Copyright (c) Society of Critical Care Medicine and Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Crit Care Med 2012 Vol. 40, No. 12 1

Initiatives to improve care for venti­lated patients have been constrained by limitations of the ventilator­associated pneumonia (VAP)

sur veillance definition. The National

Healthcare Safety Network definitions of Centers for Disease Control and Prevention (CDC) include clinically nuanced, subjective criteria such as “worsening gas exchange,” “change in

the character of sputum,” and “new or progressive and persistent infiltrates” (Table 1)(1). Growing pressures to publicly report VAP rates have exposed three major shortcomings to these definitions: they correlate poorly with histology, they are highly subjective, and they are time­consuming and complicated to implement (2, 3). These limitations preclude meaningful benchmarking between institutions, complicate the evaluation of prevention initiatives, hamper the construction of an ideal prevention bundle, and raise questions about the authenticity of zero VAP rates (4–6).

These shortcomings reflect the clini­cal challenge diagnosing VAP. A plethora of studies have documented poor correla­tion among clinical signs, microbiologi­cal cultures, radiographs, biochemical markers, and histological VAP (2, 7–15).

From the Department of Population Medicine (MK, KK, RP), Harvard Medical School and Harvard Pilgrim Health Care Institute, Boston, MA; Department of Medicine (MK, DSY, RP), Brigham and Women’s Hospital, Boston, MA; Division of Healthcare Quality Promotion (SM), Centers for Disease Control and Prevention, Atlanta, GA; Department of Medicine and Department of Quality (AR), Northshore University Health System, Evanston, IL; Veterans Affairs Boston Healthcare System (JMS), Boston, MA; Department of Biomedical Informatics (RSE), University of Utah School of Medicine, Salt Lake City, UT; Department of Medical Informatics (RSE, JFL), Intermountain Healthcare, Salt Lake City, UT; Department of Medicine (YK, KS), The Ohio State University Medical Center, Columbus, OH.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text

and are provided in the HTML and PDF versions of this article on the journal’s Web site (http://journals.lww.com/ccmjournal).

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Supported by Prevention Epicenters Program of the Centers for Disease Control and Prevention.

The authors have not disclosed any potential conflicts of interest.

For information regarding this article, E-mail: mklompas@partners.org

Copyright © 2012 by the Society of Critical Care Medicine and Lippincott Williams and Wilkins

DOI: 10.1097/CCM.0b013e318260c6d9

Objectives: The subjectivity and complexity of surveillance defi-nitions for ventilator-associated pneumonia preclude meaningful internal or external benchmarking and therefore hamper quality improvement initiatives for ventilated patients. We explored the feasibility of creating objective surveillance definitions for venti-lator-associated pneumonia.

Design: We identified clinical signs suitable for inclusion in objective definitions, proposed candidate definitions incorporating these objective signs, and then applied these definitions to retro-spective clinical data to measure their frequencies and associa-tions with adverse outcomes using multivariate regression models for cases and matched controls.

Setting: Medical and surgical intensive care units in eight U.S. hospitals (four tertiary centers, three community hospitals, and one Veterans Affairs institution).

Patients: Eight thousand seven hundred thirty-five consecutive episodes of mechanical ventilation for adult patients.

Interventions: We evaluated 32 different candidate defini-tions composed of different combinations of the following signs: three thresholds for respiratory deterioration defined by sustained increases in daily minimum positive end-expiratory pressure or Fio2 after either 2 or 3 days of stable or decreasing ventilator settings,

abnormal temperature, abnormal white blood cell count, purulent pulmonary secretions defined by neutrophils on Gram stain, and positive cultures for pathogenic organisms.

Measurements and Main Results: Ventilator-associated pneumonia incidence, attributable ventilator days, hospital days, and hospital mortality. All candidate definitions were significantly associated with increased ventilator days and hospital days, but only definitions requiring objective evidence of respiratory deterioration were significantly associated with increased hospital mortality. Significant odds ratios for hospi-tal mortality ranged from 1.9 (95% confidence interval 1.2–2.9) to 6.1 (95% confidence interval 2.2–17). Requiring additional clinical signs beyond respiratory deterioration alone decreased event rates, had little impact on attributable lengths of stay, and diminished sensitivity and positive predictive values for hospi-tal mortality.

Conclusions: Objective surveillance definitions that include quantitative evidence of respiratory deterioration after a period of stability strongly predict increased length of stay and hospital mortality. These definitions merit further evaluation of their utility for hospital quality and safety improvement programs. (Crit Care Med 2012; 40:0–0)

Objective surveillance definitions for ventilator-associated pneumonia

Michael Klompas, MD, MPH; Shelley Magill, MD, PhD; Ari Robicsek, MD; Judith M. Strymish, MD; Ken Kleinman, ScD; R. Scott Evans, MS, PhD; James F. Lloyd, BS; Yosef Khan, MD, MPH; Deborah S. Yokoe, MD, MPH; Kurt Stevenson, MD, MPH; Matthew Samore, MD; Richard Platt, MD; for the CDC Prevention Epicenters Program

Copyright (c) Society of Critical Care Medicine and Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

2 Crit Care Med 2012 Vol. 40, No. 12

In the absence of accurate signs or tests, we are unlikely to be able to improve the accuracy of the surveillance definition. There is a significant potential, however, to improve the objectivity and efficiency of surveillance by shifting to a definition that can be defined quantitatively and electronically. Preliminary work in this direction suggests that objective defini­tions are feasible and can capture similar patients to those identified by traditional VAP surveillance (16–18).

Shifting to an objective electronic defi­nition promises important benefits. An objective definition will make it possible to rigorously assess quality improvement initiatives without fear that apparent gains are due to observer bias rather than true advances in care. Objective definitions will also permit more meaningful compari­sons between institutions if coupled with adequate case­mix adjustments. Objec­tively determined differences in event rates between comparable institutions are more likely to help identify best practices and drive further improvements in care compared to subjective alternatives.

Given these pressing needs, we under­took an exploratory analysis of pos­sible objective definitions for VAP. We hypothesized that clinically significant, objective surveillance definitions can be created, defining “clinically significant” as a complication statistically associated with increased length of stay and hospital mortality. We recognized from the outset that creating an objective definition is con­strained by the absence of pathognomonic signs or tests for VAP. In the absence of cri­teria that can assure diagnostic accuracy, we chose to focus definition development instead on objectivity, practicality, effi­ciency, and most importantly, correlation with adverse outcomes. We reasoned that an objective and efficient surveillance defi­nition closely linked to adverse outcomes will still be a substantial step forward even if it does not correlate perfectly with histo­logical pneumonia. For the same reason, we intentionally focused on potential defi­nitions amenable to electronic analysis to simplify data collection, increase consis­tency between observers and institutions, and facilitate the creation of automated surveillance systems. We did not attempt to develop new definitions for clinical care because the onus in clinical care is on sen­sitivity when diagnosing patients and spec­ificity when managing patients whereas the onus in surveillance is on objectivity to facilitate reproducibility and comparisons over time and between institutions.

The candidate definitions explored in this article build upon prior work by the CDC Prevention Epicenters (16–18). The Epicenters demonstrated that establish­ing quantitative criteria for worsening oxygenation, purulent sputum, fever, and abnormal white blood cell counts makes surveillance faster, improves interob­server agreement, and predicts outcomes as well as conventional surveillance (17). The Epicenters’ definition, however, con­tinued to require radiographic evidence of VAP and closely mirrored the struc­ture and content of the current NHSN definitions of VAP. Retention of chest radiographs is problematic because they are subjective, nonspecific, and cannot be analyzed electronically (9, 19–22). We sought to extend the Epicenters’ work by evaluating alternative combinations of clinical signs and specifically exclud­ing chest radiographs to render entirely objective definitions. We assessed clini­cal significance by calculating event rates and attributable ventilator length of stay, hospital length of stay, and hospital mor­tality for each candidate definition.

METHODS

We analyzed the classical clinical signs of VAP for their suitability for objective surveil­lance definitions (Table 2). We sought signs and investigations that are 1) routinely as­sessed on most patients on most days, 2) ob­jective (quantitative rather than qualitative), and 3) electronically collectable (routinely re­corded as structured data in electronic systems in hospitals with comprehensive electronic health record systems). An analysis of common clinical signs using these three criteria is pre­sented in Table 2. The following signs met all three criteria: worsening oxygenation defined as a sustained increase in the daily minimum Fio2 after 2 or more days of stable or decreasing

daily minimum Fio2 or a sustained increase in the daily minimum positive end­expiratory pressure (PEEP) after 2 or more days of stable or decreasing daily minimum PEEP, fever, hy­pothermia, leukocytosis, leukopenia, purulent pulmonary secretions, and pathogenic cul­tures from pulmonary specimens.

We considered three different thresholds for respiratory deterioration: 1) rise in the daily minimum PEEP of ≥2 cm H2O or rise in the daily minimum Fio2 of ≥15 points sustained for at least 2 calendar days, 2) rise in the daily minimum PEEP of ≥3cm H2O or rise in the daily minimum Fio2 of ≥20 points sustained for at least 2 calendar days, and 3) rise in the daily minimum PEEP of ≥5 cm H2O or rise in the daily minimum Fio2 of ≥20 points sustained for at least 2 calendar days. All definitions of respi­ratory deterioration required evidence of stabil­ity prior to deterioration to identify trajectory changes in respiratory support rather than the continued evolution of preexisting respiratory disease. We evaluated two different thresholds for stability prior to respiratory deteriora­tion: ≥2 days or ≥3 days of stable or decreas­ing daily minimum PEEP or daily minimum Fio2. We defined fever as temperature >38°C, hypothermia as temperature  <36°C, leukocy­tosis as white blood cell count ≥12,000 cells/mm3, leucopenia as <4,000 cells/mm3, puru­lent pulmonary secretions as ≥25 neutrophils and ≤10 epithelial cells per low power field on Gram stain of endotracheal aspirate or bron­choalveolar lavage specimen, and pathogenic pulmonary specimen culture as recovery of any organism other than oral flora, Enterococcus species, coagulase­negative Staphylococcus species, or Candida species from endotracheal aspirate or bronchoalveolar lavage. We did not specify quantitative growth thresholds for cul­tures because these are not routinely calculated by many hospital laboratories and therefore of limited suitability for a generalizable surveil­lance definition.

We combined these signs to create candi­date objective definitions for VAP (Table 3 and eTable I; see Supplemental Digital Content 1,

Table 1. Centers for Disease Control and Prevention’s clinical surveillance definition for ventilator­associated pneumonia (PNU1)

Radiologic criteria (two or more serial radiographs with at least one of the following)

1. New or progressive and persistent infiltrate2. Consolidation3. Cavitation

Systemic criteria (at least one) 1. Fever (>38°C or >100.4°F)2. Leukopenia (<4,000 WBC/mm3) or leukocytosis (≥12,000 WBC/mm3)3. For adults ≥70 yrs old, altered mental status with no other

recognized cause

Pulmonary criteria (at least two) 1. New onset of purulent sputum, or change in character of sputum, or increased respiratory secretions, or increased suctioning requirements

2. Worsening gas exchange (e.g., desaturations, increased oxygen requirements, or increased ventilator demand)

3. New onset or worsening cough, or dyspnea, or tachypnea4. Rales or bronchial breath sounds

Patients must fulfill radiologic, systemic, and pulmonary criteria.

Jdivya 09/14/12 4 color fig(s); F# 08:16 Art: CCM203921

Copyright (c) Society of Critical Care Medicine and Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Crit Care Med 2012 Vol. 40, No. 12 3

http://links.lww.com/CCM/A489). We selected combinations that mirror classic clinical defi­nitions for VAP such as purulent pulmonary secretions with or without a pathogenic cul­ture, worsening oxygenation with purulent se­cretions, worsening oxygenation with systemic signs and purulent secretions, or worsening oxygenation with systemic signs and purulent secretions and pathogenic cultures.

Each candidate definition was evaluated for clinical significance using retrospective clini­cal data from eight hospitals (four academic hospitals, three community hospitals, and one Veterans Affairs tertiary hospital). Each participating hospital provided a minimum of 6 months of daily clinical data for every adult patient on mechanical ventilation during the study period. Clinical data included daily mini­mum PEEPs and Fio2s, minimum and maxi­mum temperatures, minimum and maximum white blood cell counts, and endotracheal as­pirate and bronchoalveolar lavage Gram stain and culture results. Patients’ comorbidities were derived from diagnosis and claims codes using the methods of Charlson et al (23) and Elixhauser et al (24). Six of the eight hospitals also provided Acute Physiology and Chronic Health Evaluation II scores from the day of intubation (or hospital admission day if intu­bated prior to arrival) when available.

We calculated incidence density, attrib­utable ventilator days, attributable hospital length of stay, and attributable mortality for each candidate definition. The incidence den­sity was defined as the event rate per 1,000 ventilator days. We designed this study using

standard surveillance principals which include all patients on mechanical ventilation regard­less of the duration of ventilation, mode of ventilation, underlying conditions, or “do not resuscitate” status. We therefore included all such patients when calculating the size of the study population, the number of episodes of mechanical ventilation, and the total number of ventilator days. All candidate definitions, however, did require patients to be on the ven­tilator for at least 2 days. Patients were eligible for more than one VAP event per definition so long as the events were >10 days apart.

Attributable ventilator days, attributable hospital days, and attributable mortality were calculated by first matching each case patient to between one and five control patients. Patients were matched by hospital, unit type, Charlson score (<3 vs. ≥3), and duration of mechanical ventilation prior to VAP onset. We defined the date of VAP onset as the date of increased venti­lator settings, but we matched control patients’ minimum duration of mechanical ventilation to cases’ second day of increased ventilator sup­port because all candidate definitions of respi­ratory deterioration required at least 2 days of increased ventilator settings. We then applied negative binomial regression models to deter­mine the ratio of ventilator days and hospital days postevent and logistic regression models to estimate the relative odds of hospital mortality in cases compared to controls. We used mixed effects models to account for matching. We as­sessed ventilator days from event date to extu­bation date and hospital days from event date to discharge date to focus on days potentially

attributable to VAP. Models were adjusted for hospital, unit type, age, sex, and Charlson score. Separate models were fitted for each potential definition (hence the same patient could serve as a case for some analyses and as a control in other analyses). We repeated all analyses among the subset of patients with Acute Physiology and Chronic Health Evaluation II scores where­in the Acute Physiology and Chronic Health Evaluation II score was added as a covariate to the regression models. If patients had more than one VAP event, all events >10 days apart were included in incidence­density calculations but only the first event was included in the at­tributable length of stay and hospital mortality analyses. We conducted sensitivity analyses to assess whether the attributable mortality of se­lected definitions differed for surgical vs. medi­cal patients. We also plotted the sensitivity and positive predictive values of selected definitions for hospital death to facilitate comparisons be­tween the definitions.

We used SAS version 9.3 (SAS Institute, Cary, NC) for all analyses, including the ap­plication of candidate definitions. The study was reviewed and approved by the Institutional Review Board of each participating institution.

RESULTS

The eight study hospitals contributed data from 8,123 patients with 8,735 epi­sodes of ventilation spanning a combined total of 50,324 ventilator days. Clini­cal characteristics of the population are presented in Table 4. Potential objective definitions of VAP, their event rates, and attributable ventilator days, hospital days, and hospital mortality are presented in Table 3 and eTable 1 (see Supplemental Digital Content 1, http://links.lww.com/CCM/A489). The definitions for respiratory deterioration in Table 3 are based on  ≥2 days of stable respiratory status prior to deterioration, whereas those in eTable  1 (see Supplemental Digital Content 1, http://links.lww.com/CCM/A489) presume ≥3 days of stability prior to deterioration.

The incidence of VAP ranged from 26.3 events per 1,000 ventilator days with the least restrictive definition (purulent pul­monary secretions alone) to 0.2 events per 1,000 ventilator days with the most restrictive definition (highest threshold increase in ventilator settings after ≥3 days of stable settings + abnormal temper­ature or white blood cell count + purulent pulmonary secretions + positive pulmo­nary culture for pathogenic organism). All candidate definitions were significantly associated with increased ventilator days and hospital days compared to matched controls but only definitions requir­ing sustained evidence of respiratory

Table 2. Suitability of selected clinical signs of ventilator­associated pneumonia for inclusion in an objective surveillance definition

Routine Objective Electronic

Radiographic infiltrates ✓Temperature ✓ ✓ ✓White blood cell count ✓ ✓ ✓Sputum quantity ✓Sputum quality ✓Sputum gram stain neutrophils ✓ ✓ ✓Cough ✓Dsypnea ✓Breath sounds ✓Mental status changes ✓Worsening oxygenation (rising Fio2) ✓ ✓ ✓Increasing ventilator support (rising positive end­expiratory pressure) ✓ ✓ ✓Pathogenic organisms from pulmonary secretions ✓ ✓ ✓Pao2/Fio2 ratioa +/− ✓ +/−Procalcitonin ✓ ✓

Definitions: 1) “Routine” defined as clinical signs or studies measured or ordered for most patients on most days. 2) “Objective” defined as data routinely measured using a quantitative or dichotomous scale. 3) “Electronic” defined as structured data elements routinely present in electronic form in hospitals with full­featured electronic medical record systems and amenable to electronic analysis with widely available software packages such as Excel, Access, and SAS without need for special analytic technologies such as natural language processing.

aPao2/Fio2 ratios were deemed not to meet “routine” and “electronic” criteria because some clinicians favor venous blood gases for some patients, and the Fio2 at the time Pao2 is measured is inconsistently recorded in electronic systems at present (not entered at all or entered as free text rather than as structured data easily amenable to electronic analysis).

Copyright (c) Society of Critical Care Medicine and Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

4 Crit Care Med 2012 Vol. 40, No. 12

deterioration were significantly associated with increased risk of hospital mortality. The only exceptions to this pattern were the most restrictive definitions, requir­ing all of sustained respiratory deteriora­tion  + abnormal temperature or white blood cell count + purulent pulmonary secretions + positive culture for a patho­genic organism. Hospital mortality odds ratios for these definitions were not sig­nificant. In general, requiring additional clinical signs beyond sustained respiratory

deterioration lowered event rates but did not meaningfully strengthen associations with adverse outcomes. Likewise, applying stricter thresholds for respiratory deterio­ration or requiring a baseline of ≥3 days of stable ventilator settings rather than ≥2 days lowered event rates but did not sig­nificantly magnify or diminish associa­tions with length of stay and mortality. Covariates for the intermediate respira­tory deterioration threshold definition are presented in eTable 2 (see Supplemental

Digital Content 1, http://links.lww.com/CCM/A489).

Sensitivities and positive predictive values for hospital death for selected definitions are displayed in Figure 1. Definitions requiring changes in ventila­tor settings alone or ventilator changes with an abnormal temperature or white blood cell count had the best combina­tions of sensitivity and positive predictive values. Requiring purulent sputum with or without positive cultures substantially

Table 3. Incidence densities and attributable outcomes for potential objective definitions for ventilator­associated pneumonia

Models Without APACHE II Scores Models With APACHE II Scores

Ventilator­ Associated

Pneumonias per 1,000 Ventilator

Days

Excess Ventilator Day

Ratioa

Excess Hospital Day

Ratiob

Mortality Odds Ratio

Excess Ventilator Day

Ratioa

Excess Hospital

Day Ratiob

Mortality Odds Ratio

Purulent pulmonary secretions 26.3 1.7 (1.6–1.8) 1.2 (1.1–1.3) 1.0 (0.8–1.3) 1.6 (1.4–1.8) 1.3 (1.2–1.4) 1.1 (0.8–1.5) + Positive culture for pathogen 11.3 1.9 (1.7–2.1) 1.3 (1.2–1.4) 1.0 (0.8–1.2) 1.7 (1.4–1.9) 1.3 (1.1–1.4) 1.0 (0.7–1.4)

Rise in daily minimum PEEP of ≥2 cm H2O or Fio2 of ≥15 points sustained for ≥2 days after ≥2 days stable or decreasing values

15.6 2.1 (1.9–2.2) 1.3 (1.2–1.4) 1.8 (1.4–2.2) 1.8 (1.6–2.0) 1.3 (1.2–1.5) 2.9 (2.0–4.2)

+ Abnormal temperature or WBC count 14.7 2.0 (1.9–2.2) 1.3 (1.2–1.4) 1.9 (1.5–2.3) 1.7 (1.5–1.9) 1.3 (1.2–1.5) 2.7 (1.8–4.0) + Purulent secretions 5.4 1.7 (1.5–1.9) 1.3 (1.2–1.4) 1.6 (1.2–2.3) 1.5 (1.3–1.8) 1.3 (1.2–1.5) 1.9 (1.2–2.9) + (abnormal temperature or WBC count)

AND purulent secretions5.2 1.6 (1.4–1.8) 1.2 (1.1–1.4) 1.8 (1.3–2.5) 1.6 (1.3–1.8) 1.3 (1.1–1.5) 2.1 (1.3–3.3)

+ (abnormal temperature or WBC count) AND purulent secretions AND pathogenic culture

1.7 1.5 (1.2–1.8) 1.3 (1.1–1.6) 0.8 (0.5–1.5) 1.5 (1.2–1.9) 1.4 (1.1–1.7) 0.9 (0.4–2.0)

Rise in daily minimum PEEP of ≥3 cm H2O or Fio2 of ≥20 points sustained for ≥2 days after ≥2 days stable or decreasing values

12.0 2.0 (1.8–2.1) 1.2 (1.1–1.3) 1.9 (1.5–2.3) 1.9 (1.7–2.1) 1.4 (1.2–1.5) 2.4 (1.6–3.6)

+ Abnormal temperature or WBC count 11.4 2.0 (1.8–2.1) 1.4 (1.3–1.6) 1.7 (1.4–2.2) 1.9 (1.6–2.1) 1.4 (1.2–1.5) 2.1 (1.4–3.3) + Purulent secretions 3.9 1.6 (1.4–1.9) 1.2 (1.1–1.4) 1.4 (1.0–2.2) 1.7 (1.4–2.0) 1.3 (1.1–1.5) 1.6 (1.0–2.7) + (abnormal temperature or WBC count)

AND purulent secretions3.7 1.7 (1.5–1.9) 1.3 (1.1–1.4) 1.5 (1.0–2.2) 1.6 (1.3–1.9) 1.3 (1.1–1.5) 2.2 (1.3–3.7)

+ (abnormal temperature or WBC count) AND purulent secretions AND pathogenic culture

0.6 1.7 (1.2–2.4) 1.6 (1.2–2.2) 0.7 (0.2–2.0) 1.9 (0.9–3.8) 2.4 (0.8–7.1) 0.1 (0.0–2.3)

Rise in daily minimum PEEP of ≥5 cm H2O or Fio2 of ≥20 points sustained for ≥2 days after ≥2 days stable or decreasing values

8.4 1.8 (1.7–2.0) 1.2 (1.1–1.3) 1.8 (1.4–2.4) 1.7 (1.5–2.0) 1.3 (1.1–1.5) 3.1 (1.9–5.1)

+ Abnormal temperature or WBC count 7.9 1.9 (1.7–2.1) 1.2 (1.1–1.3) 2.0 (1.5–2.6) 1.7 (1.5–2.0) 1.3 (1.1–1.5) 3.2 (1.9–5.3) + Purulent secretions 2.4 1.5 (1.3–1.8) 1.2 (1.1–1.4) 2.0 (1.3–3.3) 1.6 (1.2–1.9) 1.3 (1.0–1.5) 3.1 (1.6–6.0) + (abnormal temperature or WBC count)

AND purulent secretions2.3 1.7 (1.4–1.9) 1.3 (1.1–1.4) 2.0 (1.3–3.3) 1.5 (1.2–1.9) 1.7 (1.4–2.1) 2.5 (1.3–4.9)

+ (abnormal temperature or WBC count) AND purulent secretions AND pathogenic culture

0.5 1.9 (1.3–2.8) 1.9 (1.4–2.6) 0.5 (0.1–1.6) 2.3 (1.1–4.9) 2.5 (1.3–4.9) 0.1 (0.0–4.1)

APACHE, Acute Physiology and Chronic Health Evaluation; PEEP, positive end­expiratory pressure; WBC, white blood cell.Cases matched to controls on the basis of hospital, unit type, Charlson score, and duration of mechanical ventilation prior to ventilator­associated

pneumonia (VAP) onset. Outcomes models adjusted for hospital, unit, age, sex, Charlson score, duration of mechanical ventilation prior to VAP onset, and APACHE II score when available (only six of the eight participating hospitals were able to provide APACHE II scores). Numbers in parentheses are 95% confi­dence intervals. eTable 1 (see Supplemental Digital Content 1, http://links.lww.com/CCM/A489) presents event rates and attributable outcomes for the same definitions but presumes ≥3 days of stable or improving ventilator settings prior to respiratory deterioration.

aRatio of ventilator days from VAP to extubation in cases vs. matched controls; bratio of hospital days from VAP to hospital discharge in cases vs. matched controls.

Jdivya 09/14/12 4 color fig(s); F# 08:16 Art: CCM203921

Copyright (c) Society of Critical Care Medicine and Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Crit Care Med 2012 Vol. 40, No. 12 5

decreased both sensitivity and positive predictive values for hospital death.

DISCUSSION

Meaningful surveillance is indispen­sible for rigorous quality improvement programs. Surveillance helps define the extent of the problem, gauges the effec­tiveness of current improvement efforts, and informs the need to introduce or modify prevention efforts. The subjec­tivity of current VAP surveillance defini­tions has limited their utility to monitor and improve quality. In this article, we demonstrate the feasibility of creating objective VAP definitions that can be elec­tronically applied to routine clinical data. The capacity to electronically apply these definitions is proof of their objec­tivity and reproducibility. Just as impor­tantly, selected objective definitions are significantly associated with increased

ventilator days, hospital days, and hospi­tal mortality. These features make them promising tools for quality improvement programs.

Definitions with criteria for respira­tory deterioration were most strongly associated with adverse outcomes, par­ticularly mortality. Purulent pulmo­nary secretions alone, with or without a pathogenic culture, were associated with increased ventilator days and hos­pital days but not increased mortality. By contrast, respiratory decompensation defined by an increase in the daily mini­mum PEEP or Fio2 after a period of stable or decreasing daily minimum PEEP or Fio2 strongly predicted increased ven­tilator days, hospital days, and hospital mortality. These changes are likely more than simply markers of severity of ill­ness because by definition they are only flagged by a change in respiratory trajec­tory from stability or improvement to sus­tained deterioration.

Strikingly, selecting stricter thresh­olds for respiratory deterioration or adding systemic signs (abnormal temper­ature or white blood cell count), purulent secretions, and/or pathogenic cultures to worsening oxygenation diminished appar­ent VAP rates but had little impact on attributable outcomes. Indeed, requiring positive cultures for pathogenic organ­isms weakened the association between respiratory deterioration and mortality to the point that it was no longer statistically significant. These findings are consistent with previous studies demonstrating that most clinical signs do little to improve clinical certainty about the presence or absence of VAP (25).

The paradoxical weakening of the association between respiratory dete­rioration and mortality in definitions requiring pathogenic cultures may have been due to their low frequencies and consequent loss of power, but this finding does mirror previous work noting poorer outcomes among culture­negative VAP patients compared to culture­positive patients (26). Possible reasons for this are manifold. Negative bacterial cultures may be due to difficult to culture and less treatable organisms such as viruses and fungi, prior antibiotic exposure (which in turn is a possible marker for severity of illness), or undiagnosed noninfectious conditions, such as pulmonary emboli, hypersensitivity reactions, cryptogenic organizing pneumonia, or cancers. Fail­ure or delay in recognizing and treating these alternative conditions may account

for increased mortality in this subgroup. Conversely, positive cultures for patho­genic organisms likely facilitate better outcomes by helping affirm an infectious etiology for the patient’s decompensa­tion and guiding the appropriate choice of antimicrobials. We did not collect suit­able data in this study to indicate which of these factors might have accounted for the higher attributable mortality we observed among culture­negative patients.

The clear and consistent association between the objective definitions evalu­ated in this article and increased length of stay and mortality bespeak their clini­cal significance but their correlation with histological pneumonia, clinically diag­nosed pneumonia, or antibiotic starts is unknown. Prior qualitative analyses of patients that meet either objective or con­ventional surveillance definitions suggest that both flag similar sets of complications in similar proportions, primarily pneu­monia, acute respiratory distress syn­drome, pulmonary edema, and atelectasis (16). This is consistent with the extensive literature showing that patients clinically diagnosed with pneumonia actually have a range of underlying diagnoses that may or may not include pneumonia (2, 25, 27). These include diffuse alveolar dam­age, acute respiratory distress syndrome, pulmonary edema, pulmonary hemor­rhage, hypersensitivity reactions, contu­sion, atelectasis, pulmonary infarction, and others (15, 19, 28). Objective defini­tions are likely similar to current clinical diagnosis and surveillance definitions in this regard.

The poor specificity of clinical signs for VAP and consequent tendency of all VAP definitions to flag a range of under­lying conditions have prompted CDC to propose a new paradigm for ventila­tor­associated event surveillance (29, 30). CDC convened a working group of representatives from U.S. critical care, nursing, respiratory therapy, infectious disease, and healthcare epidemiology societies to inform the development of a new surveillance definition for VAP. The working group recommended shifting the focus of surveillance from pneumo­nia alone to complications of mechani­cal ventilation in general, defined as sustained deteriorations in ventila­tor settings after a period of stable or improving settings (specifically ≥2 days of stable or decreasing daily minimum PEEP or Fio2 followed by a rise in daily minimum PEEP of ≥3 cm H2O or rise

Table 4. Patient characteristics

Frequency, n = 8735

Male 4976 (57.3%)Age group, n (%) <40 611 (7.0) 40–59 2252 (25.9) 60–79 4005 (46.1)  ≥80 1824 (21.0)Age (mean ± sd) 65.4 ± 15.9Intensive care unit type, n (%) Cardiovascular 942 (10.8) Medical 1221 (14.0) Neuroscience 242 (2.8) Surgical 1254 (14.4) Mixed 5052 (57.8)Comorbidities, n (%) Chronic lung disease 2035 (23.3) Congestive heart failure 2451 (28.1) Myocardial infarction 1770 (20.3) Valvular disease 1298 (14.9) Peripheral vascular disease 917 (10.5) Neurological disease 1327 (15.2) Renal failure 1404 (16.1) Liver disease 328 (3.8) Acquired immunodeficiency

syndrome25 (0.3)

Lymphoma 212 (2.4) Solid cancers 1047 (12.0) Diabetes mellitus 2499 (28.6) Obesity 642 (7.4) Alcohol abuse 515 (5.9) Depression 875 (10.0)Acute Physiology and Chronic

Health Evaluation II Score (median, IQR)

16 (13–19)

Ventilator length of stay (median days, IQR)

3.0 (2.0–7.0)

Hospital length of stay (median days, IQR)

12.0 (7.0–20.0)

Hospital mortality 1843 (21.1%)

IQR, interquartile range.

Copyright (c) Society of Critical Care Medicine and Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

6 Crit Care Med 2012 Vol. 40, No. 12

in daily minimum Fio2 ≥20 points sus­tained for ≥2 calendar days). Shifting the focus of surveillance in this direction more accurately describes what can and cannot be discerned using surveillance definitions and emphasizes the impor­tance of preventing all complications in mechanically ventilated patients, not just pneumonia. Our work affirms that this paradigm is a strong predictor of adverse outcomes. Our findings that inflammatory signs, purulent secre­tions, and pathogenic cultures add little in terms of attributable length of stay or mortality and substantially decrease sen­sitivity and positive predictive values for hospital death further affirm the wisdom of surveillance for respiratory deteriora­tion in general rather than pneumonia in particular.

Work is now needed to determine the clinical correlates of events flagged by objective definitions. Better understand­ing of the association between objective definitions and clinical impressions, risk factors, antimicrobial usage, and con­comitant diagnoses (e.g., pulmonary edema, atelectasis, thromboembolic disease, sepsis, and abdominal catastro­phes) is needed to inform the interpreta­tion of event rates and the development of prevention strategies. It remains to be seen whether existing ventilator care bundles can prevent objectively defined events or whether bundles will need to be expanded to better target pulmonary edema, atelectasis, barotrauma, the acute respiratory distress syndrome, and other

complications of mechanical ventilation in addition to pneumonia.

Future evaluations should also con­sider how variability in clinical practice patterns impacts case rates for these def­initions. Parameters to consider include differences in ventilator management strategies, different clinical thresh­olds to send pulmonary specimens for culture, routine vs. on­demand Gram staining of pulmonary specimens, and the impact of preferential use of endo­tracheal aspirates vs. bronchoalveolar lavage Gram stains and/or cultures to evaluate possible respiratory infections. There is a room to consciously or sub­consciously “game” these definitions by varying these practices, particularly the definitions that require purulent sputum and/or positive cultures. This is another argument in favor of focusing external reporting on definitions that require respiratory deterioration alone. The fact that respiratory deterioration is contingent upon either a PEEP or Fio2 increase should decrease temptation to selectively manipulate one of these parameters rather than the other to manage hypoxemia. Some might still be tempted to tolerate sustained hypoxemia to limit PEEP and Fio2 increases. If so, the thresholds for significant changes in PEEP and Fio2 might need to be lowered and/or the current decision to exclude Pao2/Fio2 ratios from surveillance algo­rithms revisited. No surveillance defi­nition is likely to be completely free of risk of gaming. Objective surveillance

for respiratory deterioration alone at least mitigates the problem of surveyor subjectivity.

The incidence densities of ventilator­associated events captured by objective definitions fall within the general range of VAP rates currently reported to NHSN of CDC (31). The most liberal defini­tion (purulent sputum only) generated an incidence density of 26.3 events per 1,000 ventilator days, a rate substan­tially higher than current medians. By contrast, the strictest objective defini­tion (highest threshold for respiratory deterioration, abnormal temperature and white blood cell count, purulent sputum, and pathogenic pulmonary cul­ture) yielded only 0.2 events per 1,000 ventilator days, a rate well below the 50th percentile of VAP rates reported to CDC. The intermediate respiratory deterioration definition corresponding to CDC’s proposed new definition for ventilator­associated conditions gener­ated an incidence density of 12.0 events per 1,000 ventilator days. This is higher than most VAP rates currently reported to NHSN, but this definition explicitly seeks more than just VAP. Furthermore, the appropriateness of using current NHSN rates to benchmark objectively determined rates is unclear given the susceptibility of current definitions to gaming and the increasing disconnect between clinical and surveillance pneu­monia rates (32).

The generalizability of objective surveillance methods for widespread,

Figure 1. Sensitivity and positive predictive values for hospital death among selected objective definitions for ventilator­associated pneumonia. WBC, white blood cell count; fever/WBC, abnormal temperature or white blood cell count; PEEP, positive end­expiratory pressure.

Jdivya 09/14/12 4 color fig(s); F# 08:16 Art: CCM203921

Copyright (c) Society of Critical Care Medicine and Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

Crit Care Med 2012 Vol. 40, No. 12 7

prospective use also requires further evaluation. Our success applying this method to large numbers of patients drawn from diverse hospitals, including both medical and surgical patients and both community and academic insti­tutions, augurs well. All the data ele­ments included in our definitions are routinely recorded by clinical staff on a daily basis and almost all are routinely stored in electronic medical record sys­tems with the possible exception of daily minimum PEEP and Fio2. Nonetheless, it should be straightforward for respira­tory therapists or other staff to routinely tabulate these data on bedside charts or in electronic spreadsheets or databases that can then be analyzed manually or with electronic macros and/or statistical code that could be distributed by public health agencies. An Excel spreadsheet that automatically applies a selection of these definitions is available as an online supplement to serve as an example for readers (see Supplemental Digital Con­tent 2, http://links.lww.com/CCM/A546).

Our estimates of attributable lengths of stay and mortality should be tempered by limitations of our study. Retrospective matching of controls to cases can bias the control population towards being healthier (because the controls did not develop VAP) or sicker (if the controls die before they can develop VAP) compared to cases. We attempted to adjust for these possibilities by matching controls to cases on multiple parameters. We also used multivariate regression models to adjust for additional confounders. Some important confounders, however, were not available to us such as reason for admission, recent surgery, and history of trauma.

CONCLUSIONS

In sum, we demonstrate the feasibility of creating objective surveillance defini­tions for mechanically ventilated patients that can be electronically applied to rou­tinely collectable electronic data. These definitions strongly predict increased ventilator days, hospital days, and hospi­tal mortality. In particular, surveillance for deteriorating ventilator settings after a period of stability or improvement is a potent predictor of adverse outcomes. Requiring concurrent temperature or white blood cell abnormalities, purulent secretions, and pathogenic pulmonary cultures has little impact on attributable

ventilator days or hospital days and paradoxically weakens the association between respiratory deterioration and hospital mortality. Surveillance for venti­lator­associated complications in general rather than pneumonia per se sidesteps the limited accuracy of clinical defini­tions for VAP and broadens the focus of quality improvement from pneumonia alone to additional morbid complications in mechanically ventilated patients. The strong association between this surveil­lance paradigm and adverse patient out­comes, its objectivity and efficiency, and its emphasis on tracking all complica­tions in patients on mechanical ventila­tion rather than pneumonia alone makes it a better candidate for benchmarking within and between institutions (pre­suming adequate risk adjustment) and therefore a more useful metric to iden­tify best practices and catalyze improved care for ventilated patients. This defini­tion now merits evaluation of its utility in hospital quality and safety improvement programs.

REFERENCES

1. Horan TC, Andrus M, Dudeck MA: CDC/NHSN surveillance definition of health care­associated infection and criteria for specific types of infections in the acute care setting. Am J Infect Control 2008; 36:309–332

2. Tejerina E, Esteban A, Fernández­Segoviano P, et al: Accuracy of clinical definitions of ventilator­associated pneumonia: Compari­son with autopsy findings. J Crit Care 2010; 25:62–68

3. Klompas M: Interobserver variability in venti­lator­associated pneumonia surveillance. Am J Infect Control 2010; 38:237–239

4. Klompas M, Platt R: Ventilator­associated pneumonia—The wrong quality measure for benchmarking. Ann Intern Med 2007; 147:803–805

5. Bonten MJ: Healthcare epidemiology: Ventilator­associated pneumonia: Prevent­ing the inevitable. Clin Infect Dis 2011; 52:115–121

6. Klompas M: Ventilator­associated pneumo­nia: Is zero possible? Clin Infect Dis 2010; 51:1123–1126

7. Chastre J, Viau F, Brun P, et al: Prospective evaluation of the protected specimen brush for the diagnosis of pulmonary infections in ventilated patients. Am Rev Respir Dis 1984; 130:924–929

8. Rouby JJ, Martin De Lassale E, Poete P, et al: Nosocomial bronchopneumonia in the criti­cally ill. Histologic and bacteriologic aspects. Am Rev Respir Dis 1992; 146:1059–1066

9. Wunderink RG, Woldenberg LS, Zeiss J, et al: The radiologic diagnosis of autopsy­proven

ventilator­associated pneumonia. Chest 1992; 101:458–463

10. Torres A, el­Ebiary M, Padró L, et al: Valida­tion of different techniques for the diagnosis of ventilator­associated pneumonia. Com­parison with immediate postmortem pulmo­nary biopsy. Am J Respir Crit Care Med 1994; 149(2 Pt 1):324–331

11. Papazian L, Thomas P, Garbe L, et al: Broncho­scopic or blind sampling techniques for the diagnosis of ventilator­associated pneumonia. Am J Respir Crit Care Med 1995; 152(6 Pt 1): 1982–1991

12. Kirtland SH, Corley DE, Winterbauer RH, et al: The diagnosis of ventilator­associated pneumonia: A comparison of histologic, microbiologic, and clinical criteria. Chest 1997; 112:445–457

13. Papazian L, Autillo­Touati A, Thomas P, et al: Diagnosis of ventilator­associated pneumo­nia: An evaluation of direct examination and presence of intracellular organisms. Anesthe-siology 1997; 87:268–276

14. Fàbregas N, Ewig S, Torres A, et al: Clinical diagnosis of ventilator associated pneumo­nia revisited: Comparative validation using immediate post­mortem lung biopsies. Thorax 1999; 54:867–873

15. Petersen IS, Aru A, Skødt V, et al: Evalua­tion of pneumonia diagnosis in intensive care patients. Scand J Infect Dis 1999; 31:299–303

16. Klompas M, Khan Y, Kleinman K, et al; CDC Prevention Epicenters Program: Multicenter evaluation of a novel surveillance paradigm for complications of mechanical ventilation. PLoS ONE 2011; 6:e18062

17. Klompas M, Kleinman K, Khan Y, et al; CDC Prevention Epicenters Program: Rapid and reproducible surveillance for ventilator­associated pneumonia. Clin Infect Dis 2012; 54:370–377

18. Klompas M, Kleinman K, Platt R: Develop­ment of an algorithm for surveillance of ven­tilator­associated pneumonia with electronic data and comparison of algorithm results with clinician diagnoses. Infect Control Hosp Epidemiol 2008; 29:31–37

19. Meduri GU, Mauldin GL, Wunderink RG, et al: Causes of fever and pulmonary densi­ties in patients with clinical manifestations of ventilator­associated pneumonia. Chest 1994; 106:221–235

20. Winer­Muram HT, Rubin SA, Ellis JV, et al: Pneumonia and ARDS in patients receiving mechanical ventilation: Diagnostic accuracy of chest radiography. Radiology 1993; 188: 479–485

21. Albaum MN, Hill LC, Murphy M, et al: Interobserver reliability of the chest radio­graph in community­acquired pneumo­nia. PORT Investigators. Chest 1996; 110: 343–350

22. Loeb MB, Carusone SB, Marrie TJ, et al: Interobserver reliability of radiologists’ inter­pretations of mobile chest radiographs for

Copyright (c) Society of Critical Care Medicine and Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

8 Crit Care Med 2012 Vol. 40, No. 12

nursing home­acquired pneumonia. J Am Med Dir Assoc 2006; 7:416–419

23. Charlson ME, Pompei P, Ales KL, et al: A new method of classifying prognostic comorbid­ity in longitudinal studies: Development and validation. J Chronic Dis 1987; 40:373–383

24. Elixhauser A, Steiner C, Harris DR, et al: Comorbidity measures for use with adminis­trative data. Med Care 1998; 36:8–27

25. Klompas M: Does this patient have ventila­tor­associated pneumonia? JAMA 2007; 297: 1583–1593

26. Muscedere JG, McColl C, Shorr A, et al: Cana­dian Critical Care Trials Group: Determi­nants of outcome in patients with a clinical

suspicion of ventilator­associated pneumonia. J Crit Care 2008; 23:41–49

27. Rea­Neto A, Youssef NC, Tuche F, et al: Diag­nosis of ventilator­associated pneumonia: A systematic review of the literature. Crit Care 2008; 12:R56

28. Marquette CH, Copin MC, Wallet F, et al: Diagnostic tests for pneumonia in ventilated patients: Prospective evaluation of diagnostic accuracy using histology as a diagnostic gold standard. Am J Respir Crit Care Med 1995; 151:1878–1888

29. Magill SS, Fridkin SK: Improving surveil­lance definitions for ventilator­associated pneumonia in an era of public reporting and

performance measurement. Clin Infect Dis 2012; 54:378–380

30. Centers for Disease Control and Preven­tion: Improving surveillance for ventilator­ associated events in adults. Available at: http://www.cdc.gov/nhsn/PDFs/vae/CDC_VAE_CommunicationsSummary­for­compli­ance_20120313.pdf. Accessed April 26, 2012

31. Dudeck MA, Horan TC, Peterson KD, et al: National Healthcare Safety Network (NHSN) report, data summary for 2009, device­asso­ciated module. Am J Infect Control 2011; 39:349–367

32. Klompas M: Is a ventilator­associated pneu­monia rate of zero really possible? Curr Opin Infect Dis 2012; 25:176–182