How can we tell if a patient is hypoxemic before giving oxygen?

Pulse oximetry is the standard bedside tool, but it has real limits that the evidence makes clear.

The most direct way to tell if a patient is hypoxemic before giving oxygen is a pulse oximeter reading (SpO2). The evidence shows it correlates well with arterial blood gas measurements in many settings, with correlations around r=0.9 or higher [1, 2, 3]. In critically ill ICU patients, the mean difference between SpO2 and the true arterial value (SaO2) was only 0.02%, with a standard deviation of 2.1% [3]. That means in about 95 out of 100 readings, the oximeter is within roughly 4% of the real number.

But the limits matter. The accuracy drops when perfusion is poor. In patients with low peripheral temperatures, the fraction of low-quality signals rises sharply: at skin temperatures of 26-30°C, 38% of readings were low quality, and below 22°C, 100% were [2]. Vasoactive drugs also reduce reliability [3]. Skin pigment affects readings too: in one study of children under anesthesia, non-Caucasian children had a significantly larger average drop in measured saturation (9.0% vs 5.8%) [4], suggesting the oximeter may read differently on darker skin.

The clinical signs are far less reliable. In one study of adults in the emergency department, abnormal vital signs had 97% sensitivity for pneumonia but only 19% specificity [6] — meaning almost everyone with pneumonia had abnormal vitals, but so did most people without it. Altered breathing sounds on auscultation had 78% sensitivity and 38% specificity [6]. In infants under six months, rapid breathing (rate ≥50 per minute) had 77-86% sensitivity for pneumonia but only 38% specificity [5]. Chest indrawing was more specific at 80% but only caught about 45% of cases [5].

So the practical answer: use the oximeter, but know when to distrust it. If the signal is good and the patient has warm, well-perfused fingers, the reading is reliable. If the patient is cold, on vasopressors, or has dark skin, the margin of error widens. In those cases, an arterial blood gas is the gold standard, but it is invasive and not always available. The clinical exam alone — watching for retractions, nasal flaring, cyanosis, altered mental status — catches the worst cases but misses many.

My call: pulse oximetry is the best non-invasive tool, but its accuracy degrades with poor perfusion, cold, and darker skin, and clinical signs alone are too insensitive to rule out hypoxemia. Confidence: high.