Background The proportion of avoidable hospital deaths is challenging to estimate, but has great implications for quality improvement and health policy. Many studies and monitoring tools are based on selected high-risk populations, which may overestimate the proportion. Mandatory reporting systems, however, under-report. We hypothesise that a review of an unselected sample of hospital deaths will provide an estimate of avoidability in-between the estimates from these methods.

Methods A retrospective case record review of an unselected population of 1000 consecutive non-psychiatric hospital deaths in a Norwegian hospital trust was conducted. Reviewers evaluated to what degree each death could have been avoided, and identified problems in care.

Results We found 42 (4.2%) of deaths to be at least probably avoidable (more than 50% chance of avoidability). Life expectancy was shortened by at least 1 year among 34 of the 42 patients with an avoidable death. Patients whose death was found to be avoidable were less functionally dependent compared with patients in the non-avoidable death group. The surgical department had the greatest proportion of such deaths. Very few of the avoidable deaths were reported to the hospital’s report system.

Conclusions Avoidable hospital deaths occur less frequently than estimated by the national monitoring tool, but much more frequently than reported through mandatory reporting systems. Regular reviews of an unselected sample of hospital deaths are likely to provide a better estimate of the proportion of avoidable deaths than the current methods.