Objectives: To evaluate the reproducibility with which microbiologists inte
rpret Gram-stained sputa and examine the effect of the smear preparation me
thod.
Methods: Two hundred and ten coded slides prepared directly from a purulent
portion of sputum (DS) and 140 slides prepared after homogenization of the
same sputum (HS) were examined by three experienced microbiologists. A pro
portion of the slides prepared by each method were recoded and represented
to the raters. Intraobserver and interobserver variation was assessed using
the kappa statistic (kappa).
Results: Evaluation of the smear as being infected and predicting Streptoco
ccus pneumoniae as the infecting organism showed the greatest intraobserver
agreement (kappa=0.74-0.82) and interobserver agreement (kappa=0.50-0.82).
The agreement concerning the number of cells and infection with Haemophilu
s influenzae was only fair to moderate. Differences in the interpretation o
f smears prepared by the two methods could be explained by the intrinsic di
sagreement that occurs when the same smear is examined twice. The positive
predictive value of a positive S. pneumoniae smear for a positive culture (
>10(6)/ml) was 81% with the DS and 97% with the HS.
Conclusion: In this laboratory, a sputum Gram film interpretation suggestin
g infection with S. pneumoniae was reproducible and predictive of a signifi
cant positive culture. (C) 2000 The British Infection Society.