Underestimating the impact of ventilator-associated pneumonia by use of surveillance data.

We calculated rates of ventilator-associated pneumonia (VAP) by using surveillance data, clinical data, and coding data. Compared with the VAP rates calculated on the basis of surveillance data, the VAP rates calculated on the basis of coding data were significantly overestimated in 4 of 5 intensive care units. Efforts to improve coding and clinical documentation will address much but not all of this discrepancy between surveillance and administrative data.