We conducted a systematic review to gain an insight into the overall incidence, preventability and outcome of adverse events and added information about location, provider and type of events. Despite the enormous amount of recent attention for patient safety, such systematic compiling of all available evidence on the subject is lacking to date.
This systematic review included eight studies from the USA, Canada, the UK, Australia and New Zealand. The median overall incidence of adverse events was 9.2%, and almost half of these events were regarded as preventable. The majority of events were associated with a surgical care provider, and more than half of events were operation- or drug-related.
Although all included studies used the same definition, incidences of adverse events varied considerably. In 2000, a comparison was made between the Utah/Colorado study (incidence 3.2%) and the Australian study (incidence 16.6%), and a number of possible reasons for this difference were provided.47 48
There were a number of methodological differences between the studies, such as a lower threshold for defining causation in the Australian study and inclusion of some types of events in one study that were excluded in the other. Aside from these differences, the authors suggest that the disparity might be due to differences in quantity and methods of documentation between Australia and the USA, and different perspectives of the two studies (medicolegal versus quality-improvement). These considerations may also apply to the other studies included in this review. For example, both studies that were performed from a medicolegal point of view2 32
reported considerably lower incidences than the other studies, performed from a quality-improvement point of view.3 28–31 33
Furthermore, not all studies employed the same time frame for included events. Out-of-hospital events were included in only a few studies. Some studies only recorded one adverse event per patient record, whereas others did not enforce this restriction. These methodological differences may well, at least in part, account for the difference between the recorded incidences.
Retrospective record review has been criticised for a number of reasons. As it relies heavily on patient records, it is dependent on the quality of documentation. If adverse events are not documented properly, they will not be detected by this method. Furthermore, only those adverse events are detected that result in one of the trigger criteria of the review method. Finally, in retrospective record review, the interobserver variability is very high, especially with regard to the judgements on causality and preventability.49
The studies included in this review show moderate interobserver agreement scores, illustrating this drawback of retrospective record review.
Aside from the fact that the conclusions from this review are based solely on retrospective record review studies, and as such, most likely represent an underestimation of the problem, there are a number of other limitations of this systematic review, the most important one being the heterogeneity of the included studies. Although most studies used roughly the same methods, the details differed considerably. For example, the studies from Australia and New Zealand applied a lower threshold for causation than the other studies. The time frame of included events also differed between studies: the studies from the United States did not include events discovered after discharge, whereas the other included studies did. Because differences in methodology and perspective may lead to different numbers of recorded AEs,47 48
we must proceed with caution when drawing conclusions from the combined data from these studies. Apart from differences in methodology between the included studies, our strict inclusion criteria potentially may have caused us to exclude interesting studies. The three studies we excluded because of an insufficient number of patient records evaluated, when combined, 1607 records, amounting to 2% of all records included in this systematic review (more than 74 000 records). The two excluded studies that used a computer-assisted approach7 8
reported incidences of adverse events that were slightly lower (8.3% and 6.9%) than the median incidence of the studies included in the present review. This approach is much less time-consuming than the retrospective record review, but a drawback is that it cannot make judgements on causation or preventability.
Much attention is being devoted to finding solutions to improve patient safety. In 2005, the authors of some of the largest adverse event studies advocated the implementation of selected evidence-based practices that have a potential for large impact.50
When looking at the classification of events as demonstrated in this review, operation-related and drug-related events together comprise the majority. It would thus be logical to concentrate funds and efforts on evidence-based interventions aimed at reducing these events. In addition to the evidence-based interventions reviewed here, there are a number of other interventions that seem promising but warrant further research to prove their value. This includes interventions derived from the aviation industry, such as crew resource management and the use of checklists in the operating room.
In conclusion, adverse events during hospital admission are a serious problem, occurring in approximately 9% of all admitted patients and leading to a lethal outcome in 7% of cases. Since a large portion of the adverse events are operation- or drug-related, and almost half of these events are preventable, funds and efforts should be concentrated on interventions aimed at reducing these types of events.