Objective: To evaluate a novel surveillance paradigm for complications of mechanical ventilation based upon unexpected increases in patients’ ventilator settings after a period of stable ventilator support.
Methods: We defined a ventilator-associated complication (VAC) as an increase in a patient’s daily minimum positive-end-expiratory pressure by 2.5cm H2O or fraction of inspired oxygen by 15 points, occurring after ≥48 hours of stable or decreasing ventilator settings, and persisting for ≥48 hours. We retrospectively applied this definition to randomly selected medical and surgical patients in three large hospitals: each hospital included 100 patients ventilated 2-7 days and 100 patients ventilated for >7 days. All patients were also assessed by infection preventionists for VAP using the CDC surveillance definition. We compared patients’ length-of-stay on the ventilator, in the intensive care unit, and in hospital using the Wilcoxon signed rank test and hospital mortality using Fisher’s exact test to determine how well each definition predicted patients’ outcomes.
Results: Data were available for 597 patients. Median durations of mechanical ventilation, length-of-stay in the ICU, and length-of-stay in the hospital are summarized below for patients meeting criteria for VAC and CDC criteria for VAP. Differences in length-of-stay for patients defined by either method were highly significant (P<.0001) but only VAC was significantly associated with hospital mortality.
VAC+ VAC- P VAP+ VAP- P Number of patients 136 461 -- 56 541 -- Duration of ventilation(median days) 13 6 <.0001 13.5 7 <.0001 ICU length-of-stay(median days) 16.3 8 <.0001 18 9 <.0001 Hospital length-of-stay(median days) 21 16 <.0001 24.6 17 <.0001 Hospital mortality
(% of pts)38% 23% .0005 27% 26% NS
Conclusions: A simple, rapid, objective, and electronically collectible surveillance definition predicted patient outcomes for ventilated patients more robustly than the conventional suveillance definition for VAP: both methods distinguished between patients with long and short lengths of stay but only the novel method predicted hospital mortality.