Automated oxygen titration kept ward patients in range 85% of the time, against 63%

Design

Multicentre, unblinded, parallel-group randomised clinical trial

Population

300 hospitalised adults on supplemental oxygen at four US hospitals

Primary outcome

Proportion of time with SpO2 90–96% over 72 hours

Effect

85% vs 63% (adjusted difference 21 points, 18–25); time above 96%: 9.2% vs 29.1%

SAVE-O2 AI was an unblinded randomised trial at four US hospitals, published in JAMA Internal Medicine on 1 October. It enrolled 300 adults on supplemental oxygen for acute respiratory illness, trauma, burns or acute surgery. For 72 hours, oxygen flow was adjusted either by a closed-loop device reading SpO2 continuously or manually by clinicians.

The device kept patients in the 90 to 96% target range 85% of the time, against 63% with usual care. The bigger story is the other side: with manual titration, patients spent 29% of their time above 96%, against 9% with automation. Time below 88% was small in both groups but lower with automation. About a fifth of patients had dark skin, and results were consistent across skin tone categories.

The trial measured saturations, not outcomes such as mortality or length of stay, so it does not show automation improves recovery. What it does show is how far routine manual care drifts from the prescribed target, mostly towards too much oxygen.

Closed-loop devices are not widely available, including in most Indian wards. The practical lesson is achievable without them: write a target range, and ask nurses to turn oxygen down as well as up.

• Write an SpO2 target range on every oxygen prescription.
• Ask nursing staff to wean oxygen when saturations sit above the range, not only to increase it when below.
• Review oxygen at every ward round; a patient at 99% on oxygen usually needs less.
• Remember that pulse oximetry can overestimate saturation in darker skin; check a blood gas when in doubt.

Why it matters

Hyperoxaemia on the ward is common and largely invisible, because no one is prompted to turn the oxygen down.

Don't overread it

The trial measured time in range, not clinical outcomes, so it does not show automation improves recovery.

The statistics, in plain English

Time in target rose by 21 percentage points (95% CI 18 to 25), a large and precise difference. Time below 88% fell by 1.3 points (−2.0 to −0.5): statistically clear but small in absolute terms, because hypoxaemia was uncommon in both groups.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.