Reproductive pattern, perinatal mortality, and sex preference in rural Tamil Nadu, South India: community based, cross sectional study

Introduction

During the past three decades, death and birth rates in India have declined considerably.1 2 This decline is more pronounced in Tamil Nadu than in many of the other Indian states.3 The decline in fecundity is in line with India's current policy of two children per family. Depending on reproductive failure (spontaneous abortion, stillbirth or infant death), women go through varying numbers of pregnancies to achieve the desired number of surviving children. In India, most couples desire at least one son, and many couples may thus go beyond the recommended or desired family size.4

Considering the decline in birth rate in India, it is uncertain how many pregnancies a woman has to go through to achieve the desired number and sex of children. Few community based studies describe the reproductive pattern in relation to perinatal mortality and sex preference among rural women in a developing country. This paper presents such data for a rural population in Tamil Nadu, South India.

Subjects and methods

The field study was carried out in the rural parts of Salem district in Tamil Nadu, South India; its main objective was to investigate aspects of antenatal care among women in rural Tamil Nadu. At the time of the study Salem district had a total population of 3 896 382 (1991 census), of which 24% was urban and 76% rural (R Balasubramaniam, personal communication). In rural areas, the government provides primary health care through 35 community health centres, 117 primary health centres, and 666 health subcentres. A health subcentre serves an average population of 5000. In the rural areas of Salem district, most people earn their living from agriculture or from smallscale industry.

To obtain a representative sample of the rural population in Salem district, a sample of 30 health subcentres was selected from a list of 615 (excluding those in semiurban and tribal areas) using a list of random numbers. The sampled centres served an estimated population of about 156 000.

We developed a questionnaire containing questions about reproductive history, antenatal care, delivery care, health status of the woman after the birth, health status of the youngest child, nutritional practices during pregnancy and postpartum, breast feeding, family planning, and socioeconomic status. The questionnaire was field tested by 50 women outside the study population who had just given birth; this resulted in a series of modifications. The final questionnaire was translated into Tamil. Three different people translated it back to English to ensure that the Tamil translation was in line with the original English version. A detailed set of guidelines to the questionnaire (including an inclusion guide) was developed and translated into Tamil.

Data were collected between 18 August and 27 September 1995. Through a meticulous house-to-house survey, 15 specially selected and trained local female interviewers identified and interviewed all eligible women in the catchment area of the 30 health subcentres. The interviewers were under close supervision of two experienced field leaders. Women were included in the study if they had delivered within the preceding six months; spoke Tamil; had stayed in the uptake area for more than two days; and were not mentally retarded. All questionnaires were reviewed daily by the two field leaders, and forms with missing information or visible inconsistencies were returned. If an eligible woman was not at home or if the time was not convenient for her to be interviewed, the interviewer came back later the same day or on the following day. In total, 1396 eligible women were identified, 1321 (95%) were interviewed, 68 (5%) were not at home on either of the two visits, 4 (0.3%) chose not to be interviewed, 3 (0.2%) had died, and data on socioeconomic factors and obstetric history were missing for one woman (0.1%).

To estimate interobserver and intraobserver reliability, double data collection was made in 11% of the cases selected randomly. In 69 (5%) of the cases the interviewer went back to the same informant twice to present the same questionnaire, and in 72 (6%) of the cases the interviewer presented the same questionnaire to women formerly interviewed by another field worker. Six key variables were chosen from the questionnaire; κ measures between 0.72 and 1.0 were found for interobserver reliability and between 0.76 and 1 for intraobserver reliability. Double data entry was made. The data were processed with EpiInfo 6.03.

Bivariate analysis of categorical data was based on the χ² test. Mean (SD) values are presented. For comparison of mean values of continuous data, a two sample t test was used. Rate ratios were used to compare mortality for boys and girls. Confidence intervals for rate ratios were computed as described by Rothman.5 Statistical significance was defined as P<0.05, and for rate ratios 95% confidence intervals are presented.

Results

Table 1) shows maternal and socioeconomic characteristics of the 1320 women with valid information on these matters. Of these, 261 women (20%) had experienced the death of one or more children. The women were aged between 14 and 54 years (mean 22 years). Primiparous women had a mean age of 20.2 (SD 3.3) years and multiparous women had a mean age of 23.9 (3.9) years. All women were married.

The women had had an average of 2.3 pregnancies and had 1.8 living children, based on a reported total of 2618 live born babies, 68 stillborn babies, 275 postnatal deaths, 280 spontaneous abortions, and 60 induced abortions. The rate of stillbirth was 13.5 per 1000 births, the perinatal mortality rate was 42.0/1000, the early neonatal mortality rate was 28.9/1000, and the late neonatal mortality rate was 6.6/1000.

Of the 1321 index pregnancies, 1309 (99%) were singleton pregnancies and 12 (1%) were multiple. Overall, 690 (51.7%) infants were boys and 644 (48.3%) were girls (107:100; P=0.68 for comparison to expected value). In total, 1291 (98.6%) singletons were liveborn and 18 (1.4%) stillborn. Of the liveborn singleton infants, 53 (4.0%) died postnatally: 34 (64%) in the early neonatal period (before 7 days), 7 (13%) in the late neonatal period (8-28 days), and 12 (23%) one month or more after birth. The multiple pregnancies (11 twins and one set of quadruplets) were all live births. The quadruplets all died in the early neonatal period; one twin died in the late neonatal period and one died later.

More boys than girls were born (table 2). Stillbirths and postneonatal deaths were equally distributed with regard to sex but a higher proportion of neonatal deaths occurred among girls (rate ratio 3.42; 95% confidence interval 1.68 to 6.98). The proportion of girls who died in the late neonatal period was non-significantly higher (6.62; 0.80 to 54.99) but the proportion of girls who died in the early neonatal period was significantly higher (3.07; 1.43 to 6.57).

Daughters of multiparous women had a relative risk of neonatal death of 4.36 (1.78 to 10.71) compared with sons, but daughters of primiparous women had no excess mortality (2.04; 0.61 to 6.78) (table 3). This association was stronger for neonatal mortality in daughters of multiparous women who did not have living sons (15.48; 2.04 to 177.73) in comparison to multiparous women with at least one living son (1.87; 0.63 to 5.58) (table 4).

The mean age of living children born before the index child was 2.80 (SD 1.70) years for the 312 sons and 2.87 (1.68) years for the 367 daughters (P=0.80).

Discussion

This study showed that women in this area of southern India have a controlled reproductive pattern. They give birth to their first child relatively late and have a limited number of children. Female infants have an increased risk of early neonatal death in families who already have daughters but no living sons.

As the sampling frame included only women who had just given birth, trends in birth and death rates as well as measures of infertility cannot be described. As well, the average number of pregnancies and children per woman applies only to a population of women who just gave birth. Due to the sampling frame, abortion rates may be underestimated.