Centralised incident reporting in a Dutch collaboration of nine out-of-hours services yielded very few incident reports. To improve incident reporting and the awareness of primary caregivers about patient safety issues, a local incident-reporting procedure was implemented.To compare the number and nature of incident reports collected in a local incident-reporting procedure (intervention) versus the currently used centralised incident-reporting procedure.Quasi experiment.Three GPs' out-of-hours services (OHSs) in the centre of the Netherlands participated over 2 years before and 2 years after the intervention.A local incident-reporting procedure was implemented in OHS1, in which participants were encouraged to report all occurring incidents. A local committee with peers analysed the reported incidents fortnightly in order to initiate improvements if necessary. In OHS2 and OHS3, the current centralised incident-reporting procedure was continued, where incidents were reported to an advisory committee of the board of directors of the OHSs collaboration and were assessed every 2 months. The main outcome measures were the number and nature of incidents reported.At baseline, participants reported fewer than 10 incidents per year each. In the follow-up period, the number of incidents reported in OHS1 increased 16-fold compared with the controls. The type of incidents reported did not alter. In the local incident-reporting procedure, improvements were implemented in a shorter time frame, but reports in the centralised incident-reporting procedure led to a more systematic addressing of general and recurring safety problems.It is likely that a local incident-reporting procedure increases the willingness to report and facilitates faster implementation of improvements. In contrast, the central procedure, by collating reports from many settings, seems better at addressing generic and recurring safety issues. The advantages of both approaches should be combined.