Severity scoring of atopic dermatitis (SCORAD) was introduced as a standard tool but has not been used in a population-based epidemiological study; the objective of the present study was to determine the practicability of this instrument in this setting. We assessed the distribution of the severity of atopic eczema in the community and investigated differences between east and west Germany. A factor analysis was then carried out to characterize the variables of this scoring system and to analyse possible relationships within them. A multicentre cross-sectional study was carried out in five east German and two west German locations in 1994; pre-school children (5-6 years old) were investigated and cases of atopic eczema identified by a dermatological examination. The SCORAD was used to determine the severity of atopic eczema and the results assessed using analysis of variance and principal component analysis (varimax rotation). In all, 1511 (76.2%) of the children originally contacted participated and 11.3% were diagnosed with atopic eczema at the time of examination. The median severity scores was 21.4 (interquartile range 13.5) and there was a tendency to higher scores in west Germany for the mean overall score, the intensity score and the extent. 'Erythema' (1.30 vs. 1.06; P = 0.006) and 'excoriation' (0.77 vs. 0.36; P = 0.002) were significantly more prominent in children with eczema from west Germany (adjusted for observer). Interobserver variabilities of the SCORAD parameters were calculated, adjusted for location and were in accordance with earlier findings. Principal component analysis identified three independent factors accounting for 54.1% of the total variance. A severity factor, characterized by 'extent', 'lichenification', 'excoriation' and 'pruritus', was separated from a factor with an acute eczema-type profile ('erythema', 'oedema', 'oozing') and a factor whose major characteristics were 'extent', 'dryness', and 'sleep loss'. We conclude that atopic eczema is frequent in pre-school children. The SCORAD proved to be readily applicable and useful in epidemiological studies, but further validation is needed.