Dependability of Brazelton neonatal behavioral assessment cluster scales

The capacity of measures of neonatal behavior to differentiate individual neonates reliably across raters and dependably across the first days of life is a prerequisite to any attempt at predicting individual behavior, particularly developmental dysfunction. In a methodological program addressing this issue in the context of Generalizability Theory, we examined the variability of observed cluster scores using the Brazelton Neonatal Behavioral Assessment Scale (NBAS). The clusters were: Orientation, Autonomic Regulation, Motor Performance. Range of State, Regulation of State, and Reflex Performance. With two different repeat-measure designs and three samples, we estimated: the variance component due to true individual differences; variance components attributable to facets of observation—examiner, rater, occasion; and variance components attributable to their interactions.Results indicated that, for most dimensions of newborn behavior, there is a great deal of unpredictable variability from occasion to occasion within each individual and not much developmental change in performance over the first days of life. Also, while there are, in general, no systematic examiner or rater biases, except for the aversive reflex maneuvers, a relatively large amount of variance is attributable to specific judgments by a particular rater of a particular baby on a specific occasion. These results should be viewed in the context of relatively small true differences between individual neonates: that is, for the samples studied at least, only a small portion of the scale is used. Unless results are based on more than one examination, conclusions about differences among individuals on the NBAS should be taken with caution, as they are mitigated by a sizeable error of measurement.