Demographic variables in fetal and child mortality: Hmong in Thailand

Sm. Sci. Med. Vol. 36, No. 9, pp. 1109-1120, 1993 Printed in Great Britain. All rights reserved

0277-9536193 $6.00 + 0.00 Copyright Q 1993 Pergamon Press Ltd

DEMOGRAPHIC VARIABLES IN FETAL AND CHILD MORTALITY: HMONG IN THAILAND*

PETER KUNSTADTER,’ SALLY LENNINCTON KUNSTADTER,’ CHAI PODHISITA~ and PRASIT LEEPREECHA~

‘Institute for Health Policy Studies, University of California, San Francisco, CA 94143, U.S.A., 21nstitute for Population and Social Research, Mahidol University, Salaya, Nakhon Pathom 73170, Thailand and

‘Institute for Social Research, Chiang Mai University, Chiang Mai 50002, Thailand

Abstract-Conventional theories would not predict the 60% decline of infant mortality which has occurred among the Hmong population of Thailand, from 123/1000 in the mid-1960s to 48 in the mid-1980s. The Hmong population in northern Thailand has sustained high fertility and low use of modern health services. Most Hmong live in relatively remote rural villages and earn their living by self-employed farming. They have low levels of education, especially for women. They live in multi-gen- erational patrilineal-patrilocal extended family households. Women’s status is low. These characteristics contrast strongly with the majority ethnic Thai population, among whom a comparable mortality decline has been accompanied by widespread use of family planning, rapidly declining fertility, widespread use of modern health facilities, rapidly increasing levels of education for both sexes, rapid economic development, and a predominance of nuclear-based family households.

Distributions of Hmong pregnallcies by birth order and maternal age have remained relatively constant while fetal and young child death rates have declined for each level of parity and all maternal ages in recent cohorts. As predicted by conventional theories, infant mortality rates are highest among higher order births and for births to mothers of the highest ages, however there is relatively little effect on risk of infant mortality of first order pregnancies, or births to very young (10-14 year old) women. Fetal and infant mortality have declined steadily in recent cohorts at each parity level and all maternal ages.

Modern medical care and decline in a surplus of female deaths associated with low status of women might explain the declines in fetal and child deaths regardless of parity or maternal age. Use of modern medical care for delivery is recent and accounts for less than 10% of all recent Hmong births, but survival rates are not consistently or significantly higher for children born with a modern birth attendant. Sex-specific mortality rates calculated from reproductive histories show no surplus of female deaths in the past, but females have benefitted more from recent mortality declines than males.

Ethnographic evidence suggests that Hmong have customs which act to protect the health of mother and child (‘chicken soup theory’), and that they are predisposed to accept innovations (including use of modern medicine) which they see as beneficial. This may allow them to respond especially quickly to small opportunities for improving their children’s survival, as compared with other ethnic groups.

Key wordsAemographic factors, child survival, birth outcome, Hmong

INTRODUCITON-RISK FACTORS FOR BIRTH OUTCOME AND CHILD SURVIVAL

Correlational studies have suggested a large variety of factors may be associated with birth outcome (fetal death vs livebirth) and child survival, including demographic variables (e.g. parity, maternal age, child’s sex), physiological factors (e.g. birthweight), child care (e.g. breastfeeding, nutrition), health and medical factors (e.g. antenatal care, birth attendant, immunization), and socioeconomic factors (e.g. maternal or parental education, status of women) [l-12]. All these factors have been topics of interven- tion policies designed to improve child survival.

Family planning efforts in Thailand have resulted in widespread use of effective family planning methods in the predominantly ethnic Thai population since the early 1960s. About two-thirds of all eligible couples are current users of modern contraception or

*Revised version of paper prepared for Association for Asian Studies meeting, 5-8 April 1990, Chicago.

sterilization. National surveys show that total fertility rate (TFR) has fallen from over 6 to about 2.4, and simultaneously, Thailand has enjoyed a decline in infant mortality from about 100 to the low 40s (13, 141. Advocates of family planning suggest that declines in vital rates are not merely statistically correlated, but that they are causally linked and that the observed correlations offer guides to policy. Thus family planning and reduced fertility are proclaimed as keys to lowering infant mortality and increasing child survival. Advocates of health care stress the importance of safe birth practices and immunizations as essential to lowering fetal and infant mortality rates. At the same time as fertility has been reduced, modern medical care has become widely available and widely used among the ethnic Thai population, to the extent that as many as 75% of all births in predominantly rural Chiang Mai Province take place in hospital. The increased use of modern birth attendants may also be a cause of reduced mortality. Other authors have pointed to the strong association

1109

1110 PFTER KUNSTADTER et al.

between measures of socioeconomic development (especially education) and decline in death rates. The rapid recent increase in average income, and widespread education and literacy rates approaching 90”/0 in Thailand suggest correlations of socioeconomic development with improved birth outcome and child survival. Intercorrelations between fertility, health care, and socioeconomic variables make it difficult to test these hypotheses in most circumstances.

This paper briefly describes the health care and socioeconomic situation of one of Thailand’s ‘hilltribe’ minority groups, the Hmong, and considers in detail evidence relating to the widespread conventional wisdom concerning the relationship of fertility patterns with fetal and young child deaths. Because fertility patterns fail to explain the observed changes in mortality we also consider status of women. health care variables and socioeconomic changes.

BACKGROUND-RECENT HISTORY OF HMONG IN THAILAND

There are about 550,000 ‘hilltribe’ people in Thailand, roughly 1% of the total population. Thai- land’s highland minority groups have participated unevenly in recent demographic and socioeconomic developments. Most still live in relatively remote locations, and depend largely on slash and burn agriculture for their livelihood. In comparison with the majority ethnic Thai lowland population. they have been reached more recently and often less intensively by roads, schools, health stations or public health and family planning programs. Gener- ally their income is far below that of the ethnic Thais.

Hilltribes, especially the Hmong, have been ex- posed to considerable economic and physical stress in recent years. Many Hmong were involved in insur- gencies of the 1960s and 1970s with great disruption of normal life as they left their villages to flee from, or join with, the communists. About half the communities in our survey were more or less forcibly relocated by the government during this period, in an attempt to prevent or control insurgency and to increase access to government services. Some relocated villages are in accessible locations with ample land of good quality, roads, markets and other facilities. Other communities were relocated into un- favorable environments with insufficient land of good quality, poor transportation, few infrastructural services, and, in some places, heavy exposure to malaria.

In the past, because of income from opium, many Hmong were relatively wealthy compared with other highlanders, and were better off economically than some rural ethnic Thai. Since the early 1980s Hmong have been under especially great economic pressure as a result of rapid population growth, combined with government enforcement of regulations against opium poppy cultivation (their chief traditional source of cash income), against the cutting of forested

areas for new swidden fields (their major traditional method of acquiring and using land), and against spontaneous relocation to new sites (their traditional method of balancing population and land resources). Government policies removed the bases of the traditional Hmong economy. and opium crop re- placement schemes rarely provided adequate income substitution [l5-171.

There is considerable variation in birth and death rates between majority and minority ethnic groups and among the minority groups in Thailand. The Hmong have been found to have among the highest fertility and the lowest death rates of the hilltribes. Death rates have declined slowly among some groups (e.g. Karen and Lua’) and rapidly among others (especially the Hmong) [18-221. Studies of risk factors such as tobacco smoking, alcohol consumption and postpartum feeding of women suggests why Hmong infant mortality may be low in comparison with other ethnic groups [23], but this does not indicate why Hmong death rates have fallen rapidly in the past two or three decades.

The data reported here allow examination of the circumstances under which mortality declined among the Hmong since the early 1960s as related to birth order, maternal age, and sex of child as risk factors for high infant mortality. The Hmong data do not support the idea that fertility or ‘high risk’ births must be reduced or that female status must be raised, or that a high proportion of births must be attended by modern practitioners in order to achieve a major, sustained, rapid decline in infant mortality.

MATERIALS AND METHODS

Estimates of fetal death and child death or survival rates are derived from reproductive histories collected in a household survey 1988 and 1989. Communities were selected for the household survey following a survey of 198 Hmong communities in IO provinces in 1987788 (over 90% of all Hmong communities in Thailand which were not under military adminis- tration at the time of the survey). This survey revealed the great range of conditions in which non-refugee Hmong were living, including environments ranging from remote villages high in the mountains (over 1200 m) to urban sites in lowland valleys. Some Hmong communities had no infrastructural development, while others had a full range of administrative, educational, transportation and health facilities. Hmong are predominantly an agricultural people, and community level economic conditions ranged from relatively aflluent to poor, depending on access to land, roads and markets. Local economic systems ranged from bare subsistence in land-poor communities to traditional (and profitable) dependence on opium as a cash crop and rice as a subsistence crop with ample land, to more or less successful cultivation of a range of modern market crops (maize, cabbages, ginger, fruits: etc.) using modern farm technology

Demographic variables in fetal and child mortality 1111

(pesticides, sprinkler irrigation, tractors). Most urban household heads are civil servants or merchants.

For logistical reasons we limited our household survey to 26 communities in five provinces. Selection of communities for the household survey was deter- mined by the primary interest of the research project in economic variation. The household survey communities represent the range of economic, environ- mental, infrastructural and community history characteristics described above, and allow examination of variability within different types of community. Because we have not applied weights to the data from each community proportional to the represen- tation of the sample communities, the statistical results in this paper apply only to the study population. However, because the study population is large and covers a range of community variation, and because the demographic results are similar to an earlier study of randomly selected communities [21], we believe that the results are generally applicable to the Hmong population of Thailand.

The household survey covered 2037 women living in 1014 households in 26 different communities located in five northern Thai provinces. Except for the largest communities, in which households were selected at random from a sampling list constructed during the community level survey, all currently resident ever-married women of all ages were inter- viewed in every household in the community, along with any never-married women who were reported to have ever been pregnant. Interviews were conducted in Hmong and recorded in Thai by native speakers of Hmong with highschool or more education, who were fluent in Thai. Reproductive histories were collected by women interviewers. The authors checked questionnaire forms in the field for completeness and logical consistency, and returned them for reinterview if necessary.

Reproductive history data include woman’s date of birth and current age, pregnancy number, date of the end of each pregnancy, birth outcome (fetal death or livebirth), vital status of child at time of census, and date of death if the child died before census. Data were coded and checked by a subteam of the interviewers, two of whom specialized in coding the reproductive history data. Coded data were entered twice by two people, and the two files were compared by computer for consistency. Data from inconsist- ently entered records were checked against the code sheets. Frequency distributions for each variable were checked for out-of-range codes and cross tabulations were run to check for logical inconsistencies. Appar- ent errors were corrected by reference to original interviews. Data referred to in this paper were based on the cleaned data files.

This survey yielded reports of 9397 liveborn children and 349 fetal deaths through 1986, excluding 4 induced abortions, 5 children whose vital status at birth was unknown, and 1 whose survival to time of survey was unknown. Data on fetal deaths were

tabulated for all pregnancies completed prior to survey, but 416 births in 1987, 181 in 1988, and 1 in 1989 were excluded from tabulations of infant and child mortality because of the short interval between birth and interview.

Reproductive history data were collected from women of all ages at time of survey, regardless of when their births occurred. Age of mother at the end of each pregnancy was computed from the mother’s age at time of census. Data were grouped for analysis by date of end of pregnancy (< 1956, 1957-1961, 1962-66, 1967-71, 1972-76, 1977-81, 1982-86).

Infant mortality rates (IMRs) from before 1957 (108.9) and from 1957761 (111.7) are lower than for 1962266 (123.0); likewise there is a much higher proportion of first, second and third order births (implying lower fertility) in the data from before 1957 than in later cohorts. The apparent increases in fertility and mortality might be due to loss of memory of pregnancies that ended in fetal or early child death in the distant past, or due to a real peak in fertility and mortality in the 1962266 period, followed by a steady decline. The apparent change in proportion of first, second and third order births is probably a result of the fact that older women (who would have been having higher order births during the earlier period) were more likely to have died in the years after 1961 than were the younger mothers who were having the lower order births. We tabulated the data from the earliest cohorts for completeness, but they have not been used for most of our analysis of mortality decline.

RESULTS--BIRTH OUTCOME AND CHILD SURVIVAL BY COHORT

Hmong IMR declined steadily, by about 60% from about 123/1000 in 1962-66 to 48jlOOO among those born in 1982-1986. Proportion of liveborn children surviving to age 5 increased from 0.822 to 0.893 between 196266 and 1977-81, a 40% decline in mortality for this age group. Fetal mortality rates (FMR) declined fairly steadily from a high of 45.8 in 1962-66 to a low of 33.5 in 1982-86 (Table ]A). The fact that both FMRs and IMRs declined simultaneously suggests that the apparent improvement in child survival is not due to a systematic change in reporting of fetal deaths as postnatal deaths in the earlier cohorts. Instead there appears to have been a real decline in mortality in recent years, before, during, and after birth.

Hmong fertility appears to have declined every year since 1982, but the average Hmong total fertility rate (TFR) for live births in 1982-86 was 7.4, three times higher than the TFR of 2.4 reported for the general Thai population. In rural areas Hmong use of contraception remains low (20-30% of eligible couples), and appears to be influenced primarily by perceived resource shortages. Among the 38 urban Hmong households in our study population use of

1112 PETER KUNSTADTER et al.

Table l(A). Hmong fetal mortality, infant mortality and child survival to exact ages, by year of child’s birth

Year of birth

Q 1956 1957-61 1962.66 1967-71 1972-76 1977-81 1982-86

Liveborn (N) 1038 573 764 1037 1369 1827 2212 Fetal mortality (%) 34.7 34.9 45.8 37.6 40.2 40.0 33.5 Infant mortality (%) 108.9 III.7 123.0 100.3 92.0 75.5 47.9 Proportion surviuing

at least /o age:

8 days 0.9759 0.9738 0.9607 0.9643 0.9737 0.9748 0.9837 I month 0.9605 0.951 I 0.9463 0.9547 0.9591 0.9606 0.9792 I year 0.891 I 0.8883 0.8770 0.8997 0.9080 0.9206 0.9521 2 years 0.8564 0.8621 0.8482 0.8756 0.8919 0.91 I9 3 years 0.8256 0.8412 0.8390 0.8602 0.8795 0.8993 4 years 0.8141 0.8290 0.8259 0.8476 0.8714 0.8955 5 years 0.8092 0.8255 0.8220 0.8409 0.8612 0.8927 _

Table I(B). Hmong livebirths, fetal mortahty and infant mortality by child’s sex and year of child’s birth

Year of birth

< 1956 1957-61 1962-66 1967-71 1972-76 1977-81 1982 86

Liueborn

M (?&) F (%) Total N

Feral mortalif~

M (s) F 0% Total (o/m)

Infant morralif,t

M WI

F (%) Toval (%)

51.5 51.3 51 3 50.6 52.5 49.4 51.5 48.5 48.7 48.7 49.4 47.5 50.6 48.5 1038 573 764 1037 1369 I x27 2212

24.3 20.4 25.5 26.7 22.3 26.6 14.0 21.8 17.9 29.6 19.5 20.0 9.7 20.5 34.7 34.9 45.x 37.6 40.2 40.0 33.5

129.0 122.4 125.0 104.8 107.1 88.7 56.2 87.3 100.4 121.0 94.7 75.4 62.7 39 I

108.9 III.7 123.0 100.3 92.0 75.5 47.9

Table l(C). Hmong child survival to exact ages. by child’s sex and year of child’s birth

Proportion surviving Year of birth at least to age < 1956 1957-61 1962-66 1967-71 1972-76 1977-81 1982X6

1 month M 0.9514 0.9490 0.9515 0.9505 0.9527 0.9545 0.9737 F 0.9702 0.9534 0.9409 0.9590 0.9662 0.9665 0.9851

Total 0.9605 0.951 I 0.9463 0.9547 0.9591 0.9606 0.9792 I year M 0.8710 0.8776 0.8750 0.8952 0.8929 0.91 I3 0.9438

F 0.9127 0.8996 0.8790 0.9043 0.9246 0.9373 0.9609 Total 0.891 I 0.8883 0.8770 0.8997 0.9080 0.9206 OYSZI

2 years M 0.8336 0.8605 0.8546 0.8743 0.8789 0.8969 F 0.8810 0.8638 0.8414 0.8770 0.9062 0.9265

Total 0.8564 0.8621 0.8482 0.8756 0.8919 0.91 I9 _

3 years M 0.8056 0.8435 0.8495 0.8610 0.8679 0.8880 F 0.8472 0.8387 0.8280 0.8594 0.8923 0.9103

Total 0.8256 0.8412 0.8390 0.8602 0.8795 0.8993 4 years M 0.7888 0.8367 0.8240 0.8476 0.8609 0.8836

F 0.8413 0.8208 0.8280 0.8477 0.8831 0.9070 Total 0.8141 0.8290 0.8259 0.8476 0.8714 0.8955

5 years M 0.7850 0.8299 0.8214 0.8400 0.8526 0.8814 F 0.8353 0.8208 0.8226 0.8418 0.8708 0.9038

Total 0.8092 0.8255 0.8220 0.8409 0.8612 0.8927

contraception or sterilization is very high (84%) and is related to perceived costs of children and inter- ference with work [24].

Gender bias, infant mortality and child survival

Male dominance in Hmong society is seen in traditional, strongly adhered to marriage customs such as patrilocal postmarital residence, polygyny, ease of divorce for men and difficulty of divorce for women, and occasional bride capture, although it is now frowned upon by most Hmong. Other indi- cations of continued male dominance include par-

ental reliance on sons for old age support (a major reason for stated preference for sons vs daughters). ancestor worship by males and through the male line, plus beliefs such as that a woman’s childbearing will be made more difficult if she talks back to her husband or his relatives, and that she cannot be reincarnated as a man until she has had all the children her fate requires. Many Hmong parents now say that sons and daughters should be treated equally, but low status of Hmong women is also seen in the current practice of withholding daughters from school. In our sample girls age 5-19 who had no


schooling outnumbered boys 3 to 1, because many parents believe that daughters will contribute nothing to their old age support. This is in strong contrast with the pattern in Thai society where access to education at all levels is virtually the same for males and females.

Given the male bias of Hmong society we might expect higher female death rates at all ages, as have been reported in other populations with strong gender preference based on religion and household economic structure [25,26]. If the traditional Hmong preference for sons were associated with a strong sex difference in mortality in the past, and if there were a systematic decline in ‘excess’ deaths of females, this could have produced a general decline in infant mortality rates at all parities and maternal ages. There have been fluctuations in sex ratio at birth and fetal mortality rates (Table lB), but no consistent trend which might suggest either a change in differential reporting of births by sex, or that there was ever a surplus of female over male deaths. In fact, female IMR is lower than male in every cohort (averaging about 20 per 1000 livebirths lower), and the decline in female IMR has reached a low point of 39.1 in 1982-86 vs 56.2 for males in the same cohort. When we examine child survival by sex we find that female survival at ages 1 month to 5 years is higher than male for most ages and cohorts (Table 1C). The distribution of the population by age and sex shows no consistent ‘deficit’ of females any time in the recent past. In sum, female mortality has been consistently lower than male, and females appear to have benefitted more than males from the recent improvement in child survival. This result raises questions about conventional definitions of gender bias in relation to sex differences in child survival, and implies the importance of looking at mechanisms which may be involved [27]. One possible explanation is that Hmong view daughters as economic assets because of the work they contribute to the household before they marry, and because of the brideprice paid for them when they leave home.

Birth order, birth outcome and infant mortality

FMRs and IMRs are generally thought to follow a U-shaped curve in relation to birth order. First births and high birth orders are usually reported to have higher (statistical) risk for poor outcome and survival than intermediate birth orders, due to physiological factors. Thus it has been argued that reducing fertility, thereby reducing the number and proportion of highest order births, will reduce poor fetal outcome and infant deaths.

Tabulations show no consistent pattern of change in proportions of Hmong pregnancies by birth order. In cohorts born in 1962 and thereafter, first order births account for 16.9-20.3% of all births, second order births account for 16.1-17.7% of births, third order for 13.1-14.8%, fourth order from 1 l&12.4%, and ninth order and higher for 569.7% (Table 2). The proportion of lower order births has not increased, nor has the proportion of high order, high risk births declined. These data indicate that decline in Hmong infant and fetal deaths cannot be explained by a decline in proportion of the ‘risky’ high order births.

The relatively high risk of fetal death for first order births is clear only in cohorts born since 1977; very high order births show fairly consistent, relatively high death rates which have declined substantially since 1967 (Table 3A). The absence of high rates for lowest and highest order births prior to these dates might be due to selective memory with regard to fetal deaths, but there is no evidence for such a trend in the recent data. If there has been selective memory loss, and thus there were more unreported fetal deaths in the past than in recent cohorts, fetal death rates would actually have declined more rapidly in recent years than is shown in the tabulations.

The higher infant death rates for first vs second order births which has been reported for other populations appears in all but one Hmong cohort. IMR for first order births from 1966 to 1986 averaged 94.9/1000 vs 74.8 for second order births. IMRs

Table 2. Cumulative percentage of Hmong births by pregnancy number and birth year

Year of birth

Pregnancy <I956 1957-61 1962-66 1967-71 1972-76 1977-81 1982-86 All years number Cum % Cum % Cum % Cum % Cum % Cum % Cum % Cum %

1 26.3 20.4 16.9 20.3 19.3 19.3 16.9 19.5 2 48.1 36.8 33.0 36. I 37.0 35.9 33.0 36.6 3 65.2 49.9 47.8 49.4 51.4 50.0 47.8 51.2 4 76.7 65.1 58.8 61.0 62.7 62.8 60.2 63.4 5 85.2 77.5 69.4 70.8 72.4 73.8 70.3 73.5 6 90.7 86.6 79.2 78.8 80.9 82.0 78.6 81.7 7 94.4 90.6 88.7 85.0 86.9 88.1 86. I 88.0 8 97.1 93.7 94.4 90.4 91.3 92.7 91.8 92.7 9 98.8 95.8 96.5 95.4 94.5 95.6 94.9 95.7

IO 99.7 97.6 98.3 98.0 97.4 97.3 97.1 97.7 II 100.0 98.6 98.8 99.3 98.8 98.6 98.6 98.9 I2 99.7 99.3 99.6 99.6 99.5 99.3 99.5 13 100.0 99.6 99.9 99.9 99.7 99.6 99.8 14 100.0 100.0 100.0 100.0 99.8 99.9 15 99.9 100.0 16 100.0 100.0 Total N 1038 573 764 1037 1369 1826 2212 8819


Table 3(A). Fetal death rate by pregnancy number and year of birth

Year of birth Pregnancy ~---~~ number < 1956 1957-61 I962 --66 1967-71 1972-76 1977-81 1982-86

I 25.6 42.7 62.0 14.3 37.9 42.6 40.2 2 35.4 10.6 32.5 42.1 24.8 16.4 25.3 3 22.5 66.1 26.5 50.7 50 5 38.9 21.3 4 58.8 34.5 71.4 66. I 30.3 38.6 21.9 5 22.1 28.2 37.0 19.8 37.9 39.8 31.4 6 105.3 0.0 26.7 36. I 8.5 33.3 38.0 7 25.6 8.7 27.4 15.6 49.4 54. I 42.2 8 35.7 5.6 69.X 35.7 65.6 23.8 23.6 9 0.0 0.0 125.0 76.9 113.6 37.0 101.4

IO 0.0 0.0 71.4 74.1 25.6 96.8 21.3 II 0.0 166.7 250.0 0.0 52.6 41.7 60.6 I2 00 0.0 0.0 181.8 133.3 117.6 I3 00 0.0 0.0 0.0 1000.0 0.0 14 0.0 0.0 0.0 166.7 333.3 15 _~ 0.0 16 _ 0.0 Total 34.7 34.9 45.X 37.6 40.2 40.0 33.5

Table 3(B). Infant mortality rate by pregnancy number and year of birth

Year of birth Pregnancy ~~~~ -~~ ~~~~ -~~~ number < IY56 1957-61 I962Z66 1967-71 I972m 76 1977-81 1982-86

I 150.2 162.4 17x.3 13X.1 106.1 73.9 53.6 2 110.6 106.4 xv.4 I22 0 78.5 x2.2 39.3 3 118.0 x0.0 150.4 115.9 x5.9 X5.6 42.7 4 42.0 80.5 x3 3 X2.6 77.4 77.3 54 7 5 90.9 112.7 49.4 69.3 113.6 79.6 31.4 6 70.2 76.9 160.0 60.2 I IY.7 66.7 5Y.8

1 51.3 0.0 123.3 93.X 61.7 63. I 36. I 8 35.7 166.7 162.X 53.6 115.X 35.1 39.4 Y 166.7 x3.3 62.5 134.6 113.6 III.1 IO1 4

IO 250.0 200.0 71.4 37.0 25.6 64.5 63.X II 250.0 0.0 0.0 0.0 52.6 X3.3 121.2 I2 -. 500.0 250.0 0.0 90.1 0.0 0.0 13 500.0 0.0 0.0 200.0 250.0 0.0 14 _ 333 3 0.0 0.0 0 0 0.0 15 0.0 I6 0.0 Total 108.9 107.9 123.0 100.3 92.0 75.6 47.9

which are consistently higher than the cohort mean for the highest order (9 and higher) births, appear only in cohorts born after 1977. Over time there is a consistent trend to lower risks for all birth orders (Table 3B).

In contrast with IMRs, the association of high risk of fetal mortality in high order pregnancies is consistent for all cohorts, but the consistent pattern of relatively high risk for fetal mortality for first pregnancies has appeared only recently. This comparison of historical patterns of change for IMR and FMR suggests that risk factors for fetal death may not be the same as for infant mortality in this population.

In sum, the Hmong data indicate that substantial reduction in infant and fetal mortality rates can be accomplished without reduction of a high birth rate. As total mortality rates decline, the rate of infant mortality for first and highest order births declined more rapidly than for birth orders conventionally considered to be ‘low risk’.

Maternal age, birth outcome and infant mortality

Very young or very old age of mother is another commonly cited demographic risk factor for fetal

death. Family planning advocates have argued that high risk births to very young and very old women must be reduced in order to lower FMR and IMR. The distribution of Hmong births by maternal age suggests there has been no systematic change since the early 1960s when the decline in infant and fetal mortality began (Table 4).

The proportion of Hmong pregnancies by age of mother has fluctuated for very young mothers, but is lower in recent years than in 1962-66. Proportions of pregnancies which ended when mother’s age was IO-14 have ranged from a maximum of 3.9% in 1962-66 to a minimum of 2.3% in 1967-71; pregnancies of women under age 20 were at a minimum in 1957-61 (19.6%) rose to 27% in 196771971, and declined to 23.1% in 1982286. Proportion of pregnancies ending when maternal age was 50 and above reached a maximum in 1977-81. but accounted for only 0.5% of pregnancies for those years. They were at a minimum (0.1%) for the following cohort; pregnancies of women age 40 and above were at a maximum in 1967-71 (5.5%) and at a minimum in 1957-61 (3.7%) and 1982286 (3.8%). Thus declines in percentages of births to women of ‘high risk’


Table 4. Percentage of pregnancies by maternal age at time of child’s birth, controlled for year of child’s birth

Maternal age at childbirth

Year of birth

c 1956 1957.61 I 962-66 1967-l I 1972-76 1977-81 1982-86

IO-14 3.1 3.2 3.9 2.3 3.2 2.4 2.4 15-19 26.4 16.4 17.9 24.7 22.8 24. I 20.7 20-24 32.4 29.2 22.0 24.3 29.6 29.0 30.8 25-29 20.6 25.5 25.8 17.2 19.4 22.6 20.8 30-34 9.7 14.5 16.6 17.1 10.8 11.1 14.3 35-39 4.8 7.6 a.4 8.8 9.1 6.2 7.2 40-44 2.0 3.4 3.9 3.4 3.9 3.1 2.7 4549 0.7 0.3 1.5 1.7 1.0 I.1 1.0 250 0.4 0.0 0.0 0.4 0.3 0.5 0.1

Total number of pregnancies includes livebirths plus fetal deaths

youngest and oldest ages are associated with the total decline in infant mortality for the most recent cohort, but cannot explain the steady decline over the past 20 years.

FMRs for the youngest mothers fluctuated from 0.0 in 1957-61 to 69 in 1962-66, then declined to 22.7 in 1972-76 and increased to 78.4 in 1982-86 (Table 5A). Some of this variation is due to the small numbers involved (only 3% of all births were to mothers age 10-14). IMRs for children born to mothers age l&l4 declined consistently from 210.5 in 1957-61 to 39.2 in 1982-86 (Table 5B). IMRs for infants born to women age 15-19 also declined steadily from 159.1 in 1962-66 to 60 in 1982-86. FMRs show a steadier downward trend for mothers in this age group than for the youngest mothers, from 83.3 in 1962266 to 34.9 in 1982-86. Decline in IMR

for the teenage mothers has made a major contri- bution to reduction in the total IMR.

DEMOGRAPHIC AND NONDEMOGRAPHIC FACTORS IN DECLINE OF HMONG INFANT MORTALITY

Reproductive history data from a survey of 2037 Hmong ‘hilltribe’ women in Thailand shows a steady decline in infant mortality for all birth orders and maternal ages. Infant mortality declined a total of 60% while there was a 27% fall in fetal mortality from 1962-66 to 1982-86. The reduction in total Hmong IMR is comparable to the IMR decline in the lowland rural ethnic Thai population during the same period. Among the ethnic Thais decline in IMR was accompanied by a rapid drop in birth rate from a TFR from over 6 to about 2.4, associated with

Table 5(A). Fetal mortality rates by maternal age at childbirth, controlling for child’s year of birth

Maternal Year of birth age at child’s Q 1956 1957-61 1962-66 1967-71 1972-76 1977-81 1982-86 birth FMR FMR FMR FMR FMR FMR FMR

l&l4 31.3 0.0 69.0 41.7 22.7 46.5 78.4 15-19 51.9 102.3 83.3 43. I 38.3 43.3 34.9 20-24 32.6 5.8 53.9 43.8 39.5 47.5 29.3 25-29 23.1 27.2 30.0 33.5 43.0 31.2 35. I 3&34 29.7 23.8 7.6 33.7 40.5 9.6 21.9 35-39 19.6 46.5 30.8 21.5 40.0 44.6 18.5 4&44 0.0 111.1 148.1 27.8 57.7 92.6 109.1 4549 166.7 0.0 0.0 58.8 76.9 0.0 45.5 >50 0.0 - - 0.0 250.0 500.0 Total 34.7 34.9 45.8 37.6 40.2 40.0 33.5

FMR = fetal deaths per 1000 livebirths.

Table 5(B). Infant death rates by age of mother at time of childbirth, controlling for child’s year of birth

Maternal age at child’s birth

l&l4 15-19 2&24 25-29 30-34 35-39 40-44 4549 250 Total

Q 1956 1957-61 I 962-66 1967-71 1972-76 1977-8 I IMR IMR IMR IMR IMR IMR

62.5 210.5 172.4 166.7 136.4 116.3 159.3 181.8 159.1 152.9 105.4 102.7 103.9 69.8 149.7 107.6 76.5 60.8 78.7 88.4 80.0 83.8 101.9 79.1 69.3 154.8 128.9 61.8 108.1 57.4 78.4 69.8 76.9 53.8 56.0 53.6

142.9 166.7 III.1 83.3 57.7 55.6 333.3 0.0 166.7 0.0 230.8 0.0

0.0 - - 0.0 250.0 108.9 107.9 123.0 100.3 92.0 75.6

Year of birth

1982-86 IMR

39.2 60.0 39.5 43.9 50.0 61.7 72.7 0.0 0.0

47.9

IMR = infant deaths (deaths under age I year) of livebom children per 1000 livebirths

1116 PETER KUNSTADTER et al

widespread use of modern contraception and sterilization, use of modern medical facilities for most childbirths, major increases in average income, and near-universal education and literacy. Among the Hmong, however, use of contraception has remained low, TFR is still high, most births still take place at home attended by relatives, and since the early 1960s the lives of many Hmong villagers have been dis- rupted by insurgency and relocation, and government regulations have interfered with the bases of the traditional economy.

Also, unlike the Thais, Hmong family structure, religious beliefs and current behavior patterns stress male superiority and the inferiority of women, especially young women. Contrary to expectations based on studies of other male-dominated Asian societies, we found no evidence of ‘surplus’ deaths of females. Instead, female child survival was better than male in the past, and has improved more rapidly than male survival in recent years.

Rates of Hmong fetal and infant deaths are associated with maternal age and birth order according to the same general U-shaped pattern found in cross- sectional studies of other populations. However, there has been no consistent reduction of proportions of Hmong births to women of extremely young or old age, or, until very recently, of high parity. Rather than reducing the proportion of high risk pregnancies, the decline in IMR has been accomplished with a drop in rates for all parities and maternal ages, including a reduction in relative risk of the high risk early and late parities. These results show that a major reduction in child mortality can be accomplished with little reduction in fertility.

Bongaarts disagrees with the idea that family planning will reduce mortality by reducing high order births [28], and has thereby stimulated an important controversy [29, 301. In an earlier analysis of Thai national census data we found little evidence to support the idea that high infant mortality was ‘caused’ by high birth orders [31]. The Hmong data also appear to support Bongaarts position in that the proportion of births of ninth and higher birth order (where the ‘risky’ effect on outcome is clear in recent cohorts) is only 7.8% of all births in 1977-1986. These births accounted for only 10.2% of the infant deaths during this period, so that even eliminating all births of order 9 and above would have little effect on the average IMR for the period, especially because of the high IMR for first order births.

Bongaarts’ argument can be extended with respect to risk of fetal deaths which also generally follow a U-shaped curve in association with parity, as shown in data from 1962 and later. Mean FMR during this period is 38.3; the rate for first pregnancies is 38.4, for second pregnancies is 26.1, for third through eighth pregnancy is 36.2, and for ninth and higher (8% of all pregnancies) is 78.3 (16.3% of all fetal deaths). If births had been limited to order eight or lower, total FMR would be reduced by 3.5 percentage points to

34.8, and fetal deaths for first pregnancies would be a more important contributor to the total rate. This comparison shows that for the Hmong population, us FMRs decline and a greater proportion of pregnancies results in livebirths (as is the case in the recent cohorts as compared with the earlier ones), the role of first births in determining the total FMR increases. Thus if births were truncated at some birth order where risk increases radically, it would not have a major effect on FMR unless the risk for first pregnancies was simultaneously reduced enough to compensate for the increase in proportion of higher risk first order pregnancies.

Geronimus argues that in the U.S. the apparent relationship between very young age at childbirth and high risk has emerged from use of overly-aggregated data without considering socioeconomic conditions and social support for teenage mothers [32]. Our data tend to support her contention because we have shown that Hmong FMR and IMR have fallen with very little change in proportion of births to very young mothers (10-14 or l&19), and that the rates for mothers in these age groups have fallen at least as fast as the decline for other maternal ages. Despite the apparent psychological stress on Hmong mothers (e.g. due to the belief that their childbirths will be made difficult by talking out of turn to their in-laws), this may be because of the traditionally strong social and material support for the youngest mothers: children are strongly desired by almost all Hmong parents, who look at sons as their future source of support and daughters as helpmates in domestic tasks and sources of brideprice for their sons [24]. There is an average of more than 9 persons and more than 2 ever-married women per household in our current study population. An earlier study showed over 75% of the population was living in extended family households [33]. Because of the ubiquity of extended family households, and short generation span, it is very likely that a Hmong woman will bear her first child and her highest order children in a household with one or more experienced mothers who can assist and advise with regard to pregnancy, childbirth and childcare.

Health care, hygiene, cultural factors and decline in infant mortality

If the steady decline in IMR since 1966 is not the result of a decrease in proportions of high risk first or high order pregnancies, nor is it due to a change in the pattern of maternal age at childbirth, what might be the explanation? Whatever has caused the reduction in risk has operated over all birth orders and all maternal ages. We have already shown although women’s status remains low, data on child survival by sex indicates no historical bias against survival of female children. Another factor which might affect all maternal ages and parities more or less simultaneously is use of modern medical care (especially antenatal care, births attended by modern

Demographic variables in fetal and child mortality

Table 6. Birth attendant’ and child survival to exact age by year of child’s birth

Survive to Survive to Survive to 7 Days 28 Days I Year

Birth Attendant Modem Other Modem Other Modem Other

Born 1977-8 I Die 1 45 I 71 2 136 Survive 73 1705 73 1679 12 1614 Ppn. Survive 0.986 0.974 0.986 0.959 0.973 0.922 X2 0.077 0.750 I.933 P 0.782 0.386 0.164 Odds ratio 0.519 0.324 0.330 Born 1982-86 Die 2 34 5 41 6 100 Survive 176 1997 173 1990 172 1931 Ppm. Survive 0.989 0.983 0.972 0.980 0.966 0.951 X2 0.061 0.189 0.557 P 0.804 0.664 0.455 Odds ratio 0.667 1.403 0.674

*Birth attendant, modem: physician, nurse, trained midwife; birth attendant, other: husband, mother-in-law, other relative, traditional midwife.

1117

practitioners or conducted in sterile environments). These types of modern health care would be expected to reduce risk for fetal deaths and deaths at the time of, or shortly after birth for all maternal ages and parities.

Modern antenatal care was given for only 10.9% of the Hmong pregnancies in the 5 years prior to survey. The first Hmong live birth attended by a modern practitioner (doctor, nurse or trained midwife) was reported by a woman in our surveyed population in the 1967-71 cohort. The proportion of births attended by modern practitioner was 0.5% in that cohort, and increased to 1.8% in 1972-76, 3.8% in 1977-81 and to 8.0% in 1982-86. This increase paralleled, but followed, the decline in Hmong child death rates. Proportions of births attended by modern practitioner were much lower than reported for the general Thai population at the same times [34]. Likewise, distribution of ‘safe birth’ kits (sterile razor and string for cutting and tying the umbilical cord) is recent and not very widespread. In the recent cohorts, the vast majority of births still took place at home, attended by the husband, mother-in-law or other relatives, and even within the past 5 years the cord is usually cut and tied with locally available materials (e.g. soot-blackened bamboo, cotton thread) which may be unsterile.

Use of modern birth attendants is concentrated among a few Hmong women in a few households, and is highly negatively correlated with travel time to a hospital (linear regression r = -0.445, P = 0.022; modern and traditional birth attendant vs travel time 30 min or less and more than 30 min, Fisher exact test P = 0.0022). These are not births which would be classed as ‘risky’ on the basis of maternal age or parity. They are predominantly births to younger mothers of low parity who live in or close to town. Generally, a higher proportion of children whose birth was attended by a modern practitioner survived to at least exact ages 8 days, 29 days and 1 year, as compared with those whose birth attendant was ‘other’ (husband, mother-in-law, etc.), but the relationship is not completely consistent. A higher

proportion of those born with ‘other’ birth attendants survived to age 29 days in the 1982-86 cohort, and the differences in proportions surviving are not statistically significant. Moreover, the trend for reduced mortality is found among those born with ‘other’ attendants, even when those born with modern attendants are removed from the analysis (Table 6). The recent increase in proportion of births attended by modern practitioners cannot explain the major, consist - ent, widespread decline in Hmong child mortality since 1962, though modern birth attendants may account for a small portion of the decline in the most recent cohorts.

Successful immunization against common childhood illnesses can be expected to reduce late infant and young child mortality. The expanded program of immunization (including BCG, DPT, polio and measles vaccines) has been made available in most Hmong communities in recent years. Only about 10% of Hmong parents with children born in the 5 years prior to survey knew about these immunizations by name, but three-quarters of the children have had one or more immunizations of types unknown to their parents. Widespread immunizations are too recent to account for the steady decline in child mortality at all ages since 1962, but may have contributed to the mortality decline of children born since the early 1980s.

Meanwhile, hygienic conditions remain poor: 80.8% of the houses in the surveyed communities have dirt floors, 80.9% of the households get their water at home from unprotected sources, 97.6% get their water in the fields from unprotected sources, about 55% of the people at home and 97% in the fields defecate in the forest or around the field shelter, rather than using a sanitary pit or water seal latrine (Table 7).

Taken together these results rule out many conventional explanations of mortality decline among the Hmong. We believe several things contributed to this decline. Traditional Hmong behavior may have pre- adapted Hmong for rapid decline in infant and fetal mortality, and amplified the effects of modern medi-


Table 7. Hmong household hygiene

Households facilities and behavior: Percentage of households

Type of floor Packed earth Cement, bamboo, or wood

Source of water-at home Protected sonrced Unprotected source

Source of water---in tields Protected sonrct? Unprotected source

Boil drinking water--at home Always or most times Rarely, never, only if someone is ill

Boil drinking water-in fields Always or most times Rarely, never, only if someone is 111

80.8 I9 2

18.1 80.9

2.4 91.6

37. I 62.9

17.9 82. I

Place of defecation, by age (“Q

Place of defecation--at home Water seal or pit latrine In forest “T around house

Place of defecation-in fields Water seal or pit latrine In forest or around house

44.5 46 5 48.6 55.5 53.5 51.4

2.8 2.9 4.4 91.2 97.1 95.6

“Protected sources: piped from protected source, from protected well, rain water from roof, buy bottled water. Unprotected sources: piped from unprotected source. unprotected well, creek, stream, irrigation ditch.

cal care when it became accessible as roads penetrated the highlands.

Hmong practices which might contribute to successful pregnancies and child survival include near- universal and prolonged breastfeeding, the presence in most households of one or more experienced parents and child caretakers even at the time of first births, the expectation and desirability that women will marry and bear children at an early age, major investment (in the form of brideprice) of the household into which a newlywed young woman moves. One example of the solicitous care given to Hmong mothers is the tradition of ‘staying by the fire’, usually observed for one month following birth. During this period the mother restricts her activities and movements, is semi-isolated from contact with non-household members, and custom requires that her husband give her a chicken, or at least an egg every day in order to assure her recovery and her flow of breastmilk. We might call this the ‘chicken soup theory’ of child survival since the Hmong postpartum practice assures that the woman will have a high protein, high fat diet during recovery from childbirth. Post-partum confinement is common among Chinese and Southeast Asian ethnic groups, but a postpartum diet rich in fat and protein is not found among most Southeast Asian groups. Karen women’s post-partum diet, for example, is confined to rice, salt and peppers. Hmong women, in contrast with many Southeast Asian ethnic groups (but consistent with Chinese traditions), rarely smoke tobacco or drink liquor. Also, in contrast with many Southeast Asian groups, Hmong generally use spoons to eat freshly cooked food, rather than using their hands to eat meals which may have been cooked several hours

previously. These customs would be expected to support good health, but would not account for a recent decline in child mortality [35].

Another difference between Hmong and many other hilltribe groups is their willingness to accept innovations, and their ability to take advantage of those innovations, as has been noted, for example, with respect to agricultural change (e.g. [17, p. 401). Over 37% of the households in our surveyed population have a motorcycle or a pickup truck. Wealthier Hmong used money saved from past opium harvests to purchase pickup trucks, even if they had to park them several kilometers away from their roadless villages. At the anecdotal level Hmong tell us that ordinarily births take place at home, but that if transportation is available, they take women to hospital when they perceive signs of difficulty of delivery. They aIso seek modern medical treatment for sick women and children in town, and buy and administer modern medicines to their sick children.

It is difficult to determine the extent to which these practices have contributed to improvement in birth outcome and child survival in a cross-sectional so- ciodemographic survey, but Hmong themselves agree with this interpretation. In a recent series of 242 extended interviews with Hmong parents in 13 of the 26 household survey communities, 84% of the respondents said they thought that child survival had increased in the past 20 years. Most of their explanations referred to the effectiveness of modern medicine (43% of answers), better access to modern medical care due to availability of roads and vehicles and health stations in many communities (16%) and to modern preventive medicine (11%). Respondents in general believe that modern medicine is faster, more effective and cheaper than traditional medicine or spirit worship [36].

Roads and vehicles also improve access to markets. This has probably led to a decline in seasonal fluctuations in food supply which are common in isolated rural areas, and has contributed to improved nutrition. For example, protein can be bought and eaten more frequently than in the past, when people usually waited for a ceremonial occasion to slaughter an animal.

Hmong infant mortality in Thailand declined in parallel with the fall in mortality among the majority ethnic Thai population. In contrast with conventional wisdom, and with the experience of the majority Thai, the Hmong have sustained high fertility, strong bias in favor of males, and low use of antenatal care or modern health care for childbirth. The Hmong mortality transition does not support many common hypotheses linking those variables to fetal and infant mortality. The Hmong data show that under some circumstances it is not necessary to wait for basic social changes (e.g. improved status of women, improved education), or for major reductions in ‘high risk’ births (e.g. to very young, very old or high parity mothers), improvements in local hygiene, or wide-


spread use of modern medical personnel for births, in order to achieve a major sustained reduction in infant and fetal mortality, if effective modern curative and preventive medicine is accessible and if modern medicine is seen by potential users as beneficial. This is not a suggestion to abandon programs to improve women’s status, increase education, or reduce fertility. Such programs can be adequately justified on their own merits, without implying they are necessary or sufficient to decrease infant mortality.

Acknowledgements-Work supported primarily by National Institute of Child Health and Human Development Grant ROlHD22686, and by the Pacific Rim Program of the University of California. Assistance of the Research Insti- tute for Health Sciences, Chiang Mai University, and especially the help of Antika Tansuhaj and Chamnong Kingkeow, are gratefully acknowledged.

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14.

REFERENCES

Caldwell J. C. Education as a factor in mortality decline: an examination of Nigerian data. Pop. Stud. 33, 395413, 1979. Caldwell J. C. Influence of maternal education on infant and child mortality: levels and causes. Hlfh Policy Educ. 2, 251-267, 1982. Caldwell J., Gajanayake I., Caldwell P. and Peiris I. Sensitization to illness and the risk of death: an explanation for Sri Lanka’s approach to good health for all. Sot. Sci. Med. 28, 365-379, 1989. Cramer J. C. Social factors and infant mortality: iden- tifying high-risk groups and proximate causes. Demo- graphy 24, 2999322, 1987. Frisancho A. R., Matos J. and Bollentino L. A. Influ- ence of growth status and placental function on birth weight of infants born to young still-growing teenagers. Am. J. clin. Nurr. 40, 801-807, 1984. Hobcroft J. N., McDonald J. W. and Rutstein S. 0. Socioeconomic factors in infant and early child mortality: a cross national comparison. Pop. Stud. 38, 193-223, 1984. Hobcroft J. N., McDonald J. W. and Rutstein S. 0. Demographic determinants of infant and early child mortality: a comparative analysis. Pop. Stud. 39, 363-386, 1985. Kessel E., Sastrawinata S. and Mumford S. D. Corre- lates of fetal growth and survival. Acta paediat. stand. Suppl. 319, 120-127, 1985. Mosley W. H. and Chen L. C. An analytical framework for the study of child survival in developing countries. Pop. Dev. Rev. 10, Suppl., 1984. Palloni A. and Tienda M. The effects of breastfeedina and pace of childbearing on mortality at early ages. Demography 23, 31-52, 1986. Pebley A. R. and Stupp P. W. Reproductive patterns and child mortality in Guatemala. Demography 24, 43360, 1987. Roman E. Fetal loss rates and their relation to pregnancy order. J. Epidem. Community Hlth 38, 29-35, 1984. Chayovan N., Kamnuansilpa P. and Knodel J. Thailand Demographic and Health Survey. Chulalongkom Uni- versity, Institute of Population Studies, Bangkok and Institute for Resource Development/West%rghouse, Columbia, MD, 1988. Knodel J., Chamratrithirong A. and Debavalaya N. Thailand’s Reproductive Revolution: Rapid Fertility Decline in a Third World Setting. University of Wiscon- sin Press, Madison, 1987.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

Lee G. Y. The effects of development measures on the socio-economy of the white Hmong. Ph.D. thesis, De- partment of Anthropology, University of Sydney, 198 1. Cooper R. W. Resource Scarcity and the H’mong Response: A Study of Settlemem and Economy in North- ern Thailand. Singapore University Press, Singapore, 1984. Tapp, N. Sovereignty and Rebellion: The White Hmong of Northern Thailand. Oxford University Press, Singa- pore, 1989. Kunstadter P. Natality, mortality and migration of upland and lowland populations in northwestern Thai- land. In Culture and Population (Edited bv Polaar S.), pp. 4660. Schenkman, Cambridge, MA, 1971.- ’ Kunstadter P. Highland populations in northern Thailand. In Highlanders of Northern Thailand (Edited by MacKinnon J. and Bhruksasri W.), Chap. 1. Oxford University Press, Kuala Lumpur, 1983. Kunstadter P. Demographic differentials in a rapidly changing mixed ethnic population in northwestern Thailand. Nihon University Population Institute, NUPRI Research Paper Series 19, Tokyo, 1984. Kamnuansilpa P., Kunstadter P. and Auamkul N. Hilltribe Health and Family Planning. Family Health Division, Department of Health, Ministry of Public Health, Bangkok, 1987. Kunstadter P. Ethnicity, ecology and mortality in northwestern Thailand. In Anthropology and Epidemiol- OPV (Edited bv Janes C.. Stall R. and Gifford S. D.), p:. 1255156. Reidel, Dordrecht, 1987. Kunstadter P. Health of Hmong in Thailand: risk factors, morbidity and mortality in comparison with other ethnic groups. Culture, Med. Psychiat. 9,329-351, 1985. Podhisita C., Kunstadter P. and Kunstadter S. L. Evidence for early fertility transition among the Hmong in northern Thailand. J. Pop. sot. Sfud. 2, 137-156, 1990. Choe M. K. Sex differentials in infant and child mortality in Korea. Sot. Biol. 34, 12-25, 1987. Levine N. E. Differential child care in three Tibetan communities: beyond son preference. Pop. Dev. Rev. 13, 281-304, 1987. See, e.g., Basu A. M. Is discrimination in food really necessary for explaining sex differentials in childhood mortality. Pop. Stud. 43, 193-210, 1989. Bongaarts J. Does family planning reduce infant mortality rates? Pop. Dev. Rev. 13, 212-334, 1987. Potter J. E. Does family planning reduce infant mortality? Pop. Dev. Rev. 14, 179-187, 1988. Trussell J. Does family planning reduce infant mortality? An exchange. Pop. Dev. Rev. 14, 171-178, 1988. Kunstadter P. Child mortality and maternal parity: some policy implications. Inr. Fam. Planning Perspect. Digest 4, 75-85, 1978. Geronimus A. T. On teenage childbearing and neonatal mortality in the United States. Pop. Den. Rev. 13, 245-279, 1987. Kunstadter P. Cultural ideals, socioeconomic change and household composition: Karen, Lua’, Hmong and Thai in northwestern Thailand. In Households: Com- parative Hisrorical Studies of the Domestic Group (Edited by Netting R. McC., Wilke R. R. and Arnould E. J.), DD. 299-329. Universitv of California Press. Berkeley; -1984. Kunstadter P. and Kunstadter S. L. Health transitions in Thailand. What We Know about Health Transition: The Cultural, Social and Behavioural Determinants of Health (Edited by Caldwell J., Findley S., Caldwell P., Santow‘G., Cosford W., Braid J. and Broers-Freeman D.). Chan. 11. DD. 213-250, Tables 9. 11. Health Transition Centre,* Australian National University, Canberra, 1990.


35. Kunstadter P., Kesmanee C. and Photi-art P. Hmong 36. Kunstadter P., Kunstadter S. L., Leepreecha P.. Karen Health Behavior and Family Planning: Cultural Podhisita C., Laoyang M., Sae Thao C., Sae Thao and Other Factors Affecting Use of Modern Health and R. and Sae Yang W. Causes and consequences of Family Planning Services in Northern Thailand, increase in child survival rates: ethnoepidemiology pp. 111-112. Family Health Division, Department of among the Hmong of Thailand. Hum. Biol. 64,821-841, Health, Ministry of Public Health, Bangkok, 1987. 1992.

Documents

Demographic variables in fetal and child mortality: Hmong in Thailand