Methods: We developed ten candidate definitions based on different combinations of 6 clinical indicators and applied them to 60,725 adult admissions of ≥3 days to Brigham and Women’s Hospital between July 2015 and June 2017. Potential indicators included worsening oxygenation, new antibiotics given for ≥3 days, fever, abnormal white blood cell count, chest imaging orders, and respiratory cultures on hospital day ≥3. Worsening oxygenation was defined as ≥2 days of decreased oxygen saturation or escalation in supplemental oxygen following ≥2 days of stable oxygenation. We calculated incidence and mortality rates for each definition. We then matched each case with up to 4 controls on the basis of clinical service and duration of hospitalization and measured associations between each definition and increased mortality and length of stay, adjusting for patients’ demographics, comorbidities, and severity of illness.
Results: The incidence of NV-HAP ranged from 7.6 events per 100 admissions with the least restrictive definition (worsening oxygenation alone), to 0.7 events per 100 admissions (worsening oxygenation, fever or leukocytosis, and new antibiotics), to 0.2 events per 100 admissions (all signs present). Crude mortality rates ranged from 17% to 30%. Odds ratios for mortality in cases vs controls ranged from 4.3 (95% CI 3.8-5.0) to 8.5 (95% CI 6.3-11.4). Odds ratios for days-until-discharge in cases vs controls ranged from 1.7 (95% CI 1.5-1.9) to 2.1 (95% CI 2.0-2.2).
Conclusion: We demonstrate the feasibility of applying electronically computable objective surveillance definitions for NV-HAP. These definitions yield incidence and mortality rates comparable to existing estimates based on manual surveillance methods. Further work is needed to better understand the clinical correlates of these events and their potential preventability.
Z. Zhang, None
A. Ochoa, None
J. Young, None
C. Rhee, None
M. Klompas, None