To the Editor: Girl child marriage (ie, <18 years of age) affects more than 10 million girls globally each year and is linked to maternal and infant morbidities (eg, delivery complications, low birth weight) and mortality.1,2 Half (46%) of child marriages occur in South Asia.1,2 This study assessed whether prevalence of girl child marriage has changed over the past 2 decades in 4 South Asian nations with a girl child marriage prevalence of 20% or greater.1-3
All available population-based Demographic and Health Surveys (DHS) data from Bangladesh, India, Nepal, and Pakistan between 1991 and 2007 were analyzed. The DHS are nationally representative surveys that measure demographics, health, and nutrition with standard measures across nations and over time. Data collection and management procedures are described in detail elsewhere.3 Briefly, cluster randomized samples are selected.4 After stratification by rural or urban area and geographic or administrative regions, random clusters of approximately 25 households are selected from each area, and an eligible woman is identified from each household. All data were collected from women in or near households but not necessarily in a private setting.
The DHS procedures were approved by ICF Macro International institutional review board and the ethics review boards of each nation included in the study. Oral informed consent was obtained from all respondents. The University of California at San Diego institutional review board ruled this study to be exempt from full review due to use of secondary analysis of data with no identifiers.
The age at marriage variable was based on the difference between the date of start of first marriage or union and the respondent's date of birth (items provided via self-report). Analyses were restricted to women aged 20 to 24 years to allow for the inclusion of all women married or in union by age 18 years within the closest period for which data were available.
Prevalence estimates and 95% confidence intervals were calculated for girl child marriage and subgroups using DHS-calculated individual weights4 to take into account the multistage sampling design and provide results for all (not just ever married) women. Cochran-Armitage tests5 were used to test linear time-trend data by country; χ2 tests were used for nonlinear trends with tests adjusted for complex survey design.6 Significance was set at P <.05 using 2-sided tests. Analyses were conducted in SAS version 9.2 (SAS Institute Inc) and Microsoft Excel.
Sample sizes ranged from 1064 to 22 807 (Table 1). The prevalence of girl child marriage decreased in all countries from 1991-1994 to 2005-2007 (Table 2). Significant relative reductions occurred in marriage of girls prior to age 14 years across all 4 nations, ranging from −34.7% (95% CI, −40.6% to −28.1%) to −61.0% (95% CI, −71.3% to −46.9%). Little or no change over time was seen in marriage of 16- to 17-year-old adolescent girls for any nation except Bangladesh, where such marriages increased by 35.7% (95% CI, 18.5% to 55.3%).
Reductions in girl child marriage in South Asia have occurred but are largely attributable to success delaying marriage among younger but not older adolescent girls. Improvements in education of girls and increasing rural to urban migration may have supported these reductions,1,2 but many schools graduate students at the 10th standard (about 15-16 years), maintaining vulnerability to early marriage for 16- to 17-year-old girls. Laws against early marriage have existed for decades, setting the legal age for girls at marriage as 18 years in Bangladesh, India, and Nepal, and 16 years in Pakistan, but appear inadequate to affect this issue. Increased prevalence of marriage among 16- to 17-year-old girls in Bangladesh requires further study.
Study limitations include possible social desirability or recall bias and potential inaccuracies reporting age at marriage. Focus on young women reduces risk for recall bias. Differential survey time points allow greater time for change to be assessed for Pakistan and less time for Nepal. Analyses are restricted to time trends and lack consideration of variables (eg, changes in education) to explain findings.
Author Contributions: Dr Raj had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Raj, McDougal.
Acquisition of data: Raj, McDougal.
Analysis and interpretation of data: Raj, McDougal, Rusch.
Drafting of the manuscript: Raj, McDougal.
Critical revision of the manuscript for important intellectual content: Raj, McDougal, Rusch.
Statistical analysis: McDougal, Rusch.
Obtained funding: Raj, McDougal, Rusch.
Administrative, technical, or material support: Raj, McDougal.
Study supervision: Raj.
Conflict of Interest Disclosures: The authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Raj reported having grants pending with the National Institutes of Health and the Kellogg Foundation. Dr Rusch reported having grants pending with the National Institutes of Health; receiving compensation for travel and meeting expenses from the British Columbia Centre for Excellence in HIV/AIDS; and receiving an honorarium from the Ontario HIV Trails Network. Ms McDougal did not report any disclosures.
Funding/Support: This work was funded by a grant from the David and Lucile Packard Foundation.
Role of the Sponsor: The David and Lucile Packard Foundation was not involved in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
Raj A. When the mother is a child: the impact of child marriage on the health and human rights of girls. Arch Dis Child
. 2010;95(11):931-93520930011PubMedGoogle ScholarCrossref
Rutstein S, Rojas G. Guide to DHS Statistics. Calverton, MD: ORC Macro; 2006
Armitage P, Berry G, Matthews JNS. Statistical Methods in Medical Research. 4th ed. Oxford, UK: Blackwell Science; 2002
Rao JNK, Scott AJ. On chi-square tests for multiway contingency tables with cell proportions estimated from survey data. Ann Stat
. 1984;12(1):46-60Google ScholarCrossref