Multilevel Logistic Regression Analysis Applied to Binary Contraceptive Prevalence Data (original) (raw)
Related papers
Determinants of Contraceptive Use in Bangladesh: A Hierarchical Modeling Approach
2011
This study examines the selected determinants of contraceptive prevalence among 10-49 aged ever-married women in Bangladesh and their true impact on the contraception prevalence rate (CPR). It applies a multilevel logistic regression analysis to draw valid conclusions about the effects of the selected determinants on CPR using the 2004 Bangladesh Demographic and Health Survey (BDHS) contraceptive binary data which is a multistage stratified cluster data. Instead of standard single level logistic model, multilevel logistic regression model has been utilized since the data follow a hierarchical structure. Also the comparison between single and multilevel model has been done to investigate the necessity of multilevel effects. The findings suggest that age of the women, number of living children, education, religion, media, place of residence and wealth index have significant multilevel effects on CPR. The study has finally suggested integrating a strong awareness program that targets the 10-49 aged currently married women in Bangladesh in those divisions (level-3) and clusters (level-2) where the particular determinant of contraceptive use has been found to be less effective.
Application of Multilevel Models on Reproductive Health Statistics
Ethiopia is the second-largest country in Africa, with an estimated population of nearly 79 million and a growth rate of 2.6 percent per year. It is believed that fertility decline continues if the wider use of contraception continues in all levels and groups of peoples. The present data set has a two-level hierarchical structure, with over 16,700 women nested within eleven geographical regions in Ethiopia. In bivariate analysis it was found that Place of residence ,Working status, exposure to media messages, educational status, and women religion had shown a significant variation. Women use of contraceptive was also considerably varied among regions in the preliminary bivariate analysis.
The present data set has a two-level hierarchical structure, with over 16,700 wo men nested within eleven geographical regions in Ethiopia. The bivariate analysis result showed that Place of residence, Working status, exposure to med ia messages, educational status, and women relig ion had shown a significant variation. Women use of contraceptive was also considerably varied among regions in the preliminary bivariate analysis. In Eth iopia, Women desire for more children was found to be the main reason that woman's do not practice contraceptives. Because of Our response variable is a binary indicator of whether a wo man uses modern contraceptives we are restricted to model the probability of a wo men's use of contraceptive with its odd. Accordingly, fro m the mu ltilevel logistic regression model it was found that all the three models are found to be significant indicating that there is real mu ltilevel variation among contraceptive users in Ethiopia. The deviance-based chi-square value is significant for mu ltilevel random intercept model imp lies that in comparison to the model with mu ltilevel random intercept and fixed slope model the mu ltilevel random intercept and random coefficients model has a better fit. This further implied that mult ilevel logistic regression model is best fit over the ordinary mu ltip le logistic regression models, further fro m the model fit diagnostics statistics, Deviance, AIC and BIC presented on . We can see that the model fit statistics values for random intercept are much s maller than the other mu ltilevel models therefore the random coefficient model best fits the data in comparison with other mu ltilevel models. Fro m the random coefficient estimates for intercepts and the slopes vary significantly, which implies that there is a considerable variation in the effects of religion, p lace of residence and rad io messages, these variables also found to differ significantly across the regions. The variance component for the variance of intercept in the random slope model is large compared to its standard error. Thus, there remains some regional-level variance unaccounted for in the model. Generally, this study revealed that socio-economic, demographic and pro ximate variab les are important factors that affect contraceptives use in Ethiopia. In line with this regional differentials shows that women in more urbanized regions such as Addis Ababa, Dire Dawa, and Harari are more likely to use modern contraceptives than respondents in regions that are more rural. The effect of regional variations for religion, p lace of residence and radio messages further implies that there exist considerable deference in modern contraceptive use among regions and a model with a rando m coefficient or slope is more appropriate to exp lain the regional variation than a model with fixed coefficients or without random effects. As there is variat ion and differences in use of Modern contraceptives across regions in Ethiopia, it is reco mmended to balance the effect of those factors across rural regions in Ethiopia. Researchers should use mult ilevel models than traditional reg ression methods when their data structure is hierarchal as like in EDHS data.
Journal of Mathematics and Statistics
The present data set has a two-level hierarchical structure, with over 16,700 wo men nested within eleven geographical regions in Ethiopia. The bivariate analysis result showed that Place of residence, Working status, exposure to med ia messages, educational status, and women relig ion had shown a significant variation. Women use of contraceptive was also considerably varied among regions in the preliminary bivariate analysis. In Eth iopia, Women desire for more children was found to be the main reason that woman's do not practice contraceptives. Because of Our response variable is a binary indicator of whether a wo man uses modern contraceptives we are restricted to model the probability of a wo men's use of contraceptive with its odd. Accordingly, fro m the mu ltilevel logistic regression model it was found that all the three models are found to be significant indicating that there is real mu ltilevel variation among contraceptive users in Ethiopia. The deviance-based chi-square value is significant for mu ltilevel random intercept model imp lies that in comparison to the model with mu ltilevel random intercept and fixed slope model the mu ltilevel random intercept and random coefficients model has a better fit. This further implied that mult ilevel logistic regression model is best fit over the ordinary mu ltip le logistic regression models, further fro m the model fit diagnostics statistics, Deviance, AIC and BIC presented on . We can see that the model fit statistics values for random intercept are much s maller than the other mu ltilevel models therefore the random coefficient model best fits the data in comparison with other mu ltilevel models. Fro m the random coefficient estimates for intercepts and the slopes vary significantly, which implies that there is a considerable variation in the effects of religion, p lace of residence and rad io messages, these variables also found to differ significantly across the regions. The variance component for the variance of intercept in the random slope model is large compared to its standard error. Thus, there remains some regional-level variance unaccounted for in the model. Generally, this study revealed that socio-economic, demographic and pro ximate variab les are important factors that affect contraceptives use in Ethiopia. In line with this regional differentials shows that women in more urbanized regions such as Addis Ababa, Dire Dawa, and Harari are more likely to use modern contraceptives than respondents in regions that are more rural. The effect of regional variations for religion, p lace of residence and radio messages further implies that there exist considerable deference in modern contraceptive use among regions and a model with a rando m coefficient or slope is more appropriate to exp lain the regional variation than a model with fixed coefficients or without random effects. As there is variat ion and differences in use of Modern contraceptives across regions in Ethiopia, it is reco mmended to balance the effect of those factors across rural regions in Ethiopia. Researchers should use mult ilevel models than traditional reg ression methods when their data structure is hierarchal as like in EDHS data.
Identifying factors influencing contraceptive use in Bangladesh: evidence from BDHS 2014 data
BMC Public Health, 2018
Background: Birth control is the conscious control of the birth rate by methods which temporarily prevent conception by interfering with the normal process of ovulation, fertilization, and implantation. High contraceptive prevalence rate is always expected for controlling births for those countries that are experiencing high population growth rate. The factors that influence contraceptive prevalence are also important to know for policy implication purposes in Bangladesh. This study aims to explore the socioeconomic , demographic and others key factors that influence the use of contraception in Bangladesh. Methods: The contraception data are extracted from the 2014 Bangladesh Demographic and Health Survey (BDHS) data which were collected by using a two stage stratified random sampling technique that is a source of nested variability. The nested sources of variability must be incorporated in the model using random effects in order to model the actual parameter effects on contraceptive prevalence. A mixed effect logistic regression model has been implemented for the binary contraceptive data, where parameters are estimated through generalized estimating equation by assuming exchangeable correlation structure to explore and identify the factors that truly affect the use of contraception in Bangladesh. Results: The prevalence of contraception use by currently married 15-49 years aged women or their husbands is 62.4%. Our study finds that administrative division, place of residence, religion, number of household members, woman's age, occupation, body mass index, breastfeeding practice, husband's education, wish for children, living status with wife, sexual activity in past year, women amenorrheic status, abstaining status, number of children born in last five years and total children ever died were significantly associated with contraception use in Bangladesh. Conclusions: The odds of women experiencing the outcome of interest are not independent due to the nested structure of the data. As a result, a mixed effect model is implemented for the binary variable 'contraceptive use' to produce true estimates for the significant determinants of contraceptive use in Bangladesh. Knowing such true estimates is important for attaining future goals including increasing contraception use from 62 to 75% by 2020 by the Bangladesh government's Health, Population & Nutrition Sector Development Program (HPNSDP).
PLOS ONE, 2020
This study advances current knowledge on contraceptive use in Bangladesh by providing new insights into the extent of regional variations in contraceptive use across rural and urban areas of Bangladesh. We examined the regional variations in contraceptive use among 15,699 currently married women ages 15–49 years using data from the 2014 Bangladesh Demographic and Health Survey (BDHS). Multivariate logistic regression models of contraceptive use were calibrated with sociodemographic attributes and cultural factors. Based on the aggregate sample (i.e., rural and urban combined), we found significant regional variations in contraceptive use across the administrative divisions in Bangladesh. Based on a disaggregate sample (i.e., rural and urban separately), we found that there were significant differences in divisional variations in contraceptive use in rural areas. In contrast, no significant variation in contraceptive use across divisions in urban areas of Bangladesh was found. More specifically, among women living in rural areas, the Rajshahi and Rangpur divisions had higher odds of contraceptive use than the Barisal division, whereas the Chittagong and Sylhet divisions had much lower odds of contraceptive use even after adjusting for selected sociodemographic attributes and cultural factors. A separate analysis of the divisional variations in usage of modern methods of contraception also revealed similar findings with only one exception. Findings of this study provide an evidence-based direction for adapting a pragmatic approach to reducing the divisional disparity of contraceptive use in rural areas of Bangladesh.
2022
The determinants of contraceptive method choice in Sri Lanka are examined during a period in which contraceptive prevalence increased by over 60% and involved substantial use of sterilization and traditional methods. Data are from the 1975 World Fertility and 1982 Contraceptive Prevalence Surveys. Polytomous logistic regression is used to model four current contraceptive choices: non-use, use of a traditional method, use of a modern temporary method and use of sterilization. The analysis shows, in 1975 and 1982, strong socioeconomic as well as demographic effects on whether any method is used. However, in both years the type of method chosen is primarily a function of demographic considerations related to the couple's family-building stage rather than social status, implying that in Sri Lanka there are few socioeconomic barriers limiting access to different contraceptive methods. The family planning programme, however, has emphasized sterilization rather than birth spacing methods.
Spatial Variation in Contraceptive Use in Bangladesh: Looking Beyond the Borders
2000
In this paper, an analysis is made of spatial patterns of contraceptive use in Bangladesh and the contiguous state of West Bengal in India in order to promote a more complete understanding of social change. Earlier analysis found strong evidence of higher contraceptive prevalence in districts of Bangladesh that border Bengali-speaking districts of India. Moreover, analysis of past fertility decline in Europe revealed that shared language played a critical role. Using multilevel analysis to control for variations in individual-and household-level correlates and mapping districts that deviate considerably from their regional averages, the present analysis highlights an important role for cross-border influences only in districts that share the same language across the border. Both in Bangladesh and in West Bengal, the districts that are positive outliers in contraceptive use hug the Bangladesh-West Bengal border. Maps show that the positive outliers disregard international borders and form a contiguous band in a manner that suggests a common-cause explanation of fertility decline.
2021
Background: Ethiopia is one of the most populous countries in Africa and has the highest number of new HIV infections. Therefore, the main objectives of this study were to demonstrate the joint predictors and association between modern contraceptive use and knowledge of HIV prevention in women aged 15-49 across clusters in Ethiopia.Methods: This study considered a weighted sample of 10,223 married women aged 15-49 years, extracted from the 2016 EDHS data. The bivariate multilevel logistics regression model was used in this study.Results: In Ethiopia, 64.75%, 54.58% of the women didn’t use the modern contraceptive method and didn’t know about HIV prevention, respectively. A positive relationship between the use of modern contraceptives and knowledge of HIV prevention of women across the clusters was found, which implies that modern contraceptive use has positive implications on women's knowledge of HIV prevention techniques across the clusters.Conclusions: Predictor visiting a he...
Determinants of Adolescent Childbearing in Bangladesh: A Multilevel Logistic Regression Analysis
Journal of Child and Adolescent Behavior, 2018
Background: In South Asian region particularly in Bangladesh, adolescent pregnancy is still a major public concern. Therefore, identification of potential risk factors of adolescent pregnancy and childbearing may help to formulate policy to prevent premature delivery, death and disability. Objectives: The main objective of this study was to identity the determinants factors of adolescent pregnancy in Bangladesh. Study design: Retrospective two-stage stratified sample design. Methods: Data were extracted from nationally representative Bangladesh Demographic and Health Survey conducted in 2011. In this study, age of childbearing was considered as dependent variable and classified as adolescent (≤19 years) and adult (>19 years). Several individual, household and community level factors were considered as the independent variables. Multilevel logistic regression model with household and community random intercept was used to identify the association between dependent and independent variable. Results: The study found that about 28% adolescent had given birth in Bangladesh. Higher educated women (OR=0.65, 95% CI, 0.49-0.96), husband with higher education (OR=0.69, 95% CI, 0.56-0.84) and age at first marriage after age 15 years (OR=0.50, 95% CI, 0.45-0.55) were found to be protective for adolescent pregnancy. Secondary educated women (OR, 1.21; 95% CI, 1.03-1.39) showing the higher adolescent pregnancy than women with no education. The risk is high among women in Rangpur (OR=1.19; 95% CI, 1.12 1.25), followed by Khulna (OR=1.16; 95% CI, 1.09 1.23), Rajshahi (OR=1.16; 95% CI, 1.09 1.23) division showing among in Barisal division. Conclusion: Around one fourth of the total pregnancy occurred adolescent age in Bangladesh which is caused by various individual, household and community-level factors. Proper education and increase awareness about the adverse effect of adolescent pregnancy are necessary to reduce occurrence of adolescent pregnancy in Bangladesh.