OBJECTIVES: Frequent attenders (FA) generate a large proportion of clinical workload, referrals and prescriptions in the primary care (PC). The aim of this study was to examine factors, which predict frequent attendance of the PC services in a long-term follow-up. Factors explaining the long-term frequent use of PC services have not been previously explored. METHODS: A prospective cohort study without intervention was carried out in the primary healthcare centre in Tampere, Finland. From a random sample of 200 PC FAs, 85 patients participated in the study. All participants were PC FAs in the first study year. After four years follow-up the patients were classified as persistent or temporary FAs. A patient was considered as a persistent FA, if he visited the health centre at least 8 times a year for at least 3 out of 4 follow up years. In addition to clinical assessment, also the patient reported outcome assessment was comprehensive including e.g. 15D, BDI, SOC-13, TAS-20, SCL (somatisation part), Whitely Index, patient satisfaction, fear of death, and alcohol consumption. 59 different variables were examined as potential predictors using P-course, a web-based Bayesian prediction tool. The models were assessed with accuracy, and predictions with posterior odds and credibility intervals. RESULTS: According to the predictions, the most influential predictive factors related to persistent FAs were female gender, body mass index above 30, former frequent attendance, fear of death, alcohol abstinence, low patient satisfaction and irritable bowel syndrome. New observations related to FAs were high body mass index, alcohol abstinence, fear of death and irritable bowel syndrome. CONCLUSIONS: Our Bayesian model can be used for proactive modelling of persistent FAs in uncertain situations. However, before its use in practice, the external validity of the model will need to be defined, because we only used partitioning in the accuracy assessments and not independent data.