HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		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 Citation Map 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 Moursi, M. M Articles by Delpeuch, F. Search for Related Content PubMed PubMed Citation Articles by Moursi, M. M Articles by Delpeuch, F. Agricola Articles by Moursi, M. M Articles by Delpeuch, F.

Assessment of child feeding practices using a summary index: stability over time and association with child growth in urban Madagascar^1,2,3,4

¹ From Research Unit 106 "Nutrition, Food, Societies" (World Health Organization Collaborating Centre for Nutrition), Institut de Recherche pour le Développement (IRD), Montpellier, France (MMM, Y-MP, SE-D, GC, BM, and FD); Doctoral School 393 "Public Health: Epidemiology and Biomedical Information Science, " Université Pierre et Marie Curie, Paris, France (MMM); and Research Unit 106 "Nutrition, Food, Societies" (World Health Organization Collaborating Centre for Nutrition), Institut de Recherche pour le Développement (IRD), IRD Representation at Madagascar, Antananarivo, Madagascar (ST)

² Supported by the Nutrimad project and the Institut de Recherche pour le Développement.

³ Address reprint requests to F Delpeuch, Research Unit 106 "Nutrition, Food, Societies," Institut de Recherche pour le Développement (IRD), BP 64501, F-34394 Montpellier Cedex 5, France. E-mail: francis.delpeuch{at}ird.fr.

⁴ Address correspondence to MM Moursi, Research Unit 106 "Nutrition, Food, Societies, " Institut de Recherche pour le Développement (IRD), BP 64501, F-34394 Montpellier Cedex 5, France. E-mail: mourad.moursi{at}mpl.ird.fr.

	ABSTRACT

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 
Background: Previous studies investigating the association between an infant and child feeding index (ICFI) and length-for-age were based on a cross-sectional design and on the assumption that data collected with brief recalls could provide information about more enduring processes.

Objectives: The objectives were to test the stability of the individual ICFI values over time and to investigate how they relate to length-for-age z score (LAZ) and weight-for-length z score (WLZ) at the end of the study.

Design: This prospective cohort study included 363 children aged 6–17 mo who were visited 3 times over 6 mo. A cross-sectional ICFI (CS-ICFI) was constructed for each visit by using data on feeding practices and data from quantitative 24-h recalls. A longitudinal ICFI (L-ICFI) was constructed with use of the 3 CS-ICFIs. The stability of the CS-ICFI was assessed by using the variance of the repeatability coefficient (s²r).

Results: Stability of the CS-ICFI was shown by the value of 0.704 (95% CI: 0.625, 0.805) of the s²r, which differed significantly from 1 (P < 0.0001). There was no significant association between the CS-ICFIs and LAZ or WLZ at visit 3. In contrast, when moving from low to high L-ICFI, there was a highly significant 0.5 z score difference in mean LAZ at visit 3 (P = 0.0008). The L-ICFI was not associated with WLZ.

Conclusions: The ICFI constructed by using data collected with brief recalls can provide information about feeding in the long term. However, the absence of association with LAZ suggests a lack of precision that can be reduced by using an ICFI based on repeated measurements.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 
The importance of child feeding practices for child nutrition and health is well recognized in the scientific literature (1), and this topic has received increased attention over the past few years, especially after the publication of updated guidelines in 2002 (2). This increased attention has led, among other things, to innovative ways of measuring and assessing the adequacy of child feeding practices with use of a summary index as shown in several recent studies (3-7). This previously developed (6) cross-sectional infant and child feeding index (CS-ICFI) included information on current breastfeeding and bottle feeding, dietary diversity (consumption of various food groups in the past 24 h), food group frequency (frequency of use of various food groups in the previous 7 d), and feeding frequency (in the past 24 h). Combining various aspects of child feeding into an age-specific CS-ICFI provides a first answer to the question of how best to deal with recommended practices that vary within narrow age ranges. Most importantly and contrary to the study of single behaviors, the CS-ICFI is thought to be more likely to detect associations with various outcomes if positive (or negative) practices tend to cluster or if there is a minimum number of positive practices required to achieve benefits as suggested by previous research (8). A cross-sectional analysis of our data showed that the infant and child feeding index (ICFI) was positively correlated with energy intake from complementary foods and dietary quality estimated by mean micronutrient density adequacy (9). The CS-ICFI was intended for use in the context of developing countries, and, to our knowledge, no such composite indices were developed nor used for children younger than age 2 y in industrialized countries. Even if the CS-ICFI can be used for several purposes, recent studies focused primarily on analyzing its association with child growth measured by height-for-age. However, results were contrasted, with some studies finding a positive association (4-6), whereas others did not (6, 7). In all of these cross-sectional studies with the CS-ICFI, a major assumption has been made that summary measures of infant feeding constructed by using data collected with a brief recall (eg, 24 h or 7 d) could provide information about more enduring processes, ie, that mothers who follow good infant feeding practices at the time of the study are likely to have been feeding their children adequately and, hence, the attempts to study the relation between the CS-ICFI and linear growth, which is a summation of a long process.

In this context of epidemiologic studies reporting conflicting results, we tested this major assumption by studying the stability of the CS-ICFI at different time points using data from a prospective cohort study conducted in Madagascar. In this analysis we specifically sought to address the question of whether there are significant changes over time in the individual CS-ICFI values. Concurrently with that first objective, we wanted to investigate how cross-sectional and longitudinal assessments of feeding practices [with use of the CS-ICFI or a longitudinal ICFI (L-ICFI), which summarizes the information from several CS-ICFIs, respectively] relate to child length-for-age and weight-for-length at the end of the study.

	SUBJECTS AND METHODS

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 
Study participants and study design
This study was conducted within the Nutrimad project, the main objective of which was to contribute to the prevention of early childhood malnutrition by combining nutrition education with marketing of quality complementary foods. The study population was composed of mother-infant dyads (with infants aged 6–18 mo) living in the districts of Sahalava and Antsororokavo, town of Fianarantsoa, Madagascar. The 2 districts were selected by the Nutrimad staff as future intervention sites on the basis of available information on population size, health infrastructure, sociodemographic factors, and prevalence of childhood malnutrition. This information also showed that the 2 districts shared many characteristics with other poor urban districts of Fianarantsoa and did not present any exceptional features. There was no intervention of any kind, either of the Nutrimad project or another project, during the study.

Of the 532 infants aged 6–17 mo identified during the census before the study, all those who met the following eligibility criteria were invited to participate: 1) child born in the district, 2) no serious infant illness that required regular hospitalization and no serious infant malformation that could affect anthropometric measurements, and 3) no intention of moving out of the district in the next 6 mo. Fourteen mothers refused to participate and a total of 440 eligible infants aged 6–17 mo were included in the study at baseline, of whom 51 were lost to follow-up because their parents left the town of Fianarantsoa (n = 44) or because of infant death (n = 7). An additional 26 infants were excluded from this analysis because of missing information at the second visit that fell within the period of summer vacation (absence after 3 attempts of home visits). The analysis was therefore conducted on a final sample size of 363 infants and young children. The study was performed over 7 mo, from mid-March to mid-November 2004 and comprised 3 visits at 3-month intervals ± 15 days.

Data collection
Family demographic and socioeconomic information and feeding history were ascertained at baseline with a questionnaire. Current feeding practices and consumption of food groups during the week preceding each visit were also assessed with questionnaires. Child morbidity in the week preceding each visit was recorded for fever, nasal discharge, cough (mild, moderate, or severe), lower than normal food intake due to illness, and diarrhea. We defined lower or upper respiratory infection as cough and nasal discharge with or without fever; and diarrhea as 3 or more soft, watery stools per day. Data on mother and child anthropometry were collected at each visit with the exception of the mother's standing height which was measured only once at the baseline visit to the nearest millimeter by using locally produced boards. Infant and young child recumbent length was measured to the nearest millimeter by using Shorr infantometers (Shorr Productions, Olney, MD) by 5 trained individuals (MM and 4 investigators). Children were double weighed in the mother's or caregiver's arms. Weight was collected by using electronic scales with a maximum capacity of 150 kg (precision 100 g; Tefal, Rumilly, France). Anthropometric indices were calculated with Stata macros provided by the World Health Organization with use of the new growth reference (10). Stunting, wasting, and underweight were defined as length-for-age z score (LAZ) –2, weight-for-length (WLZ) z score –2, and weight-for-age z score –2, respectively.

Dietary intake of infants was estimated by using a quantitative 24-h recall. Portion sizes were determined using standardized household measures (11). All recalls were performed by a team of 5 investigators who were extensively trained and regularly supervised. Meals and snacks were self-defined by mothers. The number of feeding episodes was defined as the sum of meals and snacks per day.

Economic index
An economic index was constructed as suggested previously (12). Multiple correspondence analyses were performed by using an asset index along with house size, type of floor, water source, electricity, and type of sanitation facilities. The first axis represented 37% of the variance and the coordinates of the individuals on this axis were used as an economic index. This continuous index was then cut into tertiles representing low, middle, and high economic status.

Cross sectional infant and child feeding index
The CS-ICFI was constructed as described previously (6). The variables and scoring system used are shown in Table 1. Scores for dietary diversity and food group frequency were attributed in a manner that reflects the age-specific distributions. Dietary diversity was calculated with use of the 24-h recalls. The 7 food groups used were based on guidelines proposed by the Food and Nutrition Technical Assistance Project (13). The dairy group did not include breast milk. Mixtures were separated into their ingredients before food group consumption was determined. A minimum of 1 g consumption was imposed for all food groups for them to count. Dietary diversity was calculated by summing the number of food groups consumed in the 24-h period. For food group frequency (past 7 days), which was assessed separately by using questionnaires, each food group was scored as 0 if not consumed during the previous week, +1 if consumed on 1–3 days, and + 2 if consumed on 4 days. These scores were summed to give a possible range of 0–14, and then new scores were assigned as described in Table 1 to reflect age-specific distribution. The attribution of the feeding frequency scores followed international recommendations (2) by attributing, for each age range, a score of +1 for children who meet the lower end of the recommendation, ie, 2 times for 6–8 mo and 3 times for 9 mo or more, and a score of +2 for those who meet or exceed the higher end. The exception was children in the second year of life for whom a score of +3 was attributed for those who exceed the recommendation to obtain a possible maximum score of 9 for the CS-ICFI in the higher age range, to keep it comparable to the maximum score at other ages. The CS-ICFI was calculated for each age group by adding up the scores obtained, giving a possible range of 0–9. The CS-ICFI was divided into 3 categories based as closely as possible on tertiles with all age groups together in the following manner: a score of 0–5 was considered low, 6–7 was considered medium, and 8–9 was considered high.

Statistical analysis
The normal distribution of the CS-ICFI was checked by using normal quantile plots. To examine the longitudinal pattern of CS-ICFI changes at the individual level (ie, for each child) we first had to take into account 2 disturbing facts: 1) that the CS-ICFI scoring system does change across age categories and 2) that all children were not enrolled at the same age. In other words, we faced the double issue of the effect of child age and of the so-called cohort-effect (which can be seasonal or linked to the children participating in the survey, for example). Therefore, to avoid these effects and to make CS-ICFI values comparable over time we decided to normalize their distribution and then to use the z scores of the specific age x group distribution instead of original CS-ICFI values. Second, we were challenged by the choice of a statistical model to test whether these CS-ICFI z values were stable for each child from one visit to the other. Statistically speaking, this stability means that the covariance between CS-ICFI z values measured at the different visits is positive and large enough. To estimate the strength of this covariance, we adapted statistics of repeatability from the tests used to judge the reliability of laboratory measurements (14, 15). The variance of the repeatability coefficient (s²r) for the 3 visits of the same child is conceptually an equivalent of the intralaboratory (intrasubject) part of the total variance of the measurement, which is the recommended way to estimate the measurement error (16). Because we used normalized values (z scores), s²r equals 1 when the covariance equals zero and s²r tends to zero as the strength of the covariance grows. In other words, this coefficient measures the size of room that can be occupied by CS-ICFI values visit after visit. A coefficient of 0 means that there is no room for different values, ie, the value cannot change from visit to visit, and indicates perfect stability. In contrast, a coefficient of 1 implies that CS-ICFI values have all the room to take different values and indicates the absence of stability. It is possible to test whether s²r is significantly different from 1.

Changes in the distribution of CS-ICFI components over time were examined at the group level. Multinomial logistic regression models that accounted for diarrhea, respiratory infections, and correlations owing to repeated measurements were used. Interaction terms of the distribution of the components over time with age range at baseline (6–8, 9–11, 12–14, and 15–17 mo) were introduced to explore whether the relation of the CS-ICFI components with time differed between younger and older children, but all interaction terms were not statistically significant at the P 0.20 level. Therefore, results are displayed for all ages combined.

The relation between the CS-ICFI or the L-ICFI with LAZ and WLZ was analyzed with use of anthropometric data from the third visit. Confounding factors of this relation were identified with use of chi-square tests and analysis of variance with basic adjustment for sex and age. Linear models with LAZ or WLZ as the outcome and the CS-ICFI or L-ICFI as explanatory variables were used. Linear trend was tested by including orthogonal polynomials in the regression model of the relation between the continuous L-ICFI and LAZ. Linear models were also used to separately model the association of each individual component of the CS-ICFI at visit 3 with LAZ and WLZ at visit 3. For each model, adjustments were made for the same confounding factors identified previously and for the effect of other respective components whenever possible (ie, when collinearity was not too strong). All analyses were performed by using Stata 9.2 (Stata Corp, College Station, TX) and statistical significance was set at 0.05.

Ethics
The study protocol was reviewed and approved by the Direction Provinciale de la Santé, the health department of Fianarantsoa province. Oral informed consent to participate was obtained from all subjects after they were given oral and written information on the objectives and content of the study. The study complies with the principles of the revised Helsinki Declaration.

	RESULTS

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 
Loss to follow-up
The 77 infants who were lost to follow-up or excluded from the analysis did not differ significantly from the remaining sample with respect to mean LAZ (P = 0.71), mean CS-ICFI (P = 0.52), or any sociodemographic characteristic.

Sample characteristics at baseline
Main sample characteristics at baseline are shown in Table 2. Child age distribution was relatively homogeneous. Breastfeeding is very common in Madagascar and is prolonged well into the second year of life; 96% of infants in our sample were breastfed at baseline.

CS-ICFI distribution at different time points
The raw values of the CS-ICFI mean and distribution according to children's age group and to the number of the visit are shown in Table 3. It illustrates the age effect and the cohort effect that were anticipated in Subjects and Methods. The approach with use of CS-ICFI z scores instead of original CS-ICFI values led to an estimate of the variance of the repeatability coefficient (s²r) of 0.704 (95% CI0.625, 0.805), which was significantly different from 1 (P < 0.0001). In addition, a Cochran test was performed to test the hypothesis of homogeneity of the repeatability variance between subjects, and this hypothesis was not rejected (P > 0.05).

Table 4

Figure 1

Figure 2

Figure 3

View larger version (30K): [in this window] [in a new window]   FIGURE 1.. Adjusted mean LAZ (95% CI) at visit (v) 3 by CS-ICFI groups at each visit and by L-ICFI groups. LAZ was adjusted for child age, sex and morbidity, mother's height, BMI at visit 3, and level of schooling and household economic level using a linear model. *Differs significantly from the low group (reference group).

View larger version (76K): [in this window] [in a new window]   FIGURE 2.. Adjusted mean LAZ (95% CI) at visit 3 by continuous L-ICFI values. Mean LAZ was adjusted for child age, sex and morbidity, mother's height, BMI at visit 3, and level of schooling and household economic level using a linear model. *Differs significantly from the 0/1 group (reference group).

View larger version (20K): [in this window] [in a new window]   FIGURE 3.. Adjusted mean WLZ (95% CI) at visit (v) 3 by CS-ICFI groups at each visit and by L-ICFI groups. WLZ was adjusted for child age, sex and morbidity, BMI at visit 3 and level of schooling and household economic level using a linear model.

Table 5

	DISCUSSION

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

To test stability of the CS-ICFI over time at the individual level, we derived our method from the statistical examination of the reliability of laboratory measurements because we did not find any straightforward model in the literature to estimate the concordance between several measures over time. For example, the kappa coefficient is not adapted for >2 judgment occasions and the Fleiss kappa coefficient does not account for ordered categorical data (17). The null hypothesis of other tests usually used to examine changes in the distribution of any values over time is that there is independence between 2 or more series of values rather than the absence of covariance. Alternatively, statistics such as the intraclass correlation coefficients proposed by Bland and Altman (18), for example, could be used. In this case the interpretation would be that measurements from a given subject are closer to other measurements from the same subject than to measurements made at random within the sample. Again this is not the same construct as the one of the concordance between several measurements. Nevertheless, we calculated such coefficients, and, unsurprisingly, they were highly significantly different from zero.

The relative stability of the CS-ICFI involved important underlying changes in its components at the group level. As one would expect, infants were breastfed less as they grew older, whereas feeding frequency increased over time. The results on food group diversity were contrasted, with percentages of children in both the low and high categories of dietary diversity decreasing between visit 1 and visit 3. Therefore, it seems that the natural tendency in the decline of the CS-ICFI as breastfeeding drops with age was compensated for by a higher feeding frequency and somewhat higher diversity, leading to CS-ICFI being unchanged over time overall. These findings highlight both the advantage and the main weakness of a summary measure such as the CS-ICFI when it is used as a monitoring and evaluation tool. Its weakness evidently is that it can mask important changes in individual practices. On the other hand, it is a useful tool for keeping track of changes in all practices simultaneously and, as in our case, remain unchanged when some practices change positively, whereas others change in the opposite direction, thus suggesting that there could be no global benefit. Whatever the case, this finding emphasizes the importance of associating the analysis of the CS-ICFI with that of its components. Some components such as dietary diversity or food group frequency score, which is another way of measuring dietary diversity, have the potential of being valid indicators of specific aspects of complementary feeding and are also linked to child growth (19-22).

There was no significant association between any of the 3 CS-ICFIs corresponding to the 3 visits and child LAZ or WLZ at visit 3. In contrast, the L-ICFI, which summarizes information from the 3 CS-ICFIs, was very strongly correlated with LAZ at visit 3 either in tertiles or in its "continuous" form, with a clear linear trend in the latter case. This result provides very strong evidence for the effect of feeding practices on child growth which, despite being widely recognized, was rarely as clearly demonstrated as in this study. The difference in mean LAZ when one moves from the low to the high tertile of the L-ICFI was 0.5, which is considered to be large and biologically meaningful and of a magnitude comparable to findings from other studies, albeit they were obtained with use of the CS-ICFI (4-6). We conducted the same analysis using the National Center for Health Statistics references to calculate LAZ and reached similar conclusions. To our knowledge, a longitudinal assessment of child growth with use of a feeding index was conducted in only one study (23). The authors found that the feeding index at 6 mo was strongly associated with impaired growth and an increased probability of stunting in the following 12 mo. Unfortunately, these results cannot be directly compared with our findings because of methodologic choices related to the fact that their study was conducted within an intervention targeted at HIV-infected mothers. For instance, the feeding index used attributed lower scores for breastfeeding and positive scores for bottle-feeding because breastfeeding carried a risk of HIV infection.

Reasons for absence of an association between LAZ and the CS-ICFI include the inverse relation between breastfeeding and LAZ: nonbreastfed infants scored 0 for the breastfeeding component of the CS-ICFI but had higher mean LAZ than breastfed children. However, we must acknowledge that this result is also the case with the L-ICFI, notably because the inverse relation holds true over all age categories. Therefore, a more probable explanation is that, because of potential systematic and random reporting errors (24-26), the CS-ICFI may simply lack precision and induce misclassification in some cases. In turn, any misclassification affects the hypothesis that current ICFI reflects past feeding practices, whereas, by construction, there are fewer misclassifications within the L-ICFI. In fact, previous studies that succeeded in demonstrating a relation between the CS-ICFI and LAZ were performed with large sample sizes (ranging from 1487 to 4834 subjects), whereas those that failed to demonstrate this association always included <1000 subjects. We are therefore inclined to think that in the case of the former studies, lack of precision of the CS-ICFI may have been compensated for by larger sample sizes that translated into higher statistical power. In the case of our study, the precision seems to have been improved with repeated measurements.

We recognize that the empirical approach used to construct the L-ICFI is one of the main limitations of our study. First, it relies on the CS-ICFI cut into tertiles with the threshold effect and possible initial misclassification that this categorization implies. We tested several alternative ways of constructing the L-ICFI, notably with use of the CS-ICFI in its continuous form, and reached similar conclusions. Then, there is the problem of how to deal with children who move between the 2 extreme categories of the CS-ICFI (ie, low and high) at any given time point and for which a pattern cannot be distinguished. We performed our analysis with and without these children (n = 66), and results were similar overall. Another limitation of this study is the lack of adjustment for birth weight as this information was missing for 64% of children.

In sum, this study showed that the CS-ICFI constructed with use of data collected from a brief recall can provide information about feeding in the long term. This assumption, however, was not as robust as hypothesized, and indeed the CS-ICFI was not associated with child LAZ in the context of our study. These findings suggest that the CS-ICFI may lack precision as an analytic tool in some cases. Repeating CS-ICFI measurements in narrow time intervals and combining that information as proposed in this article is a first way to address this issue. However, additional studies testing alternative ways of constructing the CS-ICFI, for instance, by attributing scores or choosing components differently, are still needed to improve the usefulness of the CS-ICFI as an analytic tool.

	ACKNOWLEDGMENTS

 
The Nutrimad project, which is jointly led by the Groupe de Recherche et d'Échanges Technologiques (GRET), the Institut de Recherche pour le Développement, and the Department of Fundamental and Applied Biochemistry of the University of Antananarivo (LABASAN), was funded by the French Ministry of Foreign Affairs and the European Union. We thank Luc Arnaud and Chantal Monvois from the GRET and Charlotte Ralison from the LABASAN for their continued support during the study. We gratefully acknowledge the contribution of Edwige Landais to the development of the food composition table used in this study.

The authors' responsibilities were as follows—MM: collected the data and performed the statistical analysis; MM and YMP: wrote the first draft of the manuscript; ST: initiated and supervised data collection; and YMP, ST, BM, and FD designed the study. All authors contributed to the interpretation of the results and to the preparation of the final version of the manuscript. None of the authors had a personal or financial conflict of interest.

	REFERENCES

TOP ABSTRACT INTRODUCTION SUBJECTS AND METHODS RESULTS DISCUSSION REFERENCES

 

1. World Health Organization. Complementary feeding of young children in developing countries—a review of current scientific knowledge. WHO/NUT 98.1. Geneva, Switzerland: World Health Organization, 1998.
2. Dewey KG. Guiding principles for complementary feeding of the breast-fed child. Washington, DC: PAHO/WHO, 2003.
3. Armar-Klemesu M, Ruel MT, Maxwell DG, Levin CE, Morris SS. Poor maternal schooling is the main constraint to good child care practices in Accra. J Nutr 2000;130:1597–607.[Abstract/Free Full Text]
4. Ruel MT, Menon P. Child feeding practices are associated with child nutritional status in Latin America: innovative uses of the demographic and health surveys. J Nutr 2002;132:1180–7.[Abstract/Free Full Text]
5. Sawadogo PS, Martin-Prével Y, Savy M, et al. An infant and child feeding index is associated with the nutritional status of 6- to 23-month-old children in rural Burkina Faso. J Nutr 2006;136:656–63.[Abstract/Free Full Text]
6. Arimond M, Ruel MT. Progress in developing an infant and child feeding index: an example using the Ethiopia Demographic and Health Survey 2000. FCND discussion paper no. 143. Washington, DC: IFPRI, 2002.
7. Ntab B, Simondon KB, Milet J, et al. A young child feeding index is not associated with either height-for-age or height velocity in rural Senegalese children. J Nutr 2005;135:457–64.[Abstract/Free Full Text]
8. Arimond M, Ruel MT. Assessing care: progress towards the measurement of selected childcare and feeding practices, and implications for programs. FCND discussion paper no. 119. Washington, DC: IFPRI, 2001.
9. Moursi MM, Trèche S, Martin-Prével Y, Maire B, Delpeuch F. Association of a summary index of child feeding with diet quality and growth of 6-23 months children in urban Madagascar. Eur J Clin Nutr (Epub ahead of print 13 Feb 2008).
10. WHO Multicentre Growth Reference Study Group. WHO Anthro 2005 software and macros. Internet: http://www.who.int/childgrowth/software/en/ (accessed 5 March 2007).
11. Dop MC, Gomis MC, Gourdon M, Lesauvage S. Outils d'enquête alimentaire par entretien: élaboration au Sénégal. [Tools for dietary surveys using interviews: development in Senegal.] Paris, France: IRD Editions, 2003 (in French).
12. Traissac P, Delpeuch F, Maire B, Martin-Prével Y, Cornu A, Trèche S. Construction d'un indice synthétique de niveau économique des ménages dans les enquêtes nutritionnelles: exemples d'application au Congo. [Building a synthetic index of the household's economic level in nutritional surveys: application examples in the Congo.] Rev Epidemiol Santé Publ. 1997;45:114–5 (in French).
13. FANTA Working Group on Infant and Young Child Feeding Indicators. Developing and validating simple indicators of dietary quality and energy intake of infants and young children in developing countries: summary of findings from analysis of 10 data sets. Washington, DC: Food and Nutrition Technical Assistance Project, Academy for Educational Development, 2006.
14. Vessereau A. Essais interlaboratoires pour l'estimation de la fidélité des méthodes d'essais. [Interlaboratory tests for estimating the reliability of experiments.] Rev Stat Appl 1974;22:5–48 (in French).
15. Bland JM, Altman DG. Precision of test methods 1: guide for the determination and reproducibility for a standard test method (BS 597, part 1). London, United Kingdom, British Standards Institutions, 1975.
16. Bland MJ, Altman DG. Statistics notes: Measurement error. BMJ 1996;313:744.[Free Full Text]
17. Maclure M, Willett WC. Misinterpretation and misuse of the kappa statistic. Am J Epidemiol 1987;126:161–9.[Free Full Text]
18. Bland JM, Altman DG. Statistical methods for assessing agreement between measurement. Biochim Clin 1987;11:399–404.
19. Kennedy GL, Pedro MR, Seghieri C, Nantel G, Brouwer I. Dietary diversity score is a useful indicator of micronutrient intake in non-breast-feeding Filipino children. J Nutr 2007;137:472–7.[Abstract/Free Full Text]
20. Steyn NP, Nel JH, Nantel G, Kennedy G, Labadarios D. Food variety and dietary diversity scores in children: are they good indicators of dietary adequacy? Public Health Nutr 2006;9:644–50.[Medline]
21. Hatloy A, Hallund J, Diarra MM, Oshaug A. Food variety, socioeconomic status and nutritional status in urban and rural areas in Koutiala (Mali). Public Health Nutr 2000;3:57–65.[Medline]
22. Onyango A, Koski KG, Tucker KL. Food diversity versus breastfeeding choice in determining anthropometric status in rural Kenyan toddlers. Int J Epidemiol 1998;27:484–9.[Abstract/Free Full Text]
23. Becquet R, Leroy V, Ekouevi DK, et al. Complementary feeding adequacy in relation to nutritional status among early weaned breastfed children who are born to HIV-infected mothers: ANRS 1201/1202 Ditrame Plus, Abidjan, Côte d'Ivoire. Pediatrics 2006;117:e701–10.[Abstract/Free Full Text]
24. Piwoz EG, Creed de Kanashiro H, Lopez de Romaña G, Black RE, Brown KH. Potential for misclassification using 24-hour dietary assessment methods. J Nutr 1995;125:57–65.[Abstract/Free Full Text]
25. Beaton GH, Milner J, Corey P, et al. Sources of variance in 24-hour dietary recall data: implications for nutrition study design and interpretation. Am J Clin Nutr 1979;32:2546–59.[Free Full Text]
26. Block G. A review of validations of dietary assessment methods. Am J Epidemiol 1982;115:492–505.[Free Full Text]

This article has been cited by other articles:

				M. M. Moursi, M. Arimond, K. G. Dewey, S. Treche, M. T. Ruel, and F. Delpeuch Dietary Diversity Is a Good Predictor of the Micronutrient Density of the Diet of 6- to 23-Month-Old Children in Madagascar J. Nutr., December 1, 2008; 138(12): 2448 - 2453. [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 Citation Map 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 Moursi, M. M Articles by Delpeuch, F. Search for Related Content PubMed PubMed Citation Articles by Moursi, M. M Articles by Delpeuch, F. Agricola Articles by Moursi, M. M Articles by Delpeuch, F.