Factors associated with unwanted pregnancies in India through a hierarchical model

Issue title: Statistical Models and Applications in Biomedicine

Guest editors: Boris Zaslavsky

Article type: Research Article

Authors: Begum, Shahina^{a; *} | Dwivedi, Sada Nand^b | Pandey, Arvind^c | Mittal, Suneeta^d

Affiliations: [a] Shahina Begum, Asst Prof, Department of Community Medicine, R.D. Gardi Medical College, Surasa, Ujjain, Madhya Pradesh, India | [b] Sada Nand Dwivedi, Addl Prof, Department of Biostatistics, All India Institute of Medical Sciences, New Delhi, India | [c] Arvind Pandey, Director, National Institute of Medical Sciences, ICMR, Ansari Nagar, New Delhi, India | [d] Suneeta Mittal, Prof & Head, Department of Obstetrics & Gynaecology, All India Institute of Medical Sciences, New Delhi, India | Division of Biostatistics, Food and Drug Administration, Rockville, MD, USA

Correspondence: [*] Corresponding author: Shahina Begum, PhD, Asst Prof (Biostatistics), Department of Community Medicine, RD Gardi Medical College, Surasa, Ujjain, Madhya Pradesh-456006, India. Tel.: +91 7368 261315 (o); Fax: +91 7368 261235; E-mail: shahina05@gmail.com.

Abstract: Objective:The first objective is to find out the factor associated with unwanted pregnancies and the second is to validate the developed model using bootstrapping techniques to get the biased corrected estimates. Methods:Data of pregnant women have been extracted from second round of National Family Health Survey (1998–1999), India. The hierarchical structures of data comprising women (at lower level) nested within state (at higher level). Multilevel logistic regression analysis was carried out to achieve first objective and for the second objective, bootstrapping technique is used. Results:Multilevel logistic regression analysis revealed that age, educational status, number of surviving sons, interval from last live birth to index pregnancy, ever contraceptive use, ever physically mistreated by husband were significant factors associated with unwanted pregnancies at individual level. Only 8 percent of the variability in unwanted pregnancy could not be explained by the considered set of covariates. Also, for this model the state level variance was not significantly different from zero. Conclusion:Effective use of contraceptive method may decrease the prevalence of unwanted pregnancies.

Keywords: Contextual variable, bootstrapping, multilevel analysis, smoothing, unintended pregnancies

DOI: 10.3233/MAS-2010-0144

Journal: Model Assisted Statistics and Applications, vol. 5, no. 3, pp. 197-202, 2010