Respiratory Rate as a Digital Biomarker

Respiratory rate is the number of breaths taken per minute. It is a core vital sign with an unusual reputation: it predicts serious deterioration better than most of the measurements taken alongside it, and it is the vital sign most often estimated, rounded or simply not recorded in routine care, which is why it has been described as the neglected vital sign.

Wearables change that arithmetic. A device that estimates breathing rate every night produces thousands of measurements where clinical practice would produce a handful, and it produces them under standardised conditions with the person at rest. That shifts the measure from a snapshot taken during an assessment towards a personal baseline against which deviation can be detected.

Two approaches dominate in wearables. Movement based estimation uses an accelerometer to detect the small chest or wrist motion produced by breathing. Waveform based estimation extracts breathing from the pulse signal, since respiration modulates the amplitude, frequency and baseline of the photoplethysmogram in ways an algorithm can separate.

Both work best when the person is still, which is why consumer devices generally report an overnight average rather than a continuous daytime figure. Clinical grade estimation uses chest bands, impedance measurement or dedicated contact sensors, which are more accurate but less tolerable for extended wear. Reference standards for validation are capnography or respiratory inductance plethysmography, and reported accuracy at rest is considerably better than accuracy during activity.

The main research use is deterioration detection. A sustained rise in overnight breathing rate is an early physiological signal in respiratory infection and in exacerbation of chronic lung disease, and it typically moves before oxygen saturation does, which makes the pairing of the two more informative than either alone.

Respiratory rate is also collected as a safety parameter in trials where respiratory depression is a concern, and as a covariate in sleep and cardiovascular studies. In respiratory disease it is reported alongside a breathlessness instrument such as the Borg CR10 or the mMRC scale, because how fast someone is breathing and how hard breathing feels are different quantities that respond to different things.

Daytime estimates from wrist wearables are considerably less reliable than overnight ones, and most consumer devices do not attempt a continuous daytime figure for that reason. Absolute accuracy is modest, so the measure is best used for within person change rather than for comparing values between people or devices.

Breathing rate also responds to a great deal that has nothing to do with disease, including speech, posture, emotion, room temperature and alcohol. Without context, a night with an elevated rate is uninterpretable. And because different devices use different estimation methods and averaging windows, values are not portable across devices even when reported in the same units.

• Cretikos MA, et al. Respiratory rate: the neglected vital sign. Med J Aust. 2008. pubmed.ncbi.nlm.nih.gov
• Fieselmann JF, et al. Respiratory rate predicts cardiopulmonary arrest for internal medicine inpatients. J Gen Intern Med. 1993. pubmed.ncbi.nlm.nih.gov
• Charlton PH, et al. Breathing rate estimation from the electrocardiogram and photoplethysmogram: a review. IEEE Rev Biomed Eng. 2018. pubmed.ncbi.nlm.nih.gov
• Massaroni C, et al. Contact-based methods for measuring respiratory rate. Sensors. 2019. pubmed.ncbi.nlm.nih.gov