To determine relations to low-back problems (LBP), different prevalence rates are used. The disadvantage of using different selection criteria is that studies are not comparable, except where they provide the same results. The present aim was to establish whether different prevalence selection criteria lead to different answers on a newly formed set of questionnaires. Since this set is new, reliability tests were performed (test-retest and calculations of Cronbach's Alpha, Cohen's Kappa and the intraclass correlation). Results of the questionnaire should form the cornerstones of a primary prevention programme. Altogether 1783 nurses in four Flemish (Belgian) hospitals were questioned. Information was gathered on work circumstances, education, general health, psychosocial factors, leisure activities, family situation and musculoskeletal problems. Four different datasets with variables related to lifetime prevalence LBP, annual prevalence LBP, point prevalence LBP and a set with all related variables were constructed. The variables demonstrating a relation with LBP differed slightly depending on the kind of prevalence used (lifetime, annual, point). A factor analysis on each set of prevalence related data failed due to the lack of homogeneity of the variables. Fear avoidance, coping aspects and musculoskeletal problems in other regions then the lower back were, in all circumstances, the most discriminating variables. Their discriminating power, however, differed depending on the kind of prevalence used. The differences were too small to influence the construction of the prevention programme. It is concluded that in developing a primary prevention programme any of the prevalence rates can be used. The combination of the three types of prevalence rates studied provides the most complete and reliable image.