Validity of Electronic Criteria for Detection of Ventilator-Associated Pneumonia

Background: Surveillance for ventilator-associated pneumonia (VAP) in hospital critical care units (CCUs) is an important but labor intensive endeavor.  Attempts to automate detection of VAP have been limited due to the reliance of surveillance definitions on clinical signs such as chest x-ray interpretation, and subjective symptoms such as presence of rales, and the difficulty in adapting these definitions to incorporate valid, electronically available data.  We adapted a published surveillance mechanism utilizing electronic data (Klompas, et al), and performed a preliminary analysis of the feasibility of using such an algorithm to detect VAP in a Veterans Administration (VA) healthcare facility.

Objective: The goal of this study is to determine the feasibility of substituting an algorithm based on electronic criteria in place of traditional manual methods for VAP surveillance.

Methods: We reviewed electronic medical records for 50 patients with VAP identified by manual surveillance using National Healthcare Safety Network (NHSN) criteria from January 2003 through July 2009 in our CCU to determine whether electronic algorithm criteria were met for each of these infections.  Electronic criteria included:  sustained increases in ventilator settings for fractional inspired oxygen (FIO₂₎ and positive end-expiratory pressure (PEEP), sputum gram stain results and/or white blood cell counts indicating infection, and temperature readings indicating fever.

Results: Only 13 of 50 (26%) VAPs identified through manual NHSN surveillance methods met one or more of the electronic criteria for changes in ventilator settings.  The electronic criteria most often met were sputum gram stain with ≥25 neutrophils per high power field  (90%), presence of fever > 38° C (84%), and white blood cell count >12,000 per high power field (64%).  Consistent with the NHSN definition requirement for VAP, chest x-ray evidence of new or progressive and persistent infiltrate was present in 100% of the manually identified VAPs.

Conclusions: Utilization of an electronic algorithm for detection of VAP would reduce the time required to identify and confirm the presence of infection and provide better standardization of criteria, thus reducing subjectivity and improving reliability.  We hypothesized that a previously published algorithm based on changes in ventilator settings could be substituted for traditional manual surveillance methods in our patient population.  However, our study found that evidence of worsening gas exchange via changes in ventilator settings was not present in the majority of our manually identified VAPs.  The reasons for this are unclear.  Future research focus will include a review of electronically available alternative measures of changes in pulmonary mechanics/oxygenation.