Previous cultures are not clinically useful for guiding empiric antibiotics in suspected ventilator-associated...

To examine the predictive validity of prior cultures at predicting the microorganism isolated at the time of suspicion of ventilator-associated pneumonia (VAP).

We performed a retrospective analysis of a randomized controlled trial of different diagnostic and antibiotic strategies. In a subset of patients with pre-enrollment cultures, we examined agreement between cultures 1 to 3 days before suspicion of VAP and enrollment cultures performed on the day of suspicion of VAP and potential antibiotic error rates (estimated using the equation 1 - crude agreement).

Two hundred eighty-one (39%) of 739 patients had pre-enrollment culture. One hundred thirty (46%) of 281 yielded a pathogenic microorganism. In patients with positive pre-enrollment cultures, crude agreement was 0.63 (95% confidence interval, 0.55-0.71) for organism, 0.84 (0.77-0.89) for Gram class, and 0.61 (0.52-0.69) for species with sensitivity. Potential antibiotic error rates ranged from 16% (11%-33%) to 39% (31%-48%). Better agreement (P = .033) occurred in isolates from patients receiving new antibiotics during the surveillance period (0.78 [0.64-0.87]) compared to those not on antibiotics (0.58 [0.45-0.69]), or on no new antibiotics (0.50 [0.32-0.68]).

There is poor agreement between prior cultures and cultures performed at time of suspicion of VAP. Prior cultures should not be used to narrow the spectrum of empiric antibiotics.