Can a Respiratory Rate Wearable Detect Patient Deterioration Early?

This piece covers hospital and clinical-setting research on continuous respiratory rate monitoring and its relationship to detecting patient deterioration. It does not cover consumer fitness tracking, home use, or exercise monitoring.

The research on continuously monitoring hospitalized patients found that continuous heart rate and respiratory rate tracking did not detect every serious deterioration event, including missing all four ICU admissions in one COVID-19 ward study. Continuous monitoring changes what gets measured and when, but the evidence does not show it reliably catches every deterioration event before it happens.

The hope behind the wearable

There's a specific hope driving a lot of interest in these devices: that a small sensor on someone's chest or arm could catch trouble building before a nurse walks in for the next scheduled check. Hospitals typically check vital signs on a schedule, every few hours, and the gap between checks is exactly where deterioration can hide. A continuous monitor closes that gap, in theory. Whether it actually catches what intermittent checks miss, or just produces more data without more warning, is the question underneath all of it.

That distinction matters because the appeal of wearable respiratory rate monitoring rests almost entirely on the idea that continuous beats intermittent. If continuous monitoring doesn't reliably flag the events that matter, the value proposition changes considerably.

Why 'continuous' doesn't automatically mean 'early'

Part of what's happening here is a gap between what continuous monitoring measures and what actually predicts deterioration. The COVID-19 ward study set out specifically to test whether continuous heart rate and respiratory rate monitoring would let clinicians catch decline sooner than the standard practice of checking vitals every few hours. The sensors ran, data was collected, and two ICU physicians reviewed it independently looking for signs that would trigger an alarm. But when that data was compared against what actually happened in the medical record, the four ICU admissions that occurred during the study period weren't flagged by the sensor data at all.

That's not the same as saying the sensors were broken or inaccurate. It's closer to saying that having continuous respiratory rate and heart rate data doesn't automatically translate into recognizing the pattern that precedes a crisis, not in the way this particular alarm and review setup was built, anyway. Whether a different algorithm, different thresholds, or additional vital signs would have caught those admissions isn't something this study answers.

There's also a separate, more basic layer underneath this: how well any given wearable measures respiratory rate at all. If the underlying number is noisy or biased, no amount of continuous monitoring built on top of it will produce a more reliable warning. I'd point anyone curious about that layer specifically to how these devices measure breathing rate compared to a reference standard, since it covers the measurement accuracy question separately from the deterioration-detection question addressed here.

What agreement studies add to the picture

A separate line of research asks a narrower question: when continuous wireless monitors run alongside standard wired hospital monitors, do they measure the same numbers? In postoperative patients recovering from major abdominal surgery, one comparison found the bias between standard and wireless heart rate monitoring was relatively small. Agreement for other vital signs, including blood pressure measures, showed wider limits of agreement, some outside what the study considered clinically acceptable. This kind of agreement testing is a precondition for trusting any deterioration alarm built on top of the data, but agreement alone doesn't confirm that a system catches deterioration early, and it's easy to conflate the two.

A related study comparing manual bedside respiratory rate counts to automated wearable readings on general hospital wards found automated readings ran noticeably higher than manual counts on average, with limits of agreement wider than what's typically considered clinically acceptable. That kind of systematic offset matters for anyone trying to use a single threshold number to trigger an early-warning alarm. A monitor that reads consistently higher could either trigger false alarms or, depending on how thresholds are set, mask a real change.

Where placement and setting change the answer

None of this means continuous monitoring is without value, only that the evidence doesn't support treating it as a reliable early-warning system on its own. The postbariatric recovery research found reasonably high agreement between an accelerometer-based wearable and bedside monitor readings for heart rate and respiration rate during postoperative recovery. That speaks to feasibility of continuous tracking in that specific recovery context, distinct from whether it flags deterioration before it happens.

Placement and body location also shape what a device can pick up, a separate axis from the deterioration-detection question but relevant to anyone comparing devices. A chest-worn sensor and a wrist-worn one aren't necessarily capturing the same signal quality, and that question is covered in more depth in how sensor placement affects respiratory rate accuracy.

Common questions