The Journal of Nutrition Vol. 128 No. 12 December 1998, pp. 2363-2368

A Scale Without Anthropometric Measurements Can Be Used to Identify Low Weight-for-Age in Children Less than Five Years Old^1,2,3

Maribel Orozco^*, Homero Martínez^{*, , 4}, Hortensia Reyes^*, and Héctor Guiscafré^*

^* Interinstitutional Health Services Research Group, Mexican Social Security Institute-Ministry of Health, Mexico City, Mexico and  National Institute of Public Health, Cuernavaca, Mexico

	ABSTRACT

Abstract Introduction Methods Results Discussion References

Malnutrition and morbidity have a synergistic association that often leads to death. However, malnutrition in children who die is largely underreported, because anthropometry of the deceased child is rarely known. This study had two purposes: i) to develop a scale that would help determine if a child had low weight-for-age (w/a), in the absence of anthropometric measures; and ii) to select an appropriate cut-off that would give the best sensitivity (Se) and specificity (Sp) of the proposed scale when contrasted with actual w/a measurement. The study was designed as a diagnostic test, and carried out in a rural area in central Mexico. We included 132 children under 5 y old with w/a under 2 Z score and 284 children with marginal or no w/a deficit as a control group. The proposed scale included potential predictive variables from clinical, socioeconomic and family factors. The best logistic regression model to predict low w/a included: birth weight less than 2,800 g, introduction of weaning foods after the sixth month of life, introduction of animal protein after the sixth month of life, low socioeconomic status, low w/a in siblings and more than three morbidity episodes in the previous 6 mon. Selecting a cut-off of 4 for this model to identify children with low w/a showed a Se and Sp of 85 and 95%, respectively. We tested the external validity of the scale in a different locale, and included 877 children under 5 y old from 10 rural communities. In this population, the scale showed Se of 84% and Sp of 81% to identify low w/a. Based on these results, we propose that the scale be included as a means of identifying low w/a in children who have died. We believe that this should be done in verbal autopsies, which, based on our previous research, the Ministry of Health adopted as part of the regular activities to monitor problems in the disease to health-seeking to death process.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

The World Health Organization (WHO)⁵ estimates that there are 192.5 million underweight children and 229.9 million stunted children in the developing world (de Onís et al. 1993), corresponding, respectively, to 35.8 and 42.7% of the world's population. In Mexico, the most recent national nutrition survey estimated that 12.6 million children under 5 y old are malnourished, which corresponds to 30% of children in this age group (Sepúlveda et al. 1990). The most frequent causes of death in Mexico of children under 5 y old are acute diarrhea and acute respiratory infection (INEGI 1993).

Although other studies showed that malnutrition is causally associated with these deaths (Pelletier et al. 1993), it is rarely reported on the death certificates for several reasons. Malnutrition is often classified on the basis of deficits of weight-for-age (w/a) or height-for-age (Gómez et al. 1955, Waterlow et al. 1977). Unless these data are available at the time of death, the physician who fills out the death certificate often overlooks nutritional status (Melgar et al. 1986, Parsons et al. 1980). Because a large number of deaths occur at home (Gray et al. 1990, Villa et al. 1994), lack of weight or height data prior to death is more the rule than the exception. Furthermore, underreporting malnutrition in the deceased child occurs more frequently when the person who certifies the death is not a physician, a common situation in rural areas of the developing world (where a large number of malnutrition-associated deaths occur) (Bustamante et al. 1990, Pelletier 1994). Another difficulty of identifying malnutrition associated with mortality is that the International Classification of Diseases has only a very broad category for malnutrition, which includes "Avitaminosis, anaemia, and other nutritional deficiencies."

Failure to identify the nutritional status of a deceased child has implications for enumeration purposes. And, when nutritional deficits are underreported in mortality statistics, less pressure is on policymakers to focus attention on malnutrition. This underreporting also limits our ability to collect data on nutritional status at death for research purposes.

The objective of this project was to develop a scale and select a cut-off to identify nutritional deficits in children who have died, in the absence of anthropometric measurements such as weight or height.

	MATERIALS AND METHODS

Abstract Introduction Methods Results Discussion References

Study design.  The study involved two phases. The first was the development and internal validation of the scale; the second was the application of the scale on a different population to test its external validity.

The first phase of the study was designed as a diagnostic test. To construct and test the scale's sensitivity (Se) and specificity (Sp), we developed it with living children, although it is meant to be used in children who have died. As the scale should be used as an indicator of children's w/a deficit, the standard to test Se and Sp was the actual anthropometry of the children studied. Children with low w/a (below 2 Z-score of the reference U.S. National Centre for Health Statistics population) (Hamill et al. 1976) were taken as true positives, and controls with w/a between 2 and +1 Z-score were defined as true negatives. We used w/a as an indicator of nutritional status for three reasons: i) Several studies have found it to be the most sensitive predictor of mortality in children under 5 y old (Chen et al. 1981, Yambi et al. 1991); ii) The recommended and most widely used anthropometric index endorsed by the Mexican government is w/a for assessing the nutritional status of children under 5 y old (Secretaría de Salud 1994); and iii) Publications demonstrating the causal effect of malnutrition and its impact on mortality used w/a as indicator of nutritional status (Pelletier et al. 1994).

The second phase of the study involved testing the external validity of the scale. For this purpose, we tested the scale in a different locale, including all children under 5 y old who lived in 10 randomly selected communities.

The study protocol was approved by the Ethical Review Board of the Instituto Mexicano del Seguro Social, Mexico. Prior to the study, all adult participants in charge of children selected for the study granted verbal consent for anthropometry to be taken in their children and to answer the questionnaire.

Study period and research areas.  The first phase of the study was carried out from August to March, thus including summer, fall and winter months; the seasonality of the study proved important for the morbidity variables. The study area for this first phase (development of the scale) was the valley of Solis, located in the central highland plateau of Mexico. Children under 5 y old living in 35 rural communities, encompassing three neighboring states (México, Michoacán and Querétaro), were included in the study. Living conditions in the valley were representative of the larger central part of Mexico. Inhabitants' main occupation was agriculture, with a heavy dependence on maize as the staple food; beans and other legumes were also grown. Hygienic conditions were poor: Most households don't have piped potable water or sewage disposal system, and domestic animals, including chicken and pigs, were often kept in or near the living quarters (Allen et al. 1987).

From an existing census of the population, we identified all children under 5 y old, including those who had w/a under 2 Z-score (n = 132). A second group, consisting of 286 children with no w/a deficit or deficit over 2 Z-score, was randomly selected from the anthropometric census of the population. One interviewer, unaware of the child's anthropometry, visited all homes of selected children to apply the interview schedule to the mother or other adult who cared for the child.

Risk factors for malnutrition.  The scale was based on a combination of variables from three general areas: clinical, nutritional and socioeconomic. Variables from the clinical area included mother's age, number of pregnancies, birth weight, gender, number of acute respiratory infections and acute diarrhea episodes, and number and timing of vaccines received. The nutritional variables included type and duration of milk feeding, timing and type of weaning, and present diet evaluated by a 24-h recall. Weaning was defined as the introduction of liquid or solid foods, different from the usual milk feeding (whether breast milk or formula). The socioeconomic status of the families was assessed by means of a scale previously developed and validated in the study area (Martínez et al. 1986), which included: i) characteristics of the house (materials of roof, walls, and floor and number of rooms, windows and doors); ii) ownership of household assets (bed, radio, TV set, gas stove, blender, iron, cabinets, sewing machine, car); iii) ownership of animals (cows, sheep, chicken, turkeys, pork, horses, donkeys); iv) occupation of the head of the family; v) years of school attendance of the head of the family and of the mother; vi) type of family (extended, nuclear); and vii) size of family. Nutritional status of siblings (based on w/a) was also included in this group of variables.

The external validity of the scale was tested in 10 communities, 200-300 km from the original study area, in the state of Hidalgo. Living conditions in this area were similar to those described for the valley of Solis. All children under 5 y old (n = 877) who lived in 10 randomly selected communities from two municipalities (Metepec and Zacualtipán) were identified by means of a house-to-house census. Children were weighed and mothers received a questionnaire schedule similar to that developed in the valley of Solís. This second phase took place between October and January, thus including the fall and winter months.

Construction of the scale.  The steps followed for constructing the scale are summarized as follows: i) Variables were selected that showed both Se and Sp larger than 50% to identify low w/a. We used a nonrestrictive P value of <0.10 to assess significant difference between groups to allow initial selection of a large number of variables. ii) These variables were included in multiple logistic regressions to predict low w/a; at this stage, only those with P < 0.05 were retained. All possible regressions were run to identify the best competing models. The criterion to choose the best model was based on a combination of the highest goodness-of-fit test, predictive power of the model (as a reflection of the area under the curve when plotting a receiver operator curve or ROC) and the smallest number of variables (Hanley and McNeil 1983). iii) The variables included in this model were taken as the components of the scale. Once the scale was defined, we calculated the Se and Sp of different cut-offs, plotting them in an ROC (Nettleman 1988). The choice of a cut-off sought to maximize Sp without a major sacrifice in Se (Metz 1978). (6) Lastly, we ran the discriminant analysis function in the Statistical Package for Social Sciences (SPSS) to identify children who had been misclassified by the selected cut-off.

Analysis software.  The database was entered in D-Base III. Epi-Info was used to calculate Se and Sp. SPSS was used for the logistic regressions, to calculate the ROC curves, and to run the discriminant analysis.

	RESULTS

Abstract Introduction Methods Results Discussion References

Four hundred and sixteen children under 5 y old were identified for the first phase of the study. This included 132 children with low w/a and 284 with w/a over 2 Z-score. No significant differences existed between the two groups in terms of age, gender and type of family, but families of children with low w/a were larger than those of the other group, had a higher birth order, and had one or more siblings under 5 y old (P < 0.10) (Table 1).

Ten variables met the Se and Sp criteria for inclusion in the logistic regression models: socioeconomic status of the family, years of school attendance of both parents, presence and number of siblings under 5 y old with low w/a, consumption of pulque⁶ by the father, birth weight under 2,800 g, occurrence of more than three acute respiratory infection (ARI) episodes in the 6 mon prior to the interview, nonbreast milk or infant formula feeding in the first 6 mon of life, weaning after the sixth month of life, introduction of animal products in the weaning diet after the sixth month of life, and habitual lack of consumption of animal products in the diet in the weekly recall diet. Results of all possible multiple logistic regressions identified three competing models that showed Se between 70 and 72% and Sp between 78 and 88% (Table 2). The best model, labeled as Model 3 in Table 2, showed one of the highest predictive powers (i.e., the area under the ROC curve was 0.87) and the highest goodness-of-fit (0.193) with the lowest number of variables (6). These included low socioeconomic status, weaning after 6 mon of age, introduction of animal protein to the diet after the sixth month of life, birth weight under 2,800 g, presence of other siblings with low w/a, and three or more episodes of ARI in the 6 mon prior to the interview. Five of these six variables were shared with Models 1 and 2. As it may be difficult to obtain precise data for birth weight, we run Model 3 omitting this variable (Table 2). Although the model lost predictive power (diminishing from 87 to 80%), it still retained reasonable Se (80%) and Sp (78%), and actually improved its goodness-of-fit (0.2996).

The presence of each variable contributed one point to the scale. The scale was applied to the study population, calculating the Se and Sp of different cut-off points to identify low w/a. As seen in Table 3 and Figure 1, taking four points as the cut-off gave the scale a Se of 85% and a Sp of 95%. This cut-off correctly identified 112 low w/a and 269 nonlow w/a children. The discriminant analysis run on this population, shown in Figure 2, enabled us to identify those children (n = 35, 8% of the total) who had been misclassified. A review of their anthropometry data showed that all of them had borderline values.

View larger version (9K): [in this window] [in a new window]   Fig 1. Relationship between sensitivity (%) and specificity (%) of different cut-off points to identify low weight-for-age. The dot over the straight line corresponds to 4 points, which was selected as the best cut-off.

View larger version (10K): [in this window] [in a new window]   Fig 2. Discrimination between control and low weight-for-age children who were correctly classified as such by the proposed scale.

Once constructed, the scale was tested for its external validity in a different population. We identified 877 children under 5 y old in the communities studied in the second study area. No statistically significant differences existed between these children and the population in Solís in terms of age and gender distribution, number of siblings under 5 y old, mean birth weight, percentage of children under 2,800 g, and percentage of mothers who didn't recall the birth weight of their infants (data not shown). Forty-one percentage (n = 350) of the children showed a low w/a (below 2 Z-score). The six-item scale was applied to this population, with the only difference being that questions were included on respiratory or diarrhea episodes in the 6 mon prior to the interview, instead of just respiratory episodes. This change was intended to allow us to identify possible effects of either of these diseases as a morbidity indicator, because of the difference in seasonality when the two studies were conducted. In this second population, the scale showed a Se of 82% and a Sp of 85% to identify low w/a children.

	DISCUSSION

Abstract Introduction Methods Results Discussion References

The impact of malnutrition on mortality is now well established (Pelletier 1994). However, malnutrition rarely appears on the death certificate, so its impact is largely underestimated. This lack of awareness is reflected in the limited attention paid to it by the government and the health-care sector and the resulting lack of efforts made to alleviate this condition. In previous research, we were concerned about the large number of children who died of acute diarrhea (AD) or ARI, and we wanted to learn what causes were associated with these deaths, focusing particularly on the disease-health-seeking process (Martínez et al. 1993). As other authors have noted, the management of fatal cases is a sensitive indicator of deficiencies in curative care (Habicht 1979). We went to the houses of children who had died where we administered to the mother or adult responsible for care of the child a semistructured interview schedule which produced descriptions of the events that occurred from the time the child got sick until he or she died, a technique known as "verbal autopsy" (Gutiérrez et al. 1994). Cases of deceased children were identified from death certificates.

Our research uncovered several situations that were amenable to interventions, and, when these were brought to the attention of the health-care sector in charge of AD and ARI programs, corresponding actions were taken (Reyes et al. 1993a, b). Throughout the country, verbal autopsies are now routinely performed to identify aspects for intervention and to monitor AD and ARI deaths, as a feedback to these programs (Secretaría de Salud 1997a, b). However, we found almost no mention of malnutrition as a cause of death or even as a circumstance present during the terminal illness of the child. We were concerned about this and decided to develop a method that would allow identification of a malnourished child, taking low w/a as an indicator of malnutrition.

Some conceptual issues related to the variables that composed the scale merit discussion. Several studies associated low birth weight, commonly defined as a birth weight under 2,500 g (Battaglia and Lubchenco 1967), with higher mortality (Garay 1990). In the area where the scale was developed, a previous study found that the average birth weight of children was 3,200 g (Allen et al. 1987); in the present study it was 3,000 g. As the proportion of children under 2,500 g at birth was quite small (7% in the first study cited and 9% in the present one), we tested different birth weight cut-offs to predict malnutrition and found 2,800 g to be the best. A second issue was the presence of a morbidity variable as part of our scale. When we developed the scale, we found a significant association only with the previous incidence of ARI. However, other studies found acute diarrhea to be highly correlated with malnutrition (Beau et al. 1987, Chen et al. 1981). When we validated the scale, we also found three or more episodes of AD to be a significant predictor of low w/a. In fact, we think that either ARI or diarrhea actually represents an indicator of the burden of morbidity on the nutritional well-being of the child.

Weaning after the sixth month of life and consumption of animal protein after the sixth month of life were two more variables in our scale. Findings from other authors seemed at odds with our findings as to the role of these variables, so they merit further discussion. We found that children who had been weaned after the sixth month of life were more likely to show low w/a than those who were weaned earlier. While this may seem to contradict WHO's recommendation to delay weaning until the fourth-sixth month of life (WHO 1985), it may be explained by the fact that children who were weaned earlier were better nourished to begin with, thus triggering the maternal reaction of feeding (Launer et al. 1990). If this were the case, then the better nutritional status at early ages would explain early weaning, not the reverse. By the same token, children were more likely to show low w/a when their weaning period was based on foods that offered mainly carbohydrates, such as tortillas, crackers and maize gruel (atole), while children with no w/a deficit had been weaned earlier, receiving larger amounts of animal products in their diets by the sixth month of their life. These relationships may be confounded by other variables, like frequency of feeding or composition of the diet, and we did not attempt to control such aspects in our study. To better understand the situation, it would be useful to refer to the role that the quality of the diet may play. For example, animal products offer key nutrients associated with better growth, including iron, zinc and vitamin A. In our study area, eggs were the most common source of animal protein, probably because families raised chickens in their yards. In other areas, different foods may be better indicators of the quality of the diet.

Another issue to consider when using the scale in other populations is the feasibility of obtaining the required information. Although most of the information for the variables that composed the scale should be easy to obtain, some may be consistently lacking. Specifically, this may be the case with birth weight and nutritional status of siblings. To assess the extent to which the absence of these data would affect the predictive ability of our model, we ran it omitting the variable on birth weight and found a predictive power (i.e., the area under an ROC curve) of 0.80 (Table 2); by omitting the nutritional status of siblings, we found a predictive power of 0.78; and by omitting both, a predictive power of 0.62. With these considerations in mind, we calculated that, by using the suggested cut-off, the scale underestimates the true numbers of low w/a children by 4% (false negatives minus false positives divided by all low w/a children). The comparable figure for the second population studied is 0.04%.

The results encourage us to propose the use of the scale to identify children with low (<2 Z-score) w/a, particularly in rural areas in developing countries, where identification of malnutrition associated with death is often overlooked. Based on our previous experience, we believe that this information could be gathered as an add-on to verbal autopsies, which are now part of regular monitoring activities of the Ministry of Health's Control of Acute Diarrhea and Respiratory Infection programs in Mexico (Secretaría de Salud 1997a, b). We hope that this strategy will increase awareness that low w/a children die more frequently than do well-nourished children with the same diseases. Increased awareness of this issue might, in turn, help focus the attention and resources of governmental and nongovernmental organizations to treat such a difficult problem. We encourage other researchers to follow the methodology presented here, tailoring the development of a similar scale to the specific situation in their own working locations.

	FOOTNOTES

Manuscript received 26 April 1998. Initial reviews completed 27 July 1998. Revision accepted 20 August 1998.

	ACKNOWLEDGMENTS

The authors acknowledge the support of Hugo Tudón in the data analysis. The participation of the field staff of the Rural Research Centre located in the valley of Solis is fully acknowledged and appreciated, as is the collaboration of the mothers with the study.

	LITERATURE CITED

Abstract Introduction Methods Results Discussion References

• Allen, L., Pelto, G. & Chávez, A. (1987) The collaborative research and support program on food intake and human function: Mexico project, University of Connecticut, Department of Nutritional Sciences, Storrs, CT.
• Battaglia F., Lubchenco L. A practical classification of newborn infants by weight and gestational age. J. Pediatr. 1967; 71:105
• Beau J., Garenne M., Diop B., Briend A. Diarrhea and nutritional status as risk factors of child mortality in a Dakar hospital. J. Tropical Pediatr. 1987; 33:4-9[Abstract/Free Full Text]
• Bustamante, P., Lezana, M., Roberto, D., Villa, R. & A. Borja. (1990) La desnutrición como causa múltiple de muerte. Salud Pública Méx. 32: 309-319.
• Chen L., Huq E., Huffman S. A prospective study of the risk of diarrheal diseases according to the nutritional status of children. Am. J. Epidemiol. 1981; 114:284-292[Abstract/Free Full Text]
• de Onís M., Monteiro C., Akré J., G. Clugston The worldwide magnitude of protein-energy malnutrition: an overview from the WHO global database on child growth. Bull WHO 1993; 71:703-712[Medline]
• Garay, A. (1990) La pobreza en México. Cuadernos de Nutrición 13: 5-11.
• Gómez F., Ramos G., Cravioto J. Malnutrition in infancy and childhood with special reference to Kwashiorkor. Pediatrics 1955; 7:135-140
• Gray, R., Smith, G. & Barss, P. (1990) The use of verbal autopsy methods to determine selected causes of death in children. Institute for International Programs, Johns Hopkins University Press, Baltimore, MD.
• Gutiérrez G., Reyes H., Martínez H., Tomé P., Guiscafré H. Study of the disease-health seeking-death process. Another use of the verbal autopsy. Int. J. Epidemiol. 1994; 23:427-428[Free Full Text]
• Habicht J.-P. Assurance of quality of the provision of primary medical care by nonprofessionals. Soc. Sci. Med. 1979; 138:67-75
• Hamill P., Drizd T., Johnson C. L., Reed R. B. Monthly Vital Statistics Report. Health Examination Survey data from the National Center for Health Statistics (NCHS) 1976; 25:1120
• Hanley A., McNeil M. A method of comparing the areas under receiver operating characteristic curves derived from the same cases 1. Radiology 1983; 148:39-843
• Instituto Nacional de Estadística, Geografía e Informática (INEGI) (1993) Información sobre las 10 principales causas de mortalidad en México, Mexico City, Mexico.
• Launer L. J., Habicht J.-P., Kardjati S. Breast feeding protects infants in Indonesia against illness and weight loss due to illness. Am. J. Epidemiol. 1990; 131:322-331[Abstract/Free Full Text]
• Martínez, H., Calva, J., Meneses, L. M. & Viais, H. (1986) Uso de alimentos y bebidas en el manejo de la diarrhea aguda en la infancia, Instituto Nacional de la Nutrición, División de Nutrición de Comunidad, Publ. L-69, Mexico City, Mexico.
• Martínez, H., Reyes, H., Tomé, P., Guiscafré, H. & Gutiérrez, G. (1993) La autopsia verbal: una herramienta para el estudio de la mortalidad en los niños. Bol. Med. Hosp. Infant Méx. 50: 57-63.
• Melgar, M., Flores, S. & Santillán, P. (1986) Valoración nutricional antropométrica del paciente hospitalizado. Bol. Med. Hosp. Infant Méx. 43: 233-266.
• Metz C. E. Basic principles of ROC analysis. Sem. Nucl. Med. 1978; 8:283-298[Medline]
• Nettleman M. D. Receiver operator characteristic (ROC) curves. Infec. Control Hosp. Epidemiol. 1988; 9:374-377[Medline]
• Parsons H., Francoeur M., Howland M., Spengler R. The nutritional status of hospitalized children. Am. J. Clin. Nutr. 1980; 33:1140-1146[Free Full Text]
• Pelletier D. The potentiating effects of malnutrition on child mortality: Epidemiologic evidence and policy implications. Nutr. Rev. 1994; 52:409-415[Medline]
• Pelletier D., Frongillo E., Habicht J.-P. Epidemiologic evidence for a potentiating effect of malnutrition on child mortality. Am. J. Public Health. 1993; 83:1130-1133[Abstract/Free Full Text]
• Reyes, H., Tomé, P., Guiscafré, H., Martínez, H. & Romero, G. (1993a) Autopsia verbal en niños con infección respiratoria y diarrhea aguda. Análisis del proceso enfermedad-atención-muerte. Bol. Med. Hosp. Infant Méx. 50: 7-15.
• Reyes, H., Tomé, P., Guiscafré, H., Martínez, H. & Gutiérrez, G. (1993b) Factores de riesgo de mortalidad por infección respiratoria aguda en niños menores de cinco años. Bol. Med. Hosp. Infant Méx. 9: 770-771.
• Secretaría de Salud (1997a) Manual de Procedimientos Técnicos del Programa de Atención a la Salud del Niño. Enfermedades Diarreicas. Mexico City, Mexico.
• Secretaría de Salud (1997b) Manual de Procedimientos Técnicos del Programa de Atención a la Salud del Niño, Enfermedades Respiratorias Agudas, Mexico City, Mexico.
• Secretaría de Salud. Norma Oficial Mexicana NOM-008-SSA-1993 (1994) Control de la nutrición, crecimiento y desarrollo del niño y del adolescente. Criterios y procedimientos para la presentación del servicio, Diario Oficial, Secretaría de Salud, México City, Mexico, April 1994.
• Sepúlveda, A., Lezana, M. A., Tapia, R., Valdespino, J., Madrigal, H. & Kumate, J. (1990) Estado nutricional de preescolares y mujeres en México: Resultados de una encuesta probabilística nacional. Gaceta Méd Méx 126: 207-225.
• Villa S., Guiscafré H., Martínez H., Urbán J. C., Reyes S., Lezana M. A., Gutiérrez G. Muertes en el hogar en niños con diarrhea o infección respiratoria aguda después de haber recibido atención médica. Bol. Med. Hosp. Infant Mex. 1994; 51:233-242
• Waterlow J., Buzina R., Keller W. The presentation and use of height and weight data for comparing the nutritional status of groups of children under the age of 10 years. Bull. WHO 1977; 55:489-498[Medline]
• World Health Organization (WHO) (1985) The quantity and quality of breast milk. Report on the WHO Collaborative Study on Breast-feeding, WHO, Geneva, Switzerland.
• Yambi O., Latham M., Habicht J.-P. Nutrition status and the risk of mortality in children 6-36 months old in Tanzania. Food Nutr. Bull. 1991; 13:271-276

0022-3166/98 $3.00 ©1998 American Society for Nutritional Sciences