Majallah-i Dānishgāh-i ̒Ulūm-i Pizishkī-i Bābul (Oct 2004)
The effect of missing data in growth curves
Abstract
Background and Objective: Applying the growth curve is the most powerful way for monitoring the growth in children and through this method it would be possible to recognize in time the deviation from the natural growth pattern in children. Falling the data and missing values are general problems in analyzing the growth longitudinal data. Therefore, it is important that by computing the missing values, the data should be completed and directed towards the proper path for analysis. Methods: This 2 year longitudinal study was done on 317 infants (153 boys and 164 girls) in Shiraz during 1996. The information related to growth (Weight, height, round the head, round the arm, and round the chest) at the birth time were collected and 11 visits from the infants’ living houses were done. In order to influence the missing values on the growth charts, 4 methods (Ignoring the missing values, general and individual models of growth curve and multiple imputation) were considered to study. Mean, 3rd, 50th, 97th centiles of raw and smooth weight were computed in boys and the smooth growth charts of their weight were determined and compared based on the four methods. Findings: There was no noticeable difference in the boys’ mean weight at age under study according to growth curve methods and multiple imputation while missing values were ignored. However, the smooth growth charts showed that applying the individual growth curve model (Second level) and multiple imputation causes the noticeable difference between the values of 3rd, 97th centiles and the traditional analysis (Ignoring missing values). Conclusion: Regarding the existence of missing values in growth longitudinal studies, ignoring the missing values for analyzing is not acceptable. Applying the growth curve model method could be considered important in making desirable the analysis and the proper growth path.
