Survival Analysis and Prognostic Factors of Time to First Domestic Violence after Marriage among Married Women in Africa

Background Domestic violence remains a public health challenge worldwide. Studies in the sub-Saharan Africa have shown high level of violence against women, especially from intimate partners. What is not known is how soon domestic violence happens after marriage among ever-married women. This study assessed the timing of first domestic violence among ever-married women and identified its determinants in Nigeria, Kenya, and Mozambique. Data and Method Data of 56440 ever-married women of reproductive age pooled from DHS conducted in Nigeria (2013), Kenya (2014) and Mozambique (2011) was used. The timing of first domestic violence was estimated as the period between marriage and the first experience of domestic violence. Ever-married women without violence experience were censored on the date of the survey. Survival analysis techniques were used to assess the timing and the factors influencing the timing p=0.05. Result The prevalence of domestic violence in Nigeria, Kenya and Mozambique was 12.1%, 7.5%, and 15.4% respectively. The median time to first domestic violence across the three countries was 3 years. We found a higher prevalence of violence among formerly married women than currently married women. Women who married before age 20 years have a significantly higher risk of experiencing violence (Nigeria: aHR= 2.36 (95% CI (1.97- 2.83), Kenya: aHR= 1.89 (95% CI (1.31- 2.73) and Mozambique: aHR=1.60 (95% CI (1.24 −2.12)) than older women. Women in poorer wealth quintile were at higher risk of violence across the three countries. Other factors associated with the timing of first domestic violence are respondents’, location of residence and educational attainment. Conclusion Domestic violence has remained high in Mozambique, Nigeria and Kenya. There was a significant relationship between timing of first domestic violence and whether or not a woman remains in a marriage. Education and delayed marriage could help reduce violence in marriage.

areas. At the first stage, local governments were selected, then EAs were selected at the second 122 stage while the households were selected at the third stage. A representative sample of 40 680 123 households was selected for the survey among which 38948 were successfully interviewed. 124 Detailed sampling details have been documented (30).  The sample for the 2011 Mozambique DHS was selected from the 11 provinces in the country. 136 Each province was divided in districts, and the districts further subdivided into administrative posts EAs and the households were subsequently drawn at the 3 rd and 4 th stages.

Sampling of respondents on Violence related questions
Due to the sensitivity of violence and to maintain confidentiality, only one woman per household 145 was administered the questions on violence while. In the one-third of the households selected for 146 the male survey, one man per household was randomly selected to respond to questions on 147 domestic violence. One woman per household was randomly selected in the remaining two-thirds 148 of households and were administered the questions on violence. The selected respondents were 149 informed that questions could be sensitive and were reassured regarding the confidentiality of their To determine whether an ever-married woman had experienced domestic violence prior to the 153 survey dates, the following questions were asked. Were you: ever been pushed, shook or had 154 something thrown by husband/partner, ever been slapped by husband/partner, ever had arm twisted 155 or hair pulled by husband/partner; ever been punched with fist or hit by something harmful by 156 husband/partner, ever been kicked or dragged by husband/partner; ever been strangled or burnt by 157 husband/partner ever been threatened with knife/gun or other weapon by husband/partner; ever 158 been physically forced into unwanted sex by husband/partner; ever been forced into other 159 unwanted, sexual acts by husband/partner; ever been physically forced to perform sexual acts 160 respondent didn't want to. 161 Any ever-married who answered in affirmative to at least one of the question is considered to have 162 experienced domestic violence and were then asked how long was the first time they experienced 163 any form of these violence after they got married. The length of time between marriage date and 164 the date of first domestic violence was used as the dependent variable in this study while age, education, employment, residence and wealth indicators were used as independent variable. They 166 have been identified in prior studies as contributing factors to domestic violence among 167 women(4,10,12,19).

168
Inclusion and exclusion criteria 169 The current study focused on experience of domestic violence among ever-married women. We 170 therefore excluded never married women either they were living with sexual partner or not. Only 171 women aged 15-49 years and who were currently married, divorced, separated or widowed were 172 included in the analysis. The effective sampling size for this study was 21,564, 4,237, and 3992 173 for Nigeria, Kenya and Mozambique respectively totaling 29,793 respondents. Of the 6440 that 174 claimed to have experienced violence before, 343 did not remember their time of first experience 175 after marriage, they were therefore excluded in further analysis except in Table 1.

177
Descriptive statistics and survival analysis techniques were used to analyze the data at a 5% 178 significance level. Cox proportional hazard regression model was used to identify the risk factors 179 associated with the timings. All data were weighed so as to ensure adequate representativeness at 180 the national, regional, and county levels because of the non-proportional allocation to the sampling 181 strata and the fixed sample size per cluster. The statistical analysis was carried out using IBM 182 SPSS 24.

184
Right from the day of marriage, there are possibilities that an ever-married woman is violated by 185 her husband. However, it is not impossible that there are ever-married women without any experience of domestic violence as of the survey date. Non-inclusion of ever-married women who 187 have not experience any domestic violence as of the day of the survey might seriously bias the 188 computation of the timing of first domestic violence after marriage as well as its risk factor. The 189 survival analysis remains the utmost data analysis procedure for lifetime data when some subjects 190 have not experience the event of interest. Therefore, the populations at risk is all married women 191 involved in the study. The duration from marriage to first domestic violence, 'T', is assumed to be

207
We used the Cox proportional-hazards to model risk factors of the timings and estimate the 208 strength of the relationship between each of the selected independent variables and censored timing of first domestic violence. The outcomes of the model are the hazard ratios. A hazard ratio (HR) 210 <1 means higher risk, <1 means lower risk while =1 suggest insignificantly different risk. 211 We tested whether the proportional-hazards assumption was violated using the significance of the 212 HRs and Wald χ2 statistics in our stratified Cox analysis. We moved the significant variables in 213 the bivariate Cox regression models into the multiple Cox regression to assess their association 214 with the outcome variable while controlling for confounders. among women whose husband takes alcohol and 52.3% among women whose husband get jealous 242 when they talk to other men. In Mozambique, the highest prevalence of domestic violence was 243 among respondents aged 25-34 years (33.3%), 20-24 years (32.7%), 44.2% among 244 separated/divorced women, 32.9% among respondents with primary education, 33.1% among 245 employed women, 37.7% among women who are of same age as their husbands, 43.5% among 246 women whose husband takes alcohol and 32.3% among women whose husband get jealous when 247 they talk to other men.

251
In    (Figures 1, 2, and 3).   Table 3 shows the unadjusted (crude) determinants of the timing of first domestic violence after 287 marriage among ever-married women in the three countries. The risk of experiencing domestic 288 violence among ever-married women was significantly higher among separated/divorced and 289 widowed women than currently married women across the three countries (HR= 3.467; 95% with those in the poorest category (HR=0.647: 95% CI=0.536-0.782). Women whose husbands take alcohol and get jealous when the woman talks to another man were more than double likely to first 309 have the risk of domestic violence than others across the three countries.

DISCUSSION
While violence is a public health problem both in the developed and the developing countries.

344
Females, particularly the younger girls, are the most vulnerable population subgroups since they 345 are daily exposed and victimized by family members, at school, at work etc. It is more worrisome 346 in the developing countries due to either non-availability of strict punishment or failure to

369
It is also a source of concern that the risk of domestic violence was significantly higher among the 370 separated and divorced women than among the widows and among the currently married women.

371
It has been reported in multiple studies that occurrence of violence against women is still high in Mozambique(4).

376
The alarming rate of domestic violence in Africa remains an issue of concern. Despite the efforts 377 geared toward reducing the prevalence, significant change has not been recorded over the years.