AI titration cut hospital oxygen overshoot by two-thirds

Clinical takeaway: In adults on low-flow oxygen, nearly a third of manually managed time ran above the target range, exposure that pooled trials have linked to higher mortality. Treat time above the SpO2 (oxygen saturation) target as a titration failure rather than a safety margin.

Oxygen is the one treatment on the ward nobody is responsible for turning down. Turning it up is urgent and obvious. Easing it back waits for the next check, which on a general floor can be hours away, and no alarm sounds in the meantime.

Excess oxygen is not always the harmless cushion it might seem. In trials that assigned acutely ill adults to a liberal target rather than a conservative one, the liberally treated patients did worse on in-hospital survival, and reflexive oxygen for suspected heart attack was dropped from practice on similar grounds. Because so many hospitalized adults are on oxygen at some point, that exposure accumulates in a population nobody is monitoring for it.

The trial randomized 300 adults at four US hospitals, hospitalized with acute respiratory illness, trauma, burns, or acute surgery and receiving 1 to 10 L of oxygen, to closed-loop or manual titration for up to 72 hours. Automation held the 90% to 96% target 85% of the time versus 63% under manual titration. Overshoot accounted for the difference, dropping to 9% from 29%, while time below range fell only 1.3 points.

Earlier closed-loop trials were run outside the US in mostly light-skinned patients, which matters because pulse oximeters lose accuracy at darker skin tones. Here 78% of participants had medium or dark skin pigmentation, and results favored automation in each group. It also held against a stronger comparator than usual: manual titration reached the target 63% of the time in this trial, against 40% to 50% in prior studies, partly because site staff kept reminding bedside teams of the target.

Usual care meant checks every one to two hours in the ICU and every four to eight on the wards. The device instead read a fingertip oximeter continuously and adjusted flow from 0 to 15 L/min, holding a saturation of 93% at the lowest flow that worked. No serious adverse events occurred in either group. The O2matic PRO100 is in clinical use in several countries but is investigational in the US.

The device is not cleared for use in the US. And no one has shown yet that holding a saturation in range shifts patient outcomes, a question this trial wasn't powered to answer. The investigators are moving next to transport and prehospital care, where manual checks are furthest apart.

"Oxygen is one of the most widely used therapies in medicine yet even in 2026, it is still managed largely through repeated manual adjustments made by clinicians," said Adit A. Ginde, MD, MPH, professor of emergency medicine at the University of Colorado Anschutz School of Medicine and the trial's principal investigator. "Our findings show that autonomous oxygen titration can help patients remain in their target range more consistently while reducing both under and over oxygenation. This technology has the potential to fundamentally change how supplemental oxygen is delivered in both civilian and military medicine."

Source: Douin DJ, et al. (2026 Aug 3) JAMA Intern Med. Autonomous Oxygen Titration for Maintaining Normoxemia in Acutely Ill Adults: The SAVE-O2 AI Randomized Clinical Trial