(See the article by See et al, on pages 818–824.)

Landmark studies from the 1970s and 1980s established surgical-site infections as the most common nosocomial infections and devices as the most common risk factors for nonsurgical-site nosocomial infections.Reference Haley, Culver, White, Morgan and Emori ^{1 –} Reference Wenzel, Osterman and Donowitz ³ Infection prevention and control programs have consequently dedicated the majority of their efforts for the past 30 years to monitoring and preventing surgical-site and device-associated infections. For pneumonia, most hospitals only conduct surveillance and prevention programs for ventilator-associated pneumonia (VAP). Little or nothing is done to track and prevent nonventilator-associated hospital-acquired pneumonia (NV-HAP).

The original decision to focus on VAP is understandable. The disease has a high mortality rate and afflicts a vulnerable, well-defined segment of the hospital population. National quality improvement initiatives implicitly encouraged hospitals to focus on VAP in particular by repeatedly proposing public reporting and pay-for-performance programs for VAP but not NV-HAP. Groups like the Institute for Healthcare Improvement promulgated well-defined prevention bundles for VAP but not NV-HAP. The National Healthcare Safety Network reported device-associated pneumonia but not NV-HAP rates.

The landscape of nosocomial infection epidemiology has significantly changed, however, since the landmark studies of the 1970s and 1980s. A recent multistate point-prevalence survey organized by the Centers for Disease Control and Prevention (CDC) found that pneumonia is now tied with surgical-site infections as the most common healthcare-associated infection and that the majority of pneumonias now occur in nonventilated patients.Reference Magill, Edwards and Bamberg ⁴ The CDC estimates that more than two-thirds of the 157,500 hospital-acquired pneumonias in the United States each year occur in nonventilated patients.

These observations suggest the possibility that there is a major gap in current surveillance and prevention programs. Most hospitals do not routinely track NV-HAP, and almost everything we know about nosocomial pneumonia comes from studies in ventilated patients. Key questions, such as the incidence, risk factors, attributable mortality, and preventability of NV-HAP, have barely been studied.

As with any infection control problem, the first step toward tackling NV-HAP is to define it. The CDC does have a longstanding surveillance definition for hospital-acquired pneumonia that is intended to apply to both ventilated and nonventilated patients (PNEU). Ample data are available on the performance of PNEU in ventilated patients, but very little data are available on its performance in nonventilated patients. This gap is important because the studies in ventilated patients found surveillance for PNEU to be time-consuming, subjective, and nonspecific. Do the same concerns also apply to nonventilated patients? The limitations of PNEU in ventilated patients helped catalyze the development of ventilator-associated event definitions as objective surveillance alternatives for this population. Is something similar required for nonventilated patients?

This month’s issue of Infection Control and Hospital Epidemiology includes a valuable report on the performance of PNEU in nonventilated patients. Dr. See et alReference See, Chang and Gualand ⁵ from the CDC conducted a retrospective analysis of 250 adults and children with either PNEU or lower respiratory tract infection (LRI) as detected by infection preventionists in 8 Pennsylvania hospitals. Expert personnel from the CDC and the Pennsylvania Department of Health reviewed the charts of these patients to confirm PNEU and LRI classifications. The reviewers also looked for clinical diagnoses assigned at the time of PNEU or LRI to gather insights into the clinical significance of these surveillance events. The report includes data on the performance of PNEU in nonventilated adults, PNEU in ventilated and nonventilated children, and the relative value of LRI subcategories for bronchitis and other infections of the lower respiratory tract among adults and children. The current discussion, however, is focused on PNEU in adults.

A total of 96 PNEU events were reported in nonventilated adults. The review team revoked 15 cases (16%) for failure to fulfill PNEU criteria, typically because of differences between observers in the interpretation of radiographic reports. Thus, PNEU may be susceptible to some of the same sources of disagreement in nonventilated patients as it is in ventilated patients. Of the remaining 81 cases, frontline clinicians documented a clinical diagnosis of pneumonia in 69 patients. The positive predictive value of PNEU for clinically diagnosed pneumonia in nonventilated patients was therefore either 69 of 96 (72%) or 69 of 81 (84%), depending on whether one includes all PNEUs reported by frontline infection preventionists in the analysis or only the subset confirmed by the CDC and health department experts. The study did not include reviews of patients without PNEU or LRI, hence the investigators were unable to estimate the sensitivity of PNEU. The report also did not quantify the time required for PNEU surveillance or the level of agreement between different infection preventionists; these are important questions for future studies.

At first blush, a positive predictive value of 72%–84% might appear quite tolerable. Notably, however, the reference standard for this calculation was clinical documentation of pneumonia. The clinical diagnosis of pneumonia is notoriously difficult, variable, and prone to error, particularly in hospitalized patients with comorbid illnesses. Most of the signs used to identify pneumonia are subjective, and all are nonspecific. Case series suggest that both overdiagnosis and underdiagnosis are common.Reference Roulson, Benbow and Hasleton ^{6 ,} Reference Russell, Koch, Laurenson, O’Shea, Sutherland and Mackintosh ⁷ Sources of false positives include pulmonary edema, sterile aspiration pneumonitis, pulmonary embolism, lobar collapse, obstructive lung disease exacerbations, vasculitis, cryptogenic organizing pneumonia, sarcoidosis, lung cancer, and others. In addition, the diagnosis of pneumonia itself can encompass a wide range of disease severities. Some aspiration pneumonias, for example, improve spontaneously in 1–2 days without treatment, whereas others might trigger multisystem organ failure culminating in death. Whether PNEU is the best way to track respiratory complications in nonventilated patients and whether we need different definitions that overcome some of these limitations are questions that require further investigation.

Notably, See et alReference See, Chang and Gualand ⁵ documented an absolute mortality rate of 31% in patients with confirmed nonventilator PNEU. This startlingly high figure matches the crude mortality rate for clinically defined VAP and ventilator-associated events. Whether these events were due to true pneumonias, mimicking conditions, or both, the high mortality rate underscores the urgent need to learn more about these events, to accurately characterize their attributable mortality, and to find ways to prevent them.

Indeed, our current understanding of how best to prevent NV-HAP is exceedingly limited. Most of our assumptions about prevention are derived from studies of VAP. Proposed strategies include elevating the head of the bed (particularly during feeding) and enhancing oral hygiene. The most robust studies of these interventions to date, however, were performed in nursing home residents and had conflicting results.Reference Juthani-Mehta, Van Ness and McGloin ^{8 ,} Reference Yoneyama, Yoshida and Ohrui ⁹ One study reported fewer pneumonias; the other did not. Clearly, much work remains to be done to better understand the potentially modifiable risk factors for NV-HAP in acute-care patients and to test new interventions. Considering lessons learned in ventilated populations, some additional practices that merit evaluation include minimizing sedating agents, preventing delirium, avoiding gastric acid suppressants, and enhancing early mobilization programs.Reference Klompas ¹⁰

Ultimately, the report by See et alReference See, Chang and Gualand ⁵ provides useful affirmation that we need to improve our understanding of respiratory complications in nonventilated patients. CDC surveillance data tells us that NV-HAP is now among the most common nosocomial infections. The current report indicates that nonventilator PNEU is strongly associated with clinically diagnosed pneumonia and carries a high mortality rate. We now need to work on better tools to track these events, identify their causes, and implement effective prevention strategies.