International Journal of Epidemiology 2000;29:300-307
© International Epidemiological Association 2000
Epidemiological models and related simulation results for understanding of contraceptive adoption in India
Department of Biostatistics, All India Institute of Medical Sciences, Ansari Nagar, New Delhi-110029, India.
Background For the first time, models using multilevel analysis of Indian data and related simulation results are reported. They take hierarchical structure into account and incorporate variables from all levels to get correct analysis and proper interpretation of data on current contraceptive use (including sterilization and modern methods).
Methods The data from an Indian State, Uttar Pradesh (UP), collected by the National Family Health Survey (NFHS) conducted during 10 October 1992 to 22 February 1993 was used. For model I, 7851 currently married women who were neither pregnant nor had continuing post-partum amenorrhoea (PPA) were considered. For model II, these women with at least one child (n = 6748) were used. Two-level logistic regression analysis was carried out for which women's level (level 1) and PSU (Primary Sampling Unit) level (level 2) variables were considered. The results were considered significant at the 5% level of significance. Simulation analysis using each model was also carried out.
Results Model I reveals that those more likely to adopt contraception were women exposed to a TV message (odds ratio [OR] = 1.3; 95% CI : 1.1–1.6); whose houses were pucca (bricks and mortar) (OR = 1.3; 95% CI : 1.1–1.5); who were educated to high school level and above (OR = 2.9; 95% CI : 2.2–3.7); whose husbands were literate with schooling of 
11 years (OR = 1.7; 95% CI : 1.4–2.1); and who had 2 living sons (OR = 2.2; 95% CI : 1.1–4.4). Muslim and other religious women were less likely than Hindu women to adopt contraception (OR = 0.5; 95% CI : 0.4–0.6). Also, the PSU level availability of all weather road was positively associated with contraceptive adoption (OR = 1.4; 95% CI : 1.1–1.7). The PSU level variance, which is the unexplained PSU level variation after controlling for the considered characteristics, was significantly higher. The simulation results revealed that public health education (a TV message) was found to be more effective among less educated women. The PSU level availability of all weather road was as effective as public health education. Similar results were evident from the analysis of second data set (model II) with the noticeable finding that those whose last child is surviving are most likely to adopt contraception (OR = 8.82; 95% CI : 1.01–77.38).
Conclusions These results reveal that the survival status of the last child has a marked effect on the adoption of contraception in UP. They further support the idea that public health education (a TV message) is more effective among less educated women. Also, the PSU level presence of all weather road is equally effective. Consideration of higher level variables provides not only more accurate results but also important public health clues to help the policy planners.
Keywords Multilevel analysis, hierarchical, currently married, post-partum amenorrhoea, significant, PSU level
Accepted 30 September 1999