By Richard Kallet, MS, RRT, FCCM

Director of Quality Assurance, Respiratory Care Services, Department of Anesthesia, San Francisco General Hospital

Mr. Kallet reports he is a major stockholder in the Asthma & Allergy Prevention Company, is a consultant for Getinge Group, and receives grant/research support from Nihon-Kohden.

SYNOPSIS: Critical care physicians reviewed more than 700 mechanically ventilated patients with acute hypoxemia and reached a consensus on the presence of acute respiratory distress syndrome (ARDS) in 15% of patients and nonconsensus on the presence of ARDS in an additional 14% of cases. Hospital mortality was not different between these cohorts (37% and 35%, respectively).

SOURCE: Sjoding MW, Hofer TP, Co I, et al. Difference between patients in whom physicians agree and disagree about the diagnosis of ARDS. Ann Am Thorac Soc 2018; Oct 15. doi: 10.1513/AnnalsATS.201806-434OC. [Epub ahead of print].

In a single-center retrospective study, a panel of 13 critical care physicians (seven advanced pulmonary critical care fellows and six critical care faculty, three of whom had been involved in ARDS clinical trials) reviewed 738 mechanically ventilated patients with acute onset hypoxemia (an arterial oxygen tension to inspired oxygen fraction [PaO2/FiO2] < 300 mmHg) for the presence of ARDS using the Berlin Definition during the first week of hospitalization. All available electronic data were reviewed, including chest radiographs and CT scans. Clinicians used a standardized evaluation tool for identifying ARDS. Overall variation in ARDS diagnosis was examined according to variations in clinician assessments and differences between subject presentations.

Based on results of physician reviews, patients received an estimated probability of ARDS depending on the degree of physician agreement; patients with greater agreement that ARDS was present had a higher estimated ARDS probability. Patients were categorized as consensus-ARDS (probability of ARDS, > 80%), disagreement (probability of ARDS, 20-80%), and no ARDS (probability of ARDS, < 20%).

The panel performed more than 1,800 reviews. Individual members varied widely in their determination of ARDS (8-47%). However, between clinicians there was 69-83% agreement on the presence of ARDS. Only 7% of the total variation in diagnosing ARDS was attributable to clinicians vs. patient attributes. Nonetheless, in half the patients diagnosed with ARDS, there was substantial disagreement between critical care physicians. The highest rate of ARDS diagnosis occurred among faculty who had participated in ARDS clinical trials (34%), followed by other critical care faculty (30%) and fellows (20%).

Statistically significant patient characteristics distinguishing consensus vs. disagreement on the diagnosis of ARDS included the presence of pneumonia (75% vs. 49%, respectively), noncardiogenic shock (62% vs. 43%, respectively), and bilateral “airspace disease” on chest radiograph (70% vs. 23%, respectively). Smaller differences in the presence of nonpulmonary sepsis (35% vs. 32%, respectively) were not significant. Overall, derived decision rules based on these results indicated that the combination of pneumonia, noncardiogenic shock, and a nadir PaO2/FiO2 < 120 mmHg revealed the highest consensus (63%) and lowest nonconsensus (16%) for diagnosing ARDS. The only significant differences in clinical outcomes between consensus and nonconsensus cohorts diagnosed with ARDS were mean days of severe hypoxemia (3.2 days vs. 2.0 days, respectively) as well as ventilator-free days (14 days vs. 16 days, respectively).


One of the most vexing and contentious issues in ARDS is that clinicians often cannot agree on which patients actually have the syndrome. This can substantially affect how these patients are managed and their ultimate outcomes. This problem has existed for as long as the syndrome has been formally recognized. One is reminded of Petty’s famous editorial “Confessions of a Lumper,” which described misuse of the designation, although he personally considered it “a desirable lumping of a variety of pulmonary insults.”1 In this context, the lack of appreciable differences in hospital mortality between consensus and nonconsensus cohorts is the most meaningful result of this study. ARDS is an acute, hyperinflammatory process that at onset almost invariably is associated with nonpulmonary organ dysfunction regardless of severity.2 Historically, 90% of ARDS mortality results from progressive multiorgan dysfunction. The emerging picture over the past two decades suggests that common clinical interventions (e.g., use of physiologic tidal volume) that reduce the inflammatory process (rather than exacerbate it) are associated with fewer organ dysfunction episodes and better outcomes. Therefore, from a practical standpoint, clinicians should be less concerned about the certainty of diagnosing ARDS. Rather, our focus should be on whether patients are in a hyperinflammatory state and whether chest mechanics suggest the presence of either stretch-related or shear-related lung injury. In essence, once engaged, the hyperinflammatory state (and associated multiorgan injury) does not cease upon reaching arbitrary cutoffs in PaO2/FiO2. Our management tactics not succumb to that illusion.

An insightful finding of the Sjoding et al study was the influence of clinician experience. The highest rates of ARDS diagnosis were made by physicians who participated in clinical trials of ARDS, whereas the lowest was among critical care fellows. This suggests that clinicians whose practices largely focus on the detection of ARDS may be more attuned to the subtlety of syndrome presentation as well as an awareness of the multitude of less common mechanisms that produce ARDS.


  1. Petty TL. Editorial: The adult respiratory distress syndrome (confessions of a “lumper”). Am Rev Respir Dis 1975;111:713-715.
  2. Kallet RH, Lipnick MS, Zhou H, et al. Characteristics of non-pulmonary organ dysfunction at onset of acute respiratory distress syndrome based on the Berlin definition. Respir Care 2018; [in press].