HIERARCHICAL MODELLING OF GENDER-BASED VIOLENCE AMONG WOMEN OF REPRODUCTIVE AGE IN NIGERIA

Abstract

Multilevel models offer analysts of large scale complex survey data a relatively new approach to understanding individual and contextual influences on outcome variables. In public health, large surveys often follows hierarchical structure as the surveys are based on multistage stratified cluster sampling. The exact method to analyze such survey dataset is therefore based on nested sources of variability which comes from different levels of the hierarchy. When the variance of the residual errors is correlated between individual observations as a result of these nested structures, the conventional logistic regression is inappropriate. Previous studies on Gender-Based Violence (GBV) has ignored this variability in the prevalence of GBV and used the conventional logistic regression without taking into consideration the hierarchical structure of the population. Also, this had led to inaccurate inference. Therefore, this study investigated factors associated with GBV taking into consideration the hierarchical structure of the population. A nationally representative sample of 14,618 women of reproductive age (15-49 years) within households in communities was obtained from the National Demographic Health Survey (NDHS) of 2013. The experience of one or more forms of GBV was the outcome variable, while explanatory variables includes age at cohabitation, residence, religion, educational status, marital status, wealth index, literacy level, alcohol intake, smoking, and region. The prevalence of violence experience and its contributing factors were summarized using frequency and proportion. Differences in the relationship between outcome and explanatory variables were measured. Multilevel logistic models was used to investigate the factors associated, and assessed the contribution of geographical variations. Adjusted odds ratios (OR), 95% confidence intervals (95% CI) as well as, Intra-class Correlation Coefficients (ICC) for each random effect were estimated. The findings from this study shows that, women who were married constituted the highest proportion (89.7%) of the total sample with (57.3%) residing in the rural areas. The distribution of these women in terms of wealth index revealed that a higher proportion of these women belong to the richest wealth quintile (23.9%). About half of these women practice Islamic religion (49.4%). In addition, the risk of experiencing GBV is approximately two times (OR= 1.49, (1.32-1.69) more likely among women whose husband drinks alcohol compared with their counterparts whose husband do not take alcohol. Women in the urban areas were 29% (OR=0.71, (0.5.4-0.92) less likely to experience GBV compared to women in the rural areas. However, women in North East, Nigeria were 68% less likely to experience GBV compared to women in North Central. Women in the South West, Nigeria were approximately five times (OR=4.68, 95% CI: 3.07-7.12) more likely to experience GBV when compared with women residing in the North Central at community-level. Gender-Based violence was approximately four times (OR=3.82, 95% CI: 2.49-5.84) more likely to occur among divorced women compared to married. The use of multilevel models revealed the true state at which various factors influenced the experience of GBV among women in Nigeria. The model showed strong presence of GBV among women in southern, Nigeria. In addition, alcohol continues to be a strong contributor to GBV.