RESEARCH
Monitoring low birth weight: an evaluation of international estimates and an updated estimation procedure
Surveillance des faibles poids à la naissance : évaluation des estimations internationales et actualisation de la méthode d'estimation
Monitoreo de la insuficiencia ponderal del recién nacido: evaluación de las estimaciones internacionales y método de estimación actualizado
Ann K. BlancI,1; Tessa WardlawII
IBlancroft Research International LLC, 260 West Broadway, #6D, New York, NY 10013, USA (email: ablanc@blancroft.com)
IISenior Programme Officer, Division of Policy and Planning, UNICEF, New York, NY, USA
ABSTRACT
OBJECTIVE: To critically examine the data used to produce estimates of the proportion of infants with low birth weight in developing countries and to describe biases in these data. To assess the effect of adjustment procedures on the estimates and propose a modified estimation procedure for international reporting purposes.
METHODS: Mothers' reports about their recent births in 62 nationally representative Demographic and Health Surveys (DHS) conducted between 1990 and 2000 were analysed. The proportion of infants weighed at birth, characteristics of those weighed, extent of misreporting, and mothers' subjective assessments of their children's size at birth were examined.
FINDINGS: In many developing countries the majority of infants were not weighed at birth. Those who were weighed were more likely to have mothers who live in urban areas and are educated, and to be born in a medical facility with assistance from medically trained personnel. Birth weights reported by mothers are "heaped" on multiples of 500 grams.
CONCLUSION: Current survey-based estimates of the prevalence of low birth weight are biased substantially downwards. Two adjustments to reported data are recommended: a weighting procedure that combines reported birth weights with mothers' assessment of the child's size at birth, and categorization of one-quarter of the infants reported to have a birth weight of exactly 2500 grams as having low birth weight. Averaged over all surveys, these procedures increased the proportion classified as having low birth weight by 25%. We also recommend that the proportion of infants not weighed at birth be routinely reported. Efforts are needed to increase the weighing of newborns and the recording of their weights.
Keywords: Birth weight; Infant, Low birth weight; Research design/standards; Bias (Epidemiology); Data interpretation, Statistical; Health surveys; Developing countries (source: MeSH, NLM).
RÉSUMÉ
OBJECTIF: Réaliser un examen critique des données servant à estimer la proportion de nourrissons présentant un faible poids à la naissance dans les pays en développement et décrire les biais attachés à ces données. Évaluer l'effet de procédures d'ajustement sur les estimations et proposer une méthode d'estimation modifiée à des fins de notification internationale.
MÉTHODES: On a analysé les rapports fournis par les mères à propos de leur accouchement récent dans 62 enquêtes démographiques et de santé (EDS), représentatives de différents pays et menées entre 1990 et 2000. On a examiné la proportion de nourrissons pesés à la naissance, les caractéristiques de ceux ayant été pesés, l'ampleur des erreurs de relevé et les évaluations subjectives des mères quant à la taille à la naissance de leur enfant.
RÉSULTATS: Dans nombre de pays en développement, la majorité des nourrissons n'étaient pas pesés à la naissance. Les enfants pesés avaient une plus grande probabilité d'être nés de mères vivant dans des zones urbaines et éduquées et d'avoir vu le jour dans une installation médicale, avec l'aide de personnel bénéficiant d'une formation médicale. Les poids à la naissance indiqués par les mères sont « arrondis » aux multiples de 500 g.
CONCLUSION: Les estimations actuelles, établies à partir des enquêtes, de la prévalence des faibles poids à la naissance comportent un biais baissier important. Les auteurs recommandent de procéder à deux ajustements sur les données rapportées : une opération de pondération combinant le poids à la naissance indiqué et l'évaluation par la mère de la taille de l'enfant à la naissance et la catégorisation d'un quart des nourrissons signalés comme ayant un poids à la naissance de 2500 g exactement comme des enfants de faible poids à la naissance. En moyenne sur l'ensemble des enquêtes, ces méthodes augmentent la proportion d'enfants classés comme ayant un faible poids à la naissance de 25 %. Les auteurs recommandent également de notifier systématiquement la proportion de nourrissons non pesés à la naissance. Des efforts sont nécessaires pour étendre la pesée des nouveaux-nés et l'enregistrement de leur poids.
Mots clés: Poids naissance; Nourrisson faible poids naissance; Projet recherche/normes; Biais (Epidémiologie); Interprétation statistique données; Enquête santé; Pays en développement (source: MeSH, INSERM).
RESUMEN
OBJETIVO: Analizar críticamente los datos empleados para generar estimaciones de la proporción de lactantes con bajo peso de nacimiento en los países en desarrollo y describir los sesgos de que adolecen esos datos. Evaluar el efecto de los métodos de ajuste en las estimaciones y proponer un procedimiento de estimación modificado a efectos de la notificación internacional.
MÉTODOS: Se analizaron las declaraciones efectuadas por madres acerca del nacimiento de sus hijos más recientes en un total de 62 encuestas de Demografía y Salud de ámbito nacional llevadas a cabo entre 1990 y 2000. Se estudiaron la proporción de lactantes pesados al nacer, las características de los niños efectivamente pesados, la magnitud del problema de la declaración de datos incorrectos, y las evaluaciones subjetivas de las madres acerca del tamaño de sus hijos recién nacidos.
RESULTADOS: En muchos países en desarrollo la mayoría de los niños no eran pesados al nacer. Entre los que sí lo fueron, se observó una mayor probabilidad de tener una madre residente en una zona urbana y con cierto nivel de instrucción, así como de haber nacido en un establecimiento médico con la ayuda de personal preparado técnicamente. Los pesos de nacimiento declarados por las madres se han «escalonado» en forma de múltiplos de 500 g.
CONCLUSIÓN: Las estimaciones encuestales actuales de la prevalencia de insuficiencia ponderal del recién nacido están sesgadas sustancialmente a la baja. Se recomiendan dos ajustes para los datos notificados: un procedimiento de ponderación que combina los pesos de nacimiento declarados y la evaluación de la madre sobre el tamaño del hijo recién nacido, y la clasificación de la cuarta parte de los lactantes que según lo declarado pesaban exactamente 2500 g al nacer dentro de la categoría de lactantes con insuficiencia ponderal. Aplicando este método al conjunto de las encuestas se obtuvo un aumento medio del 25% de la proporción de niños nacidos con insuficiencia ponderal. Recomendamos también que se notifique sistemáticamente la proporción de lactantes no pesados en el momento del nacimiento. Es necesario desplegar un mayor esfuerzo para extender la práctica de pesar a los recién nacidos y registrar su peso.
Palabras clave: Peso al nacer; Recién nacido de bajo peso; Proyectos de investigación/normas; Sesgo (Epidemiología); Interpretación estadística de datos; Encuestas epidemiológicas; Países en desarrollo (fuente: DeCS, BIREME).
Introduction
A reduction of at least one-third in the proportion of infants with low birth weight is one of the seven major goals for the current decade of the "A World Fit for Children" programme of the United Nations. Moreover, nutritional deprivation the major determinant of low birth weight is a clear obstacle to the attainment of many of the Millennium Development Goals (1). Monitoring improvements in low birth weight is thus being given high priority within the UN system, as well as by national governments and the international nutrition community.
Although the significance and interpretation of low birth weight has recently been debated (24), most experts agree that weight at birth is an indicator of a newborn's chances for survival, growth, long-term health and psychosocial development (5). Babies whose birth weight is low as a result of undernourishment face a greatly increased risk of death during their first months and years of life (57). The evidence also suggests that those children who do survive may be more likely to experience health problems throughout their lives; these include impaired cognitive development, as well as diabetes and coronary heart disease in adulthood (8, 9). Low birth weight in developing countries occurs primarily because of poor maternal health and nutrition. In addition, diseases such as diarrhoea, malaria and respiratory infections, which are common in many developing countries, can significantly impair fetal growth when women become infected during pregnancy (5, 6).
For most developing countries, estimates of low birth weight based on data compiled from health facilities are biased because the majority of newborns are not delivered in health facilities, and those who are represent a biased sample of all births. As an alternative to health-facility-based data, information on birth weight has been collected systematically since about 1990 from mothers participating in nationally representative household surveys. However, an assessment of the results from 15 countries by Boerma et al. published in 1996 (10) found that mothers were often unable to provide numerical birth weights for their infants, primarily because the infants had not been weighed at delivery. Boerma et al. proposed an adjustment procedure in which additional information obtained from the mother on her assessment of the child's size at birth is used in combination with reported birth weights to calculate the percentage with low birth weight for all births (for details, see Results).
Since this initial assessment, data on low birth weight have been collected routinely in surveys, but the data have been evaluated for only a small number (1113). Reviews of low birth weight data and estimates consistently note their limitations (6, 7, 14, 15). Thus, a comprehensive examination of the data and estimation procedures is timely. The present study had three objectives:
- to critically examine the quality of the data used to produce estimates of the proportion of infants with low birth weight and to describe biases in these data;
- to assess the effect of adjustment procedures on the estimates; and
- to propose an extension of the adjustment proposed by Boerma et al. for international monitoring purposes.
Methods
Until relatively recently, international comparative reviews of low birth weight, as well as databases maintained by the United Nations Children's Fund (UNICEF) and WHO relied primarily on health-facility-based data and routine reporting systems. For example, a review published jointly by WHO and UNICEF in 1992 included data derived from hospital studies, vital registration data, health service records and some surveys (16). The advantages of survey data are that they are likely to include information on infants who were not delivered in health facilities and that, with access to the data files, a standardized methodology can be used to derive the estimates for different countries.
The Demographic and Health Surveys (DHS) programme began including questions on birth weight in its core questionnaire in about 1990. Building on the results of a previous study in Malaysia (17), questions on birth weight and prematurity were tested in an experimental DHS survey in the Dominican Republic and questions on birth weight and birth size in Peru in 1986. Subsequent evaluations suggested that the collection of such data was feasible and that they were of reasonable quality. As a result, some combination of questions on birth weight, birth size and prematurity was included in most subsequent DHS surveys as well as in the UNICEF-sponsored Multiple Indicator Cluster Surveys (MICS), Pan Arab Project for Child Development, and the Reproductive Health Surveys supported by the US Centers for Disease Control (1820).
The analyses in this paper are based on data from 62 DHS surveys conducted between 1990 and 2000 in 42 developing countries. About half of the surveys were conducted in sub-Saharan Africa. The DHS surveys are nationally representative household surveys for which women of reproductive age (1549 years) are interviewed. These surveys were chosen because survey files containing data on individuals were readily available from the DHS data archive maintained by ORC Macro (21).
Because the purpose of this research was not to estimate the current prevalence of low birth weight, but to assess the quality of survey data and estimation methodology, the maximum possible number of available surveys was included in the analysis; where more than one survey came from the same country, these were treated as individual surveys and where averages were calculated they were not weighted by population size. Therefore, the regional averages presented in the tables are not representative of the regional population but are simple averages derived from surveys conducted in that region. Country and weighted regional and global estimates of low birth weight are reported in a UNICEF/WHO publication (22).
The estimation of the percentage of infants with low birth weight is based on mothers' answers to questions about each of their live births in either the three years or the five years prior to the survey. The mother was first asked to assess the relative size of a specific child at birth. She was asked, "When (NAME) was born, was he/she very large, larger than average, average, smaller than average or very small"? (In a few surveys, the categories used in this question were modified (e.g. small, average or large). For this analysis, surveys with non-standard categories (except India) were excluded.)
The mother was then asked whether or not the child had been weighed at birth. If the answer was "yes", then the child's birth weight was obtained. The units in which birth weights had been recorded in the questionnaire varied between countries, but were usually in grams or kilograms. In addition, beginning in surveys conducted from around 199495, interviewers were instructed to record whether the birth weight was obtained from a health card or from the mother's recall.
Low birth weight is defined as a weight at birth of less than 2500 grams (irrespective of gestational age) (23). Although information on gestational age would allow the separation of infants born prematurely from those who were small for their gestational age (intrauterine growth retardation), this information is rarely available from developing countries (5). For the purposes of comparative reporting by international organizations, the indicator is taken to be the proportion of infants born in a certain recent period who weighed less than 2500 grams at birth.
Results
Birth weight reporting
In many surveys, birth weights are not reported for a substantial proportion of infants because they were not weighed (Table 1, web version only, available at http:www.who.int/bulletin; Table 2). The percentage of infants not weighed at birth, was extremely variable from less than 1% in Kazakhstan (1995) to 96% in Ethiopia (2000); the average percentage of infants not weighed at birth across all surveys was nearly half (48.7%). On the basis of data mainly from DHS and MICS surveys and some official statistics (weighted by population size), UNICEF and WHO (22) have estimated that 58% of all newborn infants in the developing world are not weighed.
In addition to Ethiopia, 12 surveys reported that more than 70% of infants were not weighed at birth (Burkina Faso (1999), Chad (1997), Egypt (1992 and 1995), Haiti (1994), India (1993 and 1999), Morocco (1992), Niger (1992 and 1998), Nigeria (1990) and Uganda (1995).
Overall, for infants who were weighed at birth, the mother did not know or did not remember the weight for about 10%. Of the roughly half of infants for whom a birth weight was reported, only about 29% of the weights were obtained from the child's health card; the remainder were based on the mother's recall.
Heaping
The data on numerical birth weight exhibit considerable heaping on digits that are multiples of 500 grams. Heaping refers to a pattern of misreporting in which the distribution of a number reported by respondents, such as age or birth weight, shows implausibly large frequencies of particular values, usually values ending in 0 or 5. A typical example of the frequency distribution of birth weights (from the United Republic of Tanzania) is shown in Fig. 1 in which the heaping is clearly visible. Across all surveys, about four in 10 reported birth weights were multiples of 500 grams. The heaping indicates that birth weights are often rounded, either by medical personnel who weigh the infants and report the weight to the mother, or by mothers themselves when recalling the figure. In many surveys, the magnitude of the heaping tended to increase with the time elapsed since the birth of the child. This pattern suggests that there is some diminution in mothers' ability to recall the exact weight as time since the birth increases. In addition, heaping is substantially worse (i.e. there are more birth weights that are multiples of 500 grams) for infants whose weights are reported from mothers' recall than when birth weights are recorded on a health card.
Although heaping is an indication of overall data quality, for the purposes of estimating the percentage of infants with low birth weight, it is the heaping at 2500 g the cut-off point for low birth weight that is most important. Averaged across all 62 surveys, approximately 6% of infants were reported to have weighed exactly 2500 grams at birth. Assuming that a proportion of the newborns reported as weighing 2500 grams actually weighed less than 2500 grams, some low-birth-weight babies would be misclassified as having had a normal birth weight and the prevalence of low birth weight will be biased downwards. The consequences of adjusting for this bias are examined below.
Reporting of birth size
The distribution of births by the mother's subjective assessment of the child's size at birth is shown in Table 3 (web version only, available at http://www.who.int/bulletin). Unlike birth weight, virtually all mothers provided this information about their children. With some regional variation, the distributions demonstrated a moderate tendency on the part of mothers to classify their children towards the larger end of the scale. The regions are also differentiated by the extent to which the distribution of birth size is concentrated within the "average" category; the countries in the Middle East and north Africa were the most concentrated and the countries in western and central Africa, the least concentrated. However regional patterns disguise a great deal of country-level variation; the proportion of infants of a size judged to be "average" by their mothers ranged from 27% in Nicaragua to 81% in the United Republic of Tanzania (1992). It is not clear whether this variation reflects relative differences in the actual size distributions or differences in mothers' perceptions of size. Cultural differences in the desirability of large versus small babies may also have some effect on mothers' reporting of birth size.
Characteristics of infants who were weighed compared with infants who were not weighed
Newborns who were weighed had characteristics that were substantially different from those who were not weighed (Fig. 2). Country-level results averaged across all surveys showed that 52% of infants weighed at birth had mothers who resided in urban areas whereas only 15% of infants who were not weighed had mothers who did so. Infants who were weighed were much more likely to have educated mothers and to be first births. No differences in the sex distribution of weighed versus non-weighed newborns were noted. The differences in urbanrural residence and education of the mother tended to be larger as the overall percentage of infants who were not weighed at birth increased.
Not surprisingly, infants who were weighed at birth were also far more likely to have been delivered in a medical facility and to have had medical assistance than infants who were not weighed at birth. Averaged across surveys, 85% of infants who were delivered in a medical facility were weighed whereas 12% of those not delivered in a medical facility were weighed. Moreover, 89% of newborns who were weighed had been delivered with the assistance of medically trained personnel whereas only 19% of the infants who were not weighed had such assistance.
The large differences between the infants who were weighed and those who were not weighed introduce bias into the estimates of the proportion with low birth weight because the factors associated with not being weighed overlap with some of the factors associated with low birth weight (e.g. low level of education of the mother). The exclusion from the estimates of newborns who were not weighed will bias the prevalence of low birth weight downwards. The magnitude of the bias is likely to be greater the higher the proportion of infants who are not weighed.
Relationship between birth weight and birth size
There was considerable consistency and regularity at the aggregate level in the birth weights reported across categories of size. The mean birth weight declined monotonically as birth size declined in every survey, except that conducted in Ethiopia. The mean birth weight for "very small" babies was less than 2500 grams in all except seven countries whereas the weight for the "average" birth size ranged from 2802 grams in India to 3477 grams in Bolivia. The mean birth weight across all surveys (3229 grams) was also very similar to the birth weight reported by mothers who classified their infants as "average" (3172 grams). However, although size and weight are consistent in the aggregate, there is considerable variation in consistency between the two measures at the individual level.
Adjustments and their effects
The adjustment proposed by Boerma et al. (8) is a straightforward weighting procedure. First, for each survey separately, the proportion of infants with low birth weight within each category of subjective size is calculated. Then, this proportion is multiplied by the overall proportion of births in each size category and summed to obtain the overall prevalence of low birth weight. The assumptions implicit in this adjustment are:
- that the infants for whom numerical birth weights are reported are as likely to have a low birth weight as those for whom no birth weight is reported; and
- that the relationship between birth weight and the size category in which the mother places her child is, within a given country, the same for infants weighed and not weighed at birth.
Based on our examination of the characteristics of infants weighed at birth, it seems clear that the first assumption is violated; no information on which to judge the extent to which the second assumption holds was available.
Estimates of the proportion of infants with low birth weight based on three different procedures are shown in Table 2. The third column gives the percentage of infants with low birth weight based only on those who were weighed the conventional procedure. These estimates range from 4% in Uzbekistan to 26% in India (1993) and average 11% over all surveys. In the fourth column, estimates that employ the adjustment proposed by Boerma et al. are shown. As expected, these estimates are almost uniformly higher than the estimates based only on infants for whom birth weights were reported and average 13% over all surveys.
In the fifth column, an additional adjustment was made for the heaping of birth weights on exactly 2500 grams. This adjustment was based on evidence from 88 DHS surveys. First, for each survey we tabulated the frequency distribution of reported birth weights between 2000 and 2999 grams. Then, we calculated the percentage of infants who weighed less than 2500 grams, after excluding those with weights reported as being exactly 2500 grams. That percentage, on average, was 25%. We therefore reclassified 25% of the infants reported as weighing exactly 2500 grams as having a low birth weight. This adjustment has a substantial effect on the percentage of infants with low birth weight in some countries. For example, in India (1999) the percentage increased by almost five percentage points from 25.5% to 30.4%. Averaged across all surveys, the estimated percentage of infants with low birth weight was approximately 13% higher after the adjustment for heaping than the estimate obtained using the adjustment proposed by Boerma et al. Overall, the average percentage of infants with low birth weight increased from 11.2%, when no adjustment was made, to 14.0% after making the two adjustments, an increase of 25%.
Conclusions
Although survey data yield more accurate estimates of the numbers of infants with low birth weight than health-facility-based data in countries where a large proportion of infants are not delivered in health facilities, these data have a number of limitations. First, survey data show that in many developing countries the majority of infants are not weighed at birth. We recommend that the percentage of infants who are weighed be reported whenever the percentage with low birth weight is reported. Indeed, the percentage weighed itself merits consideration as an indicator for regular monitoring.
Infants who are weighed at birth are a biased sample of all births and this bias becomes stronger the smaller the percentage of infants weighed at birth. All other things being equal, infants who are weighed at birth are less likely to have a low birth weight, so using these data alone to estimate the prevalence of low birth weight will result in an underestimate. Moreover, using infants for whom birth weights have been reported to calculate adjusted estimates based on birth size, as in the present study, also yields rates that are biased downwards because the population for which we have information on low birth weight is biased. Nevertheless, this procedure no doubt yields estimates that are more accurate than estimates based only on infants weighed at birth.
Substantial heaping of reported weights occurs on weights of exactly 2500 grams, the cut-off point for low birth weight. Some of the birth weights reported as being exactly 2500 grams were no doubt less than 2500 grams; thus, not including them biases estimates of the prevalence of low birth weight downwards. Estimates that count one-quarter of these births as low birth weight are substantially higher than those that do not account for the effect of heaping. We believe that estimates that include this adjustment are the most accurate available at present. Field-based studies that examine how and when mothers acquire birth weight information on their newborns as well as on mothers' beliefs about birth size would allow better interpretation of existing data and may enable improvements to be made in survey instruments.
One of the advantages of survey data on birth weight is that they include some information on infants not born in medical facilities and on those not weighed. Paradoxically, estimates of the proportion of infants with low birth weight may rise in some countries as the proportion of newborns who are weighed increases and includes more of those infants who are likely to be born with low birth weight. The fundamental issue for accurately monitoring the prevalence of low birth weight a low percentage of infants weighed at birth cannot be solved by statistical manipulation but only by efforts to increase the weighing of newborns and the recording of their weights. 
Acknowledgement
The authors wish to thank Trevor Croft for his assistance with data processing.
Conflicts of interest: none declared.
References
1. United Nations Administrative Committee on Coordination Subcommittee on Nutrition. Mainstreaming nutrition: opportunities to improve development outcomes. Fifth report on the world nutrition situation. Geneva: United Nations Administrative Committee on Coordination Subcommittee on Nutrition; 2004.
2. Wilcox AJ. On the importance and the unimportance of birthweight. International Journal of Epidemiology 2001;30:1233-41.
3. World Health Organization. Physical status: the use and interpretation of anthropometry. Geneva: WHO; 1995. WHO Technical Report Series, No. 854.
4. Osrin D, de L Costello AM. Maternal nutrition and fetal growth: practical issues in international health. Seminars in Neonatology 2000;5:209-19.
5. Bale JR, Stoll BJ, Lucas AO, editors. Improving birth outcomes: meeting the challenge in the developing world. Washington, DC: The National Academies Press; 2003.
6. United Nations Administrative Committee on Coordination Subcommittee on Nutrition. Nutrition throughout the life cycle. Fourth report on the world nutrition situation. Geneva: United Nations Administrative Committee on Coordination Subcommittee on Nutrition in collaboration with International Food Policy Research Institute; 2000.
7. Allen LH, Gillespie SR. What works? A review of the efficacy and effectiveness of nutrition interventions. Geneva: United Nations Administrative Committee on Coordination Subcommittee on Nutrition; 2001.
8. Bhargava SK, Sachdev HS, Fall CHD, Osmond C, Lakshmy R, Barker DJP, et al. Relation of serial changes in childhood body-mass index to impaired glucose tolerance in young adulthood. New England Journal of Medicine 2004;350:865-75.
9. Barker DJP. The developmental origins of adult disease. Geneva; World Health Organization; 2003. Discussion paper for the WHO Expert Consultation towards the development of a strategy for promoting optimal fetal growth.
10. Boerma JT, Weinstein KI, Rutstein SO, Sommerfelt AE. Data on birth weight in developing countries: can surveys help? Bulletin of the World Health Organization 1996;74:209-16.
11. Robles A, Goldman N. Can accurate data on birthweight be obtained from health interview surveys? International Journal of Epidemiology 1999;28:925-31.
12. Eggleston E, Tsui AO, Fortney J. Assessing survey measures of infant birth weight and birth size in Ecuador. Journal of Biosocial Science 2000;32:373-82.
13. Magadi M, Madise N, Diamond I. Factors associated with unfavourable birth outcomes in Kenya. Journal of Biosocial Science 2001;33:199-225.
14. de Onis M, Blossner M, Villar J. Levels and patterns of intrauterine growth retardation in developing countries. European Journal of Clinical Nutrition 1998;52 Suppl:S5-15.
15. United Nations Children's Fund. Progress since the world summit for children: a statistical review. New York: United Nations Children's Fund; 2001.
16. World Health Organization and United Nations Children's Fund. Low birth weight: a tabulation of available information. Geneva: WHO and UNICEF; 1992.
17. DaVanzo J, Habicht JP, Butz WP. Assessing socioeconomic correlates of birth weight in peninsular Malaysia: ethnic differences and changes over time. Social Science and Medicine 1984;18:387-404.
18. Goldman N, Moreno L, Westoff CF. Peru experimental study: an evaluation of fertility and child health information. Columbia, Maryland: Institute for Resource Development/Macro Systems Inc.; 1989.
19. Moreno L, Goldman N. An assessment of survey data on birthweight. Social Science and Medicine 1990;31:491-500.
20. Westoff CF, Goldman N, Moreno L. Dominican Republic experimental study: an evaluation of fertility and child health information. Columbia, Maryland: Institute for Resource Development/Macro Systems Inc.; 1990.
21. MEASURE/DHS+ Project homepage. Calverton, (MD): ORC Macro. Available from http://www.measuredhs.com/accesssurveys/start.cfm
22. United Nations Children's Fund and World Health Organization. Low birthweight: Country, regional and global estimates. New York: UNICEF; 2004.
23. World Health Organization. International statistical classification of diseases and related health problems, tenth revision. Geneva: WHO; 1992.
Submitted: 10 March 2004 Final revised version received: 2 August 2004 Accepted: 9 August 2004
1 Correspondence should be sent to this author.