Rapid and Reproducible Surveillance for Ventilator-Associated Pneumonia

Transcription

1 MAJOR ARTICLE Rapid and Reproducible Surveillance for Ventilator-Associated Pneumonia Michael Klompas, 1,2 Ken Kleinman, 1 Yosef Khan, 3 R. Scott Evans, 4,5 James F. Lloyd, 5 Kurt Stevenson, 3 Matthew Samore, 4 Richard Platt, 1,2 for the CDC Prevention Epicenters Program 1 Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, and 2 Infection Control Department, Brigham and Women's Hospital, Boston, Massachusetts; 3 Department of Medicine, The Ohio State University Medical Center, Columbus; 4 Department of Biomedical Informatics, University of Utah School of Medicine, and 5 Department of Medical Informatics, Intermountain Healthcare, Salt Lake City, Utah (See the Editorial Commentary by Magill and Fridkin, on pages ) Background. The complexity and subjectivity of ventilator-associated pneumonia (VAP) surveillance limit its value in assessing and comparing quality of care for ventilated patients. A simpler, more quantitative VAP definition may increase utility. Methods. We streamlined the Centers for Disease Control and Prevention definition of VAP to increase objectivity and efficiency. Qualitative criteria were replaced with quantitative criteria, and changes in ventilator settings were used to screen patients for worsening oxygenation. We retrospectively compared surveillance time, reproducibility, and outcomes for streamlined versus conventional surveillance among medical and surgical patients on mechanical ventilation in 3 university hospitals. Results. Application of the streamlined definition was faster (mean 3.5 minutes vs 39.0 minutes per patient) and more objective (interrater reliability j 0.79 vs 0.45) than the conventional definition. On multivariate analysis, the streamlined definition predicted increases in ventilator days (6.5 days [95% CI, ] vs 6.4 days [95% CI, ]), intensive care days (5.6 days [95% CI, ] vs 6.2 days [95% CI, ]), and hospital mortality (odds ratio [OR] 0.84 [95% CI, ] vs OR 0.69 [95% CI, ]) as effectively as conventional surveillance. The conventional definition was a marginally superior predictor of increased hospital days (5.2 days [95% CI, ] vs 2.1 days [95% CI, ]). Conclusions. A streamlined version of the VAP definition was faster, more objective, and predicted patients outcomes almost as effectively as the conventional definition. VAP surveillance using the streamlined method may facilitate more objective and efficient quality assessment for ventilated patients. Ventilator-associated pneumonia (VAP) surveillance is a vexing problem for hospitals [1 3]. VAP surveillance is needed to define the prevalence of VAP and to gauge the success of prevention efforts, but the process is complicated, time consuming, costly, and subjective [4]. Received 16 June 2011; accepted 30 August Correspondence: Michael Klompas, MD, MPH, Department of Population Medicine, Harvard Medical School and Harvard Pilgrim Health Care Institute, 133 Brookline Ave, 6th Floor, Boston, MA (mklompas@partners.org). Clinical Infectious Diseases 2012;54(3):370 7 Ó The Author Published by Oxford University Press on behalf of the Infectious Diseases Society of America. All rights reserved. For Permissions, please journals.permissions@oup.com. DOI: /cid/cir832 Hospitals can invest substantial resources in surveillance yet only gain results of questionable significance. There are 2 major limitations of the standard VAP definition [5] published by the Centers for Disease Control and Prevention (CDC): (1) many of the diagnostic criteria are very subjective, and (2) the definition correlates poorly with histological pneumonia. Subjectivity makes surveillance onerous to conduct and precludes meaningful comparisons between observers. For example, the definition includes change in character of sputum, rales or bronchial breath sounds, and worsening gas exchange. Interpretation of these criteria is at the discretion of the surveyor; reasonable observers are likely to arrive at different conclusions and hence very different VAP rates [6, 7]. This subjectivity makes VAP 370 d CID 2012:54 (1 February) d Klompas et al

2 surveillance unsuitable for public reporting or benchmarking, which in turn may hinder prevention efforts. Poor correlation between clinical signs and histological pneumonia makes the significance of VAP assignments questionable. The cardinal signs of pneumonia lack sensitivity and specificity in ventilated patients. Multiple studies show poor agreement between clinical criteria for pneumonia and autopsyconfirmed disease [8, 9]. Qualitative and quantitative cultures of endotracheal aspirates, bronchoalveolar lavage specimens, and protected specimen brush samples also lack sensitivity and specificity [10 15]. Barring the development of new biomarkers or imaging modalities, there is little that can be done at present to improve the accuracy of surveillance diagnoses. It might be possible, however, to improve surveillance objectivity by limiting the definition to objective and quantifiable components. For example, purulent sputum can be defined by neutrophil counts on a sputum Gram stain rather than by subjective estimation of changes in sputum color or quantity. Worsening oxygenation can be defined by specified increases in ventilator settings rather than by a surveyor s discretion alone. Focusing on objective criteria should also improve efficiency. Many quantitative criteria are available from clinical information systems and can therefore be assessed remotely and periodically rather than solely at the bedside in real time. In addition, quantitative criteria are more amenable to rapid screening and automated analysis. We have previously shown that streamlining the VAP definition is feasible, but the efficiency, reproducibility, and validity of the method have not been evaluated [16]. We therefore conducted a blinded, multicenter comparison of VAP surveillance using streamlined and conventional versions of the CDC definition. We wrestled with how to assess validity. Mismatches between streamlined and conventional definitions may not be errors given the subjectivity and inaccuracy of the conventional definition. In the absence of an objective reference standard such as biopsy or autopsy, we believe the next best alternative is to assess how well each measure predicts patient outcomes such as duration of mechanical ventilation, length of stay in the intensive care unit (ICU) and hospital, and mortality. Our reasoning is that pneumonia is a morbid condition and we seek to prevent VAP in order to prevent adverse outcomes. We therefore propose that the extent to which a surveillance definition predicts poor outcomes is a reasonable proxy for validity in the absence of a more specific reference standard. METHODS All adult patients on mechanical ventilation in the medical and surgical ICUs of 3 academic medical centers during calendar years were eligible for inclusion. We had sufficient resources to assess 200 patients per hospital. Each hospital randomly selected 100 patients ventilated for 2 7 days and 100 patients ventilated for.7 days (we enriched the sample with patients with longer lengths of stay to increase the likely number of VAPs). Each patient was independently assessed for VAP using the conventional and streamlined definitions. Conventional and streamlined assessors were blinded to each other s conclusions. The institutional review boards of Brigham and Women s Hospital (Boston, MA), LDS Hospital (Salt Lake City, UT), and The Ohio University Medical Center (Columbus, Ohio) approved the study. Conventional Definition of VAP The CDC publishes 3 definitions of VAP: a purely clinical definition, a combined clinical and laboratory definition, and a variant definition for immunocompromised patients [5]. We used the purely clinical definition as the comparator for the streamlined definition because this is the version most commonly reported to CDC s National Healthcare Safety Network [17]. The clinical definition requires patients to fulfill 1 radiographic, 1 systemic, and 2 pulmonary criteria (Table 1). Radiographic criteria include new or progressive infiltrates, consolidation, and cavitation. Systemic criteria include fever, abnormal white blood cell count, and altered mental status. Pulmonary criteria include purulent sputum, new or worsening cough or dyspnea or tachypnea, rales or bronchial breath sounds, and worsening gas exchange. Experienced infection preventionists applied this definition to study patients. Three infection preventionists at 1 hospital did so prospectively; 1 infection preventionist at each of the other 2 hospitals did so retrospectively. Streamlined Definition of VAP Our streamlined method for applying the CDC surveillance definition is presented in Table 1. As with the conventional definition, we required patients to fulfill 1 radiographic, 1 systemic, and 2 pulmonary criteria. Radiographic criteria were identical to the conventional surveillance definition, systemic criteria were limited to abnormal temperature and white blood cell count using the same thresholds as the conventional definition, and pulmonary signs were limited to purulent sputum and worsening gas exchange. Entirely subjective and nonspecific criteria such as rales and delirium were eliminated. We defined purulent sputum as $25 neutrophils per low-power field on Gram stain of endotracheal aspirate or bronchoalveolar lavage specimen [5]. We defined worsening oxygenation as a $2.5 cm H 2 O rise in a patient s daily minimum positive end expiratory pressure (PEEP) or a $0.15 point rise in the daily minimum fraction of inspired oxygen (FiO 2 ) sustained for at least 2 calendar days and occurring after at least 2 days of stable or decreasing settings. We selected these thresholds to maximize sensitivity because they are the smallest clinically meaningful increases in ventilator support. One study investigator who was blinded to conventional assessors Rapid and Reproducible VAP Surveillance d CID 2012:54 (1 February) d 371

3 Table 1. Radiology Comparison of Conventional and Streamlined Surveillance Definitions for Ventilator-Associated Pneumonia Conventional definition Streamlined definition Two or more serial chest radiographs Two or more serial chest radiographs with at least 1 of the following: with at least 1 of the following: 1. New or progressive and 1. New or progressive and persistent infiltrate persistent infiltrate 2. Consolidation 2. Consolidation 3. Cavitation 3. Cavitation Systemic signs (at least 1) Pulmonary signs (at least 2) 1. Fever (.38 C or F) 1. Fever (.38 C or F) 2. Leukopenia (,4000 WBC/mm 3 )or leukocytosis ($ WBC/mm 3 ) 3. For adults $70 years old, altered mental status with no other recognized cause 1. New onset of purulent sputum, or change in character of sputum, or increased respiratory secretions, or increased suctioning requirements 2. Worsening gas exchange (eg, desaturations, increased oxygen requirements, or increased ventilator demand) 3. New onset or worsening cough, or dyspnea, or tachypnea 4. Rales or bronchial breath sounds Abbreviations: FiO 2, fraction of inspired oxygen; PEEP, positive and expiratory pressure; WBC, white blood cell. impressions manually applied the streamlined definition to electronic data supplied by each participating hospital. We operationalized the streamlined definition by organizing each patient s data into a line list with 1 line per day that included 2. Leukopenia (,4000 WBC/mm 3 )or leukocytosis ($ WBC/mm 3 ) 1. $25 neutrophils per low power field on Gram stain of endotracheal aspirate or bronchoalveolar lavage specimen 2. $2 days of stable or decreasing daily minimum PEEP followed by a rise in daily minimum PEEP of $2.5 cm H 2 O, sustained for $ 2 calendar days; or $2 days of stable or decreasing daily minimum FiO 2 followed by a rise in daily minimum FiO 2 of $0.15 points, sustained for $2 calendar days the date, day of ventilation, minimum PEEP and minimum FiO 2 for the day; maximum temperature; minimum and maximum white blood cell counts; Gram stain results on pulmonary specimens (if obtained); and full-text radiographic report on Figure 1. Raw surveillance data organized for rapid application of the streamlined ventilator-associated pneumonia definition. The days shaded in gray meet criteria for worsening oxygenation, systemic inflammation, and purulent sputum. Temperature, white blood cell counts, Gram stain neutrophils, and chest radiographs were only reviewed for patients that met worsening oxygenation criteria. Chest radiographic reports are not shown in this figure. Abbreviations: BAL, bronchoalveolar lavage; FiO 2, fraction of inspired oxygen; PEEP, positive and expiratory pressure. 372 d CID 2012:54 (1 February) d Klompas et al

4 Table 2. Characteristics and Crude Outcomes of Patients With and Without Ventilator-Associated Pneumonia According to Conventional Versus Streamlined Definitions Conventional definition chest studies (if obtained). The surveyor could then rapidly assess patients by vertically screening daily ventilator settings to look for sustained increases after a period of stability (Figure 1). Other variables were only assessed if the patient met ventilator change criteria. Time to Complete Surveillance We assessed surveillance time for each method. The infection preventionists at the 2 hospitals that conducted retrospective surveillance recorded their time spent applying the conventional definition. The investigator who applied the streamlined method recorded time to complete surveillance for the same 2 hospitals. The infection preventionists who conducted prospective surveillance at 1 hospital did not record their surveillance times. Agreement Between Observers We assessed interrater agreement on 85 patients from 1 of the study hospitals. An infection preventionist blinded to the conclusions of the first assessor reevaluated these patients using the conventional definition. The same 85 patients were also rereviewed for VAP by an independent infection preventionist who used the streamlined definition. Interobserver agreement was calculated using the kappa statistic. Streamlined definition VAP positive VAP negative VAP positive VAP negative Characteristic and crude outcome (n 5 57) (n 5 542) P (n 5 30) (n 5 569) P Age (mean) NS NS Male 34 (60%) 293 (54%) NS 17 (57%) 310 (54%) NS Unit type Medical 19 (33%) 278 (51%) NS 9 (30%) 288 (51%) NS Surgical 35 (61%) 255 (47%) NS 20 (67%) 270 (47%) NS Mixed 3 (5%) 9 (2%) NS 1 (3%) 11 (2%) NS Comorbidities Myocardial infarction 8 (14%) 107 (20%) NS 4 (13%) 111 (20%) NS Congestive heart failure 16 (28%) 171 (32%) NS 10 (33%) 177 (31%) NS Stroke 11 (19%) 71 (13%) NS 4 (13%) 78 (14%) NS Chronic obstructive lung disease 19 (33%) 187 (35%) NS 8 (27%) 198 (35%) NS Connective tissue disease 1 (2%) 19 (4%) NS 2 (7%) 18 (3%) NS Diabetes 7 (12%) 130 (24%) NS 5 (17%) 132 (23%) NS Renal disease 11 (19%) 121 (22%) NS 7 (23%) 125 (22%) NS Cancer 11 (19%) 121 (22%) NS 6 (20%) 126 (22%) NS Charlson Index (mean) NS NS Outcome Mechanical ventilation days, median (IQR) 18 (11 27) 7 (4 12), (11 22) 7 (4 12),.0001 Intensive care days, median (IQR) 19 (14 31) 10 (6 17), (13 27) 10 (6 18),.0001 Hospital days, median (IQR) 26 (21 40) 18 (11 28), (19 36) 18 (12 29).005 Hospital mortality 13 (23%) 128 (24%) (23%) 134 (24%) Abbreviations: IQR, interquartile range; NS, not significant (P, 0.05); VAP, ventilator-associated pneumonia. Comparison of Outcomes We compared patients unadjusted outcomes by streamlined vs conventional VAP status. In particular, we compared median duration of mechanical ventilation, intensive care length of stay, hospital length of stay, and mortality in VAP-positive and VAPnegative patients using the Wilcoxon signed rank test. We then developed multivariable regression models to assess attributable length of stay and mortality. Determining attributable length of stay in patients with VAP is complicated because length of stay is a risk factor for VAP. This creates circularity between risk factor (premorbid length of stay) and outcome (length of stay). We attempted to adjust for this circularity by (1) matching patients with VAP ( cases ) to patients without VAP ( controls ), where control patients were ventilated for at least as long as the time for case patients to go from intubation to VAP onset; and (2) subtracting each patient s duration of stay prior to VAP (for cases) or match date (for controls) in order to focus on days possibly attributable to the VAP. Up to 7 controls were matched to each case on the basis of hospital, unit type, and Charlson comorbidity index (dichotomized at,3 vs $3). We applied a generalized linear mixed model to account for the case-control design and variable numbers of controls per case [18]. Separate and independent models were developed for each surveillance method. Rapid and Reproducible VAP Surveillance d CID 2012:54 (1 February) d 373

5 Table 3. Sensitivity and Specificity of Streamlined VAP Surveillance Relative to Conventional VAP Surveillance Streamlined surveillance All statistical calculations were computed using SAS version 9.2 (SAS Institute, Cary, NC). RESULTS VAP positive VAP negative Conventional Surveillance VAP positive VAP negative Total Total Sensitivity: 17/57 (30%); specificity: 529/542 (98%); positive predictive value: 17/30 (57%); negative predictive value: 529/569 (93%).; agreement (j): Abbreviation: VAP, ventilator-associated pneumonia. A total of patients were ventilated during the study period. We included 599 patients ventilated for 6409 days. The Table 4. Characteristics and Adjusted Outcomes for Patients Included in the Multivariate Analysis Conventional surveillance chart for the 600th patient could not be located. Patient characteristics are presented in Table 2. The conventional definition flagged 57 patients. The streamlined definition flagged 30 patients. The overlap between assessment methods is presented in Table 3. Retrospective assessment of 400 patients using the conventional definition required 260 hours versus 23 hours for the streamlined definition (mean 39.0 vs 3.5 minutes per patient). The 2 infection preventionists who reviewed 85 patients using the conventional definition agreed on VAP status for 65 patients (76% agreement; j ; 95% CI,.26.64), whereas 2 surveyors who applied the streamlined method to the same patients agreed on 80 patients (94% agreement; j ; 95% CI,.62.97). The surveyors applying the streamlined method disagreed with each other over interpretation of chest radiographs in 3 of the 5 discrepant cases. On unadjusted analysis, VAP patients detected by both definitions had significantly longer durations of mechanical ventilation, intensive care length of stay, and hospital length of stay compared with patients without VAP (Table 2). There was no Streamlined surveillance VAP positive VAP negative VAP positive VAP negative Characteristic and adjusted outcome (n 5 53) (n 5 195) P (n 5 30) (n 5 179) P Age (mean) NS NS Male 60% 55% NS 57% 58% NS Unit type Medical 18 (34%) 78 (40%) NS 9 (30%) 61 (34%) NS Surgical 33 (62%) 114 (58%) NS 20 (67%) 111 (62%) NS Mixed 2 (4%) 3 (2%) NS 1 (3%) 7 (4%) NS Comorbidities Myocardial infarction 13% 17% NS 13% 14% NS Congestive heart failure 26% 28% NS 33% 30% NS Stroke 17% 15% NS 13% 19% NS Chronic obstructive lung disease 34% 29% NS 27% 30% NS Connective tissue disease 2% 4% NS 7% 5% NS Diabetes 13% 20% NS 17% 24% NS Renal disease 19% 18% NS 23% 17% NS Cancer 17% 22% NS 20% 21% NS Charlson index NS NS Outcome Ventilator days 18.1 (15.3, 21.3) 11.5 (10.0, 13.2), (14.7, 22.5) 12.3 (10.5, 14.3),.0001 Intensive care days 22.0 (18.7, 25.8) 15.6 (13.8, 17.7), (18.2, 27.0) 17.0 (14.9, 19.3).005 Hospital days 30.2 (25.6, 35.6) 25.9 (23.1, 29.1) (23.1, 34.7) 28.4 (25.6, 31.4).98 Days from match to extubation 11.1 (8.8, 14.1) 4.7 (4.1, 5.5), (8.8, 16.6) 5.6 (4.7, 6.6),.0001 Days from match to ICU discharge 13.9 (11.3, 17.0) 7.7 (6.7, 8.8), (11.6, 19.9) 9.6 (8.4, 11.0).002 Days from match to hospital discharge 19.2 (15.8, 23.3) 14.0 (12.4, 15.7) (14.6, 24.4) 16.8 (15.1, 18.8).41 Hospital mortality odds ratio 0.69 (0.30, 1.55) (0.31, 2.29)..73 Analysis adjusted for hospital, unit type, age, sex, and Charlson index. Abbreviations: ICU, intensive care unit; NS, not significant (P, 0.05); VAP, ventilator-associated pneumonia. 374 d CID 2012:54 (1 February) d Klompas et al

6 difference in crude mortality rates between VAP-positive and VAP-negative patients for either surveillance definition. On multivariable analysis, VAP status by either definition predicted significantly more days on mechanical ventilation and in the ICU compared with VAP-negative controls (Table 4). There was no significant difference in total hospital length of stay for either definition. When time prior to matching was censored, both methods remained significantly associated with excess ventilator and ICU days. Time from matching to hospital discharge was significantly longer for conventionally defined VAP patients but not for streamlined-defined VAP patients. There was no difference in the adjusted odds ratio for hospital mortality between VAP-positive and VAP-negative patients regardless of surveillance definition. DISCUSSION VAP surveillance is a critical tool in assessing quality of care for ventilated patients. The conventional CDC surveillance definition, however, is laborious to implement and yields results of questionable significance. In this study, we show that streamlining the conventional definition makes VAP surveillance simpler, faster, and more reproducible. The streamlined definition required less time to apply (average 3.5 vs 39.0 minutes per patient) and produced more agreement between surveyors (j vs 0.45). Both definitions predicted significant increases in duration of mechanical ventilation and intensive care length of stay; neither definition was associated with increased hospital mortality. The conventional definition did predict longer hospital stays, whereas the streamlined definition did not. The substantial time savings associated with the streamlined definition likely reflects the method s innovations, which are (1) emphasizing objective criteria that can be rapidly assessed using electronic medical records or bedside flow sheets, (2) organizing patients data into daily line lists to facilitate efficient analysis, and (3) using sustained increases in ventilator settings as a marker of worsening oxygenation; this criterion facilitated rapid screening of large numbers of patients because individuals with stable ventilator settings did not require further evaluation. The apparent superiority of the conventional definition as a predictor of prolonged hospitalization may be an artifact of the matching process (ie, total hospital lengths of stay for streamlined and conventional VAP patients were very similar [18.9 vs 19.2 days]), but the controls matched to streamlined VAP patients had lengths of stay that were longer than those of the controls matched to the conventional VAP patients (16.8 vs 14.0 days). This might have been because there were fewer streamlined VAP-positive patients than conventional VAP-positive patients, hence the pool of streamlined-negative patients available for matching included more individuals with longer lengths of stay compared with the pool of conventional-negative patients. Alternatively, conventional surveyors might have supplemented the technical surveillance criteria with a gestalt selection for sicker patients. All patients who fulfilled streamlined criteria technically fulfilled conventional criteria because the streamlined criteria are a subset of the conventional criteria. Failure to label all streamlined VAP cases with conventional VAP affirms subjectivity in applying the conventional definition. The lack of association between both VAP definitions and mortality mirrors the inconsistent relationship between VAP and mortality in the literature. Some studies show no relationship [19 22], while others show increased risk [23 25]. The inconsistent relationship between VAP and mortality may result because the population of patients labeled with VAP includes individuals with an array of underlying conditions that range from serious (such as pneumonia, acute respiratory distress syndrome, pulmonary embolism, and pulmonary hemorrhage) to benign (atelectasis, transient pulmonary edema, microbial colonization of the endotracheal tube and upper respiratory tract) [26 28]. The mixing of serious and benign conditions under the umbrella term VAP may dilute the mortality signal expected from true pneumonias [29]. Two major limitations to the streamlined definition could give hospitals and regulators pause. One limitation is the definition s retention of radiographic criteria, a persistent locus for complexity, subjectivity, and disagreement among observers. The second limitation is ongoing concern about whether the definition is finding true cases of pneumonia. Both concerns are attributable to root weaknesses in standard diagnostic criteria for VAP rather than being new problems introduced by streamlining the CDC definition. Radiographic changes are a compulsory component of the CDC definition. Although both systemic and pulmonary criteria can reasonably be assessed quantitatively, we are not aware of an analogous option for chest radiographs. Ironically, retaining radiographic criteria adds little accuracy to the diagnosis of VAP [26, 28, 30] or prediction of patients outcomes [31]. Interpretation of radiographs in ventilated patients is notoriously difficult: many films are portable studies of poor technical quality. Interpretation is further hindered by frequent preexisting pulmonary disease and/or acute pulmonary complications that can mimic or obscure an evolving pneumonia. Most ventilated patients with infiltrates and fever or purulent sputum do not have VAP [28, 32]. The sensitivity and specificity of new or progressive alveolar infiltrates or air bronchograms for histologically confirmed pneumonia are 46% and 42%, respectively [26]. Furthermore, interobserver agreement on the interpretation of chest radiographs is poor [7, 26, 33, 34]. Removing radiologic criteria from future iterations of the VAP surveillance definition Rapid and Reproducible VAP Surveillance d CID 2012:54 (1 February) d 375

7 will likely have little impact on accuracy but could improve objectivity and reproducibility. The superior efficiency and objectivity of the streamlined method may not translate into improved accuracy. Autopsies of ventilated patients only affirm clinical diagnosis of VAP in 43% 75% and show that the diagnosis was missed in 31% 46% of patients [9, 10, 14, 26]. Limiting the surveillance definition to its quantifiable components presumably does not overcome the inherent lack of sensitivity and specificity of these signs. The one exception may be the requirement for $25 neutrophils per lowpower field in pulmonary secretion Gram stain. The presence of.50% neutrophils in bronchoalveolar lavage fluid is 95% 100% sensitive for pneumonia [12, 13, 35]. Even this sign, however, lacks specificity (38% 94%) [12, 13, 35]. Requiring confirmatory cultures, regardless of where they are taken from (ie, endotracheal aspirate, bronchoalveolar lavage, or protected specimen brush) only diminishes sensitivity further without improving specificity [10 15]. Until new diagnostic markers are discovered, we have to accept that no surveillance method can reliably predict the presence of histological pneumonia. Inaccurate diagnosis is a major problem for clinical care of ventilated patients. However, diagnostic accuracy is less critical where the primary priorities are to ensure consistency between different observers, to enable comparisons between institutions, and to assess trends. To be sure, surveillance diagnoses must correspond to clinical diagnoses but they need not be perfect in order to be useful. Surveillance definitions are most useful if they are objective, reproducible, easy to apply (ideally electronically), able to reliably predict poor outcomes, and responsive to meaningful improvements in patient care. Alternatively, given the frequent mislabeling of pneumonia, it might make more sense to focus surveillance and benchmarking on the syndrome of ventilator-associated complications in general rather than on pneumonia in particular. This approach considerably simplifies surveillance because it removes the pretense of trying to distinguish VAP from its many mimickers and facilitates an efficient, objective definition. For example, a surveillance definition predicated solely upon sustained increases in ventilator settings after a period of stable or decreasing support predicted lengths of stay as effectively as the conventional VAP definition and was a superior predictor of mortality [31]. A collateral benefit of this approach is that it broadens the focus of prevention to all complications of mechanical ventilation rather than just pneumonia alone. There are important limitations to this work. The study was conducted at a small number of academic medical centers and predominantly assessed patients for VAP by retrospective chart review rather than prospective bedside evaluations. These limitations may impede generalizability of the results and the accuracy of surveillance time estimates. However, we believe our results are robust because they come from unrelated hospitals in different parts of the country and use surveillance impressions from multiple infection preventionists. The estimated surveillance time for conventional surveillance is likely an underestimate because retrospective chart review is more time efficient than daily bedside visits. Conversely, streamlined surveillance time will be longer in hospitals without electronic systems to automatically generate daily line lists for review. In summary, we demonstrate that a streamlined version of the CDC surveillance definition for VAP is faster and more reproducible than the conventional definition and predicts prolongation of mechanical ventilation and intensive care length of stay as effectively as the conventional definition. This streamlined definition might make VAP surveillance more feasible and useful as a tool for hospitals and regulators to monitor care and catalyze improvements for ventilated patients. Notes Acknowledgments. The authors thank the infection preventionists who assessed patients using the conventional and streamlined Centers for Disease Control and Prevention definition: Tricia Lemon, RN, MPH, Tufts Medical Center, Boston; Pamela Fox, RN, CIC, and Margaret Bikowski, MS, CIC, BrighamandWomen shospital,boston;lisahines,rn,bs,cic,theohio State University, Columbus; and Ruth Kleckner, RN, BSN; Sharon Sumner, RN, BSN; and Caroline W. Taylor, RN, MSN of Intermountain Healthcare, Salt Lake City. Financial support. This work was supported by the Prevention Epicenters Program of the Centers for Disease Control and Prevention (grant #1U01CI000344). Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed. References 1. Safdar N, Dezfulian C, Collard HR, Saint S. Clinical and economic consequences of ventilator-associated pneumonia: a systematic review. Crit Care Med 2005; 33: Beyersmann J, Gastmeier P, Grundmann H, et al. Use of multistate models to assess prolongation of intensive care unit stay due to nosocomial infection. Infect Cont Hosp Ep 2006; 27: Nguile-Makao M, Zahar JR, Francais A, et al. Attributable mortality of ventilator-associated pneumonia: respective impact of main characteristics at ICU admission and VAP onset using conditional logistic regression and multi-state models. Intens Care Med 2010; 36: Klompas M, Platt R. Ventilator-associated pneumonia the wrong quality measure for benchmarking. Ann Intern Med 2007; 147: Horan TC, Andrus M, Dudeck MA. CDC/NHSN surveillance definition of health care-associated infection and criteria for specific types of infections in the acute care setting. Am J Infect Control 2008; 36: Klompas M. Interobserver variability in ventilator-associated pneumonia surveillance. Am J Infect Control 2010; 38: Schurink CA, Van Nieuwenhoven CA, Jacobs JA, et al. Clinical pulmonary infection score for ventilator-associated pneumonia: accuracy and inter-observer variability. Intens Care Med 2004; 30: Klompas M. Does this patient have ventilator-associated pneumonia? JAMA 2007; 297: d CID 2012:54 (1 February) d Klompas et al

8 9. Tejerina E, Esteban A, Fernandez-Segoviano P, et al. Accuracy of clinical definitions of ventilator-associated pneumonia: comparison with autopsy findings. J Crit Care 2010; 25: Fabregas N, Ewig S, Torres A, et al. Clinical diagnosis of ventilator associated pneumonia revisited: comparative validation using immediate post-mortem lung biopsies. Thorax 1999; 54: Papazian L, Thomas P, Garbe L, et al. Bronchoscopic or blind sampling techniques for the diagnosis of ventilator-associated pneumonia. Am J Resp Crit Care 1995; 152: Kirtland SH, Corley DE, Winterbauer RH, et al. The diagnosis of ventilator-associated pneumonia: a comparison of histologic, microbiologic, and clinical criteria. Chest 1997; 112: 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 Resp Crit Care 1995; 151: Torres A, el-ebiary M, Padro L, et al. Validation of different techniques for the diagnosis of ventilator-associated pneumonia. Comparison with immediate postmortem pulmonary biopsy. Am J Resp Crit Care 1994; 149: 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: Klompas M, Kleinman K, Platt R. Development of an algorithm for surveillance of ventilator-associated pneumonia with electronic data and comparison of algorithm results with clinician diagnoses. Infect Cont Hosp Ep 2008; 29: Dudeck MA, Horan TC, Peterson KD, et al. National Healthcare Safety Network (NHSN) report, data summary for 2009, device-associated module. Am J Infect Control 2011; 39: Brown AJ, Prescott C. Applied mixed models in medicine. Chichester, England: John Wiley and Sons, Papazian L, Bregeon F, Thirion X, et al. Effect of ventilator-associated pneumonia on mortality and morbidity. Am J Resp Crit Care 1996; 154: Heyland DK, Cook DJ, Griffith L, Keenan SP, Brun-Buisson C. The attributable morbidity and mortality of ventilator-associated pneumonia in the critically ill patient. The Canadian Critical Trials Group. Am J Resp Crit Care 1999; 159: Rello J, Ollendorf DA, Oster G, et al. Epidemiology and outcomes of ventilator-associated pneumonia in a large US database. Chest 2002; 122: Baker AM, Meredith JW, Haponik EF. Pneumonia in intubated trauma patients. Microbiology and outcomes. Am J Resp Crit Care 1996; 153: Bercault N, Boulain T. Mortality rate attributable to ventilator-associated nosocomial pneumonia in an adult intensive care unit: a prospective casecontrol study. Crit Care Med 2001; 29: Cunnion KM, Weber DJ, Broadhead WE, Hanson LC, Pieper CF, Rutala WA. Risk factors for nosocomial pneumonia: comparing adult critical-care populations. Am J Resp Crit Care 1996; 153: Fagon JY, Chastre J, Hance AJ, Montravers P, Novara A, Gibert C. Nosocomial pneumonia in ventilated patients: a cohort study evaluating attributable mortality and hospital stay. Am J Med 1993; 94: Wunderink RG, Woldenberg LS, Zeiss J, Day CM, Ciemins J, Lacher DA. The radiologic diagnosis of autopsy-proven ventilator-associated pneumonia. Chest 1992; 101: Petersen IS, Aru A, Skodt V, et al. Evaluation of pneumonia diagnosis in intensive care patients. Scand J Infect Dis 1999; 31: 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: Klompas M. The paradox of ventilator-associated pneumonia prevention measures. Crit Care 2009; 13: Winer-Muram HT, Rubin SA, Ellis JV, et al. Pneumonia and ARDS in patients receiving mechanical ventilation: diagnostic accuracy of chest radiography. Radiology 1993; 188: Klompas M, Khan Y, Kleinman K, et al. Multicenter evaluation of a novel surveillance paradigm for complications of mechanical ventilation. PLoS One 2011; 6:e Fagon JY, Chastre J, Hance AJ, Domart Y, Trouillet JL, Gibert C. Evaluation of clinical judgment in the identification and treatment of nosocomial pneumonia in ventilated patients. Chest 1993; 103: Albaum MN, Hill LC, Murphy M, et al. Interobserver reliability of the chest radiograph in community-acquired pneumonia. PORT Investigators. Chest 1996; 110: Loeb MB, Carusone SB, Marrie TJ, et al. Interobserver reliability of radiologists interpretations of mobile chest radiographs for nursing home-acquired pneumonia. J Am Med Dir Assoc 2006; 7: Balthazar AB, Von Nowakonski A, De Capitani EM, Bottini PV, Terzi RG, Araujo S. Diagnostic investigation of ventilator-associated pneumonia using bronchoalveolar lavage: comparative study with a postmortem lung biopsy. Braz J Med Biol Res 2001; 34: Rapid and Reproducible VAP Surveillance d CID 2012:54 (1 February) d 377