CoChat
Sign Up
Cited by 10Diabetes Management and ResearchNon-Invasive Vital Sign MonitoringContext-Aware Activity Recognition Systems

Toward Detection of Nocturnal Hypoglycemia in People With Diabetes Using Consumer-Grade Smartwatches and a Machine Learning Approach

Open access
Mendez, Camilo, Kaykayoglu, Ceren Asli, Bähler, Thiemo, Künzler, Juri, Lizoain, Aritz, Rothenbühler, Martina, Schmidt, Markus H., Laimer, Markus +1 more·2026·Journal of Diabetes Science and Technology
DOI 10.1177/19322968251319800PMC PMC11851596
Open this paper in CoChat — free account

Abstract

Background: Nocturnal hypoglycemia poses significant risks to individuals with insulin-treated diabetes, impacting health and quality of life. Although continuous glucose monitoring (CGM) systems reduce these risks, their poor accuracy at low glucose levels, high cost, and availability limit their use. This study examined physiological biomarkers associated with nocturnal hypoglycemia and evaluated the use of machine learning (ML) to detect hypoglycemia during nighttime sleep using data from consumer-grade smartwatches. Methods: This study analyzed 351 nights of 36 adults with insulin-treated diabetes. Participants wore two smartwatches alongside CGM systems. Linear mixed-effects models compared sleep and vital signs between nights with and without hypoglycemia during early and late sleep. A ML model was trained to detect hypoglycemia solely using smartwatch data. Results: Sixty-six nights with spontaneous hypoglycemia were recorded. Hypoglycemic nights showed increased wake periods, heart rate, stress levels, and activity during early sleep, with weaker effects during late sleep. In nights when hypoglycemia occurred during early sleep, the ML model performed comparable or better than prior studies with an area under the receiver operator curve of 0.78 for level 1 and 0.83 for level 2 hypoglycemia, with sensitivity of 0.78 and 0.89, specificity of 0.64 for both, negative predictive value of 0.94 and 0.99, and positive predictive value of 0.25 and 0.13 for level 1 and level 2 hypoglycemia, respectively. Conclusions: Consumer-grade smartwatches demonstrate promise for detecting nocturnal hypoglycemia, particularly during early sleep. Refining models to reduce false alarms could enhance their clinical utility as low-cost, accessible tools to complement CGM.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1177/19322968251319800

Cites

43 of 43 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Features of heart rate, heart rate variability, movement, and electrodermal activity extracted from wearable data have been applied to detect hypoglycemic events in cohorts ranging from 1 to 36 individuals.

    “Features of HR, HRV, movement, and electrodermal activity (EDA; a measure of perspiration) extracted from wearable data have been applied to detect hypoglycemic events in cohorts ranging from 1 to 36 individuals5,26–28.”

    From Sleep and temperature data from wearable devices support noninvasive detection of diabetes mellitus in a large-scale, retrospective analysis (2026) · cites this paper as [28]

    The Results (Baseline characteristics; page unavailable) report 36 participants, and Methods (Data preprocessing; page unavailable) describe heart-rate, heart-rate-variability, and movement data used for hypoglycemia detection. The Introduction discusses other wearable studies using electrodermal activity, but this study did not use it, and our copy does not establish the lower bound of one individual. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

  • Prior work was generated from small cohorts (the largest cohort of 70 individuals), raising questions of the generalizability of their findings.

    “Second, prior work was generated from small cohorts (the largest cohort of 70 individuals), that raise questions of the generalizability of their findings5,22–32.”

    From Sleep and temperature data from wearable devices support noninvasive detection of diabetes mellitus in a large-scale, retrospective analysis (2026) · cites this paper as [28]

    The Results (Baseline characteristics; page unavailable) report 36 participants in this study’s primary analysis, supporting its inclusion as an example of work based on a small cohort. Our copy does not establish that 70 was the largest cohort across the cited prior work or explicitly link those cohort sizes to concerns about generalizability. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

  • Prior work chiefly assessed statistical aggregation functions of wearable data streams (primary features), which are nonspecific to the effects of DM within data streams that are constantly fluctuating in natural settings.

    “Third, prior work chiefly assessed statistical aggregation functions of wearable data streams (primary features)5,26–32, which are nonspecific to the effects of DM within data streams that are constantly fluctuating in natural settings.”

    From Sleep and temperature data from wearable devices support noninvasive detection of diabetes mellitus in a large-scale, retrospective analysis (2026) · cites this paper as [28]

    The Methods > Data preprocessing section describes derived HRV parameters, accumulated accelerometer activity, and resampling of wearable data into one-minute intervals. That supports the narrower point that this study used processed wearable features. Our copy does not establish that prior work chiefly assessed statistical aggregation functions or that those features are nonspecific to diabetes-related effects in fluctuating natural settings. No page number is available in our copy. (LLM, conf=0.72)

    Experimental — accuracy not yet measured

  • Prior work typically uses less than a day of input.

    “First, where prior work typically uses less than a day of input5,22–32, we assess the impact of longitudinal input windows ranging up to multiple weeks and demonstrate improved sensitivity with longer windows (AUROC increases by 0.15 from 1-night to 21-night models, in balanced and imbalanced cohorts).”

    From Sleep and temperature data from wearable devices support noninvasive detection of diabetes mellitus in a large-scale, retrospective analysis (2026) · cites this paper as [28]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 2

  1. [39]Ke G Meng Q Finley T, et al. LightGBM: a highly efficient gradient boosting decision tree. In: NIPS’17: Proceedings of the 31st International Conference on Neural Information Processing Systems. New York, NY: ACM; 2017: 3149- 3157.
  2. [40]Pedregosa F Varoquaux G Gramfort A, et al. Scikit-learn: machine learning in Python. Mach Learn Python. 2011; 12: 2825- 2830.

Cited by

Sorted by citations

Related

Authors
Mendez, C., Kaykayoglu, C. A., Bähler, T., Künzler, J., Lizoain, A., Rothenbühler, M., Schmidt, M. H., Laimer, M., & Witthauer, L.
Affiliations
  • University of Bern (CH)
  • Diabetes Center Berne (CH)
Corresponding author
Witthauer, Lilian
ORCID
Venue
Journal of Diabetes Science and Technology
Publisher
SAGE Publications
Published
2025-02-25
Volume
20
Issue
4
Pages
1266-1275
ISSN
1932-2968 (ISSN-L)
DOI
10.1177/19322968251319800
PMC
PMC11851596
OpenAlex
W4407923260
Cited by
10
References
43
Citations by year
2026: 7 · 2025: 3
FWCI
6.82
Type
journal article
Language
en
Open access
Licence
Locations
  • Journal of Diabetes Science and Technology (publishedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
  • Open Access CRIS of the University of Bern
Funders
  • Vontobel-Stiftung (0847/2002)
  • Diabetes Center Berne
  • University of Bern
Topics
  • Health Sciences › Medicine › Endocrinology, Diabetes and Metabolism › Diabetes Management and Research
  • Physical Sciences › Engineering › Biomedical Engineering › Non-Invasive Vital Sign Monitoring
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Context-Aware Activity Recognition Systems
Keywords
nocturnal hypoglycemia, diabetes mellitus, insulin-treated diabetes, hypoglycemia detection, machine learning, sleep, heart rate, stress, physiological biomarkers, continuous glucose monitoring, low-cost monitoring
MeSH terms
Machine Learning, Continuous Glucose Monitoring, Digital Health, Predictive Learning Models, Adult, Aged, Blood Glucose, Diabetes Mellitus, Type 1, Female, Humans, Hypoglycemia, Hypoglycemic Agents, Insulin, Male, Middle Aged, Sleep, Blood Glucose Self-Monitoring, Biomarkers

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review