OBJECTIVES: Estimating transition probabilities for Markov models is challenging when the effectiveness of the studied intervention is measured using a continuous score, and only aggregate data by treatment are available. We developed a Bayesian calibration method to estimate transition probabilities and applied it in Parkinson’s disease (PD). METHODS: A previously published Markov model with health states corresponding to Hoehn and Yahr (H&Y) stages was adapted. Patient-level datasets were simulated to replicate results of clinical trials for different drugs, using the UPDRS scale to assess severity, and transition probabilities were estimated from simulated data to provide a reference case. Two calibrations methods were tested for obtaining transition probabilities without patient-level data. Firstly, the Solver tool of Excel was used, with the mean change in UPDRS score and associated variance as targets. Secondly, a Bayesian calibration was implemented in OpenBUGS to estimate the posterior distribution of transition probabilities, assuming the change in UPDRS score has a normal distribution, with observed mean and variance. All other model input parameters were taken from the original model. RESULTS: With simulated patient-level data, the incremental cost (IC) was estimated at €-7,015 (95% credibility interval: €-23,953; €5,977) and incremental QALYs (IQ) at 0.455 (0.112; 0.950). With calibration using the Solver tool, there was an infinity of solutions resulting in IC ranging from €-10,141 to €-8,206 and IQ  ranging from 0.422 to 0.473. With calibration using OpenBugs, the IC was estimated at €-6,852 (€-24,244; €6,448) and the IQ at -0.448 (0.108; 0.959). CONCLUSIONS: Incremental costs and QALYs obtained using the Bayesian calibration and analysis of patient-level data were similarly distributed. Mean results obtained using the Solver tool were comparable, but no statistical distribution around results could be provided. This example suggests that the Bayesian calibration is a valid method to derive transition probabilities from continuous outcome measures.