Predictors of low birthweight at the community level.

The outcome of pregnancy was studied in 148 women over a two year period in a rural area of Kenya as part of a prospective longitudinal study whose main objective was to study the functional effects of mild to moderate malnutrition. Data were collected on maternal anthropometric variables monthly, haemoglobin levels were determined by blood samples taken every six months, food intake was based on two days each month of actual weight and recall. Each woman's past reproductive history was established at the beginning of the study. Birth weight was taken and recorded within seventy two hours of delivery. Discriminant analysis was used to identify predictors of low birthweight. The analysis was based on 123 cases who had complete data on all the variables used in the equation. Of those included in the analysis, 14 women (11%) delivered low birthweight babies and 109 had normal birthweight babies. Results of the discriminant analysis showed that mid upper arm circumference (MUAC), body mass index (BMI), Blood haemoglobin levels (HB) and socioeconomic status (SES), are the best predictors of low birthweight. Ranked in order of relative contribution to birthweight they are BMI, HB, MUAC and SES. Low birthweight prevalence was determined as being 11.2 per cent. Eighty per cent of all known cases were correctly classified using the four variables. As a screening tool for low birthweight this model with four variables has 93% sensitivity, 78.4% specificity, 35.13% positive predictive value and 98.98% negative predictive value. The results suggest that it is possible to identify women at high risk for delivering low birthweight babies at the community level.