Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

The impact of siblings' death on the nutrition status of children in Peru - Kees de Meer, Roland Bergman. John S. Kusner, and Willem H. P. den Brinker


The impact of child deaths on the nutrition status of surviving children in the Lake Titicaca basin in the Andes of southern Peru (altitude>3,800 m) is explored. Survey data on obstetric history and social variables were collected in a random sample of 86 households in two Aymara and three Quechua Amerindian peasant communities. Independently of social factors, deaths among older siblings under five years old were associated with improved nutrition stature in children under four. Improved head circumference for age in boys was associated with the death of an older sister. The findings point to a new hypothesis concerning nutrition status in this population: mortality in children under five can be related to improved nutrition status (as evidenced by head circumference and height) in surviving siblings under four.


We conducted an empirical study on the effect of child deaths on the nutrition status of surviving siblings in Amerindian households living in the Peruvian Andes (altitude > 3,800 m). The study was initiated after an earlier one in this population showed high child mortality rates but no significant wasting (low weight for height) [1].

According to studies in the Andes, high altitude poses a complex ecological stress of cold, low air humidity, and hypoxia that is associated with physical growth retardation in children [2-4]. A more limited effect of altitude on childhood growth has been reported in the United States [5]. Studies in Andean populations suggest that nutritional factors could be related to decreased physical growth in Amerindian children [6-8]. Children in developing countries are particularly prone to the combination of undernutrition and illness [9], and it is well known that this affects the nutrition status of young children most [10]. Malnutrition in children is associated with a higher mortality risk [11-13], and the ecological stress of high altitude may compound these problems. Mortality rates of 300 and more per 1,000 live births have been reported in Amerindian children under five years of age living in the Andes in Peru [14] and Bolivia [15].

Observations of the absence of wasting in these children [1] contrasted with the theory of concomitance of undernutrition and child mortality, and we postulated that the high child mortality rates in this region are associated with improved nutrition status in the remaining children. In the present study we tested the null hypothesis that there is no relationship between the mortality of rural highland Amerindian children and the nutrition status of surviving siblings.

We also studied the effects of other variables that may affect child growth in developing countries, such as birth order [16, 17], interpregnancy interval, duration of breast-feeding [18], source of drinking water and type of sewage disposal [19], crowding [17], education of the parents [17, 19], household cohesion [20], and household income [19, 21]. No consistent relation has been found between children's nutrition status and the amount of land held by peasant households in developing countries [22], but in Peru the nutrition status of children in Andean peasant households holding less than three hectares of land has been found to be lower [23]. The amount of land held by the parents was included as an independent variable in the present study.

Materials and methods

Research location

The high, cold Andean plain around Lake Titicaca (altitude 3,809 m) in southern Peru is located in the predominantly rural department of Puno. The majority of the population of the department (985,000 in 1984) are Aymara- and Quechuaspeaking Amerindians living in the densely populated areas around Lake Titicaca and its companion Lake Arapa, where opportunities for agriculture are somewhat better than in areas farther away.

These rural highland people are the poorest of Peruvian society [24, 25]. Subsistence agriculture and some animal husbandry are their principal means of livelihood. Native Andean crops such as potatoes and quinoa (Chenopodium quinoa) and introduced species such as barley are staple foods.

There are differences in the amount of land held and the social status of households within the rural communities, but they are relatively minor. Small and ecologically dispersed land holdings are the rule. Although land availability is limited by the mountainous environment and very large statecontrolled land holdings, as much as half of the land is in fallow at any time in the highly seasonal agricultural system.

Despite major efforts to improve the coverage of and access to health care in rural Puno [26], much remains to be desired in its use and quality. In the rural areas modern sanitation is absent, and many families get their drinking water from open wells.


Two Aymara and three Quechua communities in rural Puno were selected in 1983 within the framework of an agronomic study, taking into account the effects of geographic variations and micro-climates in the Lake Titicaca basin. After permission for the research was granted by the residents in each community, 89 households were selected at random from the total of approximately 1,000 in the five communities. At the time of the present study (post-harvest period, luly-August 1989), 86 of the households (33 Aymara, 53 Quechua) were available (nonresponse rate 3.4%). The land of 77 of the households had been accurately mapped in 1985-1987 by stereoplotting aerial photographs and measurements on the ground (mean altitude 3,825 m, range 3,810-3,840 m).

Data collection

A survey was conducted in the 86 households with the assistance of a trilingual interpreter (Aymara by origin), using a questionnaire that had been professionally translated from Spanish into Quechua and Aymara, covering three areas.

The first area covered was personal data on each member of the household, including marital status, sex, year and month of birth, and years of education. For children under eight years old, the date of birth was verified by a birth certificate or vaccination card.

The second area was obstetric history. The total number and outcome of pregnancies and the number of miscarriages were recorded, as well as the year of birth, birth order, and vital state (and, where applicable, the death date) for each child. The length of stillborn infants and of those who died in the early neonatal period was asked to estimate the degree of prematurity (deaths with a crownheel length over 30 cm were considered perinatal).

The third area was the household water supply and crowding. Water might be obtained from the lake, an open well (entrance unprotected against livestock), a semiprotected well (with a wooden cover), or a protected well (with a concrete cover and either a bucket or a pump); water from the last two sources was considered to be protected. The total number of people living in the household was taken as an indicator for crowding. Housing conditions, sanitation, and the social status of the family head were not recorded separately because they were similar for all the study households.

Data analysis

Nutrition status was assessed for 202 children under 14 years old (109 boys, 93 girls) out of the 204 eligible in the 86 households. It was measured and standard deviation scores (Z scores) were calculated as described elsewhere [27]. Briefly, weight was measured with a beam balance scale for infants under 10 kg and with a spring scale (standardized under local conditions) for the rest of the children. Height, head circumference, and left mid-upper-arm circumference were measured using standard methods [28]. The height and weight of the high-altitude children were compared with reference data from the US National Center for Health Statistics [29]. Z scores for height for age (HAZ), weight for age (WAZ), and weight for height (WHZ) were calculated. Z scores for head circumference for age (ZHDCRC) and midupper-arm circumference for age (ZARMCRC) were calculated using reference data from Oosterwolde, a representative population from the northern part of the Netherlands [30, 31].

For each living child, the sums of perinatal deaths (foetal deaths after 28 completed weeks of pregnancy and live births who died in the first seven days of life), infant deaths (deaths in live births under the age of one year), and child deaths (deaths between one and less than five years of age) [32] among siblings were calculated from the survey data. These calculations were made separately for siblings born before and after the index child. Odds ratios of standardized mortality rates with 95% confidence intervals were calculated according to the method described by Morris and Gardner [33].

Relationships among the variables were explored in a PRINCALS analysis (see Appendix, p. 302). Subsequently, bivariate analyses (Pearson correlations between optimally resealed variables in PRINCALS, and one-way analyses) were performed to explore the effects of possible confounding variables and to test the null hypothesis. The results were considered significant if p<.05. Statistical analyses were carried out with SPSS-X* [34].


Nutrition status of children

Of the 202 children in the study population, 129 (64%) and 54 (27%) fell under the US tenth percentile for height and weight for age respectively, indicating a high prevalence of stunting and underweight. In only four children did weight for height fall under the tenth percentile. The HAZ, WAZ, ZHDCRC, and ZARMCRC in both the Aymara and the Quechua children averaged significantly below those of the reference population, but the mean WHZ was significantly above that of the reference (table 1). Detailed description of the nutrition status of the children by age group and culture are published elsewhere [27].

Briefly, no differences in Z scores between boys and girls were found. The HAZ, WAZ, ZHDCRC, and ZARMCRC of children in the first year of life were significantly below the median of the reference population and decreased significantly with age. The WHZ was significantly above the United States median and increased with age. The ZHDCRC, WHZ, and ZARMCRC were significantly higher in the Aymara than in the Quechua children.

Mortality rates and social variables

The perinatal mortality rate in the study households was calculated at 71 per 1,000 births. The infant and child mortality rates were 156 and 60 per 1,000 live births respectively. Of the 202 children in the study, 90 had at least one older sibling who had died before the age of 14, and 56 had younger siblings who had died. Of the 62 children under four years old, 23 had a total of 31 older sibling deaths. Perinatal and infant mortality rates were significantly higher in the Aymara than the Quechua children (table 2). There were no significant differences between the Aymara and Quechua households in the quantity of land owned, size of household, interpregnancy interval, time of weaning, duration of breast-feeding, and years of parental education. More Aymara households had access to protected drinking water than Quechua households (80% versus 20%).

Relationships among variables

Relationships among 16 variables-age, education of both parents, crowding, source of drinking water, birth order, duration of breast-feeding, amount of land held, number of deaths among older siblings, culture (Aymara versus Quechua), HAZ, WAZ, WHZ, ZHDCRC, and ZARMCRC- were explored with PRINCALS and bivariate analyses.

TABLE 1. Means of Z scores and 95% confidence intervals (CI) for nutrition status in Amerindian children under 14 years of age

  Aymara (N = 63) Quechua (N = 193) Total (N = 202)
Mean Cl limits Mean Cl limits Mean CI limits
Height for age -1.74 -2.0,-1.5 -1.50 -1.7,-1.3 -1.57 -1.7,-1.4
Weight for age -0.67 -0.9,-0.4 -0.70 -0.8,-0.6 -0.69 -0.8,-0.6
Weight for heightsa 0.89 0.6, 1.2 0.43 0.3, 0.5 0.56 0.4, 0.7
Head circumference for age -0.96 -1.2,-0.7 -1.35 -1.5,-1.1 -1.22 -1.4,-1.1
Upper-arm circumference for age -1.31 -1.5,-1.1 -1.52 -1.7,-1.4 -1.45 -1.6,-1.3

z scores are calculated from sex- and age-matched US and Dutch reference populations.

a Total number of children is 163 because of missing data from the reference population.

TABLE 2. Standardized mortality rates for Aymara and Quechua children, with odds ratios and 95% confidence intervals




Aymara/Quechua mortality ratio



Rate per 1,000


Rate per 1,000














1.3 - 3.4







0.4 - 2.4

a Perinatal mortality is calculated on the basis of the total number of births; infant and child mortality on the number of live births.

Children under 14 years old

Associations between nutritional variables and socio-cultural variables were evident in the PRINCALS analysis (results not shown; fit, 48% of total variance). For the children under 14 years old, the Pearson correlation coefficients of optimally resealed variables in PRINCALS showed that protected water supply was associated with improved ZHDCRC (p < .05), and less crowding (i.e., small household size) with improved nutrition status (WAZ p < .005, ZARMCRC p C .01, HAZ p=.10). Age was positively associated with improved WHZ (p < .005), and the Aymara children had higher ZHDCRC and WHZ (p < .01).

Education of the father was associated with improved WAZ, WHZ? and ZHDCRC in offspring (p < .01). Subsequent one-way analyses showed that these values were about one Z score higher in the offspring of fathers with more than six years of education than of those with less than three years.

The amount of land held was positively correlated with WAZ (p < .05) and ZHDCRC (p < .001), and birth order had a negative association with WAZ (p < .05). There was an association between the number of sibling deaths and ZHDCRC in the children under 14; further analyses showed that this association was limited to children under four years old.

Children under four years old

Figure 1 depicts the two-dimensional solution of PRINCALS for 62 children under four years old (fit after the exclusion of three variables with low communalities: 50% of total variance). With the exception of the number of deaths among older siblings, social and economic factors in this age group loaded mainly on the first dimension. Similar directions of the vectors for the education of the parents and the source of drinking water indicate that more education was associated with the use of protected wells. Higher birth order was associated with a longer interpregnancy interval.

FIG. 1. Plot of component loadings of 13 variables in two-dimensional solution of PRINCALS for 62 Amerindian children under four years of age in southern Peru. (Culture and duration of breast-feeding were excluded from the variables because of low communalities. Mid-upper arm circumference for age not shown; component loadings: first dimension -0.023, second dimension +0.700.)

Coding of variables at ordinal level of measurement as follows. Parental education: 0-3 years = 1, >3-6 years = 2. >6 years=3. Source of drinking water: lake water = 1, open or semi-protected well = 2, protected well = 3. Birth order: first-born = 1, etc. Interpregnancy interval: first-born or <1.5 years = 1, 1.5 to <2.5 years = 2, etc. Number of older sibling deaths. Amount of land held: <15,000 mē= 1, 15,00130,000 mē = 2, 30,001-60,000 mē = 3, >60,000 mē = 4. Age: I + age in months. Z scores were entered at a numerical level of measurement.

Fit: 50.0% Eigenvalues: first dimension 25.9%, second dimension 24.1%. Communalities are depicted as vectors. See tables 3 and 4 for the numerical values of the communalities and Pearson correlation coefficients of optimally resealed variables in the PRINCALS solution depicted here and after controlling for age.

Opposite directions of the vectors for birth order and crowding (household size) indicate negative correlations between these variables. The results show that parents with less education had more children. less knowledge of hygiene, and more crowded households. Pearson correlation coefficients showed these associations as well (tables 3 and 4). Also, the source of drinking water was associated with ZHDCRC (p<.05). The amount of land was not significantly associated with the Z scores for nutrition status of the children under four.

The nutrition status of the children under four loaded mainly on the second dimension in PRINCALS. The number of older dead siblings had a high component loading on this dimension but not on the first dimension (0.51 versus 0.08 respectively). Pearson correlation coefficients revealed significant associations between the latter variable and ZHDCRC and HAZ, and a weak association with WAZ. Similarly to the findings for all children under 14 years old, the level of education of the father was significantly associated with nutrition status in offspring under four but not with the number of sibling deaths. Birth order was associated with the number of sibling deaths but not with Z scores. Age was associated with WHZ but not with other Z scores or independent variables. Age had a very low communality and did not confound the aforementioned associations within this age group.

Dead siblings and the nutrition status of the index child: Testing the null hypothesis

One-way analyses were carried out to test the null hypothesis for the children under four years old. The Z scores for the nutrition status variables increased with the number of older siblings who had died before the age of five. The results were significant for ZHDCRC (F= 6.7; p < .005) and HAZ (F= 3.9; p < .05) with r2=.17 and .11 respectively, but not for WAZ (F= 2.9; p = .07), WHZ, or ZARMCRC. Comparisons of the ZHDCRC and HAZ for children with and without deaths of older siblings before age five are presented in figures 2 and 3. For perinatal or infant deaths among siblings, only the relation with ZHDCRC remained significant (one-way, p < .005). Relations between the death of younger siblings and Z scores for the nutrition status of the index children were not significant.

TABLE 3. Communality values and Pearson correlation coefficients of variables before controlling for age, children under four years old(fit: 50,0%)



Pearson correlation coefficient






































































































Pearson correlation coefficient

















































EDMOT = maternal education. EDFAT = paternal education. FAMSIZ = size of the household. WATER = source of dunking water. ORDBTH = birth order. BTHSP = interpregnancy interval LDSIZ = amount of land held. SIBDD = number of older sibling deaths. HAZ, wAz, wHz, ZHDCRC. and ZARMCRC are defined in the text.

*p < .05. **p < .01. ***p < .005.


Contents - Previous - Next