PERFORMANCES OF BAYESIAN STRUCTURED ADDITIVE REGRESSION AND LOGISTIC MODELS IN ASSESSING FACTORS ASSOCIATED WITH MODERN CONTRACEPTIVE USE AMONG EVER-IN-UNION WOMEN IN NIGERIA

Abstract

Background: Recognizing the factors that contribute to the use of modern methods of contraception serves a significant role in designing successful strategies that can lead to increased use and subsequently to a reduction in the rate of population growth. However, most studies on contraception methods did not consider the spatial effect with regard to the use of modern contraceptive methods. This study therefore compared the performance of Bayesian structured additive regression with the frequentist logistic regression in the identification of factors associated with modern contraceptive use among ever-in-union women across the six geo-political zones of Nigeria. Method: This study adopted a cross-sectional data extracted from the 2018 Nigerian Demographic and Health Survey (NDHS) of ever-in-union women (15-49 years) selected in both urban and rural areas in the six geo-political zones in Nigeria. The analysis was done with SPSS software (version 25) and R-programming language with BayesX installed. The Bayesian structured additive regression and frequentist logistic regression models were used for analysis. Inference was based on Markov Chain Monte Carlo (MCMC) technique and model diagnostics was applied to examine the model that best explains the interest of this study with the use of Deviance Information Criteria (DIC). Results: A total of 31,152 ever-in-union women were included in the analysis. The prevalence of modern contraceptive use was found to be 11.5% among these women. It was revealed that repondent’s age at their first marriage is a great determinant of their usage of these methods and found that those who got married between the ages 15-19 and 20-24 have higher odds of 10.8% and 12.4% to use modern methods of contraception more than those who got married at an earlier age. It was also interesting to observe that access to media and knowledge of these methods indeed had a part to play in their use of these methods as their exposure to media and knowledge gave an opportunity for these women to use it more with odds of 11.3% and 28% respectively. The findings further showed there was a strong north-south difference, with the usage of modern contraceptive methods in the northern states being distinctly lower than in the southern states. The non-linear effect of respondent’s age at first marriage was seen to have a great contribution, as a high percentage (20.9%) of these women got married at a very young age of less than 15 years which was seen to be majorly in the northern part of Nigeria in the year 2018. It was also revealed that the Bayesian approach was a better fit for the model in this study due to the DIC (110.152) which was lesser than that of the frequentist approach (17007.148). Conclusion: These findings revealed that the Bayesian approach yielded better results as compared to that of the frequentist and so other researchers are encouraged to delve into that aspect for more accurate inferences in research work. In addition, all these being taken into proper consideration, will help the government, non-governmental agencies and policy makers to channel the interventions properly to the zones and states that are in dire need of them and also manage the limited resources available.