JMIR Publications

ABSTRACT

Background:

Infections are a major health concern in hospitalized patients. Fever is one of the earliest symptoms of underlying infection, making temperature monitoring a fundamental aspect of fever detection and infection surveillance. Wearable devices, as an innovative approach for continuous temperature monitoring, are increasingly being explored in acute care hospitals. While device performance has been evaluated, how temperature data from wearables are monitored, presented and utilized to support fever or infection management in clinical practice remains poorly understood.

Objective:

To map existing evidence on the use of temperature data from wearables for fever or infection management in acute care hospitals, focusing on wearables characteristics; data transmission, storage and presentation strategies; data preprocessing and analytic approaches; and maturity of wearables' clinical integration.

Methods:

We included 24 publications from 21 studies, published between 2018 and 2024. Fourteen wearable devices were identified, monitoring temperature at axilla, chest, wrist, or upper arm. Continuous data streams were predominantly transmitted in real-time, while 4 studies used non–real-time batch uploads, periodic synchronization, or device-memory downloads. Only 8 studies presented wearable data in clinical settings, displaying data on mobile devices and/or centralized monitoring stations; only one study reported enabling alerts through EHRs. Invalid sensor data filtering methods varied across studies, such as physiological thresholds, firmware quality scores, and statistical outlier detection. Temperature data were most frequently analyzed by descriptive analytics (n=14) to depict the frequency, timing, and duration of fever episodes; followed by diagnostic analytics (n=6) to identify risk factors or distinguish causes of fever, and predictive analytics (n=8) to forecast impending fever or infections. None of the included studies used prescriptive analytics. For clinical integration, most studies were at the clinical validation stage (n=6) or clinical research stage (n=15), corresponding to an early stage of maturity; 3 studies reached the routine clinical practice stage, and none progressed to multicenter implementation or full integration stages.

Results:

We included 24 publications from 21 studies, published between 2018 and 2024. Fourteen wearables were identified, monitoring temperature at axilla, chest, wrist, or upper arm. Continuous data streams were predominantly transmitted in real-time, while 4 studies used non–real-time batch uploads, periodic synchronization, or device-memory downloads. Only 8 studies presented wearable data in clinical settings, displaying data on mobile devices and/or centralized monitoring stations; only one study reported enabling alerts through EHRs. Invalid sensor data detection and filtering methods varied across studies, such as physiological thresholds, firmware quality scores, and statistical outlier detection. Temperature data were most frequently analyzed by descriptive analytics (n=14) to depict the frequency, timing, and duration of fever episodes; followed by diagnostic analytics (n=6) to identify risk factors or distinguish causes of fever, and predictive analytics (n=8) to forecast impending fever or infections. None of the included studies used prescriptive analytics. For clinical integration, most studies were at the clinical validation stage (n=6) or clinical research stage (n=15), corresponding to an early stage of maturity; 3 studies reached the routine clinical practice stage, and none progressed to multicenter implementation or full integration stages.

Conclusions:

While promising, the clinical integration of wearables remains at an early stage of maturity. Critical gaps exist in translating temperature data into actionable clinical insights, with limited data presentation, poor EHR interoperability, and underdeveloped analytic approaches. Future research should prioritize standardized data cleaning frameworks, workflow integration, and clinical interpretation to facilitate the active use of wearables data in clinical decision-making. Clinical Trial: This review was registered on the Open Science Framework (OSF) (https://osf.io/v6sp8).