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

BMJ 1997; 314 doi: (Published 24 May 1997) Cite this as: BMJ 1997;314:1521
  1. Birgitte Bruun Nielsena, research fellow,
  2. Jerker Liljestrand, senior lecturerb,
  3. Morten Hedegaard, registrara,
  4. Shakuntala Haraksingh Thilsted, associate professorc,
  5. Abraham Joseph, professord
  1. a Perinatal Epidemiological Research Unit, Department of Obstetrics and Gynaecology, Aarhus University Hospital, 8200 Aarhus N, Denmark
  2. b Baltic International School of Health, 37185 Karlskrona, Sweden
  3. c Research Department of Human Nutrition, The Royal Veterinary and Agricultural University, Rolighedsvej 30, 1958 Frederiksberg C, Denmark
  4. d Community Health Department, Christian Medical College, Vellore 632 002, Tamil Nadu, South India
  1. Correspondence to: Dr Nielsen
  • Accepted 11 March 1997


Objectives: To study reproductive pattern and perinatal mortality in rural Tamil Nadu, South India.

Design: Community based, cross sectional questionnaire study of 30 randomly selected areas served by health subcentres.

Setting: Rural parts of Salem District, Tamil Nadu, South India.

Subjects: 1321 women and their offspring delivered in the 6 months before the interview.

Main outcome measures: Number of pregnancies, pregnancy outcome, spacing of pregnancies, sex of offspring, perinatal and neonatal mortality rates.

Results: 41% of the women (535) were primiparous; 7 women (0.5%) were grand multiparous (>6 births). The women had a mean age of 22 years and a mean of 2.3 pregnancies and 1.8 live children. The sex ratio at birth of the index children was 107 boys per 100 girls. The stillbirth rate was 13.5/1000 births, the neonatal mortality rate was 35.3/1000, and the perinatal mortality rate was 42.0/1000. Girls had an excess neonatal mortality (rate ratio 3.42; 95% confidence interval 1.68 to 6.98; this was most pronounced among girls born to multiparous women with no living sons (rate ratio 15.48 (2.04 to 177.73) v 1.87 (0.63 to 5.58) in multiparous women with at least one son alive).

Conclusions: In this rural part of Tamil Nadu, women had a controlled reproductive pattern. The excess neonatal mortality among girls constitutes about one third of the perinatal mortality rate. It seems to be linked to a preference for sons and should therefore be addressed through a holistic societal approach rather than through specific healthcare measures.

Key messages

  • In this area of Tamil Nadu, women have a controlled reproductive pattern, with an average of 2.3 pregnancies and 1.8 living children; the average age at the birth of the first child is 20

  • The sex ratio at birth is 107 boys to 100 girls

  • Girls have a substantial excess neonatal mortality, which is most pronounced among girls born to multiparous women with no living sons

  • This excess neonatal mortality among girls seems to be linked to a preference for sons and must be addressed through a holistic societal approach


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.


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.

Table 1

Characteristics of 1320 newly delivered women in rural Tamil Nadu, 1995

View this table:

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).

Table 2

Sex distribution of liveborn and dead children (singleton pregnancies), Tamil Nadu, 1995. Values are numbers (percentages) unless specified otherwise

View this table:

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).

Table 3

Sex distribution of liveborn singleton children who died within one month born to primiparous and multiparous women, Tamil Nadu, 1995 (singleton pregnancies)

View this table:
Table 4

Sex distribution of liveborn children and children who died within one month born to multiparous women with and without living sons, Tamil Nadu, 1995 (singleton pregnancies)

View this table:

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).


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.

Sources of possible bias

Women who had a dead child may be more reluctant to be interviewed than women with a living child. This tendency may be even more pronounced among women who had a dead girl because of the current focus in the Indian media and among government officials on missing girls. But the refusal rate was low (0.3%), and the field workers and their supervisors were thorough in identifying and including all women who had just given birth, so we estimated that only a minimum of bias existed here.

Some of the eligible women (5%) were not found at home even after two visits. Some of those women may have had a dead child. Usually women who have just given birth stay indoors for several months with their infant because they fear of meeting an evil spirit, with fatal consequences for their child. Missing eligible women may thus introduce some bias and lead to an underestimation of actual mortality.

Demographic factors

In this study, few mothers were adolescents. In Tamil Nadu, the legal age of marriage–that is, of the onset of sexual activity–is 18 years. In all India, and in neighbouring countries like China and Sri Lanka, an increasing trend in age at the time of marriage has been observed.6 7 8 9 10 11 The results of this study are consistent with findings from other studies that age at marriage and thus age at first birth have increased in the past few decades.

We found a perinatal mortality rate of 42.0/1000. This is not statistically different from the official local rate of 47.5 (0.67 to 1.15).

Other studies have found that in countries with a high preference for sons, spacing of pregnancies is related to the sex and vital status of the previous children12; after the birth of a girl there is a shorter interval before the next birth. This was not found in our study.

The sex ratio at birth was as expected in a country with absence of social and behavioural interference in the sex of offspring,13 14 indicating that these women did not practise sex selective abortion or underreport the birth of daughters. In China, which also has a strong preference for sons, the proportion of boys born is higher than expected13 14 15; in 1990 it was 111.75.15

Neonatal mortality

Girls had a significantly higher risk of neonatal mortality than boys, and this was even more pronounced among girls born to multiparous women without living sons. Without this excess mortality the perinatal mortality rate would be reduced by one third and would then reflect the distribution of stillbirths and early neonatal deaths which is seen in other countries–that is, half of the perinatal deaths occuring in utero and half occuring after birth.

In developing countries, neonatal mortality is largely due to birth asphyxia, pneumonia, tetanus, congenital anomalies, birth injuries, and prematurity. It is difficult to assess the contribution of hypothermia and delayed onset of breast feeding to early neonatal death.16 Boys generally have a higher risk of neonatal mortality, possibly because of a complex combination of genetic and environmental factors.17 18 Postneonatal deaths, on the other hand, are often caused by infectious diseases, the incidence and severity of which are affected by immunisation, health care, and other factors such as nutritional status. A difference in allocation of food and health care to boys and girls will thus change the usual pattern of postneonatal mortality. An excess of deaths among girls has been found in countries with a strong preference for sons and is most evident at ages 1-5 years, when a child's health and survival depend heavily on parental care.19 20 These behavioural patterns may be a result of economic and cultural forces, as economic pressure forces parents to distribute limited resources such as food and medical care to offspring selectively.15 21 22 23 24 25

The substantially increased risk of neonatal death among girls born to multiparous women without living sons indicates that other causes of death than those mentioned earlier should be considered. Some communities in India and also elsewhere have practised infanticide for many years.16 17 18 19 20 21 22 23 24 25 26 27 28 However, primary data on this are scarce.

The significantly higher risk of neonatal mortality among girls born to multiparous women without living sons indicates that a preventive healthcare system that reflects a biomedical approach to improving perinatal mortality rates among girls would have a limited effect, as such a system is unlikely to focus on factors causing the excess mortality among girls. If perinatal mortality is to decline further, a holistic approach is required to change the complex interplay between available resources and the value of girl children.


We are grateful to the women who shared their experience and time with us, to the field staff who devoted some months to fieldwork and data collection, and to government officials in Salem and in Madras for their interest in this study and their collaboration.

Funding: Grants from Danish Council for Development Research (No 104.Dan8/635) and Novo Nordisk, Denmark. Danish International Development Assistance supported the government of India's area development programme within the health and family welfare sector from 1981 to 1994 in the data collection area.

Conflict of interest: None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
View Abstract