Wearable Alerts in Older Adults Need Accountability, Not Just More Data

Wearable devices can now track heart rate, rhythm, activity, sleep, oxygen saturation, temperature and blood-pressure-related signals continuously in people’s daily lives, far beyond the reach of the clinic. Yet according to a new perspective article published in Frontiers in Digital Health, the central problem in digital health is no longer whether these data can be collected. It is whether they can be converted into information that actually changes care, particularly for older adults, in whom frailty, multimorbidity, polypharmacy, cognitive impairment, altered vascular physiology and dependence on caregivers modify both how reliably devices measure and what the resulting signals clinically mean.

The article, authored by Jose E. Krieger of the Instituto do Coração at the University of São Paulo, proposes a trajectory-to-action framework built on five elements: an individual baseline, deviation within the same person over time, concordance across multiple signals, geriatric contextualization, and a predefined clinical response. The author is explicit that the framework is conceptual and has not been prospectively validated; its clinical utility, safety, implementation feasibility and impact on outcomes all require prospective testing before it can be considered an established care pathway.

The core argument is a shift away from population thresholds. Conventional medicine is organized largely around cut-offs: blood pressure exceeds a target, heart rate enters an abnormal range, oxygen saturation falls below a defined level. Thresholds remain useful, the paper argues, but they are incomplete for continuous monitoring. A systolic pressure of 105 mmHg may be well tolerated by one older person and represent clinically important hypotension in another. A resting heart rate of 90 beats per minute may be habitual in one patient but a substantial departure from baseline in another. A drop from 5,000 to 2,500 daily steps can carry information that neither number conveys in isolation.

Frailty makes this distinction especially consequential. The paper describes frailty as reduced physiological reserve and increased vulnerability to stressors rather than simply an accumulation of diagnoses, meaning that a modest physiological perturbation may have very different consequences in a robust individual than in a frail one. Deterioration in later life often develops gradually rather than as an abrupt crossing of a diagnostic threshold: reduced walking, fragmented sleep, a rising resting heart rate, altered blood pressure, declining oxygen saturation, weight change or loss of regular daily activity may each be nonspecific on its own, but their joint deviation from a person’s previous state may signal a loss of physiological stability.

The framework’s second pillar is multimodal concordance. Wearable platforms increasingly generate numerous variables, but the article stresses that more variables do not necessarily produce more information. A small change in resting heart rate may be meaningless; a decline in activity may reflect weather or behavior; sleep disruption may be transient; weight can fluctuate because of measurement error. Rising heart rate combined with lower activity, fragmented sleep, increasing weight, lower nocturnal oxygen saturation and new breathlessness, however, tells a different story. In the proposed model, magnitude refers to the size of change from an individual’s reference state, persistence to the duration or recurrence of that change over a prespecified observation window, multimodal concordance to agreement in timing and direction across clinically relevant signals, and signal quality to the technical reliability and completeness of the data. The author emphasizes these are not universal thresholds; their definitions, windows, weighting and action cutoffs should be prespecified and validated for each clinical use case, device configuration and target population.

Blood pressure illustrates why caution is warranted. Cuffless measurement is attractive because repeated observations could reveal variability, nocturnal patterns, medication effects and hypotensive episodes that conventional measurements miss. But calibration, inter-individual variation, vascular properties, body position, movement and generalizability remain major limitations. The author cites his own work on photoplethysmography-based blood pressure estimation, which demonstrated the importance of intra- and inter-individual variability and the limits of apparently accurate models when transferred across individuals. The paper accordingly recommends that wearable or cuffless blood-pressure trends be treated as supportive monitoring information unless the specific device and use case have been clinically validated, and that any treatment change prompted by such a trend be confirmed with a clinically validated measurement approach.

The article also insists that older adults are not generic digital-health users. Tremor, edema, altered skin properties, peripheral vascular disease, arrhythmias, gait aids, sensory impairment, arthritis, cognitive impairment and difficulty maintaining charging or synchronization routines can all alter both sensing and engagement. Studies of technology adoption consistently show that older adults may value reassurance, independence, self-awareness and communication with caregivers while simultaneously encountering discomfort, usability problems, uncertain benefit, privacy concerns and lack of confidence. The proposed contextual layer would incorporate, where relevant, frailty, cognition, medication exposure, renal function, recent treatment changes, functional capacity, caregiver availability and goals of care, so that these factors modify not merely whether a patient can operate a device but how the information is interpreted and acted upon.

Perhaps the sharpest claim in the paper concerns what happens after a sensor detects something. The weakest link in many remote-monitoring systems, it argues, is not the sensor but the response. A wearable alert should be considered a request for attention, not a diagnosis. Before implementation, every clinically relevant alert should answer four questions: who receives it, how quickly should it be reviewed, what confirms or refutes it, and what happens next. Stable patterns may require no action; small transient deviations may justify observation; persistent concordant changes may prompt nurse or clinician review; rapid deterioration may require urgent escalation. In worked examples, declining activity with rising heart rate, weight and symptoms after a heart-failure hospitalization might trigger nurse review and confirmation of congestion before medication adjustment, while an irregular-rhythm notification should usually lead to confirmatory electrocardiography rather than immediate therapeutic action.

The paper extends this logic to evaluation. Wearables are frequently judged by analytical performance such as correlation, bias, sensitivity, specificity or classification accuracy, measures the author calls necessary but insufficient. He proposes a sequential chain of questions running from analytical validity, through physiological and clinical validity, to clinical utility and implementation value. For clinical implementation, studies should report false-alert burden, the proportion of alerts leading to meaningful action, time from signal to review, number needed to evaluate, clinician workload, adherence, patient and caregiver burden, and downstream testing. The appropriate outcome, the article concludes, is net clinical benefit rather than the volume of data collected. Heart-failure studies are cited as illustrating why remote monitoring appears most useful when physiological information is embedded within structured programs incorporating clinical review, education, treatment adjustment and escalation, rather than delivered as an isolated stream of measurements.

Equity receives explicit attention as well. Wearable-enabled care could extend specialist oversight beyond hospitals, support community health teams and enable earlier intervention in geographically dispersed populations. Conversely, smartphone dependence, device costs, digital literacy requirements, proprietary algorithms and fragmented infrastructure could preferentially benefit healthier and wealthier users. The paper argues that the populations in whom continuous monitoring may be most valuable must be included in validation rather than excluded because they are technically more difficult to study. If prospectively validated, the author suggests, the value of wearable medicine in older adults should be judged not by the volume of data collected or alerts generated, but by whether meaningful deterioration can be recognized earlier, unnecessary interventions avoided, treatment made safer and independence better preserved. Until such prospective studies define use-case-specific criteria and test safety, utility, burden and equity effects, the framework remains a proposal for testable implementation strategies rather than a ready-made clinical tool.

Subject of Research: A conceptual framework for translating wearable monitoring data in older adults into accountable clinical care

Article Title: From continuous signals to accountable care: a trajectory-based framework for wearable monitoring in older adults

Article References: Krieger, J. E. (2026). From continuous signals to accountable care: a trajectory-based framework for wearable monitoring in older adults. Frontiers in Digital Health, 8, Article 1998814. https://doi.org/10.3389/fdgth.2026.1998814

Image Credits: AI Generated

DOI: 10.3389/fdgth.2026.1998814

Keywords: wearables, older adults, frailty, digital health, remote monitoring, blood pressure, heart failure, alerts, clinical validation, equity, digital biomarkers, telemedicine