Radiographic Score for ARDS Classification

Evidence

This was a post hoc analysis of the Validating Acute Lung Injury Biomarkers for Diagnosis study, a single-center, multi-unit prospective observational cohort conducted in the medical, cardiac, surgical, and trauma intensive care units at Vanderbilt University Medical Center in Nashville, Tennessee. The analysis evaluated diagnostic performance of the Radiographic Assessment of Lung Edema score for acute respiratory distress syndrome and assessed agreement between raters. The parent cohort had been designed to study biomarker panels for diagnosis of acute respiratory distress syndrome among patients at risk during critical illness; this analysis focused on radiographic scoring rather than a treatment or intervention. [pmid:42902574]

The evaluated population comprised critically ill adults who received invasive ventilation for at least two consecutive days and had chest radiographs available from the first two intensive care unit days. Patients with trauma or cardiogenic edema were excluded under the additional filters described for this analysis. Of 236 patients, 92 had acute respiratory distress syndrome adjudicated during those first two days using the Berlin definition. Thus, the comparison was between patients classified with acute respiratory distress syndrome and those without it within a selected, invasively ventilated intensive care unit cohort, not a general emergency, ward, or outpatient population. [pmid:42902574]

At least two experienced raters independently scored chest radiographs from intensive care unit days 1 and 2. They were blinded to clinical data, acute respiratory distress syndrome status, the ratio of arterial oxygen partial pressure to fractional inspired oxygen, predisposing risk factors, outcomes, and group allocation; radiographs from patients with and without acute respiratory distress syndrome were intermixed. Individual rater scores were averaged for each radiograph, after which the radiograph with the highest average score across the two study days was selected for each patient. Independent physicians adjudicated acute respiratory distress syndrome without reference to the Radiographic Assessment of Lung Edema score. [pmid:42902574]

The Radiographic Assessment of Lung Edema score had an area under the receiver operating characteristic curve of 0.93 (95% confidence interval, 0.90–0.96) for acute respiratory distress syndrome in this cohort. At the reported cutoff of 20, sensitivity was 85% (76–91%), specificity was 83% (76–89%), positive predictive value was 76% (67–84%), and negative predictive value was 90% (83–94%). Pairwise intraclass correlation coefficients ranged from 0.81 to 0.87, the reported measure of interobserver reliability. These are diagnostic-performance and rater-agreement results from the study’s scoring and adjudication procedures, rather than measures of patient outcomes. [pmid:42902574]

Analysis — Diagnostic classification pattern

The observed pattern is consistent with a radiographic score that separated independently adjudicated acute respiratory distress syndrome from non-acute respiratory distress syndrome classifications in this particular cohort. Blinding of image raters to clinical information and acute respiratory distress syndrome status reduces a direct route by which knowledge of the reference classification could influence scoring. Likewise, adjudicating acute respiratory distress syndrome without reference to the score separates the reference process from the test under evaluation. Averaging multiple raters’ assessments and selecting the highest averaged score from either of two early intensive care unit days were integral parts of the evaluated method; the reported performance therefore applies to that composite procedure, not necessarily to one rater, one radiograph, or another timing strategy. [pmid:42902574]

The numerical results describe discrimination and classification characteristics within a population already restricted to adults requiring invasive ventilation for at least two days, with further exclusions for trauma and cardiogenic edema. Positive and negative predictive values are cohort-specific results and should be read with the study’s patient selection and acute respiratory distress syndrome frequency in view. Because this was an observational post hoc analysis, the results establish an association between the score-based classification procedure and adjudicated acute respiratory distress syndrome status; they do not establish causation. [pmid:42902574]

Limitations

The retrospective design cannot establish causation. Although the underlying Validating Acute Lung Injury Biomarkers for Diagnosis cohort was prospective, this diagnostic-score evaluation was post hoc. Its rigorous inclusion and exclusion criteria may constrain generalizability, and the additional filters narrowed the evaluation to invasively ventilated patients with acute respiratory distress syndrome risk factors while excluding trauma and cardiogenic edema. The estimates were reported cohort-wide, so they do not describe performance across separately reported intensive care unit subgroups or other clinical populations. [pmid:42902574]

The selection of the highest averaged score from the first two intensive care unit days means the tested approach incorporates information from more than one image and favors the higher of two averaged assessments. The study also relied on chest radiographs that raters considered usable. These design choices define the studied method and comparator, while limiting direct extrapolation to other reader configurations, image-selection approaches, or care settings. The authors identified multicenter validation and evaluation across varied intensive care unit populations as needed to examine robustness, reproducibility, and generalizability. [pmid:42902574]