To the Editor: The information in the article by Bellomo et al,1 which documents postoperative serious adverse events in a teaching hospital, contains no surprises; nor does it support some of the authors' conclusions. In 1995, the findings of the Quality in Australian Health Care Study2 were immediately sensationalised by the press with the headline "Hospital errors kill 18 000 a year".3 The article by Bellomo and colleagues provoked similar predictable media sensation.

As acknowledged by the authors, the study addressed neither the causes of the serious adverse events, nor whether they were "preventable". Furthermore, the authors fail to show how their findings "suggest that there is much scope for improving perioperative care in our tertiary hospitals", or why "this is a 'silent' epidemic which requires urgent and systematic attention". However, in televised interviews, they made no effort to reduce the alarm aexpressed at the prevalence of errors.

They have invented a new designation of "unscheduled surgery" (which presumably refers to acute, urgent or emergency admissions), preferring a title that suggests an avoidable lack of scheduling. It is hardly surprising that this group of patients required most of the admissions to the intensive care unit for which no prior booking had been made.

It is unclear why the authors mention that "six of nine patients over 92 years of age having hip surgery died". Again, one presumes that these operations were for hip fracture, a condition with 100% mortality if untreated. And why leave out patients aged between 90 and 92 years?

No amount of statistical manipulation conceals the bias that is obvious in their article. It might provide a media story, but it has minimal value for the critical reader.

In reply: We thank Tracy and Hugh for the issues they raise. The goal of our study was to establish baseline information on the incidence of serious adverse events (SAEs) for use in subsequent intervention studies.1 The data were needed for statistical power calculations. Our inclusion criteria were predefined, as is scientifically orthodox for any study. We chose to study a population of clinical relevance to inpatient medicine. Most simple procedures at our hospital require day admission with no overnight inpatient stay, so these were not relevant to our goals. Others may wish to study different patient populations and are free to do so.

In our opinion, there was no particular bias in our study, just accurate, prospective documentation of events. We used the term "unscheduled surgery" because it is verifiable and objective. An operational definition is necessary; otherwise, judgements about what is a true emergency (like judgements about what is preventable) are very dependent on observer bias. Nonetheless, according to our judgement, only 48 of 426 "unscheduled" operations were true emergencies. We wanted to identify groups that were at particular risk of death, hence the mention of patients over 92 years of age who had had hip surgery. Up to what level of expected postoperative mortality does it remain acceptable to perform major surgery in very elderly patients?

We stand by our opinion that we are dealing with a silent epidemic. It is silent because we could find no previous prospective studies of SAEs for all major operations published (in English) in the medical literature, and there was no systematic plan to tackle them. We use the term epidemic because (in the absence of objective documentation of rates of SAEs in the past) our impression is that this is a growing phenomenon, related to the increased use of major surgery in the elderly. We also consider that only the absence of SAEs would offer no scope for improvement. A rate of SAEs of 16.9% should, logically, offer much scope for improvement. Whether such improvement can be realised remains a matter for future interventional investigations.