The aims of this study were to investigate the frequency of bronchopneumoni
a diagnosed by histological criteria among autopsied intensive care unit (I
CU) patients and to compare these with rates of pneumonia diagnosed by conv
entional clinical methods, The study material comprised 141 autopsied ICU p
atients from 7 ICUs in university hospitals in Copenhagen from a 1-y period
. A total of 20 lung tissue specimens were sampled from each patient and th
e histopathological diagnoses were classified as no, mild, moderate or se,
ere bronchopneumonia. Inter-observer variation was calculated using kappa s
tatistics. Demographic data and diagnoses of pneumonia were registered from
the patient files. Twenty-six percent of the patients had pneumonia diagno
sed \rhilst in the ICU, Histological evidence of pneumonia, found for every
second patient, was regarded as the gold standard. Diagnosis of pneumonia
in the ICU had a sensitivity, of 29% and if diagnoses of pneumonia during t
he month before ICU-admission were included, a sensitivity of 60% was found
. Specificity for pneumonia diagnosed in the ICU was 77%. The percentage of
all ICU-patients with pneumonia was calculated to be between 36% and 56%,
depending on the extent of excess mortality attributable to pneumonia, Pulm
onary segments with histologically diagnosed pneumonic lesions were distrib
uted diffusely, although the upper segments tended to be affected less. Nea
rly all patients had other histopathological findings than bronchopneumonia
, The reliability coefficient among the 6 pathologists was found to be mode
rately good (kappa = 0.45).