Md. Nuruzzaman Haque (MS in Demography)
Assistant professor
Dept. of Population Science and Human Resource Development
University of Rajshahi, Rajshahi-6205, Bangladesh.
Tel: +880-721-750041(Ext. 4121)
Fax: +880-721-750064
Mobile: +8801556621599
E-mail: nzaman_pop@yahoo.com

INTRODUCTION

Motherhood or first childbirth is the most important event in women's life. Timing of motherhood or age at first birth has different effects on birth outcome, on the health conditions of the mother herself and also on her child's health. Early childbearing, for example, first childbirth at adolescence (age 10-19 years), and first childbirth after adolescence can be seen as having different effects¹.

Comparing women who are older than 19 years, adolescent (aged 10-19 years) women are at greater risk for poor maternal conditions and birth outcomes and also young mothers are more likely to suffer pregnancy related complications and to die in childbirth, than women of age 20 years or more². Women who have their first child early in their life are more likely to have more children than those women who start childbearing later³. In Bangladesh, early marriage and early pregnancy (at adolescence) are common. Early pregnancy (in Bangladesh) means early childbirth, because 90% of pregnancies result in a live birth and the remaining 10% result in miscarriage/abortion, stillbirth and/or menstrual regulations (MRs) in Bangladesh⁴. Mean age at first marriage for adolescent girls and for women aged 10-49 years are 14.35 years and 15.01 years respectively, and mean age at first childbirth for married adolescent women and for ever married women aged 10-49 are 15.68 years and 17.40 years respectively in Bangladesh (own calculation using data from the Bangladesh Demographic and Health Survey (BDHS-2004). Proportion of adolescent women who are mothers or are currently pregnant is the highest (about 35%) in Bangladesh among the Asian countries5. According to data of BDHS-2004, only 29.1% of 10-14 years ages and 42.2% of 15-19 years ages of currently married adolescent women (adolescents who are cohabiting with their husbands) respectively, were using any contraceptive method in Bangladesh.

In a society, like Bangladesh, where childbearing is socially sanctioned after marriage, the lengths of first birth interval affect the complete family size⁶ and also infants' and children's survival depends on their mothers characteristics. First birth is considered as one of the child survival risk categories (CSRC)⁴. Women, whose first births were early (<18 years) are more likely to give 2nd birth within very short birth intervals (less than 24 months)⁷. Births at age less than 18 years and birth intervals less than 24 months are also considered as one of the CSRC, and child's risks of dying and other health complications are further raised for the child born to a mother who has a combination of CSRC⁴. Complications related to pregnancy and childbirth found the important cause of death among girls aged 15-19 years in developing countries, and early marriage for females and early motherhood are very common in Bangladesh compared to other south Asian countries⁸. But, mothers aged 14 years and/or less than 14 years face the greatest risks. Findings from one study in Bangladesh by Chen Lincoln C. et al. concluded that mothers aged 10-14 years may face the risk of maternal mortality five times more compared to mothers aged 20-24 years⁹.

Considering high incidence of first birth at early ages and the negative consequences of early first birth, in Bangladesh, there is a need to unfold differentials in timing of EMAW's first birth regarding various covariates possessed by them. This study expects to estimate the extent of early first birth and pattern of giving first childbirth (since marriage) by ever married adolescent women and then to examine some selected covariates' impact on timing of the event (first birth) of EMAW and to mark the disparities in the event (first birth) of EMAW with respect to socio-demo-cultural characteristics (covariates) possessed by them. All of which, stated above, may be helpful for policy makers, program managers/ donor agencies to make/support appropriate programs for the well being of women (which is also advantageous for infants' and children's good health) and also for decreasing family size, in Bangladesh.

For analyzing fertility behavior such as pattern of giving first birth, in the society of Bangladesh, it is evident to consider married women exposed to first birth after their first marriage. In this study, ever married adolescent women (EMAW) - married adolescents including widowed, divorced and separated- are included for analyzing their first childbirth pattern from their marriage date to the end of the study period. For analyzing time to the event (first birth) of EMAW, data comes from Bangladesh Demographic and Health Survey (BDHS) - 2004. After excluding missing cases and pre-marital first birth (only one in this study), 1,629 EMAW were included in the study. Among those 1,629 EMAW, 874 EMAW have faced the event of first birth from their marriage date to the end of study period (as of May 2004).

Extent of early first birth in Bangladesh
For developing ing EMAW's (who have given childbirth) profile with age and age at first birth, data have been extracted from Bangladesh Demographic and Health Survey- 2004. Among the ever married adolescent women (EMAW) more than 53% of them had a first childbirth in Bangladesh in 2004. The percentage of EMAW who gave birth is lower in urban areas than rural areas. Percent distribution of EMAW with first birth is provided in Table 1. About 12% of EMAW of the study population have given first birth in their early adolescence (10-14 years) in Bangladesh. The percentages of EMAW who gave birth were highest at the age of 15 years, 16 years and 15 years for urban areas, rural areas and national (Bangladesh as a whole) respectively (Figure 1).

Pattern of first birth intervals
The pattern of first birth intervals (since first marriage), according to residence, urban or rural, with corresponding medians of intervals are provided in Table 2. The estimated median length of first birth intervals was 20 months at national level, and those of first birth interval were 18 months and 20 months at urban and rural areas respectively in Bangladesh. About 70% of urban EMAW and 63% of rural EMAW gave their first birth within two years of their marriage (Table 2 and Figure 2).

Source: Author's estimations based on BDHS-2004.

Statistical analysis
To explain why certain individuals are higher or lower risks of experiencing one event of interest than others, can be accomplished by the method of Cox proportional hazard model. The Cox proportional hazard model10 for censored non-occurrence event /survival data specifying time specific hazard rate or failure rate or occurrence rate of the event as (t) = 1im h 0 Pr(T< t + h| T> t)/h for non-occurrence or survival time T of individuals with covariates x_1j, x_2j,………, xnj (e.g. binary, categorical or continuous and these covariates may depend on time or not) to have the following form
(t; x) = ₀(t) exp (B_1j X_1j + B₂X_2j + ... ... ... ... ... ... + B_njX_nj), for t > 0 (1)

Where, B_1j, B_2j, ... ... ... ... ... ... ... ... ... , Bnj are non-standardized regression coefficients, j represents the number of groups or categories of the respective covariate,₀(t) is the baseline hazard function. Let us consider survival time for any event of interest with only one covariate, for example, x_1j with two categories (j = 1, 2) and values of that two categories are 0 (x₁₁ = 0) and 1 (x₁₂ = 1), equation (1) becomes
for x₁₂ = 1
(t; x = 1)=₀(t) exp(B₁₂) ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... (2)

and for x₁₁ = 0
(t; x = 0)= ₀(t) ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... (3)

Dividing (2) by (3) we get
(t; x = 1)/ (t; x =0) =RR (risk ratio or relative risk) = exp(B₁₂) ... ... ... (4)

which tells us the relative risk of occurrence (within the risk period) of that event of interest with covariate x = 1 compared to that with covariate x = 0 (considered as reference category). A RR > 1 means the group of interest (individuals possessing the category (or value) of covariate) comparing to the reference group is likely to have a shorter time to the occurrence of the event. A RR<1 means the group of interest comparing to the reference group is less likely to have shorter time to the occurrence of the event. The above model (1) can be fitted for estimating regression coefficients as well as risk ratio or relative risk, by using SPSS (Statistical Package for Social Sciences). For analyzing time to the event (first birth) data, Cox proportional hazard model (described above) has been used, which provides an opportunity to explore the effects of various time independent covariates on the timing of first birth.

In this analysis, interested event is the first birth of ever married adolescent women (EMAW), and all EMAW, after first marriage, are considered as at exposure to the risk of first birth. Observation starts at the time of first marriage and ends when a first child is born or, for right-censored cases after the month of May 2004 (end of study period). The Individual Recode Data File of BDHS-2004 included the variables, age at first marriage and age at first birth, in century month codes (CMC) format and also in years. Century month codes are the measures of number of months from the beginning of the century to the occurrence of the interested event, and January 1, 1900 is considered as the beginning of century¹¹. This CMC form of variables facilitates the calculationinterval between events. In this analysis, the durations (in month(s)) of the event (first birth) of EMAW were calculated by subtracting age (in CMC format) at first marriage from the age (in CMC format) at first birth. For Cox proportional hazard analysis, the dependent variable is the likelihood (or risk) of first birth and the status variable considered in this study as having a first childbirth which coded as 1 if ever married adolescent woman (subject of study sample) had a first birth up to the end of the study period.

Interpreting the Cox proportional hazard model involves examining the regression coefficients (B's) for each category of the covariates and also examining the relative risk/hazard rate for each category compared to the reference category of the covariates. A positive regression coefficient for a category of covariate means the relative risk/hazard for first birth is higher than the reference category, and a negative regression coefficient for a category of covariate means the relative risk/hazard for first birth is lower than the reference category. A relative risk greater than unity indicates a higher first birth risk (i.e. lower the duration/length from first marriage to first birth), and vice versa.

Covariates for Cox proportional hazard model
Place of region, place of residence, EMAW's educational attainment, educational attainment of EMAW's husband, age at marriage, age difference between the spouses, and religion are included in the Cox proportional hazard model as independent variables (covariates). Percent distribution of EMAW with covariates included in the Cox proportional hazard model, and the categories of each time independent covariate are provided in Table 3. Bangladesh is divided into six administrative divisions (place of region) where respondents lived, display a variation in contraceptive method use, in unmet need for contraceptives, in receiving maternal health care services. To control for these regional differences, region is included in this study as a categorical covariate. Also, the usual place of residence, urban or rural, where EMAW lived, displays some variation in cultural, socio-economic and in demographic features. To examine the disparities in urban and rural areas regarding timing of first birth of EMAW, residence is included as a categorical covariate in the Cox proportional hazard model. Education of EMAW at first birth is not used because this information is not available in the BDHS 2004 data. Instead, education at interview is used in place of education at first birth. For most EMAW it is reasonable, since current age (age at interview) and age at first birth of EMAW is positively correlated (the value of Pearson's correlation coefficient is 0.46 at 0.01 level of significance). Educational attainment of EMAW is generally higher compared to their husbands' educational attainment. Over half (53%) of EMAW are educated beyond the primary level and less than 15 percent of EMAW have no education; the figure contrasts with the 27 percent of husbands who have no education. Less than 43% of the husbands are educated beyond the primary level.

Source: Author's estimations based on BDHS-2004.

Age at marriage is divided into three categories as younger than 15 years (< 14 years), 15-17 years, and 18-19 years. The distribution of EMAW's age at first marriage shown in Table 3 reflects that Bangladeshi women marry in relatively very early adolescence, despite the law of minimum age at marriage (18 years for women). To examine the disparities of EMAW's timing pattern of first birth regarding beliefs in religion, religion, categorized as Islam and other than Islam, is included in the Cox proportional hazard model. To explore whether the impact of other covariates varies among urban and rural areas, the Cox proportional hazard model is used separately for urban and rural areas (Model II and Model III are respectively for urban and rural areas, in Table 4).

Note: Levels significant at a= 10%, b=5%, and c=1%; RC= Reference category.

The major findings from the Cox proportional hazard analysis appear in Table 4. The results are presented as regression coefficient (B) and relative risk/hazard (exp(B)) of having a first birth relative to reference group for a selected covariate. Model I, model II and model III in Table 4 are based on EMAW for Bangladesh as a whole (national), urban, and rural areas respectively. There are regional variations in the length of first birth intervals: EMAW from Chittagong division had the shortest first birth intervals among the six divisions in Bangladesh. According to place of residence, urban or rural, EMAW residing in urban areas have shorter first birth intervals than those EMAW who resided in rural areas. EMAW with lower educational attainment are less likely to give birth (longer first birth interval) compared to EMAW with complete secondary and/orhigher education in Bangladesh. In contrast to the effects of EMAW's education, their husband's education displays reverse results (in model I and model II). Compared to the reference group (husband with complete secondary and/orhigher education) in whole of Bangladesh (national) and in urban areas, the risk of first birth increased (had shorter first birth intervals) for EMAW with husband's education lower than complete secondary. But in rural areas (Model III), the variations in the risk of first birth of EMAW with husband's education in different levels are not statistically significant. Adolescent women who married in late adolescence had higher risk of first birth than those who married in early adolescence.

All categories for the covariate (age difference between spouses) had lower risk of having first birth compared to reference group (husband 16 years or more than 16 years older).
About 53% of EMAW had given first birth, and about 90% of them gave their first birth within three years of their marriage in Bangladesh. It may be caused because of low utilization of contraceptives at adolescence or before first birth.

EMAW with no education, incomplete primary, complete primary or incomplete secondary had lower risk of first birth (longer first birth interval) than EMAW with complete secondary and/or higher education in Bangladesh. It may have happened because, in Bangladesh, adolescent girls with more education get married later (and want child very soon after marriage) compared to adolescent girls with less education.

Adolescent women's age at marriage had significant effect on risk of first birth. According to the results of hazard regression analysis in the three models, women, who got married at less than 15 years of age and at between 15-17 years of age, had lower risk of first birth i.e. longer first birth intervals. It indicates that Bangladeshi adolescent women, who got married at age 18 or 19 years, had their first childbirth very soon after their marriage.

It should be conveyed to currently married adolescent women about the perils of early first birth, by making appropriate communication programs. This study shows that lengths of EMAW's first birth are very short in Bangladesh. Chittagong division and urban areas should be paid more consideration for lengthening duration between marriage and first birth. Early childbearing can be postponed by delaying marriage. Due to prevailing cultural and social norms favoring early marriage, only legislation for age at marriage is an effective way of delaying marriage. There are some other ways, such as policies and programs, to increase opportunities for education and more education for adolescent girls, and to increase parents (of adolescents) awareness through social campaigns about negative consequences of early marriage and early childbearing, all of which may be likely result in delayed marriage (which may also be likely to shorten the spouses' age difference). It is necessary to emphasize reproductive health education at secondary level in the country's education system. More important is the need to lengthen the interval between marriage and first birth, thus delaying first birth. There is a need to encourage newly married adolescent women (also their husbands) to use contraceptives which are relevant to spacing births through reinforcing supplies of contraceptives by family planning program efforts.

Acknowledgments: This work has been conducted in my stay at Center for Northeast Asian studies of Jilin University, China. I am grateful to Center for Northeast Asian studies of Jilin University for providing me with computer facilities and a good research environment. There was no financial support for this study.

1. Steven D. M and Michael M, 1983. The Timing of the First Birth and Changes in Personal Efficacy, Journal of Marriage and the Family, Vol. 45, No. 1, (feb., 1983), pp 47-55.
   
2. McCauley A.P and Salter C, 1995. Meeting the needs of young adults, Population Reports, 1995, Series J, No.41.
   
3. Tawiah E. O, 2002. Adolescent Fertility and Reproductive Health in Four sub-Saharan African Countries, African Population Studies, Vol. 17, No. 2, 2002, pp. 81-98.
   
4. NIPORT (National Institute of Population Research and Training), ORC Macro, Johns Hopkins University and ICDDR, B., 2003. Bangladesh Maternal Health Services and Maternal Mortality Survey 2001, Dhaka, Bangladesh and Calverton, Maryland (USA): NIPORT, ORC Macro, Johns Hopkins University and ICDDR, B.
   
5. Heidi W. Reynolds, Wong E and Tucker H, 2006. Adolescent's Use of Maternal and Child Health Services in Developing Countries, International Family Planning Perspectives, 2006, 32(1):6-16.
   
6. Dilip C. Nath, Kenneth C. Land and Giti Goswami (1999). EFFECTS OF THE STATUS OF WOMEN ON THE FIRST-BIRTH INTERVAL IN INDIAN URBAN SOCIETY. Journal of Biosocial Science, 31, pp 55-69
   
7. Saumya RamaRao, John Townsend and Ian Askew, 2006. Correlates of Inter-birth Intervals: Implications of Optimal Birth Spacing Strategies in Mozambique, Population Council, March 2006.
   
8. Save the Children, 2004. State of the World's Mothers: 2004, Westport, CT, USA: Save the Children, 2004.
   
9. Chen Lincoln C., Melita C. Gesche, Shamsa Ahmed, A.I. Chowdhury and W.H. Mosley, 1974. Maternal Mortality in Rural Bangladesh, Studies in Family Planning, Vol. 5, No. 11, (Nov., 1974), pp 334-341.
   
10. Cox D.R., 1972. "Regression Models and Life-Tables", Journal of the Royal Statistical Society, Series B (Methodological), Vol. 34, No. 2, pp 187-220.
    
11. Banerjee S., 2006. Higher Education and Reproductive Life Course: A cross-cultural study of women in Karnataka (India) and the Netherlands, Doctoral thesis, University of Groningen, Thela Thesis, Amsterdam, available at www.rug.nl/prc/publications/completedPhdDissertations