The Randomized Educational Acute Respiratory Distress Syndrome Diagnosis Study

Goddard. Critical Care Medicine 2018; published first online: 12 February 2018. doi: 10.1097/CCM.0000000000003000

Clinical Question

  • For healthcare professionals interpreting chest radiographs, does an educational intervention compared to no educational intervention improve the diagnosis or exclusion of acute respiratory distress syndrome (ARDS)


  • ARDS is common, affecting nearly 25% of mechanically ventilated patients in Critical Care
  • The Berlin Consensus Definition updated the definition of ARDS in 2013 and included the interpretation of a chest radiograph (CXR)
  • It is known that the reliability and consistency of CXR interpretation is poor, and this study attempts to improve that within the context of a large observational study investigating the epidemiology of ARDS: LUNG SAFE


  • Randomised, controlled trial using research study coordinators as participants
  • Participants were required to register with the READS website as part of the LUNG SAFE study
  • Randomisation took place immediately as part of registration but was concealed from the participant
  • Consent was then required, with the opportunity for participants to undergo the LUNG SAFE educational intervention without their data recorded for the READS trial
  • Investigators were blinded to group allocation until statistical analysis was complete
  • Subgroups were defined a priori


  • The LUNG SAFE study was conducted in 459 Intensive Care Units across 50 countries


  • Inclusion: registration as study coordinator for the LUNG SAFE study
  • Exclusion: none
  • 632 participants screened; 463 were enrolled into this READS substudy (212 intervention arm vs 252 control arm)
  • Despite uneven allocation numbers, baseline demographics were well-balanced
    • 80% physicians vs 6% nurses vs 14% ‘other’
    • Spread of experience from < 1 year to > 10 years
    • Majority ‘somewhat familiar or very familiar’ with American-European Consensus Conference definition
    • Slightly less familiarity with Berlin definition


  • Educational Intervention
    • An educational video created by an experienced educator with expert multimedia skills
    • Participants undertook 11 question test after educational video


  • No Educational Intervention
    • Participants undertook 11 question test before being shown same educational video

Common to Both Groups

  • The test required a decision for 11 CXRs, which have previously been ratified and published by the Berlin Definition group
  • Participants were asked “Does this chest radiograph have bilateral opacities consistent with pulmonary oedema (may be very mild, patchy and asymmetric) that are not fully explained by pleural effusions, pulmonary nodules or masses, or lobar/lung collapse (i.e., radiographically consistent with the diagnosis or ARDS)?”
  • Answers could be
    • Consistent
    • Equivocal
    • Inconsistent


  • Primary outcome: There was no difference in the mean proportion of correct scores between the groups
    • Intervention: 58%
    • Control: 56%
    • Mean difference 2%
    • P-value: 0.15
      • By Student T test comparing two means
      • Insufficient data in article to confirm this
  • Secondary outcome:
    • Sensitivity analysis where by the “equivocal” answers were considered with “consistent” answers resulted in no change to the outcome (70% vs 69% correct; p = 0.72)
    • Sensitivity analysis where by the “equivocal” answers were considered with “inconsistent” answers changed the significance of the outcome (73% vs 66% correct; p < 0.001)
    • Subgroup analyses did not identify any subgroups-intervention interactions

Authors’ Conclusions

  • This online educational intervention did not improve the accuracy in the identification of ARDS according to radiographic criteria


  • Important research question set within an important epidemiological study
    • Clinicians must be able to correctly diagnoses ARDS if we are to know the prevalence
  • Appropriate RCT methodology for hypothesis
  • Good concealment of allocation
  • Good blinding of investigators
  • Appropriate statistical tests and sensitivity analysis


  • Participants were recruited from active research community, so may not be representative of broader clinical community
  • Randomisation took place before consent was obtained, causing some imbalance in the numbers between the groups
  • No power calculation was conducted
    • This was not possible as it was a substudy and the sample population was already defined
  • The means and standard deviations for the primary outcome have not been published, so independent verification of statistical analysis cannot be performed
  • The 11 test CXRs had been previously published and may have been known to the participants, but this would have biased towards a more accurate outcome in both groups equally
  • It is possible that the test was insufficiently large to draw out any small benefit from the intervention
  • It is possible that the study population was too small to demonstrate any small benefit from the intervention (under-powered)

The Bottom Line

  • The ability of clinicians from various backgrounds and years of experience to correctly identify ARDS according to CXRs is extremely poor, bringing the current ARDS definition into question
  • An educational intervention did not improve this
  • Better diagnostic tools for ARDS need to be developed, which may include biomarkers, ultrasound or CT imaging

External Links


Summary author: Duncan Chambler
Summary date: 21 February 2018
Peer-review editor: Charlotte Summers

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.