QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Morris, S. S. Articles by Zúniga, M. Search for Related Content PubMed PubMed Citation Articles by Morris, S. S. Articles by Zúniga, M.

---

Article

Geographic Targeting of Nutrition Programs Can Substantially Affect the Severity of Stunting in Honduras¹

Food Consumption and Nutrition Division, International Food Policy Research Institute, Washington, DC 20006 and ^* Programa de Asignación Familiar, Colonia Palmira, Tegucigalpa, Honduras

²To whom correspondence should be addressed.

	ABSTRACT

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 
The effect of nutrition intervention programs in developing countries is likely to vary with the degree to which the program can be successfully targeted at the most vulnerable. In Honduras, the existence of a recent census of the height of first-grade children makes it possible to assess a priori the effect of different targeting strategies, holding constant other features of a hypothetical program. We simulate a nutrition intervention with 20% national coverage and uniform gains of 0.5 Z-scores for all beneficiaries, with a number of different approaches to targeting. The VIIth National Census of First-Graders’ Heights provides the baseline scenario and permits identification of priority departments, municipalities, schools and individuals, for a total of six alternative targeting mechanisms. Effect is assessed on the basis of changes in the prevalence of stunting (less than -2 Z-scores) and in two different measures of the severity of stunting adapted from the economics literature (the malnutrition gap and the quadratic malnutrition gap). We find that the simulated program has the potential to substantially improve the severity, but not the prevalence of stunting in Honduras. Household targeting with an imperfect indicator of vulnerability could reduce the malnutrition gap by >20% and the quadratic malnutrition gap by >30%, but would be very expensive to implement. "Broad stroke" geographic targeting could reduce the same measures by 15 and 20%, respectively, and would be much less expensive to implement. We conclude that geographic targeting has the potential to substantially enhance the effect of nutrition programs on the severity of stunting in Honduras.

KEY WORDS: • nutrition programs • targeting • Honduras • impact measures • children

	INTRODUCTION

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 
With 38% of children between 12 and 71 mo of age stunted (Government of Honduras 1997a ), Honduras has one of the highest rates of malnutrition in the Western hemisphere. It is also one of the poorest countries, with a Gross National Product per capita of only $740 per annum (World Bank 1999 ). Although the Honduran government has previously indicated a desire to bring the prevalence of stunting down to 27% by the year 2015 (Government of Honduras 1997b ), the large number of competing social sector priorities and the high costs of reaching a dispersed rural population make it unlikely that either the government or the nongovernmental sector will be able to provide effective nutrition interventions to more than a fraction of all children at risk. Indeed, the government’s own "Strategic Framework for Food and Nutrition Security Policies for the Medium and Long Term" (Government of Honduras 1997b ) has explicitly recognized that successful interventions would have to be carefully targeted to the most vulnerable.

Although much has been written about the targeting of poverty alleviation programs in Latin America (Grosh 1994 ), the expected benefits of targeting nutrition interventions are less well understood. In particular, there is very little evidence that can be used to assess the extent to which malnutrition clusters within subnational regions, communities or even households. It is a truism that geographical targeting is likely to be successful only when the divisions selected are both internally homogeneous and highly differentiated one from another. On the other hand, if administrative costs are to be kept from swamping the entire program budget, household-or individual-level targeting would probably require identifying indicators that could adequately predict malnutrition at the household level without requiring repeated anthropometric assessments of individual children.

In this paper, we examine the potential gains associated with different types of administrative targeting of large-scale nutrition interventions in Honduras. The magnitude of such gains can be evaluated a priori in Honduras because of the existence of regularly updated national censuses of the nutritional status of first-grade school students, which permit the accurate determination of the prevalence of malnutrition at a local level. We take the most recent of these censuses as a "baseline" scenario against which we assess the potential gains from a variety of large-scale, hypothetical nutrition interventions. Specifically, for each intervention, six different approaches to beneficiary targeting are simulated, ranging from the random selection of beneficiaries through to perfect individual-level targeting. Methods for assessing effect are adapted from the economics literature because standard nutritional measures are highly insensitive to changes of a level conceivable in a program context.

	SUBJECTS AND METHODS

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 
This analysis is based on data from the 7th National Census of First-Graders’ Height, conducted by the Government of Honduras in March 1997 (Government of Honduras 1997c ). The census covered 97.3% of all primary schools in the country and was estimated to have a total coverage of nearly 95% of all children registered in first grade. Although enrolment in primary school is often delayed in Honduras, recent figures 1993 indicate that >95% of children aged 8–9 y of age are enrolled in school (United Nations Educational, Scientific and Cultural Organization 1998 ), suggesting that the census of first-graders likely captures nearly all children of the relevant age group.

Data recorded in the 1997 Census included school identity, type (government vs. private) and location, and the birth date, height, sex and grade repetition status of each child. Height measurement was carried out by the regular schoolteachers, using a standardized methodology. For the purposes of the current analysis, age- and sex-standardized measures of height were calculated using the National Center for Health Statistics (NCHS) growth reference (Hamill et al. 1977 ). The sample sizes in the current analysis differ somewhat from the published report of the Census results. This is because all children with dates of birth between Jan 1, 1985 and December 31, 1991 were included in this analysis, assuming an interview date of March 4, 1997 (the modal date) where this was not recorded. Children (n = 1980; <1%) with height-for-age Z-scores (HAZ) outside the permissible range of -6 to + 4 were excluded, as were the approximately one fifth of children who were repeating first grade rather than being first-time enrollees. It was considered that these children were highly select, could not be deemed representative of all children in their school catchment areas and probably should not have been included in the Census in the first place if the intention was to approximate population-based indicators. The final sample for analysis was therefore 191,110 children, distributed among 8014 schools, in 297 municipalities (municipios) and in 18 departments (departamentos).

Even after standardizing measured heights using the NCHS reference, it was still considered problematic to compare rates of malnutrition across locations because of a complex of the following three factors: 1) children of markedly different ages ranging from 62 to 146 mo were included in the sample; 2) there were systematic differences in the age profiles of these children from one location to another; and 3) older children had systematically lower HAZ in this population. To the extent that nutritional status does routinely deteriorate with age between 62 and 146 mo in Honduras, areas with delayed enrolment will appear to have worse nutritional problems because the children in first grade are relatively old. To address this problem, the measures were restandardized by using a polynomial linear regression model to estimate the empirical relationship among HAZ, age and sex, and by adding the residual from this model to the predicted value for a male child aged exactly 7 y. This procedure resulted in a <0.5% reduction in the overall national prevalence of stunting (HAZ less than -2), but did alter the ranking of the 18 departments, specifically those in the middle of the distribution.

The analysis of the potential of targeted interventions to reduce the prevalence and severity of malnutrition (stunting) was based on three different measures. The first of these, P₀, is the prevalence of stunting, which is calculated as follows:

where i indexes the child and x_i takes the value 1 when the child’s HAZ is less than -2, and 0 otherwise. Although this measure is widely used to evaluate the effect of nutrition interventions, it is highly sensitive to small changes around the threshold value of -2 and at the same time not at all sensitive to large changes occurring at lower values. For the purposes of impact evaluation, this is undesirable because a change from -4 to -2.5 Z-scores is of much greater functional significance than a change from -2.1 to -1.9 Z-scores. In many respects, a better measure is the malnutrition gap, P₁, which can be interpreted as a per capita measure of the total shortfall of individual HAZ below the threshold value of -2; it is the sum of all shortfalls divided by the population, calculated as follows:

where z_i = HAZ if HAZ is less than -2, and z_i = -2 if HAZ is -2 or greater.

Thus, by way of illustration, P₁ increases by a quantity equal to 2/n for each child with HAZ = -4, by 0.5/n for each child with HAZ = -2.5 and by 0 for each child with HAZ = 1. Our final measure, P₂, is referred to as the quadratic malnutrition gap; it is similar to P₁ except that the difference between a stunted child’s Z-score and the threshold value of -2 is squared before averaging over the population. The measure is calculated as follows:

where z_i = HAZ if HAZ is less than -2, and z_i = -2 if HAZ is -2 or greater.

Thus P₂ increases by a quantity equal to 4/n for each child with HAZ = -4, 0.25/n for each child with HAZ = -2.5 and 0 for each child with HAZ = 1. This means that improvements in the Z-scores of severely malnourished children are weighted much more heavily than comparable improvements for the marginally malnourished. The three measures, P₀, P₁ and P₂, are adaptations of a series of poverty measures familiar to economists as the Foster-Greer-Thorbecke poverty measures (Foster et al. 1984 ).

For this analysis, we assume initially that a nutrition program can be implemented that would achieve a coverage of 20% of all children nationwide. Although this is clearly ambitious, it is consistent with the current policy framework being implemented in Honduras [see, for example, Interamerican Development Bank (1998) ]. We further assume that the hypothetical program would result in all beneficiary children registering an improvement of 0.5 Z-scores of height-for-age. We recognize that such a large effect would actually be difficult to achieve in a program context, given that continuous supplementary feeding from 0 to 36 mo resulted in cumulative effects at 36 mo of <1 Z-score of length-for-age in a small, carefully controlled, randomized trial in Colombia (Lutter et al. 1990 ). However, for this analysis, we consider it important to evaluate the potential impact of a highly effective intervention. Finally, we assume that program nonbeneficiaries are unaffected by the program, and that factors external to the program affect beneficiaries and nonbeneficiaries in equal measure. We then simulate six different targeting strategies as follows: 1. Perfect targeting. Beneficiaries are the C% of children with the lowest preintervention HAZ (where C is equal to the program coverage). Perfect targeting is, of course, impossible in a program context because it requires constant monitoring of child anthropometric status. It is included in our analysis purely for comparative purposes. 2. Individual targeting using an imperfect screening indicator (proxy). A proxy indicator is developed for which 50% of the total variance is explained by true HAZ and the remaining 50% is random noise. Beneficiaries are the C% of children with the lowest scores on this proxy indicator. Such an approach might be effected by identifying a set of household-level variables that are highly correlated with child anthropometric status, and then conducting a rapid census of all households in the country and scoring their risk level on the basis of these indicators. 3. Targeting by school. Beneficiaries are the children attending schools with the lowest mean preintervention Z-scores. 4. Targeting by municipality. Beneficiaries are the children living in municipalities with the lowest mean preintervention Z-scores. 5. Targeting by department. Beneficiaries are the children living in departments with the lowest mean preintervention Z-scores. 6. Untargeted. Beneficiaries are a random C% selection of children. This situation, or something similar, might arise in the case of a program with selection criteria totally uncorrelated with child anthropometric status.

To test the sensitivity of the results to different assumptions about the magnitude of the intervention effect and the coverage of the program, all analyses were repeated with intervention effects ranging from 0.0 through to 0.5 Z-scores gained per beneficiary, and program coverages of 10, 20 and 30%.

Parents in Honduras are fully aware that their children’s heights will be determined and recorded during the course of the school year. The current research thus complies with the Helsinki Declaration as revised in 1983.

	RESULTS

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 
Effect of targeting on the prevalence of stunting.

Table 1 shows the effect on the prevalence of stunting (P₀) of a program with a 20% coverage rate and gains of 0.5 Z-scores for all beneficiary children, using six different approaches to targeting. By design, program coverage was virtually identical, regardless of the method of targeting used. The most striking result of this analysis was that perfect targeting would leave the prevalence of stunting in Honduras completely unchanged. This is because all of the 20% of children with the lowest HAZ have preintervention values below -2.5, and would thus fail to exceed the -2 threshold even if they gained the full 0.5 Z-scores as a result of the program. All other targeting mechanisms, including random selection of beneficiaries, were associated with better outcomes because some children with preintervention Z-scores in the range -2.5 to -2.0 were included. Individual targeting using a proxy indicator with equal parts noise and information, and geographic targeting by school, municipality or department were all associated with similar outcomes, with the rate of stunting showing a decline from 42.1 to 38.5–38.6%. Random selection of beneficiaries was minimally less effective.

Table 2 shows the effect of the hypothetical program on the severity of malnutrition as measured by P₁ (the malnutrition gap) and P₂ (the quadratric malnutrition gap). Perfect targeting reduced the malnutrition gap from 0.355 Z-scores to 0.255 Z-scores. This follows logically from the fact that 20% of all children, all of them initially stunted, gained 0.5 Z-scores as a result of the program, whereas the remainder were unaffected, giving an average gain of 0.2 x 0.5 = 0.1 Z-scores for the whole population. In relative terms, this scenario leads to a reduction in the malnutrition gap of 28.1%. Geographic targeting was less effective, leading to reductions of 14.8% (targeting by department, the least effective geographic strategy) to 18.6% (targeting by school, the most effective strategy). With respect to effect on the malnutrition gap, there was virtually no difference between targeting by municipality and targeting by department. Individual targeting using a 50%-noise proxy indicator was somewhat more effective than the most effective geographic targeting; this was associated with a 21.2% reduction in the P₁ measure. As expected, random selection of beneficiaries resulted in the lowest effect (a 9.5% reduction in P₁).

Sensitivity of results to different assumptions about the coverage and effectiveness of the intervention.

To determine the generalizability of the findings reported above to situations in which the intervention was less effective (resulted in a <0.5 Z-score gain for program beneficiaries), or achieved greater or < 20% coverage, all analyses were repeated for a range of different assumptions. Figure 1 shows that for the case of individual targeting using an imperfect screening indicator, increasing the effectiveness of the intervention (from 0 up to 0.5 Z-scores gained per beneficiary) was associated with an almost perfectly linear increase in effect, measured as the percentage reduction in malnutrition (any measure). This linear association between the effectiveness of the intervention and program effect was in fact observed for all six targeting strategies, regardless of program coverage (data not shown). Figure 2 shows the reduction in malnutrition (each of the three measures) associated with individual targeting using an imperfect screening indicator, at three different levels of program coverage and assuming in each case a program-induced gain of 0.5 Z-scores per beneficiary. Some nonlinearity was observed, so that there are increasing returns to scale when effect was measured as the prevalence of stunting, but decreasing returns to scale when the effect measure chosen was more sensitive to changes in the severity of malnutrition (malnutrition gap or quadratic malnutrition gap). The same pattern was observed at lower levels of intervention effectiveness (data not shown). However, when a less effective targeting strategy such as large-area geographic targeting was adopted, the association between coverage and effect approached linearity. This is illustrated in Figure 3 , which shows the reduction in malnutrition associated with geographic targeting at the department level, at three different levels of program coverage and assuming gains of 0.5 Z-scores per beneficiary in each case.

View larger version (26K): [in this window] [in a new window]   Figure 1. Association between the magnitude of the intervention effect and program effect, using three alternative measures of the burden of malnutrition. Simulation study based on individual targeting using an imperfect screening indicator and 20% program coverage.

View larger version (25K): [in this window] [in a new window]   Figure 2. Association between program coverage and effect, using three alternative measures of the burden of malnutrition. Simulation study based on individual targeting using an imperfect screening indicator and an intervention effect of 0.5 Z-scores gained per beneficiary.

View larger version (23K): [in this window] [in a new window]   Figure 3. Association between program coverage and effect, using three alternative measures of the burden of malnutrition. Simulation study based on departamento-level geographic targeting and an intervention effect of 0.5 Z-scores gained per beneficiary.

	DISCUSSION

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

In this paper, we propose two novel measures of effect for population-based nutrition interventions, based on the Foster-Greer-Thorbecke poverty measures (Foster et al. 1984 ). The first of these measures, which we call the malnutrition gap, can be interpreted as a per capita measure of the total shortfall of individual HAZ below the threshold value of -2. This measure values equally a child moving from -4 Z-scores to -3.5 Z-scores and another moving from -2.5 to -2, but disregards changes above the threshold value of -2. The second measure, which we call the quadratic malnutrition gap, overweights gains among the severely malnourished relative to similar gains among the moderately malnourished. We find this measure to be more consistent with the epidemiologic literature on the associations between poor anthropometric status and health outcomes. This literature clearly indicates that a child at -4 Z-scores is far more likely to die than one at -2.5 Z-scores (Pelletier 1994 ).

We find that a perfectly targeted program with 20% coverage and 0.5 Z-score gains would reduce the malnutrition gap by >28% and would reduce the quadratic malnutrition gap by nearly 50%. We recognize, however, that perfect targeting is not a realistic option for nutrition programs because of the expense of screening, the rapid evolution of nutritional status in young children and the fact that such a system creates incentives for parents to withhold food and other resources from their children in order to qualify for the program benefits. At best, then, a system of individual screening would be based on indicators of vulnerability to stunting, which might or might not correlate highly with achieved nutritional status. We have simulated a proxy indicator of vulnerability that is highly correlated with actual nutritional status because 50% of the total variance in the proxy measure is explained by true status (equivalent to a correlation of r = 0.71). If our hypothetical program were implemented using this proxy indicator for targeting, the malnutrition gap would be reduced by 21% and the quadratic malnutrition gap by 32%. These effects are (predictably) lower than those that would result from perfect targeting, but it is very likely that this reduction would be justified by substantially lower administrative costs. Our recent experience in Honduras would suggest that implementing a screening tool based on proxy indicators would cost US$10/household. This cost is based on a single application of a two-page questionnaire, with 15 questionnaires completed per enumerator per day and one supervisor (plus one cartographer) to every four enumerators. With a total of approximately one million households in Honduras, the screening cost-effectiveness ratio using household-level indicators would be <US$500,000 per percentage point decrease in the malnutrition gap, or US$312,500 per percentage point decrease in the quadratic malnutrition gap (again assuming 20% program coverage and a 0.5 Z-score gain for all beneficiaries).

Geographic targeting is logistically simpler than individual targeting and also reduces the costs of program implementation. In this analysis, we have shown that targeting by school could lead to a 27% reduction in the P₂ measure of the severity of stunting, which we feel would be a very significant achievement in a country such as Honduras. In addition, it is only slightly lower than the benefit expected with individual targeting by proxy indicator. Targeting by municipality or by department would be associated with somewhat lower benefits, whether measured by P₁ or P₂. We believe that a school height census provides sufficient information to implement municipality-level (or department-level) targeting. The total cost of the 1997 school height census was <US$40,000, implying that for municipality-level targeting, the screening cost effectiveness ratio would be only US$2532 per percentage point decrease in the malnutrition gap, or US$1843 per percentage point decrease in the quadratic malnutrition gap.

It is interesting to note that in Honduras, targeting at the municipality level appears to be no more effective than targeting at the department level, regardless of the criterion used to quantify effect. Targeting at the school level does appear to be more effective, but this higher potential effect must be weighed carefully against the greater costs of delivering a program targeted at the school rather than at the department level, i.e., the 2514 schools selected under the former strategy (assuming 20% coverage) are distributed across 17 of the 18 departments in the country, compared with just 5 under the department-level targeting scenario. Furthermore, the school-level estimates derived from the school height census are in some cases based on quite small numbers of students, and could possibly be quite unstable from one year to the next. This could lead to the highly undesirable phenomenon of "shifting targets" in a program of several years’ duration.

The worst-case scenario is the random selection of program beneficiaries. This would be considerably less effective than even "broad stroke" geographical targeting if performance is judged by the P₂ measure, although the relative advantage of "broad stroke" geographic targeting is less apparent when effect is judged by the P₁ measure. We have not attempted to quantify the potential effect of mixed strategies, such as limited geographic targeting plus household screening using a proxy indicator. Such strategies might combine administrative ease (and manageable budgets) with acceptable effect. In a program context, it would be essential to incorporate into the analysis information on the administrative costs of operating over dispersed areas and of applying the screening test.

The comparison of different targeting options for nutrition programs is possible in Honduras because of the existence of a national census of the nutritional status of first-graders. Strictly speaking, the results of this analysis are applicable only to interventions focused on first-graders because we cannot be sure that the nutritional status of younger children in Honduras is similarly distributed among local communities, municipalities and departments. It is the case, however, that sample surveys of the nutritional status of children <5 y old in Honduras have consistently found regional patterns that match the departmental rankings evident in the school height census [see, for example, Government of Honduras (1997a) ]. Data presented by Parillón and co-workers (1988) indicate a rank correlation of 0.75 between school height census and nutrition survey estimates of the prevalence of malnutrition in the 65 distritos of Panama, despite the fact that the two measurements were made 2 y apart, which would presumably attenuate the true association. The authors do not present the data necessary to calculate the Pearson correlation coefficient, but this would usually be higher than the rank correlation if there are any "outlier" distritos that are particularly high (or low) on both measures. It should be borne in mind, however, that if the correlation between the two measures in Honduras is much <1, then the present results will overstate the expected effect of targeting based on the results of school height censuses on the prevalence or severity of malnutrition in preschool children.

We chose to age-adjust the data from the Honduras school height census because of regional differences in mean age at enrollment. It could be argued, however, that the age bias works in favor of targeting the poorest schools and areas because schools with the most delayed enrollment are situated in the poorest areas, where the most malnourished children are found. This point should be considered by those intending to use school height census data for targeting; it does not substantially affect the present results, however, because there was only a 25-mo difference between the 10th and 90th centiles of the age distribution in the school height census data used. Clearly, in other countries in which school height censuses are not able to achieve such high coverage as in Honduras and/or selectivity biases are more pronounced, the usefulness of this tool would be diminished, and other sources of information might have to be explored.

In conclusion, geographic targeting has the potential to substantially enhance the effect of nutrition programs on the severity of stunting in Honduras, especially if fine-level targeting is feasible. Assuming that effect is quantified using a measure that distinguishes between gains to the severely malnourished and gains to the marginally malnourished, such as the P₂ measure proposed in this paper, any kind of geographic targeting is better than none. No form of geographic targeting is able to achieve the level of effect seen with individual-level targeting, even if the indicator used does not correlate perfectly with achieved nutritional status. However, the costs of individual-/household-level screening are several orders of magnitude greater than for geographic targeting, and the administrative costs of running a geographically dispersed program are also expected to be very large. It will be important to determine whether similar results pertain in other countries with different baseline levels and distributions of childhood malnutrition.

	ACKNOWLEDGMENTS

 
The authors thank the coordinators of the VIIth National Census of First-Graders’ Height, Leonarda Munguía, Angelina Reyes, Mirian Leiva, Alejandrina de Padilla, and Martha Leiva, as well as the 36 other departmental managers and national supervisors, and the 10,200 first-grade teachers who conducted the measurements.

	FOOTNOTES

 
¹ Supported by the Government of Honduras as part of consultancy HON/99/007462–99. However, the opinions expressed are those of the authors only and should not be attributed to any other group or organization.

Manuscript received November 2, 1999. Initial review completed February 1, 2000. Revision accepted June 14, 2000.

	REFERENCES

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 

1. Foster J., Greer J., Thorbecke E. A class of decomposable poverty measures. Econometrica 1984;52:761-765

2. Government of Honduras National Micronutrient Survey, Honduras 1996. Executive Report 1997a Secretariat for Health, Subsecretariat for Population Risks Tegucigalpa, Honduras

3. Government of Honduras Strategic framework for food and nutrition security policy in the medium- and long-term 1997b Secretariat for Health/Technical Secretariat for International Cooperation/Secretariat for Agriculture and Livestock Tegucigalpa, Honduras.

4. Government of Honduras VIIth National Height Census. 1997 Report 1997c Secretariat for Education Tegucigalpa, Honduras.

5. Grosh M. E. Administering Targeted Social Programs in Latin America: from Platitudes to Practice 1994 The World Bank Washington DC.

6. Hamill, P.V.V., Drizid, T. A., Johnston, C. L., Reed, R. B., Roche, A. F. & Moore, W. M. (1977) NCHS growth curves for children birth-18 years. Vital Health Stat. Series 11: DHEW Publication no. (PHS)78–1650. USHEW-PHS, Hyattsville, MD.

7. Interamerican Development Bank (1998) Family Allowance Program—Phase II. HO- 0132: Executive Summary. <http://www.iadb.org/exr/doc98/apr/ho1026e.pdf>. Updated December 16. IADB, Washington, DC.

8. Lutter C. K., Mora J. O., Habicht J.-P., Rasmussen K. M., Robson D. S., Herrera M. G. Age-specific responsiveness of weight and length to nutritional supplementation. Am. J. Clin. Nutr. 1990;51:359-364[Abstract/Free Full Text]

9. Parillón C., Valverde V., Delgado H. Description of a methodology to identify and quantify poor and malnourished family groups in the Republic of Panama. Arch. Latinoam. Nutr. 1988;38:31-41[Medline]

10. Pelletier D. The relationship between child anthropometry and mortality in developing countries: implications for policy, programs and future research. J. Nutr. 1994;124:2047S-2081S

11. United Nations Educational, Scientific and Cultural Organization UNESCO Statistical Yearbook, 1998 1998 UNESCO Paris, France.

12. World Bank 1999 World Development Indicators 1999 The World Bank Washington, DC.

This article has been cited by other articles:

				S. S. Morris and R. Flores School Height Censuses Are Reliable and Valid Tools for Small-Area Targeting of Nutrition Interventions in Honduras J. Nutr., June 1, 2002; 132(6): 1188 - 1193. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Morris, S. S. Articles by Zúniga, M. Search for Related Content PubMed PubMed Citation Articles by Morris, S. S. Articles by Zúniga, M.