The General Health Questionnaire is widely used to measure the health status of individuals. Most studies have focused on traditional score values for one or more dimensions of psychopathology. We introduce a new analysis model that is person-centred and uses a latent structure approach to group individuals by a discrete latent variable. Data were drawn from a midlife (age 53) follow up of a national birth cohort study (n = 3035). For both men and women, three groups (latent classes) were sufficient to summarise individuals' reports of recent changes in social functioning. The groups differed in the number and nature of the reported changes. Furthermore, they were shown to differ in terms of: (1) reported general health, (2) in mean scores on the conventional GHQ factors and (3) in several other variables external to the GHQ (happiness in job, ability to express feelings and self-confidence). Latent Class Analysis of positively worded GHQ items defined groups who differ in perceptions of recent positive changes in social functioning. These groups extend the value of individual health profiles afforded by the GHQ by using distinctions between categories in the first and second responses that are usually combined. (c) 2006 Elsevier Ltd. All rights reserved.