This is the old United Nations University website. Visit the new site at http://unu.edu

Child malnutrition and feeding practices in Malawi

Abstract
Introduction
Data and methods
Univariate results
Multivariate results
Discussion
Policy implications
Acknowledgements
References

Nyovani Janet Madise and Mabel Mpoma

Nyovani Janet Madise is affiliated with the Department of Social Statistics at the University of Southampton in Southampton, UK. Mabel Mpoma is a graduate student in the Department of Human Nutrition at the University of Southampton.

Abstract

The 1992 Malawi and Demographic Health Survey data are used to assess the association between breast-feeding practices, socio-economic and morbidity variables, and the nutritional status of children under the age of five years using multilevel models. About 27% of under-five children in Malawi are underweight, and nearly 50% are stunted. The results of this study suggest that socio-economic factors, morbidity, and inappropriate feeding practices are some of the factors associated with malnutrition in Malawi. High socio-economic status, as measured by urban residence, the presence of modern amenities, and some maternal education, is associated with better nutritional status, whereas morbidity within two weeks before the survey is associated with low weight-for-age Z scores. Breast-feeding is almost universal and is carried on for about 21 months, but the introduction of complementary food starts much too early; only 3% of Malawian children under the age of 4 months are exclusively breastfed. Children aged 12 months or older who were still breastfeeding at the time of the survey were of lower nutritional status than those who had stopped breastfeeding. The analysis also showed a significant intra-family correlation of weight-for-age Z scores of children of the same family of about 39%.

Introduction

Many researchers agree that socio-economic and environmental conditions, together with feeding practices, are important determinants of malnutrition in developing countries [1-3]. Inadequate quantity and poor quality of food result in growth faltering, which is exacerbated by the high prevalence of fevers and diarrhoeal diseases. Inappropriate feeding patterns may also contribute to child malnutrition. Malnutrition is thought to be one of the factors contributing to the high level of child mortality in Malawi, where nearly one-quarter of all children die before they reach their fifth birthday [4, 5].

The results of the 1992 Malawi Demographic and Health Survey showed that about half of under-five children were stunted and about 27% were underweight [6]. These levels of stunting and underweight were identical to those obtained from the 1980-81 National Sample Survey of Agriculture, suggesting that levels of malnutrition have remained the same for at least a decade. One possible reason for the prevailing high levels of malnutrition in Malawi is poverty, since it is estimated that about 60% of the population live below the poverty line [7]. The results of a recent study support this, since they showed that Malawian children from higher socio-economic backgrounds had anthropometric measures that were comparable to the National Center for Health Statistics (NCHS) reference population [8].

The Malawian economy is largely based on agriculture, and the main exports are tobacco, tea, and sugar. About 88% of the population live in rural areas where they practise subsistence farming on small farms, using traditional labour-intensive methods. Consequently, as many as 70% of the subsistence farmers do not produce enough food [9]. The price of maize, the staple food, is high relative to wages. In 1992 the price of a bag of maize sufficient to feed an average family of about five persons for one month was between 70% and 120% of the minimum monthly wage [10]. Breastfeeding is almost universal and is carried on for an average period of about 21 months, but complementary feeding is introduced much too early. The World Health Organization (WHO) recommends that children under the age of four months be exclusively breastfed, but only 3% of such children were being exclusively breastfed during the 1992 Malawi and Demographic and Health Survey [6]. The early introduction of complementary food is undesirable, particularly in a developing country such as Malawi, since the typical weaning food is of poorer quality than breastmilk and may be contaminated.

The major diseases of children in Malawi are malaria, diarrhoea, and respiratory diseases, particularly pneumonia [6]. Illness affects the nutritional status of children, since it depresses a child’s appetite and reduces the absorption of nutrients, but increases the energy requirement. In addition, nutrients are lost through vomiting or diarrhoea. On the other hand, it can be argued that undernourished children are more prone to infection, so that the relationship between infection and malnourishment goes in a vicious cycle.

In this paper, multilevel analyses were performed on weight-for-age Z scores to determine the socio-economic, morbidity, and feeding variables that are associated with nutritional status. Multilevel analysis is appropriate for household survey data that often contain information on more than one child per mother. There is growing evidence of familial clustering of health outcomes between siblings owing to shared maternal or household characteristics [11, 12].

Data and methods

The data used in this study are from the cross-sectional Malawi Demographic and Health Survey, which was carried out in 1992 and was designed to provide information on fertility levels, family planning, breastfeeding practices, early childhood mortality, and the nutritional status of mothers and young children, among other things. Of 8,652 enumeration areas identified during the 1987 census, 225 enumeration areas were selected with probability proportional to size. The survey deliberately oversampled the northern region and urban areas so that regional and urban-rural comparisons could be made. Thus, the Malawi Demographic and Health Survey is not self-weighting. Details of the precise sampling and weighting procedures have been published elsewhere [6]. Software for multilevel modelling with weighted data are still at the experimental stage. Therefore, in this study unweighted data are used. This implies that, strictly speaking, the relationships observed in the analysis refer to the unweighted sample and not to the whole population of Malawian children. It has been suggested that the bias from using unweighted data may be minimized by the inclusion of the design variables as explanatory factors in the model [13]. Thus, region and urban-rural residence (the design variables) were included in the analysis as explanatory variables.

In this study, weight-for-age anthropometric measurements for children under five years old were compared to the NCHS reference population. The Centers for Disease Control package ANTHRO was used to calculate the weight-for-age Z scores. There were 3,173 children between 1 and 59 months of age, but the multivariate analysis includes only children aged 4 to 59 months, since low anthropometric measurements for infants under the age of 4 months could be a result of prematurity or other endogenous causes. However, the univariate analysis of breastfeeding practices includes children under the age of 4 months.

Separate multivariate analyses were conducted for children aged 4 to 11 months and those aged 12 to 59 months. Splitting the children into the two groups was necessary, for two main reasons. First, in Malawi nearly all children under one year of age are breastfed, and therefore the relationship between breastfeeding and nutritional status can be assessed only for older children. Second, the determinants of malnutrition for infants may be different from those for older children, so that if only one model is fitted, it may fail to identify these differences. Thus, the multivariate analysis involved 544 children aged 4 to 11 months and 2,379 children aged 12 to 59 months.

A multilevel approach was adopted to identify characteristics associated with malnutrition at the individual child and the mother or household level. Many household survey data tend to have a hierarchical structure, and the units of observation are often similar within a level of the hierarchy. For example, children of the same mother may have similar health outcomes (e.g., morbidity or mortality) because of shared maternal characteristics, such as feeding practices, sanitation, child care, or genetic frailty. Beyond the household level, similarities may also exist within geographic areas, so that three-level (or higher) models can be fitted if the data are sufficiently large. In this study, however, only a two-level model was considered.

For the 4- to 11-month age group, single-level regression using the stepwise selection procedure was used, because children of these ages are unlikely to have siblings within the same age group. Indeed, an examination of the data showed that only one mother had triplets in this age group. For the 12- to 59-month age group, a two-level model was fitted incorporating individual child variables and variables relating to the mother or household. The simplest two-level model with one explanatory variable X is of the form

WAZ_ij = a + bX_ij + u_j + e_ij

where

WAZ_ij is the weight-for-age Z score for the ith child belonging to the jth mother;
X_ij is the value of X for the ith child belonging to the jth mother;
u_j is the deviation from the overall mean for the jth mother; and
e_ij is the deviation of the ith child from the mean of the jth mother.

It is assumed that u_j and e_ij are normally distributed with means equal to zero and variances and , respectively [14]. This simple two-level model can be extended to one with several explanatory variables, some of which could be categorical. The presence of two sources of random variation is what distinguishes this model from an ordinary regression model. Ordinary regression models assume that the response variables (in this case, weight-for-age Z scores) are independent for any two observations, but when data are clustered, this assumption may not be valid. A technical consequence of ignoring clustering is that the standard errors of the estimates may be underestimated, so that some variables may appear to be significant when in fact they are not.

Goldstein [15] described the intra-cluster correlation coefficient using the formula: . This can be interpreted as measuring the degree of correlation between weight-for-age Z scores of children within the same family, and the higher the value of this coefficient, the greater the correlation. The statistical package SPSS was used to fit the ordinary regression model for the 4- to 11-month age group, and for the two-level model, the statistical package MLn [16] was used.

At the child level, the variables used were the sex of the child, the age to the nearest month, the length of the preceding birth interval, the birth order, whether the child had a fever or cough, and the number of days (if any) the child had had diarrhoea during the two weeks before the survey. The length of the succeeding birth interval (if any) was included for the 12- to 59-month age group only, since children under one year of age are unlikely to have a succeeding sibling. Mothers were also asked about the size of the baby at birth, that is, whether the baby was very large, normal, or small. This information was used as a rough indicator of the child’s health status at birth.

In the survey, mothers were asked the age when they started giving their child drinking water, formula, other liquids, or solids. However, information on the type of food that was given 24 hours before the survey was collected only for children who were being breastfed. Consequently, this information was included for the 4- to 11-month age group only. The study differentiated between those who had started receiving complementary food before the recommended age of 4 months and those who had started after 4 months. Thus, three categories were created for the type of breastfeeding: those who were exclusively or fully breastfed (breastmilk and plain water only), those who received complementary food but had started before the age of 4 months, and those who were given complementary food at or after the age of 4 months.

Since the majority of children received water during the first month, the age when drinking water was introduced was not included in the analysis. Nearly all children under 12 months of age were being breastfed, so breastfeeding status was considered for the 12- to 59-month age group only. A dummy variable was created to identify those who were still breastfeeding and those who were not.

At the mother level, the socio-economic and sanitation characteristics that were used in the analyses were the level of the mother’s education; the presence of modern amenities, such as electricity, a working radio, and a car or motorcycle; the source of drinking water; and the type of toilet facility.

Univariate results

Nearly all children under one year of age were still breastfeeding at the time of the survey (99%). Of those aged 12 to 23 months, about 79% were still breastfeeding, while the percentage still breastfeeding between 24 and 35 months of age was about 11%. Very few children aged 36 months or older were still breastfeeding (1%). The median duration of breastfeeding in Malawi is about 21 months [6]. Between 24 and 35 months of age, the percentages of children still breastfeeding differ significantly by the level of maternal education and also urban-rural residence. In this age group, about 14% of children whose mothers had four or fewer years of education were still breastfeeding, compared with only 6% of those whose mothers had five or more years of education. Similarly, the percentages of children aged 24 to 35 months still breastfeeding were 14% and 4% for rural and urban areas, respectively. The survey also collected information on the frequency of breastfeeding within the 24 hours before the survey, but there was little variation in the frequency of breastfeeding among children of various socio-economic subgroups.

Table 1 shows the percentage of breastfeeding children who were receiving complementary food or fluid according to age. Exclusive breastfeeding was uncommon among children under the age of 4 months. More than 90% of such young children were being given plain water, contrary to recommended infant-feeding practices. Furthermore, about 12% of such young children were being given sugar water, which has very little nutritional value. The use of baby formula is not common, even among older children. Overall, nearly half of the children under the age of 4 months had already started receiving solids or mushy food, and among infants aged 0 to 1 month, the percentage who had started on solids was about 25%. WHO recommends that children be given other food apart from breastmilk by the age of 6 months. In the sample, about 10% of those aged 6 months or older who were still breastfeeding did not receive solid or mushy food in the 24 hours before the survey.

TABLE 1. Percentage of breastfeeding children by age and type of complementary food or fluid during 24-hour period before the survey

% receiving	Age (mo)
	<4 N=249	4-11 N=537	12-23 N=541	³24 N=74
Plain water	92.4	95.7	96.1	100.0
Sugar water	11.6	7.8	6.7	5.4
Juice	4.8	12.1	14.0	10.8
Baby formula	7.2	9.3	5.4	0
Fresh milk	4.8	8.6	11.5	2.7
Powdered milk	1.6	3.9	5.9	4.1
Solid or mushy food	49.4	91.4	94.8	87.8

Source: ref. 6.

The use of feeding bottles is not very common in Malawi. Only about 5% of children under 4 months of age and 8% of children aged 4 to 11 months were bottle-fed. After two years of age, the percentage of bottle-fed children drops to about 2%. Nearly 70% of the children who were bottle-fed in the 24 hours before the survey were still breastfeeding, suggesting that bottle-feeding was a supplement to breast-feeding rather than a replacement for it. A disproportionately high percentage (11%) of children under two years of age whose mothers had five or more years of education were using bottles, compared with only 3% of those whose mothers had four years of education or less.

Mean weight-for-age Z scores for children aged 4 to 59 months classified by several background characteristics are presented in table 2. Girls have on average better weight-for-age Z scores than boys, but the difference is significant only for the younger age group. The mean weight-for-age Z score for children aged 4 to 11 months who were fully or exclusively breastfed was lower than for those who were given solids, but the difference was only significant at the 10% level.

For the length of the preceding birth interval, there were no significant differences between the means of the categories of this variable for the 4- to 11-month age group, but the results for the 12- to 59-month age group showed better weight-for-age Z scores for children who were born at least 48 months after the preceding child. In contrast, for the length of the succeeding birth interval, children with shorter intervals (less than 24 months) were associated with better nutritional status than those with longer intervals or those who had no succeeding sibling.

The major diseases of young children in Malawi are malaria, diarrhoea, and pneumonia [6]. The manifestation of malaria is fever; that of pneumonia is also fever, but accompanied with coughing and short, rapid breathing. About 41% of all the children under the age of five years were reported to have had fever during the two weeks preceding the survey. Almost 22% of the children had diarrhoea, and about 15% were reported to have been coughing with short, rapid breathing [6]. There were significant differences between the mean weight-for-age Z scores of children who had diarrhoea and those who had not been ill. For fevers and coughs, significant differences were observed for the older age group only.

The preliminary analysis showed that nearly all the socio-economic and sanitation variables studied were significant, and the general pattern was that children with higher socio-economic backgrounds had better weight-for-age Z scores than those with lower socio-economic backgrounds. However, because of the high interrelation of these variables, it was expected that some would not be significant in the multivariate analysis.

Multivariate results

Single-level regression analysis for the 4- to 11-month age group

The multiple regression analysis of weight-for-age Z scores for the 4- to 11-month age group is shown in table 3. The analysis showed that of the individual characteristics, the sex of the child, the mother’s assessment of the baby’s size at birth, the child’s age, and the type of breastfeeding were significant. Of the maternal or household characteristics, only the level of maternal education was retained in the model. The region of residence and urban-rural classification were significant. None of the morbidity variables during the two weeks before the survey were significant in this analysis. It is possible that the non-significance of these variables was a result of the small sample size rather than a true reflection of the lack of association between morbidity and nutritional status within this age group. The selected model explained only about 18% of the variation in weight-for-age Z scores.

TABLE 2. Mean weight-for-age Z scores of children aged 4-59 months by selected background characteristics

Characteristics		Age (mo)
		4-11		12-59
		Mean (SD)	N	Mean (SD)	N
Individual child characteristics
Sex			*
	male	-0.87 (1.46)	256	-1.32 (1.32)	1,205
	female	-0.58 (1.55)	288	-1.28 (1.22)	1,174
Size of baby at birth			*		*
	normal/large	-0.63 (1.50)	442	-1.24 (1.28)	1,986
	small	-1.09 (1.54)	102	-1.62 (1.18)	393
Birth order
	1	-0.82 (1.54)	108	-1.36 (1.22)	354
	2-4	-0.68 (1.50)	224	-1.28 (1.24)	1,057
	³5	-0.71 (1.51)	212	-1.29 (1.31)	968
Preceding birth interval (mo)					*
	none	-0.82 (1.54)	108	-1.36 (1.21)	354
	<24	-0.69 (1.64)	72	-1.25 (1.26)	403
	24-47	-0.70 (1.48)	271	-1.35 (1.26)	1,251
	³48	-0.69 (1.51)	93	-1.13 (1.34)	371
Succeeding birth interval (mo)					*
	none	-		-1.36 (1.31)	1,405
	<24	-		-0.88 (1.35)	203
	24-47			-1.29 (1.14)	741
	³48	-		-1.24 (1.37)	30
Type of breastfeeding
	exclusive	-1.12 (1.11)	42	-
	with food, started < 4 mo	-0.74 (1.54)	291	-
	with food, started ³4 mo	-0.60 (1.53)	211	-
Fever					*
	yes	-0.80 (1.57)	309	-1.43 (1.28)	941
	no	-0.61 (1.43)	235	-1.21 (1.25)	1,438
Diarrhoea			*		*
	yes	-0.90 (1.47)	199	-1.48 (1.34)	494
	no	-0.61 (1.53)	345	-1.25 (1.24)	1,885
Cough
	yes	-0.74 (1.51)	324	-1.37 (1.27)	1,004
	no	-0.69 (1.52)	220	-1.24 (1.26)	1,375
Maternal or household characteristics
Mother’s education			*		*
	none	-1.00 (1.49)	231	-1.48 (1.27)	1,016
	primary 1-4	-0.54 (1.54)	128	-1.32 (1.26)	518
	primary 5-8	-0.55 (1.52)	155	-1.10 (1.26)	720
	secondary+	-0.13 (1.10)	30	-0.82 (1.14)	125
Source of water			*		*
	private tap	-0.05 (1.40)	51	-0.76 (1.21)	224
	public tap	-0.46 (1.47)	122	-1.21 (1.29)	576
	protected well	-0.85 (1.60)	107	-1.36 (1.21)	482
	unprotected well	-0.97 (1.43)	187	-1.49 (1.23)	755
	river/stream/lake	-0.85 (1.52)	77	-1.25 (1.33)	342
Toilet facility			*		*
	flush	-0.23 (1.38)	22	-0.64 (1.15)	97
	pit	-0.68 (1.46)	408	-1.26 (1.26)	1,780
	none	-0.96 (1.69)	114	-1.56 (1.26)	502
Electricity			*		*
	yes	-0.03 (1.21)	27	-0.59 (1.07)	128
	no	-0.76 (1.52)	517	-1.34 (1.27)	2,251
Presence of working radio			*		*
	yes	-0.53 (1.45)	241	-1.18 (1.25)	1,102
	no	-0.87 (1.54)	303	-1.40 (1.28)	1,277
Presence of modern amenities					*
	car or motorcycle	-0.34 (1.23)	17	-0.83 (1.27)	72
	bicycle	-0.75 (1.55)	121	-1.29 (1.21)	589
	none	-0.73 (1.51)	406	-1.32 (1.29)	1,718
Area of residence
Urban/rural			*		*
	urban	-0.23 (1.50)	135	-0.93 (1.27)	593
	rural	-0.88 (1.48)	409	-1.42 (1.24)	1,786
Region		-0.47 (1.58)	*	-1.09 (1.27)	*
	north	-0.71 (1.54)	150	-1.28 (1.30)	735
	central	-0.93 (1.39)	208	-1.51 (1.21)	811
	south		186		833
Total		-0.87 (1.50)	544	-1.44 (1.24)	2,379

* F test significant at the 5% level or less.
Source: ref. 6.

The results suggested that on average girls had better nutritional status than boys. The estimated mean weight-for-age Z score of girls was about 0.33 standard deviation higher than that of boys. Children who were small at birth (according to the mother’s assessment) tended to have lower Z scores than children who were assessed to be normal or large at birth.

TABLE 3. Single-level regression for weight-forage Z scores for children aged 4-11 months

Parameter		Estimate (SE)
Constant		0.23 (0.32)
Individual child characteristics
Sex
	male	0.00
	female	0.33 (0.12)*
Size at birth
	normal/large	0.00
	small	-0.42 (0.15)*
Child’s age (nearest mo)		-0.21 (0.03)*
Type of breastfeeding
	exclusive	0.00
	with food, started <4 mo	0.29 (0.23)
	with food, started ³ 4 mo	0.56 (0.23)*
Maternal or household characteristics
Maternal education
	no education	0.00
	some education	0.25 (0.13)
Area of residence
Region
	north	0.00
	central	-0.07 (0.15)
	south	-0.40 (0.16)*
Urban/rural
	rural	0.00
	urban	0.53 (0.14)*

* Significant at the 5% level or less.

To turn to the type of breastfeeding, those who were exclusively or fully breastfed had the lowest mean Z scores, but the mean was not significantly different from that of children who were receiving complementary food but had started before 4 months of age. However, children who started receiving complementary food at or after 4 months had the highest mean, indicating better nutritional status.

There is a clear negative relationship between the child’s age and weight-for-age Z scores, indicating that the nutritional status of children under the age of one year in Malawi appears to deteriorate as the children grow older. This is reflected by a nearly straight line for the mean weight-for-age Z scores between the ages of 0 and 12 months (fig. 1). Beyond one year, the relationship between a child’s age and its weight-for-age Z score is less straightforward, but there are indications of slight improvements in the nutritional status of children after the second year of life.

FIG. 1. Mean weight-for-age Z scores by age of child (Malawi DHS)

The level of maternal education was marginally significant at the 5% level. Infants whose mothers had some education had weight-for-age Z scores that were on average 0.25 higher than those of infants whose mothers had no education. The socio-economic and sanitation variables (e.g., type of toilet facility, source of drinking water, presence of electricity) were highly correlated with urban-rural residence, so that once urban-rural differences were accounted for, these socio-economic characteristics ceased to be significant. Another reason for the lack of significance of some of the socio-economic variables could be the small size of the sample used in this study.

The regression model for the 4- to 11-month age group also showed strong regional differences, particularly between the South and the rest of the country. The weight-for-age Z scores for children in the South were, on average, between 0.30 and 0.40 standard deviation lower than those of children in the other regions.

Two-level regression analysis for the 12- to 59-month age group

For the 12- to 59-month age group, the size of the baby at birth, the length of the preceding and succeeding birth intervals, breastfeeding status, the number of days with diarrhoea, and whether or not the child had a fever in the two weeks before the survey were significant at the 5% level or less (table 4). Among the maternal or household-level characteristics, the source of drinking water, the type of toilet facility, and whether or not the household had electricity were significant. Urban-rural residence and region were also significant, although the regional effect was confounded by the presence or absence of electricity in the household. The mother-level variance () was 0.56, which was significant at the 5% level, indicating that weight-for-age Z scores for siblings were related. The child-level variance () was 0.89, so that the intra-family correlation coefficient, , works out to be 0.39. Thus, the degree of correlation of weight-for-age for children of the same family is about 39%. This model explained only about 9% of the variation in the Z scores.

TABLE 4. Two-level regression for weight-for-age Z-scores for children aged 12-59 months

Parameter		Estimate (SE)
Constant		-1.24 (0.21)*
Individual child characteristics
Size at birth
	normal/large	0.00
	small	-0.36 (0.07)*
Preceding birth interval (mo)
	none	0.00
	<24	0.09 (0.09)
	24-47	0.03 (0.07)
	³48	0.28 (0.09)*
Succeeding birth interval (mo)
	none	0.00
	<24	0.39 (0.09)*
	³24	-0.01 (0.07)
Child’s age (nearest mo)		-0.001 (0.002)
Still breastfeeding
	yes	0.00
	no	0.26 (0.08)*
No. of days with diarrhoea		-0.02 (0.009)*
Fever within 2 wk
	yes	0.00
	no	0.17 (0.05)*
Maternal or household characteristics
Source of water
	other	0.00
	unprotected well	-0.15 (0.06)*
Type of toilet facility
	flush/pit	0.00
	none	-0.18 (0.07)*
Electricity
	yes	0.00
	no	-0.10 (0.20)
Area of residence
Urban/rural
	rural	0.00
	urban	0.26 (0.07)*
Region
	north	0.00
	central	0.15 (0.29)
	south	0.29 (0.27)
Electricity/region interaction
	no electricity, central	-0.32 (0.30)
	no electricity, south	-0.76 (0.28)*
Mother-level variance()		0.56 (0.06)*
Child-level variance ()		0.89 (0.06)

* Significant at the 5% level or less.
Source: ref. 6.

The association between the baby’s size at birth and weight-for-age Z score was in the same direction as in the previous model and remained strong. Children who were small at birth continued to have lower weights than other children of similar age. Also, the results showed higher weight-for-age Z scores for children born 48 months or more after a preceding child than for those with shorter preceding birth intervals or first births. With regard to the length of the succeeding birth interval, those who had a succeeding sibling within two years appeared to have better nutritional status than the rest of the children. The relationship of breastfeeding and nutritional status showed that for each age, weight-for-age Z scores for children who were still breastfeeding were lower than the scores for those who were not breastfeeding (fig. 2).

FIG. 2. Mean weight-for-age Z scores according to whether or not breastfeeding

The number of days a child had diarrhoea in the two weeks before the survey was negatively related to nutritional status and was associated with a decrease in the Z scores of about 0.02 standard deviation for each additional day of diarrhoea. Coughing was not significantly related to nutritional status, but those who had fever had lower Z scores than those who had no fever.

The level of maternal education ceased to be significant once the other socio-economic variables were included in the model. Poor water supply, particularly from unprotected wells, and the absence of toilet facilities were associated with poor nutritional status. Urban-rural residence was highly significant, showing better nutritional status for urban children, on average. The region of residence was significant but was confounded with the effect of the presence of electricity. The region-electricity interaction showed that for the majority of the children, who live in households without electricity, those from the northern region were better nourished, followed by those from the central region, with those from the southern region having the lowest mean weight-forage Z scores. Children from households with electricity were better nourished on average, with those from the southern region having the highest mean (fig. 3).

FIG. 3. Mean weight-for-age Z scores showing region-electricity interaction

Discussion

The findings of this study suggest that breastfeeding patterns, morbidity, and socio-economic conditions are important determinants of undernutrition in Malawi. These results are in agreement with lessons learned in other developing countries [17, 18]. Further, the results study confirm the finding from other studies that the majority of Malawian children are given complementary food much too early [19, 20]. Only about 3% of children under the age of 4 months were exclusively breastfed. Solid or mushy food is introduced as early as 1 month to about 25% of children, contrary to WHO recommendations. The multivariate analysis suggested that the timing of the introduction of complementary food was important, and children who were given food according to the timing recommended by WHO tended to be better nourished than children who were started on solids too early. After 6 months of age, the infant needs other food in addition to breastmilk, and a delay in the introduction of solids may have undesirable consequences, such as growth faltering.

It is possible that because of the high awareness of the problem of undernutrition, mothers misguidedly introduce complementary food very early to prevent their children from being malnourished. Anecdotal evidence also suggests that some health educators recommend the early introduction of complementary feeding, especially when mothers express anxiety about the adequacy of their milk flow. The early introduction of complementary food is undesirable, because it interferes with breastfeeding. Breastmilk is the only properly balanced food for the human infant before the age of 4 months. Infants who are given complementary food too early may not feed at the breast vigorously because they may be too full, and therefore they may get fewer nutrients than recommended. This is especially true in many developing countries, where the complementary food is often of lower nutritional value than breastmilk. Another risk of the early introduction of complementary food is that of infection from contaminated food or feeding utensils. Some of the long-term risks of early complementary feeding are obesity, hypertension, and food allergies [21].

The survey did not collect information on the nutrient density of the food given to the children. However, the most common type of complementary food in Malawi is a thin gruel made with water and maize flour, to which salt and sugar are added. This food is very low in protein and energy; a 200-ml cup contains about 80 calories [22]. Health educators recommend that groundnut powder, eggs, milk, ground fish, or oil be added to enrich the porridge, but many families cannot afford these. Infants and young children may also be given soft fruits that are in season, such as mangoes, papayas, oranges, and bananas. At around 8 months, the child is gradually introduced to the adult diet, and by 18 months the child eats the same diet as adults, a thick maize porridge accompanied by a vegetable stew or, on rare occasions, fish or meat [22].

Another important variable not measured in the survey is the frequency of complementary feeding. Preparing a child’s food separately requires time and extra firewood for cooking, but many rural mothers cannot afford these, so the frequency of the child’s feeding depends on the number of times that the family has its meals. Traditionally, mothers used to cook food to be used by young children before leaving for their gardens, but this practice seems to have died out [4].

Breastfeeding beyond infancy was associated with poor nutritional status for these children. Similar results have been documented for many sub-Saharan African countries [1, 23, 24]. One reason could be that when mothers are busy or away from home, they may use breastfeeding as a substitute for a regular meal. Also, older babies may sometimes refuse to eat their meals and prefer instead to latch onto the breast for a long time.

Another explanation for the observed adverse association of breastfeeding and nutritional status beyond infancy is that of reverse causality, where mothers continue to breastfeed children who appear small for their age. However, the fact that children who were breastfed beyond two years were from poorer socio-economic subgroups suggests selectivity rather than reverse causality. There is evidence to suggest that urban mothers or those with some years of education breastfeed for shorter durations than do rural or uneducated mothers [19]. But even though they may breastfeed for shorter durations, educated mothers often have an economic advantage over their counterparts that enables them to provide more nourishing food, so that their children tend to be better nourished.

In contrast, in areas where breastfeeding durations are shorter, such as Latin America and some Asian countries, the duration of breastfeding seems to have a positive effect on child nutrition and survival [18, 25]. Others have found that the beneficial effect of breastfeeding on child survival tends to be stronger within poorer subgroups [25, 26]. A study in the poorer Dodoma region of Tanzania supports this; adverse effects of short durations of breastfeeding on nutritional status were found in this area, where most of the children were from low socio-economic subgroups [27].

The analysis also showed that girls had higher weight-for-age Z scores than boys, particularly for the 4- to 11-month age group. Quinn et al. [8] found similar results for Malawian pre-school children aged 24 to 59 months. However, studies from other countries found the opposite to be true [18, 28]. It is likely that the observed sex differences could be due to the biological frailty of boys, which has been observed in many countries.

Those children who were assessed by their mothers to have been small at birth had, on average, lower weight-for-age than those who were of normal weight or large. It is very likely that the mother’s assessment was an accurate indicator of whether the baby was indeed small-for-age or premature, and thus it would be expected that small babies would have low weight-for-age Z scores. The nutritional status of children deteriorated rapidly with age, especially during the first year of life. The age pattern of underweight observed in this study is similar to that of many countries, such as Ghana, Togo, Zimbabwe, and Bolivia [1, 2]. It is clear that the nutritional status of children starts to deteriorate when they are first exposed to complementary food.

Children born 48 months or longer after a preceding child had better weight-for-age Z scores, on average, than firstborn children or children born within four years of the preceding child. Children born after longer preceding birth intervals do not have to compete for maternal care and resources with another young child, and therefore they are likely to be better nourished than those born after a short preceding birth interval. Surprisingly, for the length of the succeeding birth interval, those who had a younger sibling before their second birthday had better weight-for-age Z scores. In Malawi, children who have a sibling before their second birthday are recognized to be in greater danger of becoming malnourished, so they are described by special names to highlight this fact. It is possible that such children may be given extra attention and food to make up for the early cessation of breastfeeding and the reduced level of maternal care brought about by the presence of a younger sibling. Indeed, in some areas of the country, such children are fostered by grandparents who dote on them, to compensate for the early cessation of breastfeeding.

Child morbidity, especially fever and diarrhoea, was negatively related to nutritional status. Apart from the wastage of nutrients as a result of high temperatures or repeated bowel movements, the practice of withholding food from a sick child in an attempt to “cure” diarrhoea may also contribute to the adverse association [29]. In the Malawi Demographic and Health Survey, 17% of those children who were still breastfeeding had their frequency of feeding reduced when they had diarrhoea, and about 25% of all children who had diarrhoea were given less fluids than normal [6].

For the source of drinking water, private taps, followed by public taps, were associated with better nutritional status. Surprisingly, the worst source of water appeared to be unprotected wells and not rivers or streams. Similar results also were found in connection with child mortality in Ghana [19], and one explanation may be that stagnant water around the mouth of the well may go back in, thereby contaminating the water. With regard to toilet facilities, children from homes with no toilets had lower nutritional status than those from homes with flush toilets or pit latrines. Good sanitation is important for the prevention of diarrhoeal diseases. Safe water supplies and flush toilets are also indicators of high socio-economic status in Malawi.

The socio-economic status of a family is important because it determines, in most cases, the availability and quality of food for the children. In addition, households with modern amenities, such as electricity, cars, or motorcycles, also tend to have better sanitation, so that their children are less likely to be exposed to infection. The level of maternal education is also important, since educated mothers tend to follow instructions about feeding and caring for children much better than their uneducated counterparts. Educated women are likely to use curative and preventive health services more than uneducated women [30].

The analyses showed that, in general, children from the southern region have lower weight-for-age than children from the other regions. The southern region is more densely populated than other regions and landholdings are smaller. In addition, rainfall in recent years has been more erratic, so that food production has been inadequate in this region. The region-electricity interaction, showing better nutritional status for children from households with electricity, especially those from the southern region, was not surprising, since the households with electricity in the South are likely to be in Blantyre, the largest commercial city in the country.

Another result of the study is the evidence of familial correlation of nutritional status among siblings. Since childhood undernutrition is largely a result of repeated infection and inadequate food intake, it is not surprising that the nutritional statuses of children in the same family are related. Other evidence of clustering of health outcomes, such as mortality, has been documented for Malawi [11]. This suggests that programmes to improve child survival and nutrition should be family based, targeting all young children of high-risk families.

The multivariate analyses explained only a small proportion of the variation in weight-for-age Z scores, suggesting that there are other explanatory factors that were not included in the models. Perhaps other types of studies, particularly those that observe children over a period while recording nutrient intake and illness, could shed more light on the factors associated with poor nutritional status.

Policy implications

It is clear that if the problem of malnutrition is to be reduced, Malawian families need to be educated about appropriate feeding practices for their infants and young children. To thrive, children need adequate nutrition and less exposure to infection. Thus, efforts are needed to improve the economic and sanitary conditions of Malawian families. The government of Malawi has already launched a poverty-alleviation programme, but in addition, specific policies with regard to child nutrition need to be implemented. It may be important to find out the level of knowledge of health educators who are responsible for teaching mothers about feeding and caring for their children. Properly trained educators are vital to any programme involved in improving mothers’ knowledge of appropriate feeding practices. Also, research into the processing and production of weaning food, using locally available products, should be encouraged and financed.

Very few mothers in Malawi follow the recommended practice of exclusive breastfeeding of infants under the age of 4 months, so this is one area where efforts should be concentrated. Employers can be encouraged to allow working mothers who are nursing to go home during the day to breastfeed their children. In addition, where practical, creche facilities could be made available in urban centres to allow mothers to be near their young children. When weaning commences, the importance of regular feeding should also be emphasized, and when mothers have to be away, child-minding on a rota basis can be encouraged to ensure that infants and young children are given food on a regular basis.

In conclusion, the nutritional status of children in Malawi is poor and may account for the high levels of infant and child mortality. Children who are introduced to complementary food early are more malnourished than those who are introduced to complementary food according to the WHO-recommended timetable. It is also clear that the socio-economic status of a family has obvious implications for the nutritional status of the children. Children living in urban areas, particularly those in households with modern amenities, are in general better nourished than other children. In addition, such children have clean water and their families may have good toilet facilities, thus making them less susceptible to infection. Programmes designed to reduce malnutrition and child mortality levels in Malawi will clearly need to deal with the issue of poverty as well as challenging some of the prevailing infant-feeding practices that adversely affect the nutritional status of young children.

Acknowledgements

The authors wish to acknowledge the useful comments made by Dr. Zoe Matthews and two anonymous reviewers.

References

1. Vella V, Tomkins A, Borghesi A, Migliori GB, Adriko BC, Crevatin E. Determinants of child nutrition and mortality in north-west Uganda. Bull WHO 1992;70:637-43.

2. Sommerfelt AE, Stewart MK. Children’s nutritional status. DHS Comparative Studies No. 12. Calverton, Md, USA: Macro International Inc., 1994.

3. Mwadime RKN, Baldwin SL. Relationship between household access to food and malnutrition in eastern and southern Africa. East Afr Med J 1994;71:571-9.

4. Malawi Government and UNICEF. The situation of children and women in Malawi. Lilongwe, Malawi: UNICEF, 1987.

5. Millard AV, Ferguson AE, Khaila SW. Agricultural development and malnutrition: a causal model of child mortality. In: Caldwell J, Findley S, Caldwell P, Santow G, Cosford W, Braid J, Broers-Freeman D, eds. What we know about health transition: the cultural, social and behavioural determinants of health. Proceedings of an International Workshop. Canberra, Australia: National Centre for Epidemiology and Population Health 1989;1:285-310.

6. Malawi National Statistical Office and Macro International, Inc. Malawi demographic and health survey report. Zomba, Malawi: Malawi National Statistical Office and Macro International, Inc., 1994.

7. Malawi Government. Policy framework for poverty alleviation programme. Lilongwe, Malawi: Ministry of Economic Planning and Development, 1995.

8. Quinn VJ, Chiligo-Mpoma MO, Simler K, Milner J. The growth of Malawian preschool children from different socioeconomic groups. Eur J Clin Nutr 1995;49:66-72.

9. Centre for Social Research. The characteristics of nutritionally vulnerable subgroups within the smallholder sector of Malawi: a report from the 1980/81 national sample survey of agriculture. Zomba, Malawi: University of Malawi, 1988.

10. Malawi Government. Food security and nutrition bulletin. Lilongwe, Malawi: Department of Economic Planning and Development, 1992.

11. Madise NJ, Diamond I. Determinants of infant mortality in Malawi: an analysis to control for death clustering within families. J Biosoc Sci 1995;27:95-106.

12. Curtis SL, Steele F. Variations in familial neonatal mortality risks in four countries. J Biosoc Sci 1996;28:141-59.

13. Skinner CJ, Holt D, Smith TMF, eds. Analysis of complex surveys. Chichester, UK: John Wiley and Sons, 1989.

14. Woodhouse G, Rasbash J, Goldstein H, Yang M, Howarth J, Plewis I. A guide to MLn for new users. London: University of London, Institute of Education, 1995.

15. Goldstein H. Multilevel models in educational and social research. London: Griffin, 1987.

16. University of London, Institute of Education. MLn Statistical Software, 1995.

17. Jonsson U. Millions lost to wrong strategies. In: UNICEF. The progress of nations. Oxford, UK, and New York: Oxford University Press, 1994:6-11.

18. Melville B, Williams M, Francis V, Lawrence O, Collins L. Determinants of childhood malnutrition in Jamaica. Food Nutr Bull 1988;10(1):43-8.

19. Madise NJ. Birthspacing and its impact on under five mortality in Malawi. Doctoral thesis. University of Southampton, Southampton, UK, 1993.

20. Macro International, Inc. Nutrition of infants and young children in Malawi, Africa. Calverton, Md, USA: Nutrition Chartbooks, 1994.

21. World Health Organization. Infant feeding: the physiological basis. Bull WHO 1989;67(suppl):60-5.

22. Malawi Ministry of Health. Weaning facts for Malawian families. Nutrition Education Booklet No. 1. Lilongwe, Malawi: Malawi Ministry of Health, 1992.

23. Brakohiapa LA, Bille A, Quansah E, Kishi K, Yartey J, Harrison E, Armar MA, Yamamoto S. Does prolonged breastfeeding adversely affect a child’s nutritional status? Lancet 1988; ii: 416-8.

24. Chikusa NJM. Effects of breastfeeding on infant mortality and malnourishment using the multinomial logistic model. In: Patel MS, Nokoe S, eds. Biometry for development. Proceedings of the First Scientific Meeting of the Biometric Society, Kenya Group and East/Central African Network. Nairobi, Kenya: International Centre of Insect Physiology and Ecology, 1990:187-93.

25. Retherford RD, Choe MK, Thapa S, Gubhaju BB. To what extent does breastfeeding explain birth-interval effects on early childhood mortality? Demography 1989;26:439-50.

26. Palloni A, Millman S. Effects of inter-birth intervals and breast feeding on infant and early childhood mortality. Pop Stud 1986;40:215-36.

27. Serventi M, Dal Lago AM, Kimaro DN. Early cessation of breast feeding as a major cause of severe malnutrition in under twos: a hospital based study - Dodoma Region, Tanzania. East Air Med J 1995;72:132-4.

28. Gwatkin DR, Wilcox GR, Wray JD. The policy implications of field experiments in primary care and nutrition care. Soc Sci Med 1980;14C(2):121-8.

29. Çevik N, Büyükgebiz B, Büyükgebiz A, Çevik N. Nutrition and diarrhoeal diseases. Food Nutr Bull 1988;10(1):52-5.

30. Boerma JT, Sommerfelt AE, Rutstein SO, Rojas G. Immunization: levels, trends and differentials. DHS Comparative Studies, No. 1. Columbia, Md, USA: Institute for Resources, 1990.