BackgroundThere is a need for better implementation of patient‐centred (PC) communication and shared decision‐making (SDM) in routine cancer care.ObjectiveThe aim of this study was to assess whether a programme to implement SDM in oncology had effects on PC communication in clinical encounters.DesignThis study constitutes a secondary analysis of data derived from an implementation trial applying a stepped wedge design that, among other strategies, incorporated training and coaching to enhance the PC communication skills of physicians.Setting and ParticipantsWe analysed audio recordings of clinical encounters collected in three departments of a comprehensive cancer centre in Germany before and after rolling out the implementation programme.Main Variables StudiedWe assessed the PC communication skills of physicians.Main Outcome MeasuresEach recording was rated by two researchers using the German version of the Four Habits Coding Scheme (4HCS), an observer‐based measure of PC communication. Interrater reliability of the outcome measure was acceptable but moderate. Demographic data of patients participating in audio recordings were analysed.MethodsData were analysed using descriptive statistics and linear mixed‐effects models.ResultsIn total, 146 encounters, 74 before and 72 after implementation, were evaluated. The mean age of patients was 57.1 years (SD = 13.8), 70.3% were female, the largest portion of patients had medium formal education (32.4%) and were (self‐) employed (37.8%). No statistically significant effect of the implementation programme on the physicians' PC communication skills was found.DiscussionThe results indicate that the investigated programme to implement SDM in oncology, including training and coaching, had no effects on PC communication in clinical encounters. These results are in contrast to other studies that report the effects of specific training or coaching on PC communication. Reasons for the lack of effect include the short duration of our training compared to other studies, limited reliability and moderate interrater reliability of the 4HCS scale, limited reach of the intervention programme as well as the inclusion of physicians regardless of their exposure to the interventions.ConclusionFurther research is needed to develop implementation strategies that improve physicians' PC communication skills.Patient ContributionData on patients and clinical encounters with patients and physicians were analysed. There was no other patient or public involvement.