Hospital-Acquired Pneumonia and Ventilator Associated Events

Michael Klompas

Hospital-acquired pneumonia (HAP) is the most common and morbid healthcare-associated infection.¹^,² If affects ˜1% of hospitalized patients and 5%-15% of patients on mechanical ventilation.¹^,²^,³ Although the risk of pneumonia is substantially higher among patients on mechanical ventilation, nonventilator hospital-acquired pneumonia (NV-HAP) accounts for numerically more cases in most hospitals by virtue of the greater number of nonventilated patients than ventilated patients.¹^,² Crude mortality rates for patients with either NV-HAP or ventilator-associated pneumonia (VAP) range between 15% and 30%.⁴^,⁵^,⁶ Concern for these conditions is also among the greatest driver of antibiotic utilization in hospitalized patients, much of which appears to be unnecessary.⁷^,⁸

The frequency, morbidity, and high antibiotic utilization associated with NV-HAP and VAP make them the important targets for healthcare epidemiologists. Many aspects of both conditions, however, remain uncertain. Diagnostic criteria for pneumonia are subjective, insensitive, and nonspecific.⁹^,¹⁰ This makes surveillance difficult, time-consuming, and unreliable. It also complicates the interpretation of prevention studies and quality improvement initiatives since lower rates are not necessarily indicative of true decreases in disease.¹¹

The U.S. Centers for Disease Control and Prevention (CDC) created ventilator-associated event (VAE) definitions as a potential alternative way to track safety and quality for patients on mechanical ventilation.¹² VAE definitions were designed to broaden the focus of surveillance to include multiple serious complications of mechanical ventilation in addition to pneumonia while simultaneously making surveillance more objective, reproducible, and amenable to automation using the data routinely found in electronic health record systems. VAE definitions are controversial, however, due to their limited overlap with traditionally defined VAP and limited data thus far on how best to prevent them.¹³ In addition, there is currently no analogous technique to track NV-HAP rates or to monitor the impact of NV-HAP prevention initiatives using objective clinical data, thereby creating a potential mismatch between surveillance and prevention frameworks for ventilated vs nonventilated patients.

This chapter will summarize current knowledge about the diagnosis, epidemiology, microbiology, risk factors, and prevention of VAP, VAE, and NV-HAP.

PATHOPHYSIOLOGY

Pneumonia refers to the inflammation of the lungs caused by infection with microorganisms and is characterized histologically by the accumulation of neutrophils in the distal bronchioles, alveoli, and interstitium. The histologic hallmark of pneumonia in hospitalized patients is heterogeneity, particularly among ventilated patients. The lungs of recently ventilated patients are often notable for multiple patchy areas of inflammation, particularly in the dependent regions, which vary markedly in age, inflammation, organisms, and healing.¹⁴^,¹⁵ This pattern suggests that patients may regularly aspirate oral secretions that trigger localized inflammatory reactions, some of which progress to clinically manifest disease and some of which resolve spontaneously without antibiotics. Radiolabeling studies and assays for pepsin and salivary amylase in endotracheal aspirates confirm that most patients routinely aspirate oropharyngeal and gastric secretions, but only a fraction of these aspiration events progress to clinically manifest pneumonias.¹⁶

DIAGNOSIS

Sir William Osler noted over a century ago that “the wards and postmortem room show a very striking contrast in their pneumonia statistics.”¹⁷ It appears that little has changed in the intervening century. Pneumonia remains a very challenging diagnosis to render in hospitalized patients. Patients’ presenting problems, comorbidities, and acute noninfectious complications are all potential sources of error when trying to diagnose pneumonia because many have signs and symptoms that overlap with the clinical and radiographic presentation of HAP. Potential mimics include pulmonary edema, pulmonary hemorrhage, pulmonary emboli, drug reactions, atelectasis, chemical aspiration, traumatic contusions, and prior pulmonary disease, alone or in combination. These conditions are common in hospitalized patients, particularly the frail and critically ill, thus creating frequent opportunities for misdiagnosis.¹⁸^,¹⁹

The classic clinical signs associated with pneumonia include fever, leukocytosis, purulent secretions, impaired oxygenation, and new or progressive radiographic infiltrates. In practice, however, none of these signs or
symptoms are particularly sensitive nor specific for pneumonia (Table 16-1).²⁰ The most suggestive sign to rule out pneumonia is the absence of a new radiographic infiltrate on portable chest imaging, but even this sign only marginally lowers the probability of pneumonia. When infiltrates are found, air bronchograms are specific for pneumonia but rarely present. Conversely, alveolar infiltrates are sensitive but nonspecific.²¹ Computed tomography (CT) can sometimes reveal consolidations that were not apparent on portable chest radiographs.²²

TABLE 16-1 Sensitivity, Specificity, and Positive and Negative Likelihood Ratios to Diagnose Ventilator-Associated Pneumonia Relative to Histopathology

*Clinical sign*	*Sensitivity*	*Specificity*	*Positive likelihood ratio*	*Negative likelihood ratio*
Fever	46%-67%	42%-65%	1.2 (0.8-1.9)	0.9 (0.5-4.1)
Leukocytosis	50%-77%	45%-58%	1.3 (0.8-2.4)	0.7 (0.3-1.6)
Purulent secretions	69%-83%	33%-42%	1.3 (0.9-1.8)	0.6 (0.3-1.4)
New infiltrate	78%-100%	33%-75%	1.7 (1.1-2.5)	0.4 (0.1-0.9)
Values drawn from Klompas M. Does this patient have ventilator-associated pneumonia? JAMA. 2007;297(14):1583-1593.

Combining multiple signs does little to improve accuracy. Radiographic infiltrates and ≥2 of fever, leukocytosis, and purulent secretions, for example, have a sensitivity of 65%-69% and specificity of 36%-75% relative to autopsy.¹⁰^,²³ Scoring systems have been proposed that integrate multiple clinical signs to diagnose VAP by assigning points for each sign. The Clinical Pulmonary Infection Score (CPIS), for example, assigns between 0 and 2 points for each of temperature, white blood cell count, PaO₂:FiO₂ ratio, radiographic findings, quantity of tracheal secretions, and culture results depending on the magnitude and/or severity of each factor. A total score of >6 is supposed to suggest VAP. In practice, however, the CPIS performs no better than the simple combinations of signs noted above. The net sensitivity of a CPIS score >6 is 46%-77% with a specificity of 42%-85% relative to histology.¹⁰^,²³^,²⁴

CT may help increase diagnostic certainty, particularly in patients with equivocal signs and symptoms. CT tends to confirm disease in patients with high pretest probability of disease but can often rule out pneumonia in patients with equivocal presentations.²⁵ The test is most useful, then, for borderline cases. If intravenous contrast is administered, it can help differentiate whether consolidations are more likely due to atelectasis or pneumonia. Atelectasis enhances with radiographic contrast, whereas pneumonia does not.²⁶ Even CT, however, corresponds variably with histological diagnoses.²⁷

Microbiological cultures are also prone to both false positives and false negatives. False positives are due to contamination of specimens with bacteria colonizing the oropharynx or, in ventilated patients, the endotracheal tube or tracheotomy site. False negatives are due to prior exposure to antibiotics (which is common in hospitalized patients) or failure to sample the correct segment of the lung. There is a long-standing controversy in the field over whether invasive samples such as bronchoalveolar lavage are superior to noninvasive samples such as endotracheal aspirates. European guidelines tend to favor invasive sampling, whereas American guidelines do not.²⁸^,²⁹ The American guidelines include a summary of the accuracy of various pulmonary sampling methods relative to histopathology pooled from nine different studies (Table 16-2).²⁹ In general, qualitative endotracheal aspirate cultures tend to be more sensitive than quantitative bronchoalveolar lavage cultures, but quantitative bronchoalveolar lavage cultures are more specific. The sensitivity and specificity of quantitative endotracheal aspirates, however, tend to be similar to quantitative bronchoalveolar lavage cultures.

The impact of invasive vs noninvasive sampling on patient outcomes has been evaluated in five randomized controlled trials. In a meta-analysis including 1367 patients, there were no apparent differences between these two strategies with regard to mortality, duration of mechanical ventilation, intensive care unit (ICU) length-of-stay, or changes in antibiotics.³⁰ The two largest studies contributing to the meta-analysis, however, had important methodological differences that color their interpretation.

Fagon and colleagues randomized 413 patients with suspected VAP to semiquantitative cultures of endotracheal aspirates vs bronchoscopy with microscopic examination of specimens and quantitative cultures.³¹ Clinicians in both arms were directed to initiate antibiotics only if there were signs of severe sepsis or if bacteria were present on Gram stain and then to modify antibiotics in accordance with culture results or to stop antibiotics if cultures were negative. Antibiotics were started in 91% of patients randomized to endotracheal aspirates vs 52% of patients randomized to bronchoscopy. The invasive strategy was associated with significantly more 28-day antibiotic-free days (12 vs 8 days, P < .001) and lower 14-day mortality rates (16% vs 26%, P = .02) but no difference in ventilator-free days, ICU length-of-stay, hospital length-of-stay, or 28-day mortality rates (31% vs 39%, P = .10). The Canadian Critical Care Trials Group randomized 740 patients with suspected VAP to either endotracheal tube cultures with nonquantitative cultures or bronchoalveolar lavage with quantitative cultures.³² In contrast to the study by Fagon and colleagues, however, empiric antibiotics were started in all patients regardless of Gram stain findings (but were modified or stopped at clinicians’ discretion once cultures returned). In addition, the investigators did not enroll patients known to be infected or colonized with Staphylococcus aureus or Pseudomonas species, potentially biasing the study population toward patients less likely to have drug-resistant pathogens and thus patients in whom culture findings might be less likely to inform treatment modifications. The investigators ultimately found no difference between arms in
antibiotic-free days, duration of mechanical ventilation, ICU length-of-stay, or 28-day mortality.

Given the differences in enrollment criteria and management protocols between these two pivotal trials, it is possible that invasive sampling methods may have a role to play in decreasing antibiotic utilization but only if paired with a rigorous protocol limiting antibiotic starts to patients with severe sepsis or positive Gram stains and stopping antibiotics if cultures are negative. Indeed, other studies suggest similar results can be obtained by limiting antibiotic starts to patients with septic shock or cultureproven disease without recourse to bronchoscopy.³³^,³⁴^,³⁵

All told, the weight of evidence suggests that routine clinical diagnosis of pneumonia is often wrong. Up to two-thirds of patients started on antibiotics for possible pneumonia most likely do not have pneumonia, and up to one-third of patients with pneumonia may be missed.⁸^,¹⁰^,³⁶ The limited accuracy of pneumonia diagnostic criteria has critical implications for the development, interpretation, and optimization of antibiotic stewardship, surveillance, and prevention strategies for VAP, VAE, and NV-HAP.

TABLE 16-2 Accuracy of Pulmonary Culture Sampling Methods Relative to Histopathology

*Diagnostic method*	*Sensitivity*	*Specificity*	*Positive predictive value*	*Positive likelihood ratio*	*Negative likelihood ratio*
Endotracheal aspirates (any growth)	75% (58%-88%)	47% (29%-65%)	61% (45%-76%)	1.4 (0.74-2.49)	0.56 (0.17-1.83)
Endotracheal aspirates (≥10⁵ CFU/mL)	57% (45%-69%)	83% (70%-92%)	81% (67%-91%)	3.3 (0.88-11)	0.53 (0.35-0.81)
Quantitative BAL (≥10⁴ CFU/mL)	57% (47-66%)	80% (71-88%)	77% (66%-85%)	2.4 (0.99-5.6)	0.56 (0.33-0.96)
Protected specimen brush (≥10³ CFU/mL)	48% (38%-57%)	72% (63%-80%)	60% (49%-71%)	1.9 (0.98-3.6)	0.72 (0.51-1.0)
Values in brackets are 95% confidence intervals. BAL, bronchoalveolar lavage; CFU, colony-forming units. Values drawn from Kalil AC, Metersky ML, Klompas M, et al. Management of adults with hospital-acquired and ventilator-associated pneumonia: 2016 clinical practice guidelines by the Infectious Diseases Society of America and the American Thoracic Society. Clin Infect Dis. 2016;63(5):e61-e111.

ANTIBIOTIC STEWARDSHIP

Suspected respiratory infections are the most common indication for antibiotics in hospitalized patients, yet many of these courses are likely unnecessary given the high rate of overdiagnosis for both VAP and NV-HAP as well as increasing recognition that many HAPs may be viral rather than bacterial.⁷^,³⁷ Multiple strategies have been proposed to decrease unnecessary antibiotic utilization, including strategies to decrease antibiotic starts, to limit the breadth of antibiotics when agents are started, to encourage early de-escalation, and to limit the total duration of treatment. The traditional approach to antibiotic starts over the past two decades has been to encourage early initiation of broadspectrum antibiotics even when the diagnosis of pneumonia is unclear given data associating delays in appropriate therapy with increased mortality. More recent studies suggest a more nuanced approach. The increased risk of death associated with delays in appropriate antibiotics is most evident for patients with sepsis, particularly septic shock.³⁸^,³⁹ Multiple investigations bear out the safety (and potential benefit) of gathering diagnostic data to confirm infection before starting antibiotics for patients without septic shock.³¹^,³³^,³⁴^,³⁵ The breadth of treatment should be informed by both patient-specific risk factors for antibiotic-resistant organisms and the local ecology of antibiotic resistance in the patient’s unit of care.²⁹ Evaluation of Gram stain morphology can also help inform whether to include coverage for MRSA (reserve for patients with Gram-positive cocci on Gram stain as well as risk factors for MRSA) or conversely double coverage for potentially resistant Gramnegative bacilli (reserve for patients with Gram-negative organisms on Gram stain and risk factors for drug resistant pathogens).⁴⁰^,⁴¹ Once antibiotics have been started, patients should be re-evaluated daily to determine whether there is ongoing necessity for antibiotics (stop if the diagnosis of pneumonia has not been borne out) and to focus treatment according to culture results. Observational data suggest it is safe to stop antibiotics after as few as 2 days in patients with minimal and stable ventilator settings.⁴²^,⁴³ Current guidelines from both the United States and Europe suggest treating VAP and NV-HAP for no more than 7 days so long as the patient is improving.²⁸^,²⁹ It may be possible to shorten treatment even further by monitoring serial procalcitonin values to help confirm response to treatment.⁴⁴

SURVEILLANCE

Surveillance for HAP is inherently challenging because of the complexity and subjectivity of pneumonia diagnostic criteria combined with the large number of patients that need to be evaluated. Both the European and U.S. Centers for Disease Control and Prevention (CDC) publish surveillance definitions for HAP. These criteria were similar for many years but diverged when CDC supplanted their traditional VAP definition with VAE definitions for ventilated patients. CDC does still maintain their traditional pneumonia definitions, but only for HAP surveillance in nonventilated patients and as the basis for attributing bloodstream infections to pneumonia rather than to central lines.

TABLE 16-3 CDC’s Traditional Surveillance Criteria for Pneumonia (PNU1)

Systemic criteria	At least one of the following: Temperature >38.0°C White blood cell count ≤4000 or ≥12 000 cells/mm³ Altered mental status in a patient aged ≥70 years
Pulmonary criteria	At least two of the following: New onset of purulent sputum, change in character of sputum, increased secretions, or increased suctioning requirements New onset or worsening cough, or dyspnea, or tachypnea Rales or bronchial breath sounds Worsening gas exchange, increased oxygen requirements, or increased ventilator demand
Radiographic criteria	Two^a or more serial chest imaging studies with at least one of the following: New and persistent or progressive and persistent infiltrate, consolidation, or cavitation
^aOne definitive imaging study is adequate for patients without underlying pulmonary or cardiac disease.

CDC maintains three different traditional pneumonia surveillance definitions that vary slightly in the precise number and nature of their criteria.⁴⁵ PNU1 is designed to identify pneumonias without laboratory confirmation, PNU2 is for laboratory-confirmed cases, and PNU3 is for immunocompromised patients. All three definitions require patients to meet systemic, pulmonary, and radiographic criteria although the specifics vary by definition. The criteria for PNU1 are shown in Table 16-3.

Different hospitals have operationalized the CDC surveillance definitions in different ways. Some prospectively screen all patients for all criteria, some use chest imaging as a screen and only review patients with infiltrates, and some use positive sputum cultures as a screen (perforce missing patients if sputum cultures were not sent even though some might have met one of the nonlaboratory definitions for pneumonia). In practice, however, most hospitals now conduct VAE surveillance for the ventilated population rather than PNU surveillance. Very few hospitals have surveillance systems to track HAP in nonventilated patients, but those that do still rely on PNU definitions. Some parties have proposed using administrative data (discharge diagnosis codes) to track VAP or NV-HAP, but the sensitivity and positive predictive value of diagnosis codes for nosocomial pneumonia are unacceptably low.⁴^,⁴⁶^,⁴⁷^,⁴⁸ Pilot work exploring the feasibility of automatable definitions for NV-HAP surveillance using routine electronic clinical data is underway.³

CDC’s traditional surveillance criteria for pneumonia have been criticized because they are complicated and time-consuming to apply yet subjective and nonspecific. The subjective components of the PNU definitions, such as “change in character of sputum,” “increased oxygen requirements,” and “new and progressive infiltrates,” make them prone to high levels of disagreement between different observers. Studies comparing infection preventionists routinely find that agreement is fair at best. In one study, for example, 50 ventilated patients with respiratory deterioration were independently evaluated by three different surveyors.⁴⁹ The surveyors reported an almost twofold variation in perceived VAP rates. In another study, investigators created six case vignettes of patients with possible VAP and distributed them to surveillance personnel around the country, asking how many of the six met traditional CDC criteria for VAP.⁹ Nearly equal numbers of respondents thought that one, two, three, four, or five of the vignettes were consistent with VAP. Of note, the low rate of agreement among infection preventionists evaluating patients for VAP mirrors the low rate of agreement between physicians evaluating patients for VAP.⁴⁹^,⁵⁰^,⁵¹^,⁵²^,⁵³

The subjectivity and complexity of traditional pneumonia surveillance definitions are compounded by their lack of specificity. Differences in the background prevalence of the many conditions that can mimic the clinical presentation of pneumonia can have a large effect on the perceived pneumonia rate in a unit.⁵⁴ Differences in microbiological diagnostic protocols in different units can also have a manyfold impact on perceived pneumonia rates.⁵⁵ The net consequence of the many sources of error and variability in pneumonia surveillance using traditional definitions is that infection prevention programs can invest substantial time and resources into surveillance without any assurance that rates are credible, reproducible, comparable to other institutions, or reliable proxies for true pneumonia.

VENTILATOR-ASSOCIATED EVENTS

The limitations of traditional VAP definitions prompted the CDC to develop VAE definitions (publicly released in 2013). VAE definitions were designed to meet three major objectives: (1) broaden the focus of surveillance to capture a fuller array of serious complications in ventilated patients beyond pneumonia, (2) increase the objectivity and reproducibility of surveillance, and (3) enable the possibility of fully automated surveillance using electronic clinical data.¹² Importantly, however, VAE definitions were not designed to increase the sensitivity or specificity of pneumonia surveillance. Rather, they follow from the recognition that clinical criteria for VAP are inherently inaccurate, and thus it is unrealistic to expect standard surveillance criteria to yield accurate diagnoses.

VAE definitions are predicated upon the principal that most practitioners work to minimize the degree and duration of ventilator support whenever possible in order
to ready patients for extubation and minimize the risks of barotrauma, hyperoxia, and other complications of mechanical ventilation. Usual practice is therefore to continually decrease patients’ ventilator settings to the lowest levels of ventilator support that can provide adequate oxygenation and ventilation. If a patient’s ventilator settings are stable or declining for a number of days but then rise and remain elevated for a sustained period, it is a signal that the patient may have suffered a respiratory complication. Two ventilator settings are used to identify VAEs in adults, the positive end-expiratory pressure (PEEP) and the fraction of inspired oxygen (FiO₂), while mean airway pressure (MAP) and FiO₂ are used to identify VAEs in children. VAEs are defined in adult patients as either (1) an increase in the daily minimum PEEP of ≥3 cm H₂O sustained for ≥2 calendar days following ≥2 calendar days of stable or decreasing daily minimum PEEP values or (2) an increase in the daily minimum FiO₂ of ≥20 points sustained for ≥2 calendar days following ≥2 calendars of stable or decreasing daily minimum FiO₂ values. VAEs in pediatric patients (PedVAE) are defined as either (1) an increase in the daily minimum MAP of ≥4 cm H₂O sustained for ≥2 calendar days following ≥2 calendar days of stable or decreasing daily minimum MAP or (2) an increase in the daily minimum FiO₂ of ≥25 points sustained for ≥2 calendar days following ≥2 calendars of stable or decreasing daily minimum FiO₂ values. Notably, neither adult nor pediatric VAE definitions include any radiographic criteria. The CDC elected to exclude radiographic criteria from VAE definitions given their subjectivity, complexity, and variable accuracy.²¹^,²⁷^,⁵⁶^,⁵⁷

The adult VAE framework does include two subcategories designed to identify the subset of respiratory complications that may be due to infections. The first subcategory is termed, infection-related ventilator-associated complications (IVACs). An IVAC is present if a patient meets VAE criteria and has a concurrent fever or abnormal white blood cell count, and new antibiotics are started and continued for at least 4 calendar days. The supporting criteria for IVAC must begin within 2 days of the first day of increased ventilator settings, excluding the patient’s first 2 days of mechanical ventilation. The second VAE subcategory is called possible ventilator-associated pneumonia (PVAP) and is present in a patient with IVAC who has concurrent laboratory evidence for a possible lung infection. This can include suggestive histopathology, positive cultures from pleural fluid, positive assays for respiratory viruses or Legionella, positive quantitative respiratory cultures that exceed a specified minimum growth threshold, or positive bacteria cultures with any amount of growth combined with Gram stain evidence of purulence (≥25 polymorphonuclear cells and ≤10 squamous epithelial cells per low-power field). Examples of VAE, IVAC, and PVAP are shown in Figure 16-1. PedVAE does not currently include criteria to identify the fraction of events that might be infectious although exploratory work in this arena has been published.⁵⁸^,⁵⁹

Clinical Correlates of VAEs

VAEs are nonspecific surveillance indicators of nosocomial respiratory deterioration. Multiple investigators have published case series of VAEs that included chart reviews designed to detect the clinical diagnoses that triggered the increase in ventilator settings that led to the fulfillment of VAE criteria. Most VAEs are triggered by one of the four clinical events: pneumonia (25%-40% of cases), fluid overload including pulmonary edema (15%-50% of cases), atelectasis (10%-15% of cases), and acute respiratory distress syndrome (ARDS) (5%-20% of cases).⁵³^,⁶⁰^,⁶¹^,⁶²^,⁶³^,⁶⁴^,⁶⁵ Less common triggers for VAEs include mucous plugging, abdominal compartment syndrome, pulmonary embolism, pneumothorax, radiation pneumonitis, sepsis syndromes due to extrapulmonary infections, stroke, and transfusion-associated lung injury. In some cases, there is no apparent clinical trigger that can be identified through retrospective chart review for the increase in ventilator settings that precipitated a VAE.

Overlap Between VAE and Traditionally Defined VAP

Concordance between VAE and traditionally defined VAP is limited. On meta-analysis, the estimated sensitivity of VAE criteria for traditionally defined VAP was only 41.8% (95% CI 18%-66%, 11 studies) and the positive predictive value 23% (95% CI 13%-34%, 9 studies).⁶⁶ The low positive predictive value of VAE for VAP is consistent with CDC’s intent to broaden the focus of surveillance from pneumonia alone to include a wider array of respiratory complications in mechanically ventilated patients. VAE’s limited sensitivity for traditionally defined VAP reflects the threshold effect imposed by the requirement for a sustained increase in ventilator settings to meet VAE criteria. Many clinically diagnosed VAPs do not meet VAE’s threshold for increased ventilator settings.⁶⁷ These cases present a quandary for surveyors because they are common yet the clinical significance of pneumonia without impaired oxygenation is unclear, particularly in light of the known high rate of VAP overdiagnosis. What is known is that clinically diagnosed pneumonias that do not require increases in ventilator support are associated with more benign clinical courses. They are associated with less time to extubation and lower mortality rates than VAPs that do require increased ventilator support. In addition, patients treated for VAP who have minimal and stable ventilator settings have similar outcomes if they are treated with very short courses of antibiotics (median 2 days) or more conventional courses (median 8 days).⁴²^,⁶⁸

VAE Surveillance

VAE surveillance can be conducted manually, electronically, or through a combination of both methods. A number of hospitals have successfully automated VAE surveillance using routine electronic health record data.⁶⁹^,⁷⁰^,⁷¹ Some hospitals that have automated VAE surveillance report have reported that electronic VAE surveillance is more sensitive and accurate than manual VAE surveillance.⁶⁹^,⁷¹ This is notable insofar as VAE definitions were designed to minimize subjectivity in surveillance and increase reliability no matter the surveillance method. The cases missed and misclassified by human reviewers, however, are a reminder that human error is an additional factor underlying interobserver variability in surveillance, even when using more objective definitions. Other investigators have demonstrated that VAE case detection can vary depending on how one obtains and analyzes ventilator settings.

Analyzing hourly settings recorded in nursing flow sheets will identify a different set of cases compared to analyzing an electronic feed of minute-to-minute ventilator settings.⁶² The CDC has tried to minimize this source of variability by requiring that a PEEP value must be maintained for at least 1 hour in order to be considered the daily minimum PEEP, but other variations in how ventilator settings are entered and captured may yet influence VAE rates.

FIGURE 16-1 Examples of ventilator-associated events. Panel A depicts a VAE. This patient was on stable ventilator support with a PEEP of 5 cm H₂O from January 3-5, but the PEEP was increased to 8 cm H₂O on January 6 and sustained at this higher level. The patient meets VAE criteria with an event date of January 6. Panel B depicts an IVAC. This patient met VAE criteria on March 6 but also had the additional criteria necessary for IVAC within the 5-day window beginning 2 days before the increase in ventilator settings (ie, January 4-8). These include fever, leukocytosis, and a new course of antibiotics that continues for at least 4 calendar days. Panel C depicts a PVAP. This patient met VAE criteria on April 6 on account of a sustained increase in FiO₂, IVAC criteria on the basis of fever + leukocytosis + an antibiotic start within the 5-day window spanning April 4-9, and a positive culture for S aureus that exceeded the minimum quantitative growth threshold for bronchoalveolar lavage (≥10⁴ colonyforming units/mL). BAL, bronchoalveolar lavage; EA, endotracheal aspirate; FiO₂, fraction of inspired oxygen; LPF, low-power field; PEEP, positive end-expiratory pressure; WBC, white blood cell.

Hospitals without access to automated surveillance software can operationalize VAE surveillance by constructing line lists for all patients on mechanical ventilation for ≥4 days. The line list should include one line per calendar day per patient (see Figure 16-1 for examples). Manual surveyors can take advantage of the fact that VAE definitions are nested: only patients that meet VAE criteria are eligible for IVAC and only patients that meet IVAC criteria are eligible for PVAP. This means that hospitals conducting manual surveillance can focus on VAE case finding alone for the population at large (which only requires two variables, daily minimum PEEP and FiO₂) and limit additional data collection (temperature, white blood cell count, antibiotic exposures, Gram stains, and culture results) to the subset of patients that meet VAE criteria (typically about 5% of the ventilated population). An initial patient line list then should include at a minimum target patients’ daily minimum PEEP and FiO₂ values for adults or daily minimum MAP and FiO₂ for children. This list can then be analyzed by scanning for a period of at least 2 days of stable or decreasing ventilator settings followed by at least 2 days of higher ventilator settings (increase in daily minimum PEEP of ≥3 cm H₂O or increase in daily minimum FiO₂ of ≥20 points for adults, increase in daily minimum MAP of ≥4 cm H₂O or increase in daily minimum FiO₂ of ≥25 points for children). Once a VAE has been identified, surveyors working in adult units can then assess whether the event meets IVAC criteria by adding daily minimum and maximum temperatures, white blood cell counts, and antibiotic exposures to the line list within the 5-day window surrounding the first day of increased ventilator settings (2 days before, day of increase, and 2 days following). Note that antibiotic courses can extend beyond the 5-day window immediately surrounding a VAE to meet IVAC criteria, so long as they began within the 5-day window. Patients that meet IVAC criteria can be assessed for PVAP by evaluating their concurrent microbiological data (and pathology reports if pertinent).

EPIDEMIOLOGY

It is difficult to know the precise incidence of NV-HAP and VAP due to the limited accuracy and subjectivity of diagnostic criteria. A point-prevalence survey of 199 hospitals in 10 states using CDC’s PNU definitions reported a net prevalence of 0.9 cases per 100 patients, 65% of which occurred in nonventilated patients. This corresponds to 3.5 cases of VAP per 100 patients on mechanical ventilation and 0.6 cases of NV-HAP per 100 nonventilated patients. The prevalence of pneumonia can vary markedly, however, depending on the choice of definition and population. Investigators in Belgium characterized the impact of different diagnostic criteria on observed VAP rates within a single medical-surgical ICU. Perceived VAP rates ranged from 4% for patients with an infiltrate, purulent secretions, and temperature or leukocytosis to 12% using VAE’s PVAP definition, to 23% using CDC’s traditional PNU definition, to 42% using CPIS criteria.⁷² Even using the same definition, rates can vary manyfold between hospitals. For example, one retrospective analysis of NV-HAP rates in 21 US hospitals using PNU definitions reported incidence rates ranged between 0.04 and 1.11 cases per 100 patients.⁴

VAP and NV-HAP

The European CDC recently reported on a multinational point-prevalence survey of HAP rates among 15 European countries that included data from over 230 000 patients.² The overall prevalence was 1.3 HAPs per 100 patients using a definition similar to the U.S. CDC’s PNU definitions. Rates varied approximately sixfold across countries, however, from a low of 0.6 cases per 100 patients in the United Kingdom to 3.6 cases per 100 patients in Ireland. As expected, HAP rates were much higher among intubated patients compared to nonintubated patients, with a net HAP prevalence of 15% among ventilated patients but only 1.0% among nonventilated patients. System wide, however, nonventilated patients accounted for almost three times as many HAP cases as ventilated patients on account of their greater numbers. HAP risk varied by age from a low of 0.6 cases per 100 patients in patients aged 1-44 years to 1.3 cases per 100 patients in patients aged 45-74 years and to 1.8 cases per 100 patients among patients 75 years and older. Women were approximately half as likely to get HAP compared to men. The prevalence of HAP increased by length-of-stay since admission, ranging from 0.6 cases per 100 patients among patients in hospital for ≤3 days to 1.8 cases per 100 patients among those hospitalized for ≥15 days. HAP rates were highest in ICUs (8.1 cases per 100 patients), intermediate in medical and geriatric units (1.1-1.3 cases per 100 patients), lower in surgical units (0.8 cases per 100 patients), and lowest in gynecology and obstetrics units (0.1 cases per 100 patients).

Ventilator-Associated Events

The U.S. CDC has published data on the basic epidemiology of VAEs drawn from 1824 U.S. facilities.⁷³ Median VAE rates were highest in trauma units (11.0 cases per 1000 ventilator days), major teaching hospital surgical units (9.0 cases per 1000 ventilator days), neurologic units (8.6 cases per 1000 ventilator days), major teaching hospital medical units (7.7 cases per 1000 ventilator days), and burn units (6.5 cases per 1000 ventilator days); intermediate in medical cardiac units (5.5 cases per 1000 ventilator days) and surgical cardiothoracic units (4.9 cases per 1000 ventilator days); and lowest in small nonteaching hospitals’ medical units (1.1 cases per 1000 ventilator days). The proportion of VAEs that qualified as IVACs also varied substantially by unit, ranging from 40% to 50% in burn and surgery units, to 35% in medical units, and to 30% in cardiac units. The median interval from intubation to VAE was 6 days (interquartile range 4-9 days).

Morbidity and Mortality

Crude mortality rates for HAP range from 15% to 30% in most recent series. A prospective cohort study of 14 212 patients admitted to 23 French ICUs, for example, reported 30-day mortality rates of 28% for VAP and 24% for ICU-acquired
NV-HAP.⁶ Studies that included non-ICU patients have reported lower crude mortality rates for NV-HAP. A retrospective chart review of 1300 patients with NV-HAP admitted to 21 hospitals, for example, reported a hospital mortality rate of 16%.⁴ Notably, however, mortality rates may be similar for NV-HAP and VAP when looking at both diagnoses within a common population using a common case finding strategy. The Pennsylvania Patient Safety Authority, for example, found that NV-HAP and VAP mortality rates statewide using PNU definitions were 18.7% and 18.9%, respectively.⁷⁴ Likewise, the New York City health department reported mortality rates of 20.7% for NV-HAP and 21.6% for VAP using administrative claims data from all New York City hospitals.⁷⁵ VAE mortality rates tend to be similar to or higher than VAP mortality rates. The overall mortality rate among 19 676 VAEs reported to CDC in 2014 was 31% for VAE, 27% for IVAC, and 34% for non-IVAC VAEs.⁷³ A metaanalysis of five trials comparing VAP and VAE mortality rates estimated that VAEs were associated with a 50% higher odds of death compared to traditionally defined VAP.⁶⁶

Notwithstanding the high crude mortality rates associated with NV-HAP, VAP, and VAE, the attributable mortality rate is likely much lower. The challenge is that patients at risk for these events tend to be older, to have many comorbidities, and to be hospitalized for long periods of time before they develop NV-HAP, VAP, or VAE. These are all independent risk factors for death. A number of investigators have attempted to quantify the marginal contribution of NV-HAP, VAP, and VAE to mortality risk. Case-control studies that match NV-HAP patients to non-NV-HAP patients and attempt to control for residual differences between the two groups have reported that NV-HAP is associated with a four- to eightfold higher risk of hospital death compared to non-NV-HAP patients.³^,⁵ Similar analyses for VAP, however, have yielded inconsistent results: some studies have reported higher mortality rates in patients with VAP compared to matched controls, while others have not.⁷⁶ These discrepancies may be due to the difficulty and variability defining VAP and the possibility that some cases flagged by traditional definitions may better reflect colonization rather than infection. Most studies of VAEs have found that patients with VAEs are approximately twice as likely to die as matched controls without VAEs and 50% more likely to die compared to patients with traditionally defined VAP.⁶²^,⁶⁶^,⁶⁸^,⁷⁷

More recent studies have applied more sophisticated methods to characterize the attributable mortality of NV-HAP and VAP. Competing risk analyses using advanced statistical methods to account for time-dependent confounding have estimated attributable mortality rates for VAP of about 8%.⁷⁸^,⁷⁹ Similar methods suggest that ICU-acquired pneumonia may extend ICU length-of-stay by about 4 days.⁸⁰ All told, VAP may be directly responsible for as few as 4%-6% of ICU deaths.⁸¹ Melsen and colleagues estimated the attributable mortality of VAP in a novel fashion by using data from randomized trials of prevention strategies.⁸² They calculated the average reduction in VAP rates and mortality rates using individual patient-level data from 24 randomized controls of different VAP prevention measures. The prevention measures were collectively associated with a 30% reduction in VAP rates and a (nonsignificant) 4% reduction in mortality rates. They thus extrapolated that a 100% reduction in VAP rates would have led to a 13% reduction in mortality rates and thus the attributable mortality of VAP is 13%.

MICROBIOLOGY

Among 8133 VAPs reported to the CDC between 2011 and 2014, S aureus was the most common isolate (25%) followed by Pseudomonas aeruginosa (17%), Klebsiella pneumoniae (10%), Enterobacter species (8.3%), and Escherichia coli (5.4%).⁸³ Pathogens associated with VAEs are similar. Among 2517 possible and probable pneumonias detected using VAE criteria, S aureus predominated (28%), followed by P aeruginosa (13%), K pneumoniae (7.9%), E coli (6.0%), and Haemophilus influenzae (4.7%).⁷³ In the European CDC’s 15-country point-prevalence survey of HAP, nonfermenting Gram negatives predominated: P aeruginosa (19%), Acinetobacter species (13%), Klebsiella species (12%), and S aureus (10%).² There were some differences in pathogens between VAP and NV-HAP: Gram-positive cocci were less frequent in VAP (15% vs 24%), and nonfermenting Gram-negative bacilli were more common (40% vs 30%). Acinetobacter species in particular were notably more common among intubated patients (13% vs 5%).

Only gold members can continue reading. Log In or Register to continue