Table 4 Modelled effect of exclusive use of ETA or BAL on the reported incidence of VAP, using the numbers of clinically suspected VAP episodes from infection surveillance occurring in the ICU over 12 months (fig 2; n = 110)

	BAL mean estimate (95% CI)	Qualitative ETA mean estimate (95% CI)	Quantitative ETA mean estimate (95% CI)
Positive culture	23 (11 to 35)	96 (85 to 106)	56 (41 to 69)
Negative culture	87 (75 to 99)	14 (4 to 25)	54 (41 to 69)
Proportion of “clinically suspected VAPs” reported as “confirmed VAP”	21% (10 to 32%)	87% (77 to 96%)	51% (37 to 63%)
Cases/1000 ventilator days	6.1 (2.9 to 9.2)	25.5 (22.5 to 28.1)	15.0 (11.0 to 18.3)

Numbers show mean estimate and upper and lower confidence limit using 95% CIs for proportions (z test) and were derived using the data from fig 1.
BAL, bronchoalveolar lavage; ETA, endotracheal aspirate; ICU, Intensive Care Unit; VAP, ventilator-associated pneumonia.