[Variations in the epidemiolgy of adverse events: methodology of the Harvard Medical Practice Design]

The Harvard Medical Practice (HMP) Design is based on a multi-staged retrospective review of inpatient records and is used to assess the frequency of (preventable) adverse events ([P]AE) in large study populations. Up to now HMP studies have been conducted in 9 countries. Results differ largely from 2.9% to 3.7% of patients with AE in the USA up to 16.6% in Australia. In our analysis we systematically compare the methodology of 9 HMP studies published in the English language and discuss possible impacts on reported frequencies. Modifications in HMP studies can be individualised from each stage of planning, conducting, and reporting results. In doing so 2 studies from the USA with lowest rates of AE can be characterised by their context of liability and the absence of screening for nosocomial infections. Studies with a high proportion of AE are marked by an intense training of reviewers. Further conclusions are hindered by divergences in defining periods of observation, by presenting frequencies as cumulative prevalences, and differences in the reporting of study results. As a consequence future HMP studies should go for complete, consistent and transparent coverage. Further research should concentrate on advancing methods for collecting data on (P)AE.