Differences between Patients in Whom Physicians Agree and Disagree about the Diagnosis of Acute Respiratory Distress Syndrome

Abstract

Rationale: Because the Berlin definition of acute respiratory distress syndrome (ARDS) has only moderate reliability, physicians disagree about the diagnosis of ARDS in some patients. Understanding the clinical differences between patients with agreement and disagreement about the diagnosis of ARDS may provide insight into the epidemiology and pathophysiology of this syndrome, and inform strategies to improve the reliability of ARDS diagnosis.

Objectives: To characterize patients with diagnostic disagreement about ARDS among critical-care-trained physicians and compare them with patients with a consensus that ARDS developed.

Methods: Patients with acute hypoxemic respiratory failure (arterial oxygen tension/pressure [Pa_O2]/fraction of inspired oxygen [Fi_O2] < 300 during invasive mechanical ventilation) were independently reviewed for ARDS by multiple critical-care physicians and categorized as consensus-ARDS, disagreement about the diagnosis, or no ARDS.

Results: Among 738 patients reviewed, 110 (15%) had consensus-ARDS, 100 (14%) had disagreement, and 528 (72%) did not have ARDS. ARDS diagnosis rates ranged from 9% to 47% across clinicians. Patients with disagreement had baseline comorbidity rates similar to those of patients with consensus-ARDS, but lower rates of ARDS risk factors and less severe measures of lung injury. Mean days of severe hypoxemia (Pa_O2/Fi_O2 < 100) were 3.2 (95% confidence interval [CI], 2.6-3.9), 2.0 (95% CI, 1.5-2.4), and 0.8 (95% CI, 0.7-0.9) among patients with consensus-ARDS, disagreement, and no ARDS, respectively. Hospital mortality was 37% (95% CI, 28-46%), 35% (95% CI, 26-44%), and 19% (95% CI, 15-22%) across groups. Simple combinations of specific ARDS risk factors and lowest Pa_O2/Fi_O2 value could effectively discriminate patients (area under the receiver operating characteristic curve = 0.90; 95% CI, 0.88-0.92). For example, 63% of patients with pneumonia, shock, and Pa_O2/Fi_O2 < 110 had consensus-ARDS, whereas 100% of patients without pneumonia or shock and Pa_O2/Fi_O2 > 180 did not have ARDS.

Conclusions: Disagreement about the diagnosis of ARDS is common and can be partly explained by the difficulty of dichotomizing patients along a continuous spectrum of ARDS manifestations. Considering both the presence of key ARDS risk factors and hypoxemia severity can help guide clinicians in identifying patients with diagnosis of ARDS agreed upon by a consensus of physicians.

Keywords: acute lung injury; diagnosis; epidemiology; risk factors.

Figure 1.

Estimated proportions of patients categorized as consensus-ARDS, disagreement, or no ARDS when reviewed by multiple physicians across ranges of individual physiologic variables. Physiologic variables analyzed were the worst values recorded during the 24 hours after ARDS onset or matched 24-hour periods during mechanical ventilation for patients without ARDS. Missing values were imputed using multiple imputation and the estimates were combined. (A) Pa_O2/Fi_O2. (B) Static lung compliance. (C) Plateau pressure. (D) Dead space fraction. ARDS = acute respiratory distress syndrome; Fi_O2 = fraction of inspired oxygen; Pa_O2 = arterial oxygen tension/pressure.

Figure 2.

Relationship between the severity of hypoxemia and the likelihood of an ARDS diagnosis, and how this relationship is modified by the presence or absence of pneumonia and shock. The figure was generated after fitting multilevel logistic regression of ARDS reviews nested within patient, where ARDS diagnosis was the outcome, and model covariates included the Pa_O2/Fi_O2, a squared Pa_O2/Fi_O2 term, interaction terms with pneumonia and shock, and clinical reviewer adjustment. ARDS = acute respiratory distress syndrome; Fi_O2 = fraction of inspired oxygen; Pa_O2 = arterial oxygen tension/pressure.