This study aimed to estimate the between- (HTP) and within- (TP) herd true prevalence distribution of Mycobacterium avium subsp. paratuberculosis (MAP) infection in dairy cattle herds participating in the Danish MAP control programme. All herds enrolled in the programme between 2011 and 2013 were included in the analysis, and one annual milk-ELISA test of all lactating cows present in such herds was considered. A Bayesian latent class model was used to obtain HTP and TP posterior distributions for each year. The model adjusts for uncertainty in age-specific test sensitivity and prior prevalence estimates. Bayesian posterior probabilities were computed in order to compare prevalence between the years. A total of 665,700 samples were included in the study, from 221,914, 224,040, and 220,466 cows sourced from 1138, 1112, and 1059 herds in years 2011, 2012, and 2013, respectively. In that period, HTP estimates of 0.92 (95% posterior probability interval (PPI), 0.87-0.96), 0.78 (95% PPI, 0.74-0.83), and 0.75 (95% PPI, 0.71-0.78) were recorded, respectively. Low TP were observed, with population mean estimates of 0.08 (95% PPI, 0.07-0.08), 0.07 (95% PPI, 0.07-0.08), and 0.07 (95% PPI, 0.06-0.07) for the three consecutive years. Statistically-important differences were recorded for HTP and population mean TP estimates between years, indicating a trend for a decreasing level of MAP infection at both herd and animal level. Model results showed that MAP infection was widespread among the Dairy cattle herds participating in the Danish control programme, though in general it was kept at very low levels.