Patient safety is important in primary care, as most patients and most of their health problems are treated in this setting 1. Adverse events in primary care occur between five and 80 times per 100 000 consultations 2. General practitioners (GPs) were positive about reporting adverse events 3, but the validity and usefulness of incident reporting systems remains unclear [4, page 3]. A range of methods is available for identifying adverse events, such as review of medical records and case reviews of deceased patients 567. The aim of the presented study was to compare five different methods for identifying adverse events in general practice with respect to the number and type of identified events, the patient subgroups affected, and the agreement of events across the methods.

Table 2 presents the events reported by the five methods. A total of 68 events were identified using these methods. Each of the methods provided events that were not found with other methods; no overlap between the methods was detected. Mostly women were affected, in almost two-third of the registered events. Patients over 50 years and patients under 50 year were almost equally affected, but the age group below 18 was rarely involved. An exception to this general trend was the subgroup of events based patient registration, which involved mostly patients under 50 years of age.

The pharmacists registered six events per 16320 prescriptions (including repeated prescriptions), all related to female patients. Of these there were five prescribing errors and one known allergic reaction.

Twenty-seven patients answered positive to one or more items of the patient questionnaire. In total, 78 positive answers were given, it is unclear whether these answers referred to one or more events. Patients answering positive to questions were slightly younger compared to all responders (43,8 vs. 45,5 years old). By far most often (16/78) mentioned was a breach of confidentially of their medical record during treatment by their GP. Next, a lack of respect by GP or practice assistant was reported (8/78), followed by a delay in diagnosis (5/78), inappropriate drug prescribed (5/78) or inappropriate advice given (5/78) and a wrong appointment at the practice (5/78).

There had been 4,095 patients consulting the practices during the study period, from which 150 medical records were randomly selected. Analysis resulted in the finding of 11 events, all errors of treatment and communication.

During the study period 28 patients had died. One medical record was not available for examination, as this patient had not given permission. This method generated between one or two events per ten deceased patients.

All five methods proved to identify a number of adverse events. The patient survey accounted for the highest number of events and the pharmacist reports for the lowest number. All methods resulted in a variety of events, except for the pharmacist reports, which only referred to pharmaceutical treatment. The identified events referred to adult male and female patients of all ages, but events on children were very seldom reported. Events based on patient registration mostly involved individuals aged 50 years or younger. There was no overlap between the methods regarding the identified events.

A systematic review of methods to identify adverse events in health care concluded that "the available methods have widely differing purposes, strengths and weaknesses and must be considered as complementing each other by providing different levels of qualitative and quantitave information" [4, page 4]. Incident reporting systems have the advantage of being not as time-consuming as formal studies, but they are likely to underestimate the number of adverse events (numerator) while the number of opportunities for incidents (denominator) remains unknown. A recent study in hospitals also concluded that different methods should be used in order to provide an adequate assessment of clinical adverse events 14.

Our study in general practice suggested that GPs did not report all adverse events, as compared to other sources. It was difficult to assess the validity of the GP reported events because of the lack of overlap with other methods. Event reporting by GPs is probably important for raising awareness and a safety culture, but it is unlikely to be a comprehensive method for identifying adverse events. None of the GPs had a feeling of suspicion with registration. They did not think the registered events would be used against them. They all felt it was difficult to remember to register events due to daily routine working procedures and time pressure.

Almost three in ten patients reported on health safety issues, and patients as a group reported a substantial number of adverse events. It was unclear whether the checked items of the patient questionnaire referred to one or more health safety issues; we assumed these were from 27 events (thus one event per patient), in order not to overestimate number of events. In this study almost 60% of patient reported events concerned psychological harm or harm in trust or confidence. Another study showed that less than a third of reported harm was physical 15. Patients registering events may reveal blind spots in care provided by the GP. The questionnaire seemed quite easy to complete and its analysis asked for a relatively low time investment. However, the validity and usefulness of patient reported incidents needs further research 1617.

Inaccuracies of medical record are well documented 17. This method generated different types of adverse events, although the number in diagnostic events was rather low. It took a large time-investment to go through the medical records.

The reporting of adverse events by a pharmacist generated relatively few events. A study showed that community pharmacists correct 1% of the prescriptions 19, which obviously influences the number of remaining events. The low numbers may explained by the computerised systems, which including monitoring for potential contra-indications and interaction of drugs.

The audit of medical records of deceased patients was feasible, but the number of identified adverse events was low, and therefore its yield was restricted.

The limitations of this study should be recognized. The number and type of adverse events seemed to depend on the individual who registered. One could imagine asking other (para)medical caregivers (physiotherapists, midwifes, etc) to register as well. Also, in the communication with hospitals and referrals to specialists, things may go wrong. Specialists could therefore be a registration source of adverse events for GPs and this probably goes the other way around as well. Future studies should take this into account. Registration of events stays difficult as one kept in doubt whether an event is an adverse event or something that is generally accepted as inherent to GP care. This might have negatively influenced the number of registered events. Patients may also experience not every incident in their care as an adverse event. Whether this was the case may be influenced by the response of the doctor towards the event 20.

Patient safety has attracted "a level of public interest that the rest of the quality-improvement field in health care has failed to excite" 21. This may be explained by a fascination for the accidental death, despite the prolonged uncertainty about the "true" estimates of mortality resulting from medical error 21. In our study, none of the identified adverse events was associated with mortality.

The public interest for patient safety has led to a number of potentially valuable initiatives to identify and prevent adverse events. None of the methods to identify adverse events proved to be superior. We suggest that more research is needed on the various approaches in general practice before their wide scale implementation can be recommended.

The authors declare that they have no competing interests.

RWe performed the study according to protocol and wrote the manuscript. RWe and RWo collected and analysed data. MW and CvW initiated and supervised the study, and developed the study protocol. All authors read and approved the final version of the manuscript.