500 women in Nairobi aged 18–24 years found that a quarter of them were raped before age 18 years.8 The 2008–09 Demographic and Health Surveys showed a high prevalence of gender-based violence.9 The results from the 2010 Kenya VACS and the resulting national action plan led to a cross-sectoral approach in which the Government of Kenya and partners prioritised violence prevention and responses.10,11

Kenya’s National Prevention and Response Plan was grounded on the idea that violence against children is neither justifiable nor inevitable, and that it can be prevented. Between 2010 and 2019, there was considerable legislative progress in child protection. The 2010 revision of the Kenya Constitution strengthened legal protections for child victims of violence and the revision of the Children’s Act protected children from physical, sexual, and psychological abuse, neglect, and other forms of exploitation.12,13 These policy frameworks were complemented by substantial investments in primary prevention by the Government of Kenya and organisations including PEPFAR and UNICEF.

In 2019, Kenya conducted a second nationally representative VACS, updating data on violence and assessing progress since 2010. VACS have been done in 21 countries and Kenya was the second country to repeat the VACS after Zimbabwe.14

This study examined changes in violence, contexts of sexual violence, and risk factors for violence that have implications for HIV prevention, given the HIV epidemic in Kenya.

The objectives of this study were to estimate the prevalence of lifetime physical, sexual, and emotional violence in 2010 and 2019; estimate the prevalence of service seeking, risk factors for sexual violence, and inequitable gender attitudes; and examine if these estimates significantly changed between 2010 and 2019. The consistency in methods between the 2010 and 2019 VACS makes Kenya uniquely poised to examine changes in violence prevalence and related risk factors.

Methods

Study design

We analysed the 2010 and 2019 Kenya VACS, which are cross-sectional nationally representative household surveys of young people aged 13–24 years that are designed to produce national estimates of physical, sexual, and emotional violence. VACS use a three-stage cluster sampling approach with random selection of enumeration areas as the first stage, households as the second stage, and an eligible participant from the selected household as the third stage. VACS sampling was done without replacement. Sampling design, inclusion criteria, protocol details, and data collection procedures are summarised in table 1, and additional details are available in survey final reports and other publications.7,15–17

Table 1:

Implementation of the 2010 and 2019 Kenya Violence Against Children and Youth Surveys

2010 2019
Lead and implementing agencies DCS with technical support from the KNBS, UNICEF Kenya, and the CDC, and funded by UNICEF. The CDC provided technical assistance. DCS with technical support from KNBS, UNICEF Kenya, CDC, and Ministry of Health, in collaboration with the UCSF, Population Council, and LVCT Health, and funded by the President’s Emergency Plan for AIDS Relief. The CDC provided technical assistance.
Sampling frame KNBS fourth NASSEP, based on 1999 population and housing census. KNBS fifth NASSEP, based on 2009 population and housing census.
Mode of survey administration and survey languages Face to face, using pen and paper; one of the following languages: Borana, English, Kalenjin, Kamba, Kikuyu, Kisii, Kiswahili, Luhya, Luo, Masai, Meru, Mijikenda, and Somali. Face to face using tablets with programmed Open Data Kit forms; one of the following languages: Borana, English, Kalenjin, Kamba, Kikuyu, Kisii, Kiswahili, Luhya, Luo, Masai, Meru, Mijikenda, and Somali.
Sample selection Three-stage cluster sampling: 238 EAs selected using probability proportional to size as the first stage, 35 households selected per EA using simple random sampling as the second stage, and one eligible participant per household using Kish methods of random selection as the third stage. Three attempts were made at each household to complete the interview. After the third attempt, if an interview was not completed for any reason, neither the selected participant nor the household were replaced. Three-stage cluster sampling: 264 EAs selected using probability proportional to size at the first stage, 34 households selected per EA using simple random sampling as the second stage, and one eligible participant per household using Kish methods of random selection as the third stage. Three attempts were made at each household to complete the interview. After the third attempt, if an interview was not completed for any reason, neither the selected participant nor the household were replaced.
Consenting Obtained informed consent from all participants using WHO guidelines; for minors, permission obtained from parent or guardian followed by assent from child. Interviewers read the consent and assent information to the parents and participants. Obtained informed consent from all participants using WHO guidelines; for minors, permission obtained from parent or guardian followed by assent from child. Interviewers read the consent and assent information from the tablets to the parents and participants.
Open in a new tab

CDC=US Centers for Disease Control and Prevention. DCS=Kenya Department of Children’s Services. EA=enumeration area. IRB=institutional review board. KNBS=Kenya National Bureau of Statistics. NASSEP=National Sample Survey and Evaluation Program. UCSF=University of California, San Francisco.

Both the 2010 and 2019 Kenya VACS included both male participants and female participants who spoke one of the survey languages listed in table 1. Both the surveys followed the same sampling, ethical review, and consenting procedures. For participants aged 13–17 years (minors), parental permission was obtained and assent from the minor participant, and participants aged 18 years or older and emancipated minors provided informed consent. Consent and assent were provided verbally and documented by interviewers on the paper consent document or tablets for the 2010 and 2019 surveys, respectively. Direct referrals to counselling services were offered to participants who needed and wanted services.

The main violence questions and sample selection information are shown in the appendix (pp 1–4). Sex-matched interviewers conducted in-person interviews with participants in private and facilitated direct referrals to counsellors as required. The 2010 protocol was reviewed and approved by the Ethical Review Committee of the Kenya Medical Research Institute and by the US Centers for Disease Contol and Prevention (CDC) institutional review board (IRB; protocol number 5964) and the 2019 protocol by the CDC, University of California, San Francisco, Kenyatta National Hospital/University of Nairobi, and Population Council IRBs (protocol number 6538).

Procedures

The global VACS questionnaire draws questions from validated survey tools that have been extensively used in global contexts, using measures that have been cognitively tested and validated in VACS and other studies.7,15 Published studies have demonstrated psychometric properties, reliability, and validity of the violence questionnaire tools employed in VACS.18,19 The VACS questionnaire contains sections on demographics, risk and protective factors, violence victimisation, violence perpetration, sexual behaviour, HIV testing and services, violence service knowledge and uptake, and health outcomes. For this study, the main outcome variables were violence victimisation, context of violence, and risk factors for violence. All analyses were done with the entire sample of 13–24-year-olds stratified by sex and survey year.

Both the 2010 and 2019 surveys assessed four forms of sexual violence: unwanted sexual touching, unwanted attempted sex, forced sex, and pressured sex. Lifetime sexual violence was defined as having experienced one or more of these forms of sexual violence at any age (0–24 years) by any perpetrator

In 2010, physical violence questions were asked for three perpetrator types: (a) intimate partner, (b) parents, adult caregivers, and other adult relatives, and (c) adults in the community or neighbourhood. In 2019, peer physical violence was added, but this measure was not included in the present analysis (because there was no 2010 comparator). Lifetime physical violence was defined as having experienced physical violence at any age (0–24 years) from one or more of the three perpetrator types included in both surveys.

Both 2010 and 2019 VACS assessed emotional violence by adults, with the 2010 survey asking about any adults, whereas the 2019 survey asked specifically about parents, adult caregivers, and other adult relatives. The 2010 VACS assessed whether any adult had made the respondent feel humiliated or unwanted, or if the respondent had been threatened with or experienced actual abandonment. The 2019 VACS asked whether an adult relative ever made the respondent feel unloved, said hurtful words like they wished the respondent was dead or was never born, or ridiculed the respondent. To compare emotional violence between surveys, the 2010 questions on abandonment were excluded. Any lifetime emotional violence was defined as experiencing any emotional violence at any age (0–24 years) from an adult

Any lifetime violence included those who had experienced lifetime sexual, physical, or emotional violence.

Both VACS assessed the age at the first incident of any form of sexual violence, and disclosure and service seeking (ie, told someone, knew a place to seek service, sought service, and received service, among those who experienced any lifetime sexual violence).

Both VACS assessed violence risk factors, including multiple sexual partners (two or more sexual partners in the past 12 months); infrequent condom use (sometimes or never used condoms in the past 12 months); never tested for HIV; sexually transmitted infection (ever had symptoms or diagnosis of sexually transmitted infection); and child marriage (marriage before age 18 years).

Both VACS assessed inequitable gender attitudes with the following yes or no statements: only men, not women, should decide when to have sex; if someone insults a boy or man, he should defend his reputation with force if he needs to; there are times when a woman should be beaten; women who carry condoms have sex with a lot of men; and a woman should tolerate violence to keep her family together. The surveys assessed attitudes towards acceptance of wife beating with the following yes/no statements: it is acceptable for a man to hit or beat his wife (i) if she goes out without telling him, (ii) if she neglects the children, (iii) if she argues with him, (iv) if she refuses to have sex with him, and (v) if she burns the food.

Statistical analysis

The study sample was weighted to represent the population using a three-step weighting procedure, as follows: computation of base weight, adjustment of base weight for differential non-response, and calibration of the post-stratification weight to census data. Demographic characteristics and weighted prevalence of violence, context, and risk factors for violence and HIV were assessed. The analyses to assess change were done in two parts, as follows: unadjusted difference using Pearson χ2 test, which is equivalent to the Z test for comparing two independent proportions,20 and adjusted difference using logistic regression analyses and controlling for age, educational attainment, marital status, orphan status, rural versus urban status, and previous pregnancy. Logistic regression models tested each violence variable independently while controlling for covariates and a dummy variable for year. Unadjusted logistic regression models were also run to show changes in estimates between the adjusted and unadjusted models (appendix p 5). Differences were considered statistically significant if the two-sided p value associated with the Pearson χ2 test or the logistic regression was less than 0·05.

All “Don’t know” and “Declined” responses were set to missing and were not included in any calculations. Pairwise deletion was used to handle missing data. Sensitivity analyses assessed the effects of changing variable definitions, inclusion or exclusion of some variables from the logistic regression models, and different age groupings on the trends observed. All analyses were stratified by sex.21

Analyses used SAS software version 9.4, accounting for the complex survey design (weight, cluster, and strata) of the 2010 and 2019 Kenya VACS.

Role of the funding source

There was no funding source for this study.

Results

During the 2010 VACS, 1227 female participants (overall response rate 84·4%) and 1456 male participants (overall response rate 80·4%) completed the survey. During the 2019 VACS, 1344 female participants (overall response rate 74·0%) and 788 male participants (overall response rate 66·5%) completed the survey. Only 0–3% of variables included in the study had missing data. All percentages presented are weighted to be representative of the population. The demographic distributions of the 2010 and 2019 surveys were similar for females and males with respect to age, educational attainment, and orphan status (table 2). Females were also similar with respect to rural versus urban status. For females, the two samples differed significantly on marital status (20·9% were married or lived together as if married in 2019 vs 30·0% in 2010; p<0·0001) and ever being pregnant (24·4% in 2019 vs 32·3% in 2010; p=0·0005). Males differed significantly on marital status (5·5% in 2019 vs 9·6% in 2010; p=0·0003) and rural versus urban residency (61·7% rural in 2019 vs 82·4% rural in 2010; p=0·0003).

Table 2:

Demographic characteristics of 13–24-year-olds, Kenya 2010 and 2019 VACS

2010 VACS
 2019 VACS
 p value
n Weighted % n Weighted %
(95% CI) (95% CI)
Females
Age, years 1227 ·· 1344 ·· ··
 13–17 ·· 444 (40·6–48·1) ·· 44·9 (41·4–48·4) 0·84
 18–24 ·· 55·6 (51·9–59·4) ·· 55·1 (51·6–58·6) ··
Educational attainment 1218 ·· 1343 ·· ··
 Completed primary school or lower ·· 40·4 (34·7–46·0) ·· 45·3 (41·6–49·1) 0·15
 Higher than primary school ·· 59·6 (54·0–65·3) ·· 54·7 (50·9–58·4) ··
Marital status 1224 ·· 1315 ·· ··
 Married or lived together as if married ·· 30·0 (26·6–33·4) ·· 20·9 (18·2–23·5) <0·0001
 Never married or lived together as if married ·· 70·0 (66·6–73·4) ·· 79·1 (76·5–81·8) ··
Orphan status 1227 ·· 1300 ·· ··
 Orphan ·· 25·5 (21·0–29·9) ·· 23·4 (20·1–26·6) 0·45
 Not an orphan ·· 74·5 (70·1–79·0) ·· 76·6 (73·4–79·9) ··
Rural or urban status 1227 ·· 1344 ·· ··
 Rural ·· 75·6 (65·8–85·5) ·· 65·2 (61·5–68·9) 0·072
 Urban ·· 24·4 (14·5–34·2) ·· 34·8 (31·1–38·5) ··
Pregnancy 1219 ·· 1338 ·· ··
 Ever pregnant ·· 32·3 (28·7–35·9) ·· 24·4 (21·6–27·2) 0·0005
 Never pregnant ·· 67·7 (64·1–71·3) ·· 75·6 (72·8–78·4) ··
Males
Age, years 1456 ·· 788 ·· ··
13–17 ·· 47·2 (42·9–51·6) ·· 46·0 (42·6–49·4) 0·66
18–24 ·· 52·8 (48·4–57·1) ·· 54·0 (50·6–57·4) ··
Educational attainment 1446 ·· 788 ·· ··
 Completed primary school or lower ·· 40·8 (36·4–45·1) ·· 41·9 (37·9–45·9) 0·70
 Higher than primary school ·· 59·2 (54·9–63·6) ·· 58·1 (54·1–62·1) ··
Marital status 1453 ·· 778 ·· ··
 Married or lived together as if married ·· 9·6 (7·1–12·0) ·· 5·5 (4·6–6·5) 0·0003
 Not married or lived together as if married ·· 90·4 (88·0–92·9) ·· 94·5 (93·5–95·4) ··
Orphan status 1456 ·· 744 ·· ··
 Orphan ·· 24·8 (21·0–28·5) ·· 22·0 (18·4–25·7) 0·30
 Not an orphan ·· 75·2 (71·5–79·0) ·· 78·0 (74·3–81·6) ··
Rural versus urban status 1456 ·· 788 ·· ··
 Rural ·· 82·4 (75·9–88·9) ·· 61·7 (54·1–69·2) 0·0003
 Urban ·· 17·6 (11·1–24·1) ·· 38·3 (30·8–45·9) ··
Open in a new tab

VACS=Violence Against Children and Youth Surveys.

The prevalence of lifetime sexual, physical, and emotional violence and experiencing no, one, or multiple types of violence were significantly lower for females in 2019 compared with their experiences in 2010 (table 3). The prevalence of lifetime violence in females in 2019 versus 2010 was 25·2% versus 36·2% (for any sexual violence), 45·9% versus 74·9% (for any physical violence), and 16·8% versus 28·4% (for any emotional violence). A similar pattern was observed for males with one exception—the prevalence of males who experienced only one form of violence did not change significantly. Observed declines were significant for both females and males after controlling for covariates.

Table 3:

Prevalence of lifetime violence, overlap of lifetime violence, and different forms of lifetime sexual violence among 13–24-year-olds, Kenya 2010 and 2019 VACS

Prevalence
 Adjusted models*
2010
 2019
 p value β coefficient†
 p value
n Weighted % (95% CI) n Weighted % (95% CI) 2010 2019
Females
Any lifetime sexual violence 1222 36·2 (31·5–41·0) 1344 25·2 (22·3–28·2) <0·0001 0 −0·489 0·0002
Any lifetime physical violence 1224 74·9 (68·7–81·0) 1344 45·9 (42·1–49·7) <0·0001 0 −1·196 <0·0001
Any lifetime emotional violence 1222 28·4 (24·6–32·2) 1343 16·8 (14·3–19·3) <0·0001 0 −0·658 <0·0001
Any lifetime violence 1224 82·5 (78·0–87·1) 1344 58·3 (54·2–62·4) <0·0001 0 −1·176 <0·0001
Only one form of violence 1224 40·3 (37·1–43·5) 1344 34·4 (31·3–37·6) 0·0098 0 −0·267 0·0080
Only two forms of violence 1224 27·5 (22·7–32·4) 1344 18·2 (15·1–21·3) 0·0008 0 −0·513 0·0026
All three forms of violence 1224 14·7 (11·0–18·3) 1344 5·7 (4·6–6·8) <0·0001 0 −0·969 <0·0001
Experienced no violence 1224 17·5 (12·9–22·0) 1344 41·7 (37·6–45·8) <0·0001 0 1·176 <0·0001
Experienced unwanted sexual touching 1212 25·8 (21·5–30·1) 1343 11·4 (8·7–14·1) <0·0001 0 −1·052 <0·0001
Experienced unwanted attempted sex 1216 16·2 (13·0–19·3) 1337 15·4 (12·9–17·8) 0·69 0 0·037 0·81
Experienced pressured sex 1218 10·3 (8·0–12·6) 1339 6·5 (5·0–7·9) 0·0030 0 −0·411 0·024
Experienced physically forced sex 1220 9·2 (6·5–11·9) 1342 5·4 C3·9–6·9) 0·0081 0 −0·472 0·031
Males
Any lifetime sexual violence 1444 19·7 (16·6–22·7) 787 11·4 (9·0–13·8) <0·0001 0 −0·628 0·0003
Any lifetime physical violence 1447 78·7 (75·5–81·8) 788 52·7 (45·5–59·9) <0·0001 0 −1·115 <0·0001
Any lifetime emotional violence 1445 29·8 (25·8–33·8) 788 11·5 (8·7–14·4) <0·0001 0 −1·217 <0·0001
Any lifetime violence 1447 83·0 (80·1–85·9) 788 56·6 (49·8–63·4) <0·0001 0 −1·247 <0·0001
Only one form of violence 1447 47·2 (43·4–51·0) 788 41·5 (35·0–48·0) 0·15 0 −0·156 0·33
Only two forms of violence 1447 26·5 (23·1–30·0) 788 11·2 (8·9–13·4) <0·0001 0 −1·144 <0·0001
All three forms of violence 1447 9·3 (7·4–11·1) 788 3·9 (1·4–64) 0·0105 0 −0·793 0·047
Experienced no violence 1447 17·0 (14·1–19·9) 788 43·4 (36·6–50·2) <0·0001 0 1·247 <0·0001
Experienced unwanted sexual touching 1438 13·0 (10·4–15·5) 784 4·7 (2·8–6·5) <0·0001 0 −1·101 <0·0001
Experienced unwanted attempted sex 1438 9·4 (7·4–11·4) 780 6·8 (3·9–97) 0·18 0 −0·299 0·32
Experienced pressured sex 1434 4·5 (3·1–5·9) 782 2·9 (1·3–4·5) 0·14 0 −0·575 0·077
Experienced physically forced sex 1439 1·9 (0·9–2·8) 783 1·8 (0·5–3·1) 0·95 0 −0·022 0·96
Open in a new tab

VACS=Violence Against Children and Youth Surveys.

*

The adjusted model for females controlled for age, educational attainment, marital status, orphan status, rural versus urban status, and previous pregnancy using logistic regression; the adjusted model for males controlled for age, educational attainment, marital status, orphan status, and rural versus urban status using logistic regression; the logistic regression was used to assess if changes in prevalence were significant after controlling for the important covariates.

†

β coefficients were generated using logistic regression, accounting for the complex survey design of the Kenya VACS.

Table 3 presents the prevalence of the four types of sexual violence for females and males. Among females, the prevalence of each form of sexual violence was significantly lower in 2019 compared with 2010, except for unwanted attempted sex (15·4% in 2019 vs 16·2% in 2010; p=0·69). Among males, only the prevalence of unwanted sexual touching was significantly different (4·7% in 2019 vs 13·0% in 2010; p<0·0001), with the other three forms of sexual violence not significantly changing between the surveys. Declines in the prevalence of the types of sexual violence that were significant before adjustment remained significant for both females and males after controlling for covariates.

Table 4 provides contextual information for sexual violence experiences. For females, among those who experienced sexual violence, the age at the first incident of sexual violence did not change between the surveys. In 2019, a significantly smaller proportion of males first experienced sexual violence at age 13 years or younger compared with 2010, after controlling for covariates. Significantly more female survivors sought services (11·3% in 2019 vs 4·6% in 2010; p=0·013) and received services (10·0% in 2019 vs 3·0% in 2010; p=0·0020) in 2019 compared with 2010, and these changes remained significant after controlling for covariates. The proportions of female survivors who told someone about their experiences of sexual violence or who knew a place to seek services did not differ over time. Significantly more male sexual violence survivors in 2019 knew of a place to seek services compared with 2010 (33·1% in 2019 vs 15·5% in 2010; p=0·013), and this significant difference remained after controlling for covariates. However, the proportion of male survivors of sexual violence who told someone, sought assistance from sexual violence services, and received these services did not significantly change.

Table 4:

Context of first incident of sexual violence and service seeking for any incident of sexual violence among 13–24-year-old sexual violence survivors, Kenya 2010 and 2019 VACS

Prevalence
 Adjusted models*
2010 VACS
 2019 VACS
 p value β coefficient†
 p value
n Weighted % (95% CI) n Weighted % (95% CI) 2010 2019
Females
Age at first incidence of sexual violence, years
 ≤13 400 20·5 (15·4–25·5) 296 20·9 (16·4–25·4) 0·91 0 −0·162 0·48
 14–15 400 32·8 (26·4–39·3) 296 25·4 (19·7–31·2) 0·083 0 −0·415 0·063
 16–17 400 24·2 (18·7–29·7) 296 26·0 (19·9–32·1) 0·66 0 0·149 0·56
 ≥18 400 22·5 (16·3–28·7) 296 27·7 (21·0–34·4) 0·26 0 0·456 0·12
Disclosure and service seeking
 Told someone 384 43·0 (38·1–48·0) 304 46·7 (38·7–54·8) 0·43 0 0·162 0·42
 Knew of a place 381 25·0 (18·3–31·7) 299 33·7 (27·9–39·5) 0·050 0 0·397 0·083
 Sought services 381 4·6 (1·6–7·7) 299 11·3 (7·0–15·6) 0·013 0 1·154 0·0042
 Received services 381 3·0 (0·8–5·1) 299 10·0 (5·8–14·1) 0·0020 0 1·472 0·0017
Males
Age at first incidence of sexual violence, years
 ≤13 259 30·7 (23·6–37·8) 82 24·3 (13·7–35·0) 0·34 0 −1·060 0·0070
 14–15 259 24·3 (16·9–31·7) 82 15·2 (7·5–22·9) 0·099 0 −0·351 0·33
 16–17 259 20·5 (13·9–27·1) 82 27·7 (14·2–41·1) 0·29 0 0·663 0·12
 ≥18 259 24·5 (17·5–31·5) 82 32·8 (21·0–44·6) 0·23 0 0·448 0·15
Disclosure and service seeking
 Told someone 245 31·8 (25·5–38·1) 88 27·9 (16·8–38·9) 0·54 0 −0·232 0·37
 Knew of a place 243 15·5 (9·6–21·5) 88 33·1 (17·2–49·0) 0·013 0 1·136 0·015
 Sought services 244 3·8 (0·9–6·6) 88 6·8 (0·3–13·3) 0·31 0 0·340 0·51
 Received services 244 1·9 (0·2–3·5) 88 6·0 (0·0–12·4) 0·057 0 0·963 0·069
Open in a new tab

VACS=Violence Against Children and Youth Surveys.

*

The adjusted model for females controlled for age, educational attainment, marital status, orphan status, rural versus urban status, and previous pregnancy using logistic regression; the adjusted model for males controlled for age, educational attainment, marital status, orphan status, and rural versus urban status using logistic regression; the logistic regression was used to assess if changes in the prevalence were significant after controlling for the important covariates.

†

β coefficients were generated using logistic regression, accounting for the complex survey design of the Kenya VACS.

Table 5 presents changes in the prevalence of risk factors for violence and of inequitable gender attitudes. Child marriage significantly declined among females between 2010 and 2019, after controlling for covariates. Significantly fewer females endorsed one or more attitudes justifying wife beating in 2019 compared with 2010 (61·8% in 2019 vs 75·1% in 2010; p<0·0001). Significantly fewer males in 2019 who reported having had sex had never been tested for HIV (24·2% in 2019 vs 43·2% in 2010; p<0·0001), endorsed one or more inequitable gender attitudes (47·8% in 2019 vs 59·2% in 2010; p<0·0001), and one or more attitudes justifying wife beating (68·7% in 2019 vs 81·1% in 2010; p<0·0001). These declines were statistically significant after controlling for covariates. However, there were no other significant changes in risk factors or inequitable gender attitudes in males or females. Sensitivity analyses indicated no changes in study findings or conclusions with or without peer violence being included in the 2019 physical violence definition or with or without gender attitude variables being controlled for in the logistic regression analyses (appendix pp 4–5).

Table 5:

Violence risk factors and inequitable norms about gender and intimate partner violence against women among 13–24-year-olds, Kenya 2010 and 2019 VACS

Prevalence
 Adjusted models*
2010 VACS
 2019 VACS
 p value β coefficient†
 p value
n Weighted % (95% CI) n Weighted % (95% CI) 2010 2019
Females
Risk factors
 Multiple sexual partners in the past 12 months‡ 426 4·5 (2·2–6·8) 445 6·0 (3·5–8·5) 0·37 0 −0·144 0·68
 Infrequent condom use in the past 12 month‡ 423 86·2 (82·2–90·2) 444 75·6 (69·9–81·2) 0·0020 0 −0·016 0·96
 Transactional sex in the past 12 months‡§ 448 4·5 (1·9–7·2) 446 7·6 (4·6–10·6) 0·14 0 0·424 0·32
 Never tested for HIV 584 15·0 (10·9–19·0) 512 12·1 (8·6–15·5) 0·28 0 −0·513 0·070
 Had symptoms or diagnosis of sexually transmitted infection in the past 12 months‖** 1220 8·5 (5·5–11·5) 1342 6·3 (4·7–8·0) 0·19 0 −0·264 0·29
 Child marriage‖ 607 12·8 (8·9–16·7) 669 8·7 (6·4–11·1) 0·063 0 −0·604 0·021
 Inequitable gender norms
  Endorsed at least one norm‖ 1206 51·7 (46·7–56·7) 1312 49·8 (45·8–53·8) 0·57 0 −0·041 0·76
 Norms justifying intimate partner violence against women
  Endorsed at least one norm‖ 1188 75·1 (70·7–79·5) 1310 61·8 (58·7–64·9) <0·0001 0 −0·581 <0·0001
Males
Risk factors
 Multiple sexual partners in the past 12 month‡ 355 37·7 (30·9–44·5) 244 30·6 (25·3–35·8) 0·093 0 −0·382 0·078
 Infrequent condom use in the past 12 month‡ 351 55·0 (47·4–62·6) 243 50·5 (43·6–57·3) 0·39 0 −0·118 0·54
 Transactional sex in the past 12 months‡§ 381 3·2 (1·3–5·2) 245 7·2 (2·0–12·4) 0·087 0 0·857 0·088
 Never tested for HIV 640 43·2 (37·2–49·3) 310 24·2 (19·5–28·9) <0·0001 0 −1·018 <0·0001
 Had symptoms or diagnosis of sexually transmitted infection in the past 12 months‖** 1445 5·3 (3·6–7·0) 786 3·0 C2·2–3·9) 0·011 0 0·491 0·056
 Child marriage‖ 664 2·0 (0·4–4·6) 403 0·6 (0·0–1·6) 0·18 0 −1·674 0·061
 Inequitable gender norms
  Endorsed at least one norm‖ 1434 59·2 (54·9–63·5) 775 47·8 (43·5–52·0) <0·0001 0 0·416 0·0003
 Norms justifying intimate partner violence against women
  Endorsed at least one norm‖ 1420 81·1 (78·1–84·0) 769 68·7 (65·1–72·4) <0·0001 0 −0·655 <0·0001
Open in a new tab

VACS=Violence Against Children and Youth Surveys.

*

The adjusted model for females controlled for age, educational attainment, marital status, orphan status, rural versus urban status, and previous pregnancy using logistic regression; the adjusted model for males controlled for age, educational attainment, marital status, orphan status, and rural versus urban status using logistic regression; the logistic regression was used to assess if changes in the prevalence were significant after controlling for the important covariates; for child marriage, the adjusted model for both females and males excluded marital status because of the high correlation between marital status and early child marriage.

†

β coefficients were generated using logistic regression, accounting for the complex survey design of the Kenya VACS.

‡

n includes only those who had sex in the previous 12 months.

§

Transactional sex includes exchanging sex for money, food, gifts, or other favours.

n includes only those who have ever had sex.

‖

n includes every survey participant.

**

Includes disease acquired through sexual contact, or bad-smelling abnormal genital discharge, or a genital sore or ulcer.

Discussion

Between 2010 and 2019, we observed significant declines in lifetime sexual, physical, and emotional violence experienced by male and female young people aged 13–24 years in Kenya. These declines over a 9-year period are important public health successes, and do not appear driven only by demographic changes. Most declines remained significant after controlling for demographic factors, and demographic distributions in the study populations were similar except for the proportion of females who were married and ever pregnant, and the percentage of males who were married and in urban versus rural settings. These differences might reflect demographic trends consistent with economic development (ie, declines in marriage and increases in urbanicity) and could affect service uptake, which is more accessible in urban settings.22

The declines in lifetime violence prevalence suggest that programming and policy strategies implemented in Kenya since the 2010 VACS might have contributed to the decreases. Although service seeking and receipt for sexual violence significantly improved among females, most young people who experience sexual violence do not seek or receive services. Disclosure and knowing a place to seek services did not improve for females. For males, knowledge of where to seek services improved, but service uptake did not. Improving service knowledge, access, use, and quality among violence survivors might be important given compelling evidence of the negative effect that violence can have on health across the lifespan.23

Among females, risk factors for violence were not significantly different between 2010 and 2019 except for child marriage, a documented risk factor for wife beating.24 It is important to note that other drivers related to human sexuality exist and that the fraction of sexual and HIV testing behaviour as measured by VACS attributable to sexual, physical, and emotional violence is unclear and cannot be fully unpacked from the current analysis of cross-sectional data. However, this finding underscores the need for continued or new programmes to address HIV risk among women and girls, who remain a priority population given their disproportionate risk for HIV.25 Fewer females and males endorsed attitudes justifying wife beating and fewer males endorsed inequitable gender attitudes in 2019 compared with 2010. These attitudes have been associated with intimate partner violence and sexual risk behaviours.26 The reduced endorsement of these attitudes suggests a shift towards lower acceptance of violence against women among young people in Kenya. Further monitoring is needed to determine if these trends continue to hold or improve.

For males, other than declines in unwanted sexual touching and increases in knowing where to seek services, no other forms of sexual violence or disclosure and service seeking changed significantly between 2010 and 2019. A relatively low prevalence of lifetime sexual violence was observed at both timepoints for males. Nonetheless, prevention is crucial given the link between childhood violence and poor outcomes, including risk of perpetration.2–4 Some factors associated with HIV acquisition improved among males, consistent with observed reductions in HIV incidence and prevalence in Kenya.27 The 2018 Kenya Population-based HIV Impact Assessment27 reported HIV prevalence was 4·9% among people aged 15–64 years in 2018 compared with 7·1% in 2007. Additionally, HIV testing among men improved with increased awareness of HIV status.

Although this study could not identify the reasons for declines in the prevalence of violence and risk factors for violence and HIV, programmes and policies implemented after the 2010 VACS might have contributed. The Government of Kenya developed and implemented a prevention and response plan for violence against children after the 2010 VACS. Substantial investments in this plan were made by the Government of Kenya and its partners. For example, the PEPFAR DREAMS initiative delivers a package of evidence-based programmes to prevent HIV among adolescent girls and young women, including violence prevention, and has reached more than 400 000 adolescent girls and young women since 2016.25 Kenya’s response plan also focused on equipping community and national partners to collaborate to prevent violence through multisector programmes and initiatives. All PEPFAR implementing partners have a mandate to identify and offer services to survivors of gender-based violence.

Legislative changes in Kenya might also have contributed to observed violence reductions. A new Constitution in 2010 included an article protecting children from abuse, neglect, harmful cultural practices, all forms of violence, inhumane treatment and punishment, and hazardous and exploitative labour.13 The 2014 Marriage Act set the minimum age of marriage at 18 years.28 The Children’s Act increased protection for children, and established child-focused protection units in select police stations.12 A toll-free Child Helpline was expanded to facilitate abuse reporting and to provide free services to survivors of childhood violence, including medical referrals.29 These services can be beneficial and improve outcomes but could benefit from monitoring and evaluation. Additional work is needed to make them user friendly, timely, and effective.30,31

These findings are subject to several limitations. First, this study is an analysis of two serial cross-sectional surveys and could not assess the reasons for observed declines; further monitoring is needed to identify the factors that are driving the observed changes and whether these changes hold or improve over time. Second, this study could not determine the magnitude or nature of the contributions of programmatic and policy changes to the observed declines. Third, question wording for the two surveys was not identical for some measures. For example, the 2010 emotional violence questions asked about any adult perpetrator, which could have included other adults such as teachers and strangers, whereas the 2019 emotional violence questions asked about parents, adult caregivers, or other adult relatives. However, all indicators were similarly structured for both surveys and could be meaningfully compared. Fourth, declines in response rates between the surveys could have implications for prevalence observed in 2019. Nonetheless, further analyses indicated no significant difference in the demographic distribution of the 2019 survey participants and the population. The main drivers for lower response rates in 2019 were no one being home or the selected respondent not being available. Finally, other societal shifts not directly reflected in the data, such as increased urbanisation, economics, and increased awareness about violence might have contributed to the observed declines.

In conclusion, this study documented significant declines in the prevalence of violence among children and young people in Kenya over a 9-year period. Although the study could not identify reasons for the decline, Kenya has made substantial investments in addressing violence against children since the 2010 VACS. Many of the interventions recommended by the INSPIRE32 technical package are being implemented in Kenya through PEPFAR support. Kenya has adopted a multisectoral approach involving relevant government departments and institutions, civil society organisations, and bilateral partners. Kenya’s dedication and investment in preventing violence against children and young people are worthy of recognition. Other countries in the region that have made substantial investments and implemented similar violence prevention programmes, such as the DREAMS and INSPIRE interventions, could use repeat VACS data to monitor violence and related outcomes over time.

Supplementary Material

Supplementary Appendix

Research in context.

Evidence before this study

We searched PubMed on May 22, 2020, for articles from low-income and middle-income countries (LMICs) that estimated the prevalence of violence or assessed changes in violence against children or women, without language restrictions. We used the search terms “violence against children”, “violence against women”, AND “low-middle-income country”, “Kenya”. Previous studies have noted a high prevalence of violence against young people in Kenya. Exposure to childhood violence is a known risk factor for HIV. However, these previous studies were limited by implementation in small geographical locations, or the use of narrow definitions for violence. Additionally, little information is available about temporal changes in violence among children and young people in LMICs.

Added value of this study

Our study noted a significant decline in all forms of violence in Kenya between 2010 and 2019. To our knowledge, this is the first study using data from a nationally representative survey to document a decline in violence in a sub-Saharan African country. In addition to declines in violence, the study also showed declines in some risk factors for violence and HIV. Our study presents important results on service uptake among those who experienced violence, finding that seeking and receiving services for sexual violence significantly improved among adolescent girls and young women, whereas among adolescent boys and young men, only knowledge of where to seek services significantly improved and not actual service seeking behaviour.

Implications of all the available evidence

Violence against children and young people significantly declined, as did some risk factors for violence and HIV in Kenya between 2010 and 2019. The Government of Kenya, in collaboration with partners, has made substantial investments in primary prevention and response to violence. However, service uptake remains low. Therefore, prioritising service awareness and use among survivors of violence will improve the wellbeing of children and young people in Kenya. Our results underscore the importance of continued focus on policies and programmes that facilitate the decline of violence against children and young people in Kenya.

Acknowledgments

The 2010 study was funded by UNICEF Kenya with support from the US President’s Emergency Plan for AIDS Relief (PEPFAR). The 2019 study was supported by PEPFAR through the US Centers for Disease Control and Prevention (CDC) under the terms of cooperative agreements 5NU2GGH001520 and 1U2GGH001619. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the CDC.

Footnotes

Declaration of interests

We declare no competing interests.

See Online for appendix

Contributor Information

Francis B Annor, Division of Violence Prevention, National Center for Injury Prevention and Control, US Centers for Disease Control and Prevention, Atlanta, GA, USA.

Laura F Chiang, Division of Violence Prevention, National Center for Injury Prevention and Control, US Centers for Disease Control and Prevention, Atlanta, GA, USA.

Patricia R Oluoch, Division of Global HIV and TB, Center for Global Health, Centers for Disease Control and Prevention, Nairobi, Kenya.

Vivienne Mang’oli, Department of Children Services, Nairobi, Kenya.

Marygorret Mogaka, Department of Children Services, Nairobi, Kenya.

Mary Mwangi, Division of Global HIV and TB, Center for Global Health, Centers for Disease Control and Prevention, Nairobi, Kenya.

Anne Ngunjiri, LVCT Health, Nairobi, Kenya.

Francis Obare, Population Council, Nairobi, Kenya.

Thomas Achia, Division of Global HIV and TB, Center for Global Health, Centers for Disease Control and Prevention, Nairobi, Kenya.

Pragna Patel, Division of Global HIV and TB, Center for Global Health, US Centers for Disease Control and Prevention, Atlanta, GA, USA.

Greta M Massetti, Division of Violence Prevention, National Center for Injury Prevention and Control, US Centers for Disease Control and Prevention, Atlanta, GA, USA.

Linda L Dahlberg, Division of Violence Prevention, National Center for Injury Prevention and Control, US Centers for Disease Control and Prevention, Atlanta, GA, USA.

Thomas R Simon, Division of Violence Prevention, National Center for Injury Prevention and Control, US Centers for Disease Control and Prevention, Atlanta, GA, USA.

James A Mercy, Division of Violence Prevention, National Center for Injury Prevention and Control, US Centers for Disease Control and Prevention, Atlanta, GA, USA.

Data sharing

The dataset used along with the questionnaire and data dictionaries are available to the public at the Together for Girls website at https://www.togetherforgirls.org/. Researchers can submit a request for both the 2010 and 2019 Kenya Violence Against Children and Youth Survey datasets to Together for Girls at https://www.togetherforgirls.org/request-access-vacs/

References

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Appendix
NIHMS1854030-supplement-Supplementary_Appendix.docx (52.9KB, docx)

Data Availability Statement

The dataset used along with the questionnaire and data dictionaries are available to the public at the Together for Girls website at https://www.togetherforgirls.org/. Researchers can submit a request for both the 2010 and 2019 Kenya Violence Against Children and Youth Survey datasets to Together for Girls at https://www.togetherforgirls.org/request-access-vacs/

RESOURCES