Andrea L Spray1, Brittany Eddy2, James Aaron Hipp3, Lora Iannotti3. 1. George Warren Brown School of Social Work, Washington University in St. Louis, St. Louis, Missouri, USA. andrealspray@gmail.com 2. Childrern's Nutrition Program of Haiti, Léogâtne Commune, Haiti. 3. George Warren Brown School of Social Work, Washington University in St. Louis, St. Louis, Missouri, USA.
Abstract
BACKGROUND: Over a fifth (21.9%) of children under 5 years of age in Haiti suffer from chronic malnutrition, 11.4% are underweight, and 5.1% suffer from acute malnutrition. Léogâne Commune has one of the highest under-five mortality rates in the country. Ordinary least squares regression conducted using data from household surveys to assess the impact of causal factors on child undernutrition may mask important local variations. OBJECTIVE: To characterize the nutrition and health situation of children 6 to 35 months of age in LLogdne Commune, Haiti, using geographically weighted regression. METHODS: In July 2008, the Children's Nutrition Program of Haiti conducted a representative cross-sectional household survey (N = 150) using a modified 33 x 6 alternative sampling design. Household questionnaires were administered to caregivers of children 6 to 35 months of age and anthropometric measurements were collected. Geographically weighted regression was employed to evaluate how undernutrition (weight-forage) and its household determinants vary across the region. Geographically weighted regression and ordinary least squares regression models were compared. RESULTS: The residuals of the ordinary least squares regression model were spatially autocorrelated (Moran's I = 0.08, z = 1.90, p = .058), indicating that undernutrition occurs in pockets rather than being evenly distributed throughout the population. There was no improvement in performance from the ordinary least squares regression model to the geographically weighted regression model. CONCLUSIONS: Despite some limitations, this study illustrates a promising approach for using geospatial data to improve the understanding of how a nutrition situation varies across a region and provide deeper insight into its underlying causes.
BACKGROUND: Over a fifth (21.9%) of children under 5 years of age in Haiti suffer from chronic malnutrition, 11.4% are underweight, and 5.1% suffer from acute malnutrition. Léogâne Commune has one of the highest under-five mortality rates in the country. Ordinary least squares regression conducted using data from household surveys to assess the impact of causal factors on child undernutrition may mask important local variations. OBJECTIVE: To characterize the nutrition and health situation of children 6 to 35 months of age in LLogdne Commune, Haiti, using geographically weighted regression. METHODS: In July 2008, the Children's Nutrition Program of Haiti conducted a representative cross-sectional household survey (N = 150) using a modified 33 x 6 alternative sampling design. Household questionnaires were administered to caregivers of children 6 to 35 months of age and anthropometric measurements were collected. Geographically weighted regression was employed to evaluate how undernutrition (weight-forage) and its household determinants vary across the region. Geographically weighted regression and ordinary least squares regression models were compared. RESULTS: The residuals of the ordinary least squares regression model were spatially autocorrelated (Moran's I = 0.08, z = 1.90, p = .058), indicating that undernutrition occurs in pockets rather than being evenly distributed throughout the population. There was no improvement in performance from the ordinary least squares regression model to the geographically weighted regression model. CONCLUSIONS: Despite some limitations, this study illustrates a promising approach for using geospatial data to improve the understanding of how a nutrition situation varies across a region and provide deeper insight into its underlying causes.