Acute respiratory infections in young Ethiopian children.

Dear editor

The identification of risk factors for acute respiratory infections (ARI) is crucial for designing interventions to both minimize transmission and augment the immune response, particularly in Sub-Saharan Africa where poverty-related ARI is still a major cause of preventable death in young children.1 I therefore read with interest Geberetsadik et al’s recent study of the factors associated with ARI in Ethiopian children.2 Their study uses nationally representative data on households and individuals to build a model of the social, demographic, and anthropometric determinants of ARI. A precise understanding of their model, however, requires clarification of several items in their paper. I am grateful to the authors for their willingness to respond to the following: 1) the total number of observations reported in Table 1 (10,808) differs from the total reported in the text (11,645); 2) the percentage of missing values in some of the predictor variables seem substantial (eg, ∼17% in weight-for-age Z-score) raising the possibility of biased estimates; 3) the 95% confidence interval reported in Table 2 for fathers with tertiary education (0.2–1.0) differs from the 95% confidence interval reported in the text (0.2–0.6); and 4) the note to Table 2 would seem to indicate a significance level set at P<0.005 for the multivariable model whereas the text states P<0.05.

The supplemental information will help ensure an accurate reading of the methods and findings of this valuable study.

Dear editor

Our response, using the same numbering, follows: 1) 11,645 is the total number of children included in the study but after excluding the missing data for some variables, 10,808 is the total size included in analysis. This is indicated in the text. 2) Since the sample size is very big, the bias introduced will be minimal. Moreover, there is no significant difference in the background characteristics of those children with missing values and without missing values. This ensures that the effect of the missing values on the result is insignificant. 3) This was a typing error, the 95% confidence interval is 0.2–1.0. 4) This was a typing error, the significant level is P<0.05.