15250-m-cuypers

91 PCPCC study protocol 5 outcomes from both arms will be compared. Involved healthcare professionals will be included in a survey-study to evaluate their opinion on working with the DA. A comparison will be made with procedures of usual care from the control group. The description of this design follows the CONSORT recommendation for reporting on trials (www.consort-statement.org ) with the extensions for pragmatic 27 and cluster 28 randomized trials. Randomization Nineteen Dutch hospitals have been randomized to either ‘usual care’ (arm 1) or ‘usual care + DA’ (arm 2). With this so-called pre-randomization, the conventional sequence of obtaining informed consent followed by randomization is reversed 29 . This pre- randomization is needed because we make clinicians aware (when introducing the DA) of the principles of shared decision-making and the characteristics of the DA. This could affect the control group if they were recruited within the same hospital. To prevent potential imbalance between the two arms that could arise from hospital characteristics (for example, hospital size and treatment profile), two strata were included in the randomization procedure. First, if cooperation between two hospital locations leads to overlap in medical staff or patients visiting both locations, there is a risk for contamination bias if these hospital locations are not in the same cluster. In our sample, four pairs of hospitals have this overlap in hospital staff or patient visits. To maintain variability in hospital characteristics within each cluster, only two pairs were allowed to enter the same cluster. The second criterion is related to hospital specific treatment variation. In The Netherlands, hospitals that perform robot-assisted radical prostatectomy indicate a significant larger proportion of their patients for surgery compared to other hospitals 4 . At the moment of randomization, three hospitals from our sample are known for having robotic surgery facilities to treat prostate cancer patients, and only two of these hospitals were allowed to join the same cluster. Randomization was performed by a statistician not involved in the study and blind to the identity of the hospitals, using SPSS version 19.0 (Statistical Package for Social Sciences, Chicago, IL, USA). As a first step, the four paired hospitals were randomized (block size = 2). Next, the remaining 11 hospitals were also block randomized (block size = 6, with the last position unused). A set seed was chosen that fulfilled to the criteria that only two robot facilitated hospitals were allowed into the same cluster. The generated group order was then applied to a pre-existing list of participating hospitals, which was sorted in alphabetical order. As there is an uneven number of hospitals participating in this study, the largest cluster that was formed was identified as intervention cluster.