Objective: To establish thresholds for adverse patient outcomes in the
absence of knowledge of patient illness severity indices. Outcomes: P
ulmonary embolism, unplanned return to operating rooms, unplanned read
missions, clean and contaminated wound infections, and hospital-acquir
ed bacteraemia. Design: Analysis of results of surveys of hospitals in
Australia by the Australian Council on Healthcare Standards following
the introduction of clinical performance measures into the Accreditat
ion process. Setting: Acute care hospitals in Australia undergoing Acc
reditation surveys in 1993 and 1994. Methods: Stratification of hospit
als into small (1-99 beds), medium (100-199 beds), and large (greater
than or equal to 200 beds), calculation of mean rates for the above ou
tcomes in each group, and establishment of thresholds based on two sta
ndard errors from the mean. Results: The mean rate of occurrence of in
cidents was higher for larger hospitals, Thresholds were generally low
er for smaller and higher for larger hospitals. Conclusions: Bed-size
is a useful index for ''flagging'' peer group variation, The methodolo
gical issues in establishing thresholds and their implications in moni
toring the quality of care in hospitals are discussed. Copyright (C) 1
996 Elsevier Science Ltd.