Determinants of child nutrition and mortality in north-west Uganda.

Determinants of child nutrition and mortality in north-west Uganda V. Vella,' A. Tomkins,2 A. Borghesi,3 G.B. Migliori,3 B.C. Adriko,4 & E. Crevatin5 An anthropometric survey of children aged 0-59 months in north-west Uganda in February-March 1987 indicated a high prevalence of stunting but little wasting. Use of unprotected water supplies in the dry season, prolonged breast-feeding, and age negatively affected nutrition; in contrast, parental education level improved nutrition. Mortality during the 12 months following the survey was higher among those who had low weight-for-age and weight-for-height, but children who had low height-for-age did not have higher mortality. Weight-for-age was the most sensitive predictor of mortality at specificities >88%, while at lower specificity levels weight-for-height was the most sensitive. Children whose fathers' work was associated with the distillation of alcohol had a higher risk of mortality than other children. The lowest mortality was among children whose fathers were businessmen or who grew tobacco. Introduction Several studies have found that malnutrition increases the risk of childhood mortality. For example, for Bangladesh, Sommers & Lowenstein reported that the risk of mortality for children below the 10th percentile of arm-circumference-for-height was 3.4 times that of children above the 50th percen- tile (1). Also, for India, Kielman & McCord found that mortality decreased exponentially with each 10% rise in the % median weight-for-age (2). In Bangladesh, Chen et al. reported that mid-upper-arm circumference and weight-for-age were the best pre- dictors of mortality (3), while Briend et al. showed that mid-upper-arm circumference was the best pre- dictor of mortality (4). In contrast, studies in Africa have found that anthropometric parameters are weak predictors of mortality (5, 6); however, one recent study in south-west Uganda showed that anthropo- metric indicators are good predictors of mortality, with mid-upper-arm circumference being the most sensitive.a I Primary Health Care Project Officer, UNICEF, P.O. Box 7047, Kampala, Uganda. Requests for reprints should be sent to Dr Vella at the following address: SA3PH, World Bank, 1818 H Street, N.W. Washington, DC 20433, USA. 2 Director, Centre for International Child Health, Institute of Child Health, London, England. 3 Medical Officer, Collegio Universitario Assistenti Medici Missio- nari (CUAMM), Kampala, Uganda. 4Deceased. Formerly, District Medical Officer, Arua District, Ministry of Health, Entebbe, Uganda. 5 Director, School of Hygiene and Preventive Medicine, Univer- sity of Trieste, Trieste, Italy. a Vella, V. An analysis of predictors of childhood malnutrition and mortality in south-west Uganda. Ph. D. thesis, University of London, 1990. Reprint No. 5326 Malnutrition leads to a higher risk of death, but what are the determinants of malnutrition? Frequent- ly suggested causes are poverty, low parental educa- tion, lack of sanitation, low food intake and malab- sorption, diarrhoea and other infections, poor feeding practices, family size, short birth intervals, mother's time availability, child-rearing practices, and seasonality. These variables interact to cause inad- equate consumption and assimilation of nutrients and a subsequent impairment of health and physiological functions. The present article reports on the nutritional status in a district of Uganda and on attempts to identify sensitive predictors of childhood mortality; a further aim was to determine the major causes of malnutrition and mortality in the study region. Methodology Between February and March 1987, a total of 1178 children aged 0-59 months were selected from 30 villages in the district of Arua, north-west Uganda. The villages were chosen using random sampling with probability proportional to size, based on the 1980 Ugandan census (7). The weight and height/length of the children were measured. Weight was measured to the nearest 100 g using Salter spring scales. The height of each child aged 24-59 months was determined, while length was measured for children below 2 years of age; the measurements were made to the nearest mm using locally constructed height/length boards. The ages of the children were assessed by examining birth and baptismal certificates or local calendars of events. Socioeconomic and health-related variables were collected from each household using a Bulletin of the World Organization, 70 (5): 637-643 (1992) © World Health Organization 1992 637 V. Vella et al. questionnaire that was distributed to the head of the family and to the mother of the child. The families surveyed were revisited 1 year later to record any deaths that had occurred among the children who had been measured. After being validated, the data were analysed using SPSS software (8). The anthropometric measurements were transformed into standard deviation (S.D.) scores and expressed as % deviations from those for the median NCHS referen- ce population (9) by means of a software package obtained from the Centers for Disease Control (10). Results Table 1 shows the prevalence of malnutrition among the study children, by age group. Nutritional status was relatively satisfactory in the first 5 months of life but deteriorated thereafter. The proportion of children who were underweight (low weight-for-age) was greater in the second year of life, but improved thereafter. Wasting (low weight-for-height) was commoner among those aged 6-24 months but was very low for other age groups. Stunting (low height- for-age) was low in the first 5 months of life but reached high levels subsequently. Mortality rates were around 10% during the first year of life, 3.1% in the second year, 4% in the third year, and about 0.5% thereafter. Mortality was signi- ficantly higher at low levels of weight-for-age and weight-for-height but remained the same at different values of height-for-age (Table 2). Compared to a baseline level of >-1 S.D., the relative risk for mor- tality was 3 at <-3 S.D. weight-for-age and 4.6 at <-2 S.D. weight-for-height. Table 1: Percentage prevalence of children below different cut-offs for weight-for-age, height- for-age, and weight-for-height, north-west Uganda Age (months) 0-5 6-11 12-23 24-35 36-47 48-59 Total Parametera (n = 108) (n = 142) (n = 231) (n = 187) (n = 215) (n = 183) (n = 1066) Weight-for-age <-2 S.D. 3.7 31.7 41.6 23.5 20.5 16.9 24.8 < 80% 8.3 40.8 44.2 21.9 23.3 21.3 28 Height-for-age <-2 S.D. 9.3 28.9 50.2 48.7 47.9 49.7 42.4 < 90% 6.5 16.2 28.1 29.4 33.5 38.8 27.5 Weight-for-height <-2 S.D. 1.9 6.3 7.8 1.6 0 0 3 < 80% 1.9 5.6 3.5 1.6 0 0 2 a < -2 S.D. = < -2 standard deviations from the median of the NCHS standard population; <80% and < 90% = percentages from the median of the NCHS standard population. Table 2: Mortality levels among the study children according to different cut-offs (in standard deviations (S.D.)) for weight-for-age, height-for-age, and weight-for-age 1 year after the sur- vey S.D. cut-off <-3.00 to -2.50 to -2.00 to -1.50 to <-3 -2.51 -2.01 -1.51 -1.01 >1 x2 test Weight-for-age 8/75 2/74 3/115 4/166 4/174 16/462 P <0.001 (10.7)a (2.7) (2.6) (2.4) (2.3) (3.5) Height-for-age 6/192 3/119 3/141 5/156 7/132 13/326 P >0.10 (3.1) (2.5) (2.1) (3.2) (5.3) (4) Weight-for-height 1/3 1/4 2/25 3/63 7/121 23/850 P <0.001 (33.3) (25) (8) (4.8) (5.8) (2.7) a Figures in parentheses are percentages. WHO Bulletin OMS. Vol 70 1992 Child nutrition and mortality In north-west Uganda Table 3: Percentage mortality levels, by anthropometric intervals, according to age group Age (months) Parametera 0-11 12-23 >23 Total x2 test b Weight-for-age (S.D.) < -3 17.6 (17)C 11.1 (27) 6.5 (31) 10.7 (75) P= 0.48 -3 to -2.01 15.6 (32) 0 (69) 0 (88) 2.6 (189) P <0.001 > -2.01 8.5 (201) 1.5 (135) 1.1 (466) 3 (802) P <0.001 Height-for-age (S.D.) <-3 23.1 (13) 4.3 (47) 0.8 (132) 3.1 (192) P<0.001 -3 to -2.01 10.5 (38) 2.9 (69) 0 (153) 2.3 (260) P<0.001 >-2.01 9 (199) 0.9 (115) 2 (300) 4.1 (614) P <0.001 Weight-for-height (S.D.) <-2 9.1 (11) 11.1 (18) 33.3 (3) 12.5 (32) P = 0.51 >-2.01 10 (239) 1.4 (213) 1 (582) 3.2 (1034) P<0.001 a S.D. = standard deviation from the median of the NCHS standard population. b The statistical significance relates to a x2 test for the equality of death rate across the age groups (within each anthropometric group). c Figures in parentheses are the numbers of children in each age group. Table 3 shows that below these anthropometric cut-off points mortality was higher for children aged < 12 months. For each anthropometric parameter, Fig. 1 shows the sensitivity (% of total mortality correctly identified) versus specificity (% of survivors correct- ly identified). Above 88% specificity, weight-for-age was the most sensitive indicator, while at lower spe- cificities weight-for-height was more sensitive; the least sensitive was height-for-age. The choice of an anthropometric indicator to identify children at higher risk of death therefore depends on the level of Fig. 1. Plot of sensitivity versus specificity for the anthropometric parameters in predicting mortality, 1 year after the survey was carried out. 35, 30 325 .r c 20 cD 15 0-0 101 5 0 80 81 82 83 8485 86 8788 89 90 91 92 93 94 959697 98 99100 % Specificity specificity. However, there should be as few false positives, as possible, and a high level of specificity; weight-for-age seems to be better in this respect than weight-for-height or height-for-age. The only indicator of household and family sta- tus that was significantly related to child mortality was the father's occupation. Child mortality was highest when the father's occupation was associated with alcohol distillation, while the lowest mortality was recorded when the father was a tobacco grower or a businessman (Table 4). Table 5 shows the results of the stepwise mul- tiple regression in which the dependent variables were weight-for-age, height-for-age, or weight-for- height. Father's education was positively correlated with weight-for-age. Age, breast-feeding, use of unprotected water supplies in the dry season, skin infections, and diarrhoea in the 2 weeks before the survey negatively influenced the coefficient of weight-for-age. Maternal education was positively associated with height-for-age. Age, breast-feeding, use of unprotected water supplies in the dry season, skin infections and diarrhoea were negatively associated with height-for-age. Father's education was positively associated with weight-for-height, while skin infections were negatively correlated with this parameter. Breast- feeding was negatively associated with all three anthropometric parameters. This association arose because prolonged breast-feeding was associated with lower values of the anthropometric parameters (Fig. 2). WHO Bulletin OMS. Vol 70 1992 - - Weight-for-age -* -*- Weight-for-height --u--Height-for-age -0s_ _~~~~~U- .-_ ---- As. S '"Kv,ssA~~~~~-K->~~~x' sw , 639 V. Vella et al. Table 4: Child mortality and father's occupation in the study population Father's No. of % who died 1 year after occupation children the surveya Alcohol distillation 138 7.2 Farmer 751 4.3 Businessman 119 0.8 Tobacco grower 22 0 Other 148 2 Total 1178 3.9 a %2 test, P <0.05. There was no significant correlation between the anthropometric parameters and the sex of the child, father's occupation, presence of a latrine in the house, degree of crowding, pregnancy of the mother, parents' polygamous relationships, type of storage used for drinking-water, rubbish disposal practices, distance from a health unit, management of type of feeding and of fluid intake during diarrhoea, posses- sion of a health card, and use of unprotected water supplies during the wet season. Fig. 2. Plot of mean standard deviations (S.D.) of weight-for-height for breast-fed and non-breast-fed children, by age group. 0 4) _0.2 0 .I - 0.4 q (U c¢ -0.8 I- NBrona-fed 0] Non-b -fd 0-11 12-23 Age (months) ,23 ANOVA significance of Fc0.001 Table 5: Study variables correlated with anthropometric measurements (standard deviations) through stepwise multiple regression Regression coefficient for: Weight-for-age Height-for-age Weight-for-height Morbidity in the previous 2 weeks Diarrhoea -0.462a (0.170)b -0.423c (0.201) -0.161 (0.142) Fever -0.132 (0.153) -0.148 (0.181) -0.023 (0.127) Measles -0.188 (0.406) -0.055 (0.480) -0.252 (0.338) A.R.l.d -0.287 (0.202) -0.198 (0.239) -0.144 (0.168) Skin infection -0.729e (0.229) -0.633c (0.271) -0.495a (0.191) Others -0.380 (0.274) -0.153 (0.333) -0.195 (0.234) Child was -0.856e (0.116) -0.735e (0.139) -0.357e (0.065) breast-fed Age (months) -0.025- (0.003) -0.037G (0.004) N.E.f Use of unprotected water supplies -0.223a (0.084) -0.302a (0.101) N.E. during the dry season Father's education (years) 0.030a(0.009) N.E. 0.016c (0.008) Mother's education (years) N.E. 0.039C (0.017) N.E. r 2 0.10 0.10 0.06 F test P <0.001 P <0.001 P <0.001 ap <0.o1. b Figures in parentheses are the standard errors. cp <0.05. d A.R.I. = acute respiratory infections. e1 p <0.001 . fN.E. = Not entered into the model by the stepwise process, because the coefficient did not reach the significant level of P <0.05. WHO Bulletin OMS. Vol 70 1992 , .. _- ..... . . 640 J. ,a Child nutrition and mortality in north-west Uganda Discussion Among the study population the nutritional status was relatively satisfactory among those aged < 6 months, but deteriorated thereafter. This was probably due to the onset of weaning, which is often associated with an increased prevalence of diarrhoea. The low prevalence of wasting and high prevalence of stunting could be explained by the different effects that infections have on weight and height. After an acute infection, weight growth often recov- ers relatively rapidly, but linear growth is slower to recover (11), and is prone to setbacks should further acute episodes occur. There is still lack of agreement about the causes of stunting among children in developing countries. However, it is likely to be related to chronic dietary impairments (especially protein-energy but also of micronutrients) com- pounded by frequent infections. In the study popu- lation the high prevalence of stunting suggests long-term nutritional stress. There were few wasted children, which indicates that widespread, severe short-term food shortages were not a problem. Genetic factors may be important in the development of stunting; however, the condition most commonly occurs as the end result of chronic nutritional insuffi- ciency and food infections. Stunting has a multi- factorial etiology, having a clear association with poverty and poor living conditions, with no single factor being totally responsible. Thus, we can pre- sume that a reduction in stunting will only result if there is an improvement in socioeconomic con- ditions, household food security, provision of safe water supplies, sanitation, and other elements of primary health care. Previous studies in India (2) and Bangladesh (3, 19) reported that weight-for-age was a sensitive predictor of mortality; in contrast, studies carried out in Guinea-Bissau (5) and Zaire (6) found that anthro- pometric indicators were poor predictors. These find- ings could indicate that in Africa, in contrast to Asia, anthropometry plays a relatively minor role in identi- fying children who are at a higher risk of death. In our study, low levels of weight-for-age and weight- for-height were significantly related to mortality, confirming that also in Africa anthropometric indica- tors are sensitive predictors of mortality. Child mortality was highest in families where the father eamed his living from alcohol distillation, and lowest in those where the father was a tobacco grower or a businessman. The reason for this is probably that those involved in alcohol distillation were the worst off socioeconomically, while those who cultivated tobacco or were involved in business were better off. Mosley & Chen suggest that eco- nomic wealth can improve child survival through the prevention and cure of diseases, by increasing child-spacing and hygiene practices, by provid- ing improved access to safe water supplies and sanitation, and by increasing the availability of food (12). The implication is that changes in all these variables are necessary for an improvement in child survival. Parental education level, which is usually related to child mortality (13, 14), was not so in the present study. It is useful to consider whether certain variables had a greater impact than others on nutritional status. Stepwise multiple regression failed to demonstrate any correlation between the anthropometric par- ameters and the following: the child's sex, marital status of the mother, occupation of the father, parents' polygamous relationships, type of storage used for drinking-water, use of boiled water for drink- ing, disposal of waste-water, presence of a latrine, disposal of rubbish, distance from a health unit, pos- session of a child health card, whether the child had been weighed in the previous 3 months, or whether the parents knew about the correct feeding and fluids practices to adopt during diarrhoea epi- sodes. Similarly, there was no relation between the anthropometric indices and any history of the child being ill with fever, measles, or acute respiratory infections in the 2 weeks prior to the survey. Variables that negatively influenced the anthropometric coef- ficients included age, breast-feeding, use of unprotec- ted water supplies during the dry season, skin infec- tions, and diarrhoea in the 2 weeks prior to the survey. The only variable that favourably influenced nutrition was parental education level. The associa- tion b,etween prolonged breast-feeding (>12 months) and poor nutritional status has been reported in pre- vious studies (7, 15-17). It is possible that prolonged breast-feeding is associated with a lower intake of solids. Production of breast-milk is lower after 12 months' post-partum, and children who are still suck- ling at this age receive less total protein-energy intake than those who stop breast-feeding earlier. It is also possible that children who are still being breast-fed are reluctant to accept other foods in suf- ficient quantity, as reported by Brakohiapa et al. in a study carried out in Ghana (15). Also, children who breast-fed longer may have belonged to famil- ies living under poorer socioeconomic conditions, which affected their nutritional status. However, the multiple regression analysis carried out suggests that this is not the case. The use of unprotected water supplies during the dry season was negatively correlated with weight- for-age and height-for-age. This could have arisen because, especially in the dry season, the use of polluted water supplies causes spread of infections WHO Bulletin OMS. Vol 70 1992 641 V. Vella et al. (particularly diarrhoea), leading to malnutrition. Alternatively, the association could be a proxy indicator of household wealth. The relationship between parental educational level and nutritional status could be related to health knowledge, but it has been suggested that educa- tional level is an indirect measure of socioeco- nomic status (18). Educational level remained signi- ficantly associated with nutritional status after allow- ing for a range of socioeconomic variables, and we suggest that education per se has an important influ- ence on nutrition. Parental educational level could function by lowering fatalistic attitudes to illness, increasing belief in the possibility of changing child health status and the acceptance of new ideas, generating greater confidence in dealing with health professionals, producing more direct responsibility in child-rearing practices, as well as increasing parental health knowledge. Better-educated parents might have been more able to decide on priority actions to be taken in matters such as child immunization. Also, educated parents would probably accept more easily the concept of family planning, be better informed on how to use health facilities, and have shared family resources more equitably, especially in favour of their children; furthermore, they could have improved weaning practices, thus limiting the prevalence of diarrhoea. Parental educa- tion is likely to have a major impact on curable and preventable diseases and on the nutritional intake of children. In our study, skin infections and diarrhoea negatively influenced anthropometry, and are synonymous with poor hygiene, sanitation, and socioeconomic conditions. Therefore health educa- tion directed at changing sanitation practices, hy- giene, water supplies, and behaviour could consider- ably improve nutrition. Unfortunately many sanitation programmes fail because of the difficulty in changing deep-rooted behaviours, failure to assess per- ceptions, and the use of inappropriate sanitary tech- nology. The impact of improvements to water sup- plies and environmental sanitation will be minimal if people do not change behavioural practices that pollute water and food. Educational efforts, can, however, result in a lower incidence of diarrhoea and other infections, with subsequently improved child growth. Our results show that the predominant nutri- tional problem in the study community was stunting. A range of socioeconomic and environmental factors is likely to have been responsible. Parental educa- tion, independently of other variables, appeared to be particularly important in this respect, and the find- ings suggest the need for improved, appropriate educa- tion programmes for children and their parents, in addition to primary health care projects. Acknowledgements We thank the members of the survey team, village chiefs, resistance councils, and environmental health staff who assisted in implementing the study. We extend our grati- tude to the Uganda Ministry of Health and to Collegio Uni- versitario Assistenti Medici Missionari (CUAMM), for their financial and logistic support. The views expressed in this article are those of the authors and do not necessarily reflect those of the Uganda Ministry of Health or of CUAMM. Resume D6terminants de la nutrition et de la mortalitd juv6niles dans le nord-ouest de l'Ouganda Entre f6vrier et mars 1987, 1178 enfants ag6s de 0 a 59 mois ont 6t6 s6lectionnes dans 30 villages du district d'Arua, dans le nord-ouest de l'Ougan- da. Diverses mesures anthropometriques ont ete faites sur ces enfants et des donn6es socio-eco- nomiques concernant les familles ont e recueillies. Les families ont ete revisit6es un an plus tard et on a note tous les d6ces survenus chez les enfants examines. La sensibilite des parametres anthropom6- triques en tant que predicteurs de la mortalit6 6tait la plus grande pour le rapport poids/age lorsque la sp6cificit6 6tait superieure a 88%; l'indicateur poids/taille 6tait le plus sensible lorsque la specifi- cite 6tait plus faible, et l'indicateur poids/age 6tait le moins sensible. Le choix d'un indicateur anthro- pom6trique pour rep6rer les enfants ayant le risque le plus 61ev6 de mortalit6 d6pend par consequent du niveau de sp6cificit6; toutefois, comme les ressources sont limit6es, un taux 6lev6 de sp6cificite est habituellement exig6 et dans ce cas l'indicateur poids/age semble supe- rieur aux indicateurs poids/taille et taille/age pour rep6rer les enfants chez qui le risque de mortalit6 est le plus elev6. La mortalite juvenile 6tait plus 6lev6e dans les familles ou la principale profession du pere etait en rapport avec la distillation de l'alcool, et plus faible lorsque le pere travaillait dans une planta- tion de tabac ce qui indique que, dans la zone etudi6e, ces professions correspondent a des conditions socio-economiques differentes. L'etat nutritionnel des nourrissons de moins de 6 mois etait satisfaisant, mais se d6t6riorait par la suite; la maigreur (faible poids pour l'age) 6tait plus fr6quente chez les 6-24 mois, et 6tait faible dans les autres groupes d'age, alors que le retard de croissance (faible taille pour l'age) 6tait tou- jours important chez les plus de 5 mois. Les WHO Bulletin OMS. Vol 70 1992642 Child nutrition and mortality in north-west Uganda variables suivantes etaient liees de fagon signifi- cative aux r6sultats anthropom6triques: niveau d'etudes des parents, allaitement au sein, utilisa- tion de r6serves d'eau non protegees en saison seche, presence d'infections cutan6es, et diar- rh6e. Ces variables sont probablement liees au mauvais etat nutritionnel de par leur association avec la pauvret6, la sous-alimentation et les infec- tions. L'allaitement au sein prolong6 (au-dela de 12 mois) influait n6gativement sur l'6tat nutrition- nel. Cette association peut etre due a une alimen- tation plus r6duite chez les enfants nourris au sein, ou au fait que ces enfants se trouvaient dans les familles les plus pauvres. Dans la population 6tudi6e, le principal pro- bleme nutritionnel est donc le retard de croissan- ce, associ6 a des variables traduisant des condi- tions socio-6conomiques mediocres. Le retard de croissance resulte probablement d'une serie de facteurs associ6s a des carences chroniques, notamment prot6ino-6nerg6tiques, et egalement en oligo-6l6ments, aggrav6es encore par de fr6- quents 6pisodes infectieux. II s'agit d'un probleme nutritionnel chronique, dO a un stress nutritionnel prolong6 provoqu6 par l'interaction d'une insuffi- sance alimentaire et d'infections. Son 6tiologie est plurifactorielle, mais il est clairement associe a la pauvrete et a l'insalubrit6 des conditions de vie, aucun facteur pris isolement n'6tant entierement a incriminer. Nous pouvons donc supposer qu'une amelioration du retard de croissance dans la population etudi6e ne peut avoir lieu que s'il y a am6lioration des conditions de vie et de l'approvi- sionnement de la famille, un plus large acces a une source d'eau saine, et 6galement une amelio- ration de l'assainissement, de l'education sanitaire et nutritionnelle, des services de vaccination, de la lutte contre les maladies diarrh6iques et d'autres 6l6ments des soins de sante primaires. References 1. Sommers, A. & Lowenstein, M.S. Nutritional status and mortality: a prospective validation of the QUAC stick. American journal of clinical nutrition, 28: 287-292 (1975). 2. Kielmann, A.A. & McCord, C. Weight for age as an index of risk of death in children. Lancet, 1: 1247-1250 (1978). 3. Chen, L.C, et al. Anthropometric assessment of energy-protein malnutrition and subsequent risk of mortality among preschool-aged children. American journal of clinical nutrition, 33: 1836-1845 (1980). 4. Briend, A. et al. Arm circumference and other fac- tors in children at high risk of death in rural Bangla- desh. Lancet, 2: 725-727 (1987). 5. Smedman, L. et al. Anthropometry and subsequent mortality in groups of children aged 6-59 months in Guinea-Bissau. American journal of clinical nutrition, 46: 369-373 (1987). 6. The Kasongo Project Team. Anthropometric assessment of young children's nutritional status as an indicator of subsequent risk of dying. Journal of tropical paediatrics, 29: 69-75 (1983). 7. Report on the 1980 population census, vol. 1: the provisional results by administrative areas. Census Office, Entebbe, Uganda, September 1982. 8. SPSS/PC+ V 2.0 base manual for the IBM PC/XT/AT and PS 2. SPSS Inc., Chicago, IL, 1988. 9. National Center for Health Statistics. NCHS growth curves for children: birth-18 years. Rock- ville, MD, U.S. Department of Health, Education and Welfare, 1977. 10. Jordan, M.D. Anthropometric software package tutorial guide and handbook. Centers for Disease Control, Atlanta, GA, April 1987. 11. Ashworth, A. Growth rates in children recovering from protein-energy malnutrition. British journal of nutrition, 23: 835-845 (1969). 12. Mosley, W. & Chen, L. An analytical framework of child survival in developing countries. Population development review, 10 (suppl.): 24-45 (1984). 13. Hobcraft, J.N. et al. Socioeconomic factors in infant and child mortality: a cross-national comparison. Population studies, 38: 193-223 (1984). 14. Caldwell, J. Education as a factor in mortality de- cline. An examination of Nigerian data. Popula- tion studies 1979, 33 (3): 395-413 (1979). 15. Brakohiapa, L.A. et al. Does prolonged breast- feeding adversely affect a child's nutritional status? Lancet, 2: 416-418 (1988). 16. Thoren, A. & Stintzing, G. Value of prolonged breast-feeding. Lancet, 2: 788 (1988). 17. Victora, G.C. et al. Is prolonged breast-feeding associated with malnutrition? American journal of clinical nutrition, 39: 307-314 (1984). 18. Kitagawa, E.M., & Hauser, P.M. Education differen- tials in mortality by cause of death: United States, 1960. Demography, 5: 318-353 (1968). 19. Alam, A. et al. Anthropometric indicators and risk of death. American journal of clinical nutrition, 49: 884-888 (1989). WHO Bulletin OMS. Vol 70 1992 643