Aims: To examine a representative series of adverse events in New Zealand public hospitals where death was the final outcome recorded, with a view to determining the relationship between adverse event and death.
Methods: A review was carried out of the 38 adverse events (AEs) in the New Zealand Quality of Healthcare Study where death was the outcome, and categories of relationship were established. These were identified from the total of 850 AEs determined by two-stage retrospective review of a representative sample of 6579 medical records drawn from 13 NZ public hospitals in 1998. A stricter definition of AEs, comparable with American studies, was then applied to estimate rates of death associated with AEs.
Results: There were 118 deaths at discharge identified in the sample of medical records, giving a rate of 18.0 deaths per 1000 admissions overall. A total of 30 deaths, either at or after discharge, were associated with AEs (4.6 per 1000 admissions); 19 being judged attributable to the AE either "definitely" (10) or probably (9), giving a combined rate of 2.8 AE-attributable deaths per 1000 admissions. The "definite" group had an age, comorbidity, and added-bed-days profile that was close to the average for all deaths associated with an AE. The "probable" group departed from this profile in being younger, exhibiting higher comorbidity, and having twice the added bed-days. Based on population life tables, the average years of life lost was 11.8 years for the definite group and 25.0 years for the probable group. Assessed on the preventability of the associated AE, it was estimated that 2.2 deaths per 1000 admissions were highly preventable. However, once deaths that were not judged to be attributable to the AE were excluded, the rate reduced to 1.3 AE-preventable deaths per 1000.
Conclusions: Because of the nature of the record review procedure used in the New Zealand Quality of Healthcare Study, a finding of death did not necessarily mean that an adverse event and death were causally related. Indeed, it is possible that extrapolations of mortality rates in this and other similar studies over-estimate by about a half the number of deaths caused by healthcare management.