ARDS criteria are a set of clinical rules used to diagnose acute respiratory distress syndrome, a life-threatening form of lung failure in which fluid leaks into the air sacs and makes oxygen exchange collapse. The current standard, known as the Berlin Definition, was published in 2012 and grades severity into mild, moderate, and severe based primarily on how poorly the lungs exchange oxygen. A 2023 global update expanded those criteria to account for settings where blood gas analyzers and mechanical ventilators are not available, but the core logic remains the same: acute onset, bilateral lung opacities on imaging, low oxygenation that is not fully explained by heart failure, and a known trigger.
What the Berlin Definition Actually Requires
The Berlin Definition replaced an older set of criteria from a 1994 American-European Consensus Conference, which had been in clinical use for nearly two decades but suffered from vague wording and weak ability to predict who would die. The Berlin Definition tightened things up into four pillars that must all be present for a diagnosis:
- Timing: Symptoms must appear within one week of a known clinical trigger or new or worsening respiratory complaints.
- Imaging: Chest X-ray or CT scan must show bilateral opacities (whiteness in both lungs) that are not fully explained by fluid overload, collapsed lung segments, or nodules.
- Origin of edema: The respiratory failure must not be fully explained by cardiac failure or fluid overload. If no obvious risk factor for ARDS exists, an echocardiogram or other objective assessment is needed to rule out a heart-driven cause.
- Oxygenation: The ratio of arterial oxygen pressure to the fraction of inspired oxygen (the PaOâ‚‚/FiOâ‚‚ ratio) must be 300 mm Hg or lower while the patient is receiving a minimum of 5 cmHâ‚‚O of positive end-expiratory pressure (PEEP).
That oxygenation ratio is the number that determines severity. Mild ARDS falls between 200 and 300, moderate between 100 and 200, and severe at 100 or below. In the cohort used to validate the Berlin Definition, mortality climbed with each step: roughly 27% for mild, 32% for moderate, and 45% for severe. Survivors with more severe disease also spent longer on mechanical ventilation.
1JAMA. Acute Respiratory Distress Syndrome: The Berlin DefinitionHow Severity Grading Steers Treatment
The severity tiers are not just bookkeeping. They directly inform how aggressively a clinical team manages a patient’s ventilation. In mild ARDS, standard lung-protective ventilation with low tidal volumes and moderate PEEP is usually sufficient. As oxygenation worsens into moderate territory, clinicians start considering deeper sedation, higher PEEP, and neuromuscular blockade. For severe ARDS, prone positioning becomes a high-priority intervention. Turning a patient face-down for at least 16 hours a day redistributes lung ventilation and has been shown to offer a survival advantage of roughly 10 to 17 percentage points in patients whose PaOâ‚‚/FiOâ‚‚ ratio is below 100.
2American Journal of Respiratory and Critical Care Medicine. Prone Position in Acute Respiratory Distress Syndrome. Rationale, Indications, and LimitsStepped protocols used in ICUs explicitly tie escalation to PaOâ‚‚/FiOâ‚‚ thresholds. One recent stepped-wedge trial protocol, for example, calls for neuromuscular blockade and prone positioning when the ratio falls to 150 or below and the patient is already on at least 60% inspired oxygen.
3PubMed Central. Protocol and statistical analysis plan for the identification and treatment of hypoxemic respiratory failure and acute respiratory distress syndrome with protection, paralysis, and proningA single study of 153 ARDS patients reported 30-day mortality of 10% in mild cases, 23% in moderate, and 55% in severe, reinforcing the idea that the Berlin tiers capture meaningfully different levels of risk.
4PubMed Central. Prognostic values of the Berlin definition criteria, blood lactate level, and fibroproliferative changes on high-resolution computed tomography in ARDS patientsThe 2023 Global Definition Update
A major limitation of the Berlin Definition was that it assumed the patient was intubated on a mechanical ventilator. During the COVID-19 pandemic, huge numbers of patients with obvious ARDS were managed on high-flow nasal oxygen (HFNO) and never intubated, technically falling outside the Berlin criteria’s requirement for a minimum PEEP setting. A 2023 international consensus panel addressed this and several other gaps with four main changes:
- High-flow nasal oxygen included: Patients on HFNO with a flow rate of at least 30 liters per minute can now meet the oxygenation criterion.
- Pulse oximetry allowed: Where arterial blood gas measurement is unavailable, the SpOâ‚‚/FiOâ‚‚ ratio (using a pulse oximeter reading instead of an arterial sample) can substitute, with a cutoff of 315 or below when SpOâ‚‚ is 97% or less.
- Lung ultrasound accepted: Bilateral opacities can now be identified by lung ultrasound, not just chest X-ray or CT, which is especially valuable in resource-limited hospitals.
- Relaxed device requirements in low-resource settings: In places where PEEP delivery or precise oxygen flow measurement is not possible, those requirements are waived entirely.
These changes were partly anticipated by the Kigali modification, developed from work at a Rwandan referral hospital. That earlier proposal had already dropped the PEEP requirement, substituted pulse oximetry for blood gases, and allowed lung ultrasound, demonstrating that a modified definition could identify ARDS patients who would otherwise go undiagnosed in settings without ventilators or arterial blood gas analyzers.
6PubMed. Diagnosing acute respiratory distress syndrome in resource limited settings: the Kigali modification of the Berlin definitionA multicenter study of COVID-19 patients treated with HFNO found that the Berlin Definition’s predictive ability for 28-day mortality was similar whether patients were on HFNO or mechanical ventilation, lending support to the expanded definition.
7PubMed. The Predictive Validity of the Berlin Definition of Acute Respiratory Distress Syndrome for Patients With COVID-19-Related Respiratory Failure Treated With High-Flow Nasal OxygenWhen the Oxygenation Number Lies
One underappreciated problem with ARDS criteria is that the PaOâ‚‚/FiOâ‚‚ ratio is not stable across different types of respiratory support. Moving a patient from high-flow nasal oxygen to a mechanical ventilator can cause the ratio to jump by more than 100 mm Hg, which would shift the patient from what looked like moderate or severe ARDS on HFNO to apparently mild ARDS once intubated. One study of critically ill COVID-19 patients found that the median increase in PaOâ‚‚/FiOâ‚‚ after switching from HFNO to mechanical ventilation was 127 mm Hg, making it impossible to predict severity during ventilation from the numbers observed during HFNO.
8PubMed Central. Limitations of the ARDS criteria during high-flow oxygen or non-invasive ventilation: evidence from critically ill COVID-19 patientsPEEP itself also muddies the picture. Higher PEEP recruits collapsed lung units and can improve the oxygenation ratio even though the underlying lung injury has not changed. Researchers have proposed incorporating PEEP directly into the ratio (a metric called the P/FP ratio), which better predicts mortality, especially at high PEEP levels. Using that adjusted ratio reclassified a meaningful fraction of patients: about 15% of those categorized as mild and 13% of moderate ARDS patients shifted to more severe categories, while a third of moderate and 14% of severe patients moved to milder ones.
9PubMed Central. P/FP ratio: incorporation of PEEP into the PaO2/FiO2 ratio for prognostication and classification of acute respiratory distress syndromeIn other words, two patients with identical lung injuries can receive different severity labels depending on what device is delivering their oxygen and how much pressure is applied. This is a real clinical problem, not just an academic one, because treatment escalation decisions are tied to those thresholds.
The Gap Between the Clinical Diagnosis and What Is Happening in the Lungs
ARDS is a clinical syndrome, not a pathological diagnosis. The hallmark tissue finding in ARDS is called diffuse alveolar damage (DAD), a pattern of widespread destruction of the air sac lining visible under a microscope. But autopsy studies consistently show that a substantial number of patients who meet clinical ARDS criteria do not have DAD at all, and some patients who have DAD were never diagnosed with ARDS during life.
In one autopsy series, only half of patients clinically diagnosed with ARDS actually showed DAD on postmortem examination; a quarter had pneumonia without DAD, and an eighth had invasive fungal infection instead.
10PubMed. ARDS: a clinicopathological confrontation Another study comparing clinical ARDS criteria with autopsy findings in pneumonia patients found a 37% discrepancy rate — meaning over a third of patients diagnosed with ARDS based on clinical criteria did not have the expected tissue damage.
11PubMed. Discrepancy between clinical criteria for diagnosing acute respiratory distress syndrome secondary to community acquired pneumonia with autopsy findings of diffuse alveolar damageA larger autopsy study of 382 patients found that the clinical criteria had a sensitivity of about 75% and specificity of 84% for detecting DAD.
12PubMed. Comparison of clinical criteria for the acute respiratory distress syndrome with autopsy findings Patients who did have DAD tended to be more hypoxic and had lower lung compliance, and they were more likely to die from refractory low oxygen rather than from shock.
13PubMed. Acute respiratory distress syndrome in patients with and without diffuse alveolar damage: an autopsy studyThis matters because if a third or more of clinically diagnosed ARDS patients actually have a different underlying problem (unilateral pneumonia, heart-related fluid, pulmonary embolism), then treatment strategies calibrated for DAD-type ARDS may not be optimal for them. It is one of the key motivations behind the push for biological subtyping.
Biological Subphenotypes
Even among patients who genuinely have ARDS, the syndrome is not one disease. Researchers analyzing clinical and blood biomarker data across multiple large trial cohorts have identified at least two distinct subphenotypes, labeled hyperinflammatory and hypoinflammatory. These are not visible on imaging or through standard bedside measurements. They emerge from patterns in inflammatory markers, clotting function, and metabolic disturbance.
14PubMed Central. Phenotypes in acute respiratory distress syndrome: moving towards precision medicineThe hyperinflammatory group makes up a minority of ARDS patients — roughly 18% in one recent cohort — but has dramatically worse outcomes. In that study, 60-day mortality was about 51% in the hyperinflammatory group compared with 28% in the hypoinflammatory group, and after adjusting for other variables, hyperinflammatory patients had nearly three times the odds of dying.
15PubMed. Bedside identification of subphenotypes in acute respiratory failure (PHIND): a multicentre, observational cohort studyWhat makes this clinically significant, rather than just scientifically interesting, is that the two groups seem to respond differently to therapies. Secondary analyses of older randomized trials have shown divergent responses to fluid management strategy, PEEP levels, and statin therapy depending on which subphenotype the patient belongs to. A treatment that helps one group may be neutral or harmful in the other. The challenge has been identifying subphenotype at the bedside quickly enough to guide real-time decisions. Recent work on near-patient assays using a small panel of blood biomarkers is a step toward making that feasible.
Efforts to further refine subgrouping have also used broader biomarker panels. One study using latent class modeling identified three distinct ARDS subgroups, with the top discriminating features being inflammatory cytokines (IL-8, IL-6, IL-10, TNF-alpha) and serum lactate. Mortality varied across all three groups.
16PubMed Central. Endotyping in ARDS: one step forward in precision medicineConditions That Look Like ARDS but Are Not
Because the ARDS criteria are based on clinical observations rather than a definitive biological test, several conditions can produce the same picture of bilateral lung opacities and poor oxygenation. Cardiogenic pulmonary edema is the classic mimic: fluid backs up into the lungs from a failing heart, flooding the air sacs and causing bilateral whiteness on a chest X-ray that looks nearly identical to ARDS. The Berlin Definition tries to exclude this by requiring that cardiac failure not fully explain the picture, but in practice the distinction is not always clean, especially in patients who have both heart disease and a lung injury trigger simultaneously.
17PubMed Central. Decision support tool for differential diagnosis of Acute Respiratory Distress Syndrome (ARDS) vs Cardiogenic Pulmonary Edema (CPE): a prospective validation and meta-analysisIn a study that screened over 1,200 patients who appeared to meet Berlin criteria for severe ARDS based on oxygenation, fewer than a quarter actually had true severe ARDS on manual review. The rest had unilateral pneumonia, cardiogenic edema, atelectasis, chronic lung disease, pleural effusion, pulmonary embolism, or other conditions that can mimic the ARDS picture.
18Intelligence-Based Medicine. Discriminating Acute Respiratory Distress Syndrome from other forms of respiratory failure via iterative machine learningThat enormous false-positive rate partly reflects the difficulty of interpreting bilateral opacities on imaging, which is subjective and has moderate inter-reader agreement even among experienced radiologists. It also reflects the fact that many critically ill patients have overlapping problems — a septic patient can have both ARDS and some degree of fluid overload at the same time.
Imaging Beyond the Chest X-Ray
The traditional chest X-ray has been the standard imaging tool in the ARDS definition since the beginning, but it has well-known limitations. Portable ICU films are often poor quality, taken at suboptimal angles, and difficult to interpret reliably. The 2023 global update formally added lung ultrasound as an acceptable alternative, which raised questions about how the two modalities compare.
A meta-analysis of lung ultrasound for ARDS diagnosis found high specificity (about 94%) but modest sensitivity (around 63%), meaning ultrasound is good at confirming ARDS when present but misses some cases.
19PubMed Central. Diagnostic accuracy of lung ultrasound in diagnosis of ARDS and identification of focal or non-focal ARDS subphenotypes A retrospective study comparing the two approaches found that both were equally related to mortality when used within the Berlin framework, and that ultrasound could identify high-risk patients who did not meet the X-ray-based definition. The authors concluded the two should be complementary rather than interchangeable.
20PubMed Central. Chest radiography versus lung ultrasound for identification of acute respiratory distress syndromeFor hospitals in low-income settings that lack CT scanners and struggle with portable X-ray film quality, ultrasound is a practical improvement. It is portable, repeatable, radiation-free, and can be done at the bedside by the treating clinician in minutes.
ARDS Criteria in Children
The Berlin Definition was developed and validated in adults. Children have different respiratory physiology, and the causes of acute lung injury in pediatrics skew toward different conditions (bronchiolitis, aspiration, drowning). The Pediatric Acute Lung Injury Consensus Conference (PALICC) developed separate criteria, and a 2023 update (PALICC-2) recommends that all patients under 18 years old without active perinatal lung disease be diagnosed using PALICC-2 rather than adult ARDS criteria.
21PubMed Central. Executive Summary of the Second International Guidelines for the Diagnosis and Management of Pediatric Acute Respiratory Distress Syndrome (PALICC-2)Key differences include the use of the oxygenation index (which accounts for mean airway pressure, not just PEEP and FiOâ‚‚), the acceptance of unilateral or bilateral opacities on imaging, and separate severity categories calibrated to the pediatric population. Newborns with certain types of lung disease are excluded because their respiratory failure has a fundamentally different mechanism.
Automated Screening With Machine Learning
ARDS is notoriously underrecognized. Large observational studies have documented that clinicians fail to diagnose ARDS in a significant fraction of patients who meet criteria, and even when they do, the diagnosis often comes late. Part of the problem is that reviewing imaging, calculating the oxygenation ratio, and excluding cardiac causes takes time and attention in a chaotic ICU environment.
One line of research has explored whether machine learning algorithms can screen for ARDS automatically using data already being captured by the ventilator. A random forest model trained on raw ventilator waveform data achieved an area under the curve of about 0.88 to 0.89, with sensitivity around 90% and specificity around 71 to 75%. That model could flag likely ARDS cases using just six hours of ventilator data, potentially alerting clinicians before they would have recognized the syndrome themselves.
22Critical Care Explorations. Use of Machine Learning to Screen for Acute Respiratory Distress Syndrome Using Raw Ventilator Waveform DataThese tools are still in early development and none are part of standard care. But the appeal is clear: if an algorithm running in the background can flag patients who are silently meeting ARDS criteria, it could shorten the gap between onset and recognition, which is when protective ventilation and prone positioning do the most good.
What Happens After Survival
Once someone survives ARDS and leaves the ICU, the story does not end cleanly. A study following 109 survivors over 12 months found that standard lung volume measurements and airflow tests returned to normal by about six months, but the lungs’ ability to transfer carbon monoxide (a marker of how well oxygen moves from the air sacs into the blood) remained low throughout the entire year. No patients needed supplemental oxygen at the one-year mark, though about 6% still dropped below normal oxygen levels during exercise.
23New England Journal of Medicine. One-year outcomes in survivors of the acute respiratory distress syndromeFunctional recovery was slow. The distance survivors could walk in six minutes improved from about 281 meters at three months to 422 meters at 12 months, but these values still fell below what would be expected for their age and sex. Much of the lasting disability in ARDS survivors comes from muscle wasting, nerve damage, and psychological sequelae (anxiety, depression, post-traumatic stress) rather than from the lungs themselves. For many patients, the lung injury resolves but the body never fully bounces back from the weeks of critical illness it took to get there.