The aim was to study patterns of utilization of non-psychiatric admissions over time and factors affecting the utilization. The study cohort includes all individuals born 1934–66, living in one of two Danish municipalities and admitted to a non-psychiatric department at least once in 1977 (n = 2 686). The hospitalizations of the cohort were followed during a 5-year period by means of the Danish National Patient Register. The data were analysed using a longitudinal latent class (LC) model and a longitudinal latent Markov (LM) model. The LC model suggests that among the cases in the cohort there were 4 variants of utilization patterns. The LM model adequately described the sample in only 3 variants or classes. These classes may be interpreted as a small group of “chronically illʼʼ individuals (1.9% of the cohort), a major group of “healthy” individuals, with no, or only a single, random re-admission during the follow-up period (74.4% of the sample), and finally an intermediate group of “high utilizers” (23.7% of the sample). This “chronicity” variable was markedly associated with mental illness, multiple discharge diagnoses from non-psychiatric departments and total utilization of hospitalizations during the follow-up period. Conversely, gender, age and days in hospital per admission were without importance. The study implies that the analysis of patterns of hospital admissions over time can yield important insight into health service utilization and that longitudinal latent structure analyses are powerful statistical tools in this aspect. The data strongly indicated that a high utilization of admissions for physical illnesses (i.e. chronicity) does not simply depend on the severity of a physical disorder, but mental disturbances may be a more important factor.