Respiratory Failure: The Three Prompts for ICU Admission
Admission to the ICU for respiratory insufficiency is driven by one of three conditions: acute hypoxemic respiratory failure, acute hypercapnic (ventilatory) failure, or upper airway impairment. Each has a distinct mechanism, and each points toward its own first-line therapy.
Hypoxemic respiratory failure is caused by shunt — perfusion of lung units that are not ventilated — typically because alveoli are collapsed or flooded with cells, fluid, or blood, as in pneumonia, pulmonary edema, alveolar hemorrhage, or ARDS. The two levers of therapy are raising the inspired oxygen concentration (FiO2) and applying positive end-expiratory pressure (PEEP) to reopen collapsed or flooded alveoli and keep them open through exhalation.
Hypercapnic respiratory failure reflects inadequate alveolar ventilation, either from increased dead space (ventilation of poorly perfused lung units) or from pump failure. Moving air into the alveoli requires the respiratory muscles to generate enough pressure to overcome both elastic loads (the lung and chest wall) and resistive loads (the airways). Ventilatory failure results when those loads become too great or the muscles become too weak — severe COPD or asthma exacerbations, loss of central drive from stroke or opiate overdose, or neuromuscular weakness such as Guillain-Barré syndrome. Mechanical ventilation, invasive or noninvasive, is the mainstay of treatment.
Upper airway impairment usually means either obstruction (e.g., angioedema) or inability to protect the airway (e.g., opiate intoxication). Management is securing a patent airway — endotracheal tube, tracheostomy, or cricothyrotomy.
Oxygen Therapy: Titrate to a Target — and Resist the Urge to Go Higher
Several devices deliver supplemental oxygen, each with different characteristics that suit different patients, and all are titrated against a target arterial oxygen saturation (SaO2) or pulse oximetry reading (SpO2). The critical principle is that higher saturation is not always desirable. Evidence indicates that supplementing oxygen to patients whose saturation is already 96% or higher may actually increase mortality; accordingly, patients receiving oxygen therapy should generally be kept at an SpO2 of 96% or lower, and oxygen should not be started at all in acute myocardial infarction or stroke when the SpO2 is 93% or higher. There is no clinical benefit to a supraphysiologic PaO2 (>100 mm Hg), so sustained hyperoxia should be avoided and targets individualized.
High-Flow Nasal Cannula
High-flow nasal cannula (HFNC) mixes and humidifies air and oxygen at high flows (≥30 L/min) to deliver a precise, consistent FiO2 (0.21–1.0) through a nasal cannula. The high flow generates a small amount of positive airway pressure, although the amount varies with flow rate and cannot be quantified. Physiologically, HFNC can reduce respiratory rate, work of breathing, and dead space, and it is better tolerated than many alternatives.
The evidence base continues to mature. Systematic reviews have found no difference in hypoxemic events when HFNC is compared with other oxygenation methods before and during intubation. In acute hypoxemic respiratory failure, HFNC may reduce the need for tracheal intubation, though it has no demonstrated mortality effect; current guidelines nonetheless recommend it because avoiding intubation is important to patients. In the post-extubation period, high-flow oxygen may reduce recurrent respiratory failure and reintubation — but in high-risk patients (older than 65 years or with chronic heart or lung disease), high flow alone was less effective than high flow combined with noninvasive ventilation immediately after extubation. HFNC should be started in a monitored setting (emergency department or ICU) with close observation for tolerance and effectiveness, and in acute hypoxemic respiratory failure the validated ROX index — (SpO2/FiO2) divided by respiratory rate — helps identify patients at low or high risk of needing intubation.
Noninvasive Positive Pressure Ventilation
Mechanical ventilatory support can be delivered noninvasively — by helmet, or nasal, oral, or full-face mask — or invasively through an endotracheal tube or tracheostomy. The choice depends on the clinical situation and the goals of treatment, and any form of mechanical ventilatory support should be initiated in a monitored unit. Noninvasive positive pressure ventilation (NPPV) comes in two basic types: continuous positive airway pressure (CPAP), which delivers a constant airway pressure during both inspiration and expiration, and bilevel positive airway pressure (BPAP), which delivers a higher pressure during inspiration (IPAP) than expiration (EPAP), adding inspiratory support and augmenting tidal volume. Support is titrated to gas exchange and to the patient's respiratory effort.
| Table 1. Modes of Noninvasive Positive Pressure Ventilation | ||
|---|---|---|
| Mode | Function and Physiologic Effects | Typical Indications |
| CPAP | Constant airway pressure throughout the respiratory cycle; maintains airway patency, increases functional residual capacity and mean airway pressure | Obstructive sleep apnea; pulmonary edema; excessive dynamic airway collapse; preintubation and postextubation support |
| BPAP | Two pressure levels (IPAP/EPAP); CPAP effects plus decreased work of breathing and augmented tidal volume | COPD exacerbation; obesity hypoventilation syndrome; neuromuscular disease; time-limited trial in selected “do not intubate” patients with clear goals of care |
| BPAP with S/T mode | BPAP plus a minimum set respiratory rate; continues delivering breaths if apnea occurs | Hypoventilation; central apneas |
Patient selection matters. Evidence favors NPPV for COPD exacerbations, cardiogenic pulmonary edema, neuromuscular disease, obesity hypoventilation syndrome, and for patients at high risk of extubation failure (e.g., older than 65 years or with heart failure or COPD). Benefit is greatest when the underlying process reverses quickly: in COPD, NPPV reduces intubation and shortens hospital stay most clearly for moderate respiratory acidosis (pH 7.25–7.35), while severe acidosis (pH <7.2) generally requires intubation. NPPV is not benign, however — it can increase mortality in patients who fail it and ultimately require intubation, and harm has been associated with delayed intubation and with tidal volumes that are too high. Every patient on NPPV must be monitored and reevaluated within 2 hours to judge effectiveness and the need for adjustment. Contraindications include persistent altered mental status, copious airway secretions, emesis, gastric distention, airway obstruction, recent esophageal surgery, cardiac arrest, inability to protect the airway, facial trauma or surgery (oral, nasal, or sinus), significant hemodynamic instability, and mask intolerance.
Application is a skill. Proper mask sizing and patient adaptation require time and coaching. Timing is crucial: late application, when respiratory failure is already impending, is associated with subsequent intubation and worse outcomes. After initiation, monitor tolerance and adverse effects (comfort, skin integrity, gastric distention, eye irritation), ventilatory parameters (tidal volume, respiratory rate), and clinical improvement (pH, respiratory rate, oxygen saturation, mental status).
Invasive Mechanical Ventilation
Invasive mechanical ventilation delivers positive-pressure ventilation through an endotracheal tube or tracheostomy. Indications are hypoxemic or hypercapnic respiratory failure, contraindication to NPPV, and inability to protect the airway. Timing, mode, and settings depend on disease and patient factors; the mode is chosen so that its characteristics favor the prevailing clinical goal.
| Table 2. Most Frequently Used Invasive Ventilator Modes | |||
|---|---|---|---|
| Mode | What the Ventilator Controls | Key Characteristics | Goal Served |
| Volume control (VC-CMV / assist-control) | Flow and volume of each mandatory breath; set tidal volume is delivered regardless of patient effort, compliance, or resistance | Guarantees minimum minute ventilation; tidal volume can be capped for lung protection; fixed flow can make it the least comfortable mode | Safety |
| Pressure control (PC-CMV) | Inspiratory pressure (and inspiratory time); pressure stays at the set level regardless of effort or mechanics | Guarantees pressures stay below a ceiling, but delivered tidal volume varies with effort, compliance, and resistance — tidal volume and minute ventilation must be monitored | Safety (if tidal volume stays in range) |
| Pressure support (PC-CSV) | Inspiratory pressure only; the patient triggers and cycles every breath | Patient determines breath size and timing, allowing synchrony and comfort; can be used to assess extubation readiness (spontaneous breathing trials) | Comfort |
| IMV / SIMV (PC- or VC-IMV) | Mandatory breaths (volume- or pressure-controlled) at a minimum set rate, with spontaneous breaths permitted between them (usually pressure-supported) | Mandatory breaths are synchronized to patient effort near the scheduled time; guarantees a minimum minute ventilation while allowing spontaneous breathing | Safety with spontaneous breathing — but not for weaning |
When safety is the priority, settings focus on lung-protective ventilation: adequate minute ventilation and oxygenation while maintaining the lowest effective tidal volume and plateau pressure, thereby preventing ventilator-induced lung injury. In practice this means tidal volumes around 6 mL/kg of ideal body weight and keeping the end-inspiratory plateau pressure at or below 30 cm H2O, with PEEP set high enough to prevent end-expiratory alveolar collapse. When patients are awake and the risk of lung injury has fallen, the prevailing goal becomes comfort: choose a mode that allows synchrony and adequate support, taking into account respiratory muscle weakness or fatigue, acidosis, and the level of sedation.
Ventilator-induced lung injury can occur early or late and results from alveolar overdistension (“volutrauma”) or repetitive alveolar collapse with each expiration (“atelectrauma”). High positive intrathoracic pressures can also cause barotrauma (pneumothorax, pneumomediastinum) and hemodynamic compromise by reducing venous return and increasing right ventricular afterload. One subtle hazard of pressure-control modes: a spontaneously breathing patient's own inspiratory effort adds negative pleural pressure that increases the true transalveolar pressure even though the ventilator displays only its set positive pressure — so watch the delivered tidal volumes, not just the set pressures.
Liberation from the Ventilator: Daily Spontaneous Breathing Trials
Patients should be liberated from mechanical ventilation as soon as possible. Pairing daily awakening (or targeted light sedation) with daily spontaneous breathing trials (SBTs) reduces both mechanical ventilation time and mortality. Think of liberation as an active daily decision rather than a gradual taper: approaching it as a slow “wean” can extend unnecessary ventilation substantially, whereas protocolized daily readiness screening followed by SBT cuts ventilator days and ICU length of stay.
| Table 3. Common Criteria for Spontaneous Breathing Trials and Extubation | |
|---|---|
| Criteria to perform an SBT | Cause of respiratory failure improved; FiO2 ≤40% and PEEP ≤5–8 cm H2O; pH >7.25; hemodynamic stability; able to breathe spontaneously |
| SBT technique | 30 minutes to 2 hours on low pressure support (≤8 cm H2O) or a T-piece |
| Criteria to pass — at least 30 minutes without: | Clinical evidence of respiratory distress; SpO2 <90%; respiratory rate >35/min; new arrhythmias; tachycardia; hypotension or hypertension |
| Additional considerations before extubation | Quantity of secretions (need for frequent suctioning); adequacy of cough; mental status |
Before extubation, also consider airway patency: patients with risk factors for post-extubation stridor (intubation >7 days, traumatic intubation, large tube, prior stridor, head/neck surgery) should have a cuff-leak test — deflate the cuff and assess for air passing around the tube. An absent cuff leak implies a substantial (~30%) risk of post-extubation stridor, and extubation should be delayed until the cause is treated. When patients fail SBTs, hunt for perpetuating processes: critical illness myopathy/polyneuropathy, myocardial ischemia, heart failure, volume overload, delirium, malnutrition, and electrolyte abnormalities (hypophosphatemia, hypokalemia, hypomagnesemia).
Hemodynamic Support: Monitoring and Vascular Access
ICU patients sometimes require invasive hemodynamic monitoring, but these devices carry real risks and should be used only when sufficient information cannot be obtained noninvasively — and removed as soon as possible. Routine use of pulmonary artery catheters does not improve outcomes, is associated with increased complications, and may increase mortality; they remain reasonable to consider in selected scenarios such as mixed or undifferentiated shock. Assessment of tissue perfusion starts with the physical examination (skin temperature, mottling, capillary refill, jugular venous distention, urine output, mental status), vital signs, and focused cardiac evaluation with tools such as echocardiography and cardiac output monitors — applied in the right clinical scenario to minimize erroneous results. Often, examination and vital signs alone are sufficient.
Vascular access choices depend on urgency, expected duration, and purpose. Peripheral venous access with a short, wide-bore catheter is the route of choice for rapid volume resuscitation. Intraosseous devices provide emergent access when IV access cannot be obtained, and should be removed within 24 hours of placement. All IV access should be removed as soon as possible to reduce infection and thrombosis risk, and lower-risk options (e.g., peripheral IVs) should be prioritized whenever possible; ultrasound-guided placement minimizes procedural complications such as arterial or nerve injury, bleeding, and pneumothorax.
| Table 4. Types of Central Venous Access | ||||
|---|---|---|---|---|
| Type | Indications | Duration | Potential Complications | Contraindications |
| Peripherally inserted central catheter (PICC) | Caustic medications (vasoactive agents, sedatives, antibiotics); central access | Days to 1 year | Low pneumothorax risk; lower infection risk than nontunneled catheters; clot/occlusion risk from small vessel diameter | Current or pending dialysis |
| Temporary nontunneled | Same as PICC; short-term dialysis; central venous pressure monitoring | Short-term | Infection risk increases after 7 days; site-specific complications (e.g., pneumothorax with subclavian or low internal jugular approach) | — |
| Long-term tunneled | Long-term TPN, chemotherapy, antibiotics, dialysis; Dacron cuff promotes tissue adherence to reduce infection and dislodgement | >6 weeks | Infection; same placement risks as nontunneled catheters | — |
| Totally implanted port | Long-term intermittent access (e.g., chemotherapy) | >6 weeks | Lowest infection risk, but harder and costlier to implant; occult extravasation beneath skin | — |
| Intraosseous (tibia; humeral head in adults) | Emergency access when IV access unattainable | ~24 hours | Low infection risk; flow may be slower; infusion pain can be treated with slow preservative-free lidocaine | Fracture of the target bone; osteoporosis; intraosseous attempt at the same site within 24–48 hours |
Blood Pressure Support and the Approach to Shock
A mean arterial pressure (MAP) of 65 mm Hg is considered the threshold for adequate organ perfusion in most people, and targeting higher pressures (80–85 mm Hg) in septic shock has not improved mortality. Blood pressure is usually monitored noninvasively with a cuff, but an arterial line for continuous monitoring is more useful when systolic pressure falls below 90 mm Hg, when frequent measurements are needed (especially with continuous IV vasoactive infusions), or when cuff readings are unreliable — morbid obesity, extremity vascular anomalies, or other anatomic features limiting cuff accuracy. Management of hemodynamic perturbations should focus on the underlying cause of hypotension: hypotension from decreased preload is initially treated with volume resuscitation, and persistent hypotension may require vasopressors.
Shock itself is organ dysfunction from an imbalance between cellular oxygen supply and demand — most often inadequate oxygen delivery from circulatory failure. Oxygen delivery is the product of cardiac output and arterial oxygen content (DO2 = CO × CaO2), with cardiac output determined by heart rate and stroke volume, and stroke volume by preload, contractility, and afterload (SVR). Disease processes that disturb any of these variables can produce shock, which is classified into four types by the primary physiologic derangement — and each type has a characteristic hemodynamic profile. Types are not mutually exclusive: sepsis may begin as distributive shock and acquire a cardiogenic component from sepsis-induced cardiomyopathy.
| Table 5. Physiologic Classification of Shock and Hemodynamic Profiles | |||||
|---|---|---|---|---|---|
| Type | Primary Defect | CO | SVR | Filling Pressures (CVP/PCWP) | Examples |
| Distributive | Low SVR (vasodilation) | ↑ (compensatory) | ↓ | ↓/↓ | Septic shock; pancreatitis; burns; anaphylaxis; neurogenic; adrenal crisis |
| Cardiogenic | Pump failure | ↓ | ↑ | ↑ (PCWP if LV; CVP if RV) | MI; myocarditis; arrhythmia; acute valvular regurgitation |
| Hypovolemic | Low preload | ↓ | ↑ | ↓/↓ | Hemorrhage (trauma, GI); GI losses; burns; osmotic diuresis |
| Obstructive | Extracardiac flow obstruction | ↓ | ↑ | ↑ (variable PCWP) | Tension pneumothorax; tamponade; pulmonary embolism; aortic dissection |
Early recognition is essential because organ dysfunction in early shock is often reversible; untreated shock progresses from compensated preshock through decompensated shock to irreversible multisystem failure. Evaluation and treatment proceed simultaneously: focused history and examination (the brain, kidneys, and skin are the bedside “windows” on organ dysfunction — new confusion, oliguria, delayed capillary refill and mottling), lactate and routine labs, ECG and chest x-ray, and point-of-care ultrasound to classify the shock type. Fluid resuscitation should be guided by the underlying cause and by dynamic assessment of fluid responsiveness — passive leg raise, stroke-volume or pulse-pressure variation (>12% suggests volume responsiveness in appropriately ventilated patients), or echocardiographic assessment — rather than by fixed volumes, and the need for continued fluid should be frequently reassessed as the clinical picture evolves (e.g., fluid removal may become appropriate once ARDS or renal failure develops).
Vasopressors and Inotropes
When blood pressure support is needed, norepinephrine is the most commonly used agent and has been shown to reduce mortality; other agents serve specific situations. Selection rests on receptor pharmacology and the type of shock.
| Table 6. Selecting Vasopressors and Inotropes | ||||
|---|---|---|---|---|
| Agent | Shock Types | Receptors | Primary Effect | Comments |
| Norepinephrine | Distributive, cardiogenic, hypovolemic | α1, α2, β1 | ↑ SVR, ↑ inotropy | First choice in distributive, cardiogenic, and hypovolemic shock; reduces mortality |
| Epinephrine | Distributive, cardiogenic, hypovolemic | α1, α2, β1, β2 | ↑ SVR, ↑ inotropy | First choice for anaphylaxis; may be added to norepinephrine in septic shock |
| Vasopressin | Distributive, hypovolemic | V receptors | ↑ SVR | Add-on to norepinephrine in septic shock; no role in other shock states |
| Dopamine (high-dose) | Cardiogenic | D, α1, β1 | ↑ SVR, ↑ inotropy | Increased mortality vs norepinephrine in distributive shock; reserve for severe bradycardia in septic shock |
| Dopamine (low-dose) | Cardiogenic | D, β1 | ↑ inotropy, ↑ HR | Not recommended to augment renal blood flow |
| Phenylephrine | Distributive | α1 | ↑ SVR | When norepinephrine is contraindicated (tachyarrhythmias) or first-line drugs fail; may depress cardiac output via reflex bradycardia |
| Dobutamine (inotrope, not a vasopressor) | Cardiogenic; distributive | β1, β2 | ↑ inotropy | First choice for cardiogenic shock without hypotension; add-on in distributive shock with depressed cardiac function |
| Angiotensin II | Distributive | AT II | ↑ SVR | Role still being defined; concern for increased thrombotic events |
Complications of Critical Illness: Weakness and Cognitive Decline
ICU complications can be classified as early (occurring during hospitalization) or late (persisting after the critical illness). Between 25% and 100% of critically ill patients develop muscle weakness, which may arise from the nerves (critical illness polyneuropathy), the muscles themselves (critical illness myopathy), both (critical illness neuromyopathy — perhaps the most prevalent form), or be nonspecific (ICU-acquired weakness). It may also relate to prolonged neuromuscular blockade. Identified risk factors include sepsis, multisystem organ failure, severe illness, prolonged immobility, and hyperglycemia. Often the first clue is failure to wean from mechanical ventilation. Evaluation starts at the bedside with the Medical Research Council muscle scale, though electromyography remains the gold standard; ICU-acquired weakness is a diagnosis of exclusion. It generally improves over weeks to months but may persist for years.
| Table 7. Definitions and Characteristics of ICU-Acquired Weakness | |
|---|---|
| ICU-acquired weakness | Clinically detected weakness with no explanation other than the critical illness; proximal and distal symmetrical flaccid weakness sparing cranial nerves; failure to wean from the ventilator is often the first indication; diagnosis of exclusion |
| Critical illness polyneuropathy | ICU-acquired weakness with electrophysiological evidence of axonal polyneuropathy; quadriparesis/quadriplegia with decreased tone and decreased deep tendon reflexes, sparing facial muscles |
| Critical illness myopathy | ICU-acquired weakness with electrophysiological or histological evidence of myopathy; examination similar to polyneuropathy; new sensory loss is suggestive; creatine kinase may be elevated |
| Critical illness neuromyopathy | Coexistence of polyneuropathy and myopathy; mixed features; perhaps the most prevalent form |
The long-term picture extends beyond muscle. As many as 30% to 80% of critical care survivors develop long-term cognitive impairment; one year after critical illness, the level of impairment is similar in severity to mild Alzheimer disease. Specific risk factors and interventions are not well defined, but the development and duration of delirium during the ICU stay appear to be major predictors — one more reason that delirium prevention and early mobilization are core ICU therapies, not extras.
References
- MKSAP 19 — Pulmonary and Critical Care MedicineAmerican College of Physicians (2022). "Critical Care Medicine: Principles of Critical Care," pp. 61-66.
- Harrison's Principles of Internal Medicine, 22eMcGraw-Hill (2024). Ch.313 "Mechanical Ventilatory Support" (Schissel) and Ch.314 "Approach to the Patient with Shock" (Baron/Massaro).
- UK-ROX trial — conservative oxygen therapy in mechanically ventilated adultsMartin DS, Gould DW, Shahid T, et al; UK-ROX Investigators. JAMA 2025;334:398-408.
- Conservative oxygen targets in mechanically ventilated patients — updated meta-analysisNguyen N, et al. Crit Care Med 2026;54(5).
- RENOVATE trial — high-flow nasal oxygen vs noninvasive ventilation in acute respiratory failureMaia IS, Kawano-Dourado L, Tramujas L, et al; RENOVATE Investigators/BRICNet. JAMA 2025;333(10):875-890.
- Post-extubation NIV vs HFNC in very high-risk patients — prevent, don't rescueHernandez G, Paredes I, Colinas L, et al. Intensive Care Med 2022;48:1751-1759.
- Surviving Sepsis Campaign 2026 guidelines — hemodynamic support (vasopressor sequencing, MAP targets, fluid responsiveness)Prescott HC, Antonelli M, Alhazzani W, et al. Crit Care Med 2026. doi:10.1097/CCM.0000000000007075.