Teenage pregnancy and experience of physical violence among women aged 15-19 years in five African countries: Analysis of complex survey data

Background Pregnant teenage women are prime targets of violence against women perpetrated by intimate partners, family members, and miscreants in their neighborhoods. This study estimated the prevalence of Teenage pregnancy (TP) and Physical Violence (PV) and further assessed the relationship between TP and PV in five Low-and-Middle-Income Countries (LMICs). Methods The study was conducted among five LIMCs (Burkina Faso, Kenya, Malawi, Nigeria, and Tanzania) using data from the most recent Demographic and Health Surveys conducted in these countries. Modified Poisson with the robust standard error was used to quantify the association between TP and PV. All analyses adjusted for the complex survey design structure (clustering, weighting, and stratification). Results The analysis involved a total of 26055 adolescent women aged 15–19 years across the five countries. The overall prevalence of TP was 25.4% (95%CI = 24.4–26.4) with the highest prevalence occurring among Malawians [29.0% (95%CI = 27.4–30.7)]. Meanwhile, the prevalence of TP among older adolescents (18–19 years) was approximately two-thirds significantly higher compared with young adolescents [aPR(95%CI) = 1.60[1.49–1.71)]. The prevalence of PV among teenagers across the five countries was 24.2% (95%CI = 22.3–26.2). The highest prevalence of PV was recorded among Nigerian adolescent women [31.8% (95%CI = 28.5–35.3)]. The prevalence of PV among adolescent women who were pregnant was approximately 5-folds significant compared to those who were not pregnant (adjusted prevalence ratio; aPR = 4.70; 95% CI: 3.86–5.73; p<0.0001). Conclusion There was a high prevalence of pregnancy among older teenagers aged 18–19 years. Close to a quarter of teenage women ever experienced physical violence. Pregnant teenage women ever experience of physical violence was very high compared to non-pregnant peers. Intervention should target PV and TP by adopting a gender-sensitive approach to eliminate physical violence, particularly among teenagers to prevent TP.

Introduction Every year, an estimated 21 million teenage girls aged 15-19 years become pregnant [1], of which 16 million give birth. Birth among older teenagers aged 15-19 years accounts for 11% of worldwide births, 95% of which occur in low and middle-income countries (LMIC) [2,3]. Sub-Saharan Africa has the highest teenage pregnancy and delivery rates in the world and the rates are still increasing [4]. In some parts of Africa, more than 10% of adolescent girls become mothers before the age of 16 years [5]. Teenage birth rates in Africa are highest with 115 births per 1000 women, with urban-rural disparities where there are up to three times more teenage pregnancies in rural populations than urban populations [6].
Increased rate of Teenage pregnancy (TP) rates in Africa have been influenced over the decades by reasons such as progressively lower age of menarche, initiation of first sexual activity at a progressively lower age, and low contraceptive use rates among teenagers [7]. TP is viewed as a complex socioeconomic and medical problem in itself, with immediate and longterm adverse effects. Each year, an estimated 3.9 million unsafe abortions are carried out among girls aged 15-19 years, contributing to maternal mortality and other immediate and long term health problems [1]. Among those who opt to give birth, the complications are equally worse.
Many studies from almost every part of the globe consistently and convincingly report adverse obstetric, perinatal, medical, and socioeconomic outcomes among TP women and their children. Pregnant teenage women are at an increased risk of adverse obstetric complications such as anaemia in pregnancy [8], pregnancy-induced hypertension, and associated hypertensive disorders of pregnancy [9][10][11]. Others include preterm labour, cephalopelvic disproportion, higher incidence of operative delivery [4,12,13], post-partum hemorrhage, and associated blood transfusion. Major perineal lacerations and need for episiotomy [14], obstetric fistula [15], and postnatal depression [10] also occur more frequently in teenage pregnancies. It is often cited that teenagers who give birth aged [15][16][17][18][19] years are more than twice as likely to die as those aged 20 years and above [5].
Other health and socio-economic complications include a higher incidence of HIV/AIDS among unmarried pregnant teenagers, termination of academic pursuits [4,30], teenage mothers struggle due to lack of preparedness for childbearing, negative public attitude directed towards adolescent pregnant women [31], mental health problems, and socioeconomic disadvantage [32]. Other social problems include single parenthood, with long term effects of increased risk of child abuse, child neglect, maternal suicide, and repeated suicidal attempts [19]. Additionally, offspring of teenage pregnancies are more likely to end up with teenage pregnancies, perpetuating the cycle of teenage pregnancy. The adverse effects of teenage pregnancies are often potentiated due to certain behavioural tendencies that make teenage pregnant women hide the pregnancy and they are more likely to remain un-booked for antenatal services even up to the time of delivery [33,34].
In developing countries, it is estimated that as high as 90% of teenage pregnancy occurs within marriage bringing to bear the contribution of early marriages [35]. The prevalence of intimate partner violence (IPV) during pregnancy has been reported as 40% from a review of facility-based studies in Africa [1,36]. The prevalence of IPV among pregnant teenage women aged 15-19 years in Africa is not known but believed to be relatively higher than among older age groups. Elsewhere outside Africa, studies have found out that, women aged less than 20 years had 4.3 times of experiencing violence during pregnancy compared to women aged 30 years and above [37]. Physical violence among pregnant women has been associated with miscarriage, stillbirth, premature labour and birth, foetal injury, and low birth weight [36,38]. Pregnant teenagers were significantly more likely to have experienced forced sexual initiation, experienced physical violence in the form of beating more often [39] and other physical, sexual, or domestic violence in attempts to meet the demands of pregnancy and childbearing. This abuse in pregnancy is significantly associated with adverse obstetric outcomes among teenage pregnant women [40]. The prevalence of nationwide estimates of TP in Africa and rates of experienced physical abuse among pregnant teenagers aged 15-19 years and its predictive factors are significantly lacking in extant literature.
Even when available, previous studies in the area of teenage pregnancy suffered from inherent design limitations such as small sample size, inadequate confounder information, unresolved bias, and studies limited to healthcare facilities. Even the population studies fail to cover wide geographical areas making findings from such studies to have limited generalisability and power to effect change. The main objective of this population-based study was to determine the association between TP and PV among adolescent women aged 15-19 years living in five low-and middle-income countries in Africa. The study uses recent demographic and health survey data from five African countries.

Study description
The study was conducted among five low-and-middle-Income countries namely Burkina Faso, Kenya, Malawi, Nigeria, and Tanzania. Demographic and Health Survey (DHS) data for these countries were pooled for analysis and results presented to inform targeted intervention. DHS data was obtained for Burkina Faso (2010), Kenya (2014), Malawi, and Tanzania (2015/ 2016), and Nigeria (2018). These countries recent DHS module assessed domestic physical violence which other counties did not. Our main objective was to assess the association between ever experience physical violence and teenage pregnancy. This is deemed appropriate for including these counties.

Study design
The DHS surveys collect information on a wide range of variables for a sample of countries that participate in the survey. Globally, DHS has earned a worldwide reputation for collecting nationally representative data on different aspects of health indicators including fertility issues among men and women in childbearing years, family planning, maternal and child health, gender, HIV/AIDS, malaria, and nutrition, domestic violence since 1984 [41]. The Monitoring and Evaluation to Assess and Use Results Demographic and Health Surveys (MEASURE DHS) funded by the U.S. Agency for International Development (USAID) is responsible for providing technical assistance to all DHS surveys in over 90 countries to enhance global understanding of the health and population in developing countries [42]. DHS collects data using a cross-sectional design and employs a two-stage sampling of all geographical regions in a country based on enumeration areas (EAs). Details about the study design and procedures for data collection have been published elsewhere [43].

Study participants
DHS collects information on all aspects of health including household (HH), child, men, and women individual records. However, the present study focused on the women survey (women in their reproductive age, 15-49 years) and their respective HH records. To answer the study objectives, the present study focuses on older teenagers aged 15-19 years. The data for the individual women were merged with their corresponding households to obtain complete data. The sample for all five countries is 26,055. The sample was stratified across the five countries as follows: Burkina Faso (3349), Kenya (6,078), Malawi (5,273), Nigeria (8423), and Tanzania (2,932). Missing responses in some variables are clearly defined in S1 Table.

Outcome measures
In this study, two outcome variables were considered; namely, TP and PV. We assessed the association between TP and ever experienced domestic PV. TP has been identified as a public health issue that can cause severe health problems for a mother and her born or unborn child [1]. DHS defined TP as a percentage of reproductive teenage girls who are mothers, pregnant with their first child, and have begun childbearing [44]. In the present study, TP was estimated based on: 1. Women aged 15-19 years and have had a live birth.
2. Women aged 15-19 years who were pregnant with their first child.
After generating the index variable for TP, it was categorized as a dummy variable (Yes = 1 or No = 0). A detailed process and variable definition can be found in S1 Table. Ever experience of PV was as assessed by DHS; where women aged 15-49 years were asked whether they have ever experienced physical violence during the past 12 months (often or sometimes) preceding the survey, however, our research only considered those aged 15-19 years. DHS collected information on women who have experienced any form of physical violence scaled on 12 items. The items were about; any husband/partner ever if ever-married; anyone other than any husband/partner since they were age 15 years; and anyone during any pregnancy, if ever pregnant. The 12-item scale physical violence module was naturally coded as; never, sometimes, or often. Based on the objective of our study, an artificial dummy variable was generated for all the 12 items by recoding into; as 0 "Never experienced" or 1 "Ever experienced (sometime and often)". An index variable from the raw scores (ranging from 0-12) was generated and further reclassified into a modified binary primary outcome variable as 0 "Never experienced" or 1 "Ever experienced (any of the 12-item)". For reliability and internal consistency, the Jann Stata module was used to compute Cronbach's alpha for weighted data due to DHS design [45]. The overall test of reliability for physical violence was very high and of good quality (α = 0.76). The items all tapped into the same concept of measurement (see S2 Table).

Covariate
Household (HH) and community characteristic. Age of HH head (<30, 30-39, 40-49, 50-59, 60-59 and 70+), sex of HH head (Male or Female), HH has Telephone (No or Yes), HH wealth index (Poorest, poorer, middle, richer and richest), the number of HH members (<4, 4-5, 6-8 and 9+), HH has electricity (No or Yes), HH has a radio(No or Yes), HH has color television (No or Yes), where HH food is prepared (In the household, separate household and others), HH has a mobile phone (No or Yes), HH has a watch (No or Yes), HH own land for agriculture (No or Yes), HH number of animals own (None or 1+animal), relationship structure (No+1 adult; two adults, opposite sex; two adults, same sex; three+ related adults and unrelated adults) and pace of residence (Urban or Rural). Detailed variable definition, type of variable, measurement, and scale of measurement used in this study has been clearly defined in S1 Table. Individual characteristics. Age of participant (15-17 and 19-19), educational level (None, primary, secondary or higher), marital status (Never married, married, divorced/widowed), knowledge on pregnancy (No knowledge and have knowledge), knows modern contraceptive (Knows no method and knows method), family planning awareness (aware and aware), currently abstaining (No or Yes) and currently working (No or Yes).
Based on Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendation for cross-sectional study design, implementation, and reporting, missing responses were strictly excluded in our analysis [46]. Detailed variable definition, type of variable, measurement, and scale of measurement used in this study has been clearly defined in S1 Table.

Data analysis
This study adjusted for the design of DHS since it is a complex survey. In preventing wrong estimations, adjusting for the participants' weighting, stratification, and clustering in a complex survey dataset is a key issue to consider during data analysis. DHS adopted a multistage cluster survey design and due to the complex design of DHS, this study adjusted for the primary sampling units, stratification, and the sampling weights to reduce bias and to improve data analysis in all estimates. Since the data used were pooled from five countries at different time points, the individual women and domestic violence weights were normalized due to different population sizes. The normalized sampling weight was calculated by multiplying the raw weight by the overall Sampling fraction: Where Economic and Social Affairs, Population Dynamics [47]. The population information from DHS and the UN can be found in S3 Table. Three approaches to data analysis were carried out. First, bivariate descriptive statistics were generated with Rao-Scott chi-square for testing the independence of covariate variables among the various countries and how different they are. Before performing inferential analysis, we analyzed to assess multicollinearity among the variables considered by adopting the variance inflation factor (VIF) and pairwise correlation. The analysis showed no suspected multicollinearity (overall mean VIF< 1.34) and also, none of the individual variables' pairwise correlation coefficient is �0.8.
Secondly, factors influencing TP were determined by using the modified Poison regression model. Normalized domestic violence weight was considered for the quantification of association between TP and ever experience domestic violence adjusting for significant factors influencing TP. Stata 16 was used to perform all analyses and p-value 0.05 was deemed significant.
The hypothetical idea for the study was to assess the association between TP and PV and the analytical procedure can be found in S1 Fig.

Ethical requirements
All DHS surveys have been reviewed and approved by the ICF Institutional Review Board (IRB). Ethical procedures in the overall process of the survey, including coordination of activities, were strictly followed. Data were collected after taking informed consent, and all information was kept confidential. Country-specific DHS survey protocols are reviewed by the ICF IRB and typically by an IRB in the host country. ICF IRB ensures the right of human subjects surveys which comply with the U.S. Department of Health and Human Services regulations for the protection of human subjects (45 CFR 46), while the host country IRB ensures that the survey complies with laws and norms of the nation [48]. The legitimacy to use DHS data was obtained from MEASURE DHS. The data underlying the results presented in the study are available upon request. The data is not in the public domain and acquisition of the data can be obtained from DHS https://dhsprogram.com/data/dataset_admin/login_ main.cfm.

Results
The analysis involved a total of 26055 teenage women aged 15-19 years across the five countries. The overall mean age was 16.9 years in all countries; however, the mean age is statistically different among the countries (F-test = 21.6, p-value<0.0001). The Rao-Scott test showed that there was a statistically significant association between all covariates and country understudied (p<0.05) except the HH wealth index (see Table 1).

Prevalence of teenage pregnancy among adolescent women aged 15-19 years
The overall prevalence of TP among women aged 15 S4 Table).

Individual factors influencing teenage pregnancy among adolescent women aged 15-19 years
It was quite surprising to identify from Poisson estimates that all individual demographic characteristics studied influence TP in the five countries except working status. Increasing age was estimated to have an increased PR of experiencing TP.  Table 3).

Experience of physical violence among adolescent pregnant women aged 15-19 years
TP correlation with PV was assessed to establish the relationship. The results show that there exists a positive significant correlation between TP and experience of PV (both raw scores and

PLOS ONE
Teenage pregnancy and experience of physical violence among women aged 15-19 years in five African countries  Table). Unexpectedly, segregating PV by TP status, the analysis showed that, almost all adolescents with TP have ever experienced any form PV in all the countries (p-value<0.001) (see S6 Table). In all countries, TP was statistically associated with PV and that, PV among adolescent women who were pregnant was approximately 5-folds significant compared with those who were not pregnant [aPR(95%CI  Table 4).

Discussion
This study uses the recent DHS data sets of five LMICs in Africa to obtain the prevalence of teenage pregnancy among women aged 15-19 years and the factors that are associated with it. Overall, one in four teenage women aged 15-19 years were pregnant, of which rural residence was identified to significantly influence the likelihood of an adolescent becoming pregnant in all countries studied. Overall, the log-likelihood of an adolescent becoming pregnant was 56% higher among rural teenagers compared with their urban peers. Elsewhere outside of Africa, one in five teenage women aged 15-19 years have been pregnant, with more than 64% of such pregnancies unplanned [49]. Sexual abuse, exploitation were significant associated factors identified elsewhere. However, in Africa, a recent systematic review identified low self-esteem, curiosity, cell phone usage, and substance abuse as individual determinants of adolescent pregnancy [50]. A host of socio-cultural, environmental, economic as well as health-related factors contribute to the problem of adolescent pregnancy [50]. Interestingly, the number of HH

PLOS ONE
Teenage pregnancy and experience of physical violence among women aged 15-19 years in five African countries members was found to have a decreasing TP log-likelihood ratio as the number of HH increases. TP has been established to be more common in the context of significant changes in family life including family size in these recent times [51]. The probable reason for this finding could that, lager family size, of course, older adults could serve as caretakers to the younger ones.
Studies have shown that up to 90% of teenage pregnancy occurs among married teenagers where intimate partner violence (IPV) is common [1,36]. We found out that, teenage women aged 15-19 years in Nigeria, who were either married or widowed, were more than 20 times more likely to experience violence, compared to teenage women who were never married. The probable reason could be that, in Africa, premarital pregnancy is considered a shame-based culture [52] and in other to avoid shame, these pregnant women get married when they heard about the pregnancy. This suggests that tackling the problem of teenage pregnancy and intimate partner violence among teenage women requires a closer look at early marriages. The

PLOS ONE
Teenage pregnancy and experience of physical violence among women aged 15-19 years in five African countries problem of early marriages is very common in Africa. However, they remain an underresearched human right violation, that seeks to silence and constrain the spaces within which women can speak [53,54]. Most women who marry early are unemployed children who are susceptible to violence and abuse, often truncating the developmental potential of such children [55]. Early marriages are deep-rooted in the social, cultural, and religious fabrics of communities that make eradication very difficult. However, universal female education and female child empowerment may produce sustainable results. Abstaining from sex was observed to have a high PR of pregnancy among teenagers aged 15-19 years compared with non-abstaining teenagers in all countries. For example, in Kenya, abstaining teenage women were more than 3.5 times more likely to have experienced teenage pregnancy compared with girls who are not abstaining from sex. This result, at a cursory look, seems impossible because trials have demonstrated the effectiveness of abstinence-centered health education intervention among adolescent girls [56][57][58]. However, the theoretically, fully protective abstinence intentions often fail among African adolescent girls [59,60] in the light of developmental pressure and other factors confronting teenagers in the communities in which they live such as cultural silence surrounding sexual matters, poor parental supervision or neglect within large and 'broken' families, deficient parent-child communication, inadequate neighborhood safety, poverty, inattention to (or lack of access to alternative support for) teenagers who may be sexually abused by family members and limited access to adolescentfriendly services among others. This results in teenagers claiming to be abstaining, whereas in practice they may occasionally fail to abstain. This makes such theoretical 'abstainers' unprotected as they are less likely to adopt a contraceptive method. Approximately, one in four adolescents aged 15-19 years experienced any form of PV. Interestingly, almost all adolescent pregnant women in all countries have ever experienced any form of PV. Intimate partner violence (IPV) among all adolescents from a review of facilitybased studies in Africa reported prevalence as high as 40% [1,36]. The difference in the high value of 40% by [36] compared to findings in the present study of 24.2% may be due to the difference in study settings as the former was a facility-based study, where victims of violence were more likely to present at a health facility with complaints, thereby overestimating the prevalence in the community. Physical violence among pregnant women is not without consequence for mother and child.
TP among women who have ever experience PV was over four times compared to those who have never experience PV in all countries. It was imperative to establish that, among all countries, physical violence is highly associated with TP. This makes PV an influencing factor for teenage pregnancy among women aged 15-19 years.

Limitation
This study has some significant limitations that must not be overlooked. Firstly, the vast variation in years of the most current DHS data sets available for use of concern. The data set used spanned a period of 8 years with Burkina Faso's current DHS conducted in 2010, while that of Nigeria was conducted in 2018. Kenya conducted its latest DHS in 2014 while both Malawi's and Tanzania's current DHS were conducted in 2015/2016. Secondly, datasets from only five African LMICs were used which would make generalizations of the findings to the whole of Africa problematic. Attempts were made at correcting this. However, we were faced with challenges as some countries did not have datasets that included violence which was the primary outcome of this analysis. Also, the presence of physical violence as considered in this study within the last 12 months could probably mean some of the adolescents were not yet pregnant. Although the results were derived from a robust analysis of nationally representative datasets which is a strength in itself, the authors advise in the light of the aforementioned that the results be contextualized appropriately when making inferences. Although all authors are from Ghana, DHS data from Ghana was not included because experience of physical violence was not captured.

Conclusion
There is a high prevalence of pregnancy and PV among adolescent women aged 15-19 years. About a quarter of teenage women ever experienced physical violence. Counter-intuitively, pregnant teenage women were at a higher disadvantage to ever experience physical violence. We recommend gender-sensitive approaches in addressing the issue of Sustainable Development Goal (SGD) 5.2 (Eliminate all forms of violence against all women and girls in the public and private spheres, including trafficking and sexual and other types of exploitation) among teenage women to have better lives.