Temporal trends in patient safety in the Netherlands: reductions in preventable adverse events or the end of adverse events as a useful metric?

To cite: Shojania KG, Marang-van de Mheen PJ. BMJ Qual Saf 2015;24: 541–544. Two years ago, BMJ Quality & Safety published the first example of a longitudinal national adverse event (AE) study. That study included 400 admissions from each of 21 randomly selected hospitals in the Netherlands in 2004 and 200 admissions from 20 hospitals in 2008. The authors reported an increase in AEs (ie, harm from medical care) from 4.1% in 2004 to 6.2% in 2008. Reassuringly, the preventable AE rate did not change, leaving one to wonder if the increase in non-preventable AE rates reflected better documentation in medical records (or just a chance finding). The lack of improvement in patient safety over time in the Netherlands mirrored the results of a US study that showed no improvement in preventable AEs from 2002 to 2007. Commenting on this lack of improvement over time, an editorial in BMJ Quality & Safety (including one of us as an author) suggested that, while the results at least partially reflect the paucity of effective patient safety interventions, they may also highlight limitations of AEs as a metric of improvement. AEs represent a conceptually simple but practically heterogeneous category, including medication problems, healthcare-acquired infections, postoperative complications, delayed diagnoses, fall-related injuries, pressure ulcers, and many other errors and complications. This heterogeneity of AE types presents measurement problems because a broad effort to look at all AEs will probably not capture all events within a given category of interest. Suppose institutions have generally targeted, say, surgical complications (with checklists), a few specific healthcare-associated infections (eg, catheter-associated bloodstream infections with the central line bundle) and medication-ordering errors (with clinical pharmacists and/or computerised order systems). Then, it makes more sense to capture these outcomes comprehensively than to partially capture all types of harm from medical care, including ones for which we have not implemented any effective interventions. With AEs as the metric, random error from incomplete data capture for specific outcomes of interest limits our ability to document improvements even if they have occurred, especially if reductions in one category of AE have been counterbalanced by increases in another. Interestingly, Dutch investigators have now added a third time point to their previous study and report a substantial albeit non-significant reduction in preventable AEs. After adjustment for oversampling of deceased patients and patient characteristics, the preventable AE rate fell by 30% from 2008 to 2012 (p=0.10). Despite this encouraging signal of improvement, the editorial by Vincent and Amalberti accompanying this latest study again calls for a move away from focusing on AEs and the use of more granular measurement, focusing on outcomes that capture the impacts of specific interventions. We agree. However, it may seem strange that a paper reporting possible improvements in preventable AEs should elicit critical reflections on the utility of AEs as a metric similar to those made in response to previous studies 2 that showed no improvement. EDITORIAL