In Reply Drs Siempos and Berlin are concerned that the computer algorithm may have overestimated the incidence of ARDS reported in our study. We used objective data (eg, arterial partial pressure of oxygen to fraction of inspired oxygen ratio, presence of an ARDS risk factor, presence of bilateral infiltrates) that were entered into the database by the investigators in the study intensive care units. If all the criteria for ARDS (as per the Berlin Definition1) were met, these patients were coded as having ARDS—this is what was referred to as the “computer algorithm.” We did not set any additional criterion that was not part of the ARDS Berlin Definition. The electronic case report form provided additional guidance, including what was meant by “bilateral infiltrates,” and specifically that they could not be explained by effusions, lobar or lung collapse, or nodules. Moreover, all investigators were offered web-based training on ARDS recognition and chest x-ray interpretation. Although it is possible that some patients without an ARDS risk factor may have had a clinical entity other than ARDS (eg, cryptogenic organizing pneumonia, acute hypersensitivity pneumonitis, or acute eosinophilic pneumonia), this would not have been a large proportion of patients. A separate substudy is under way to examine this in more detail.
Bellani G, Pham T, Laffey J, for the LUNG-SAFE Investigators and the ESICM Trials Group. Incidence of Acute Respiratory Distress Syndrome—Reply. JAMA. 2016;316(3):347. doi:10.1001/jama.2016.6471