Understanding the role of uncertainty in data and its application in healthcare
We have been hearing a lot of policy talk in the UK health arena about how we need to do more to incorporate wearable technology, biosensors and remote monitoring into the health service (Fit for the Future: The 10 Year Health Plan for England) , but not much about how this will happen.
However although wearables and biosensors can offer useful insights into population health issues and for consumers to track trends in their own health improvements, non-clinical grade devices may not be suitable for health professionals looking at personalised healthcare in its current form due to uncertainty about the veracity of the data and how it has been analysed.
Wearable devices are being driven from wellness gadgets towards supporting clinical workflows. From smartwatches tracking heart rate, to patches and sensors monitoring activity, recovery, or vital signs the technology is advancing rapidly. As we start this shift, one overlooked but critical concept in interpreting the data, is measurement uncertainty - the degree to which a given reading might deviate from the “true” physiological value. Looking further into understanding uncertainty in clinical care allows clinicians to be safe and informed, better understanding what can and cannot be achieved with these devices.
So, what do we mean by uncertainty? Just like a blood pressure cuff or thermometer has margin of error, so too do wearables. Uncertainty can happen in several ways:
Sensor capture – this can rely on how well a wearable attaches, how much motion or external interference affects it such as a loose watch vs a tight watch strap or skin tone.
Signal processing and algorithms – how raw data is cleaned, filtered, interpreted by the company who provide this – Garmin, Apple etc (other brands available)
Clinical context and interpretation – how the number is used in decision-making, is it stable over time? Is it likely artefact?
Wearables may offer apparently continuous streams of data, but this apparent continuousness does not always equal accurate. As clinicians, the real question is: how reliable is this measurement in this patient, in this context, right now? How confident does this allow me to feel in planning a treatment based on this data?
Why does this matter in clinical practice? Without transparency on uncertainty, wearable data risk being mis-interpreted:
A heart-rate variability (HRV) estimate from a consumer wearable might look stable at rest, but during activity its error margin may widen significantly.
Two devices might use identical sensors but very different algorithms e.g. one filters aggressively and misses short bursts; another interpolates data silently to fill in the gaps and another reports every fluctuation. Clinicians need to know algorithmic limits.
In a clinical decision-making scenario, if a wearable reading changes (say, a drop in pulse amplitude or increase in variability), is that real physiology — or simply noise from poor signal or misprocessing?
Knowing where there is uncertainty and the parameters of it means knowing when you can trust the reading or when you should treat it as a trend indicator rather than a definitive value.
In the UK, whilst there is a national call for greater use of wearables in healthcare, there is currently no consistent guidance or regulatory framework to support clinicians in interpreting or relying on data from these technologies. While wearables are increasingly being discussed in strategic plans and innovation frameworks, there are no NICE guidelines, no agreed standards for clinical validation, and no recognised grading system for the reliability of consumer or medical-grade wearables. This absence of clear national direction creates an uneven landscape where data from commercial devices may enter clinical conversations without any structured assurance of accuracy, repeatability, or clinical relevance.
For clinicians, this presents a dilemma. Patients are increasingly arriving at their GP with data from devices that promise precision but often provide little transparency about their algorithms, calibration, or data provenance. This risks triggering a clinical pathway to secondary care that could be avoided if the clinician was supported by clinical evidence on the trustworthiness of the data.
Without independent evaluation or standardised grading, clinicians are left to rely on their professional judgement and experience to decide whether wearable readings hold any diagnostic or monitoring value. In practice, this often means balancing the patient’s expectations and the potential implications of the data against clinical reasoning, including consideration of treatment options, professional accountability, and even the medico-legal risk of disregarding information a patient believes to be significant.
In doing so, wearable data introduce a new dimension of uncertainty — not about the patient’s physiology, but about the credibility and trustworthiness of the technology itself.
To move forward safely, there is a need for nationally coordinated action. Organisations such as NICE and NHS Digital could support clinicians in use of data from wearables if there were:
Formal guidance on the use and interpretation of wearable-derived data in clinical practice.
A tiered reliability framework or evidence grading system for wearable brands and devices, like existing approaches for medical devices or diagnostic tools.
More transparency and publication of validation data from manufacturers, including real-world accuracy studies.
Independent testing and accreditation to help clinicians distinguish between wellness devices and those suitable for clinical use.
In the absence of such standards, clinicians should continue to approach wearable data as contextual information rather than clinical evidence — useful for prompting discussion, but not yet robust enough to influence decision-making on its own.
Without guidance, grading, and validation, we risk allowing enthusiasm for digital tools to outstrip the certainty and safety that underpin clinical practice.
Reference:
Jamieson, A., Chico, T.J.A., Jones, S. et al. A guide to consumer-grade wearables in cardiovascular clinical care and population health for non-experts. npj Cardiovasc Health 2, 44 (2025). https://doi.org/10.1038/s44325-025-00082-6