Database Open Access
BIG IDEAs Lab Glycemic Variability and Wearable Device Data
Peter Cho , Juseong Kim , Brinnae Bent , Jessilyn Dunn
Published: March 6, 2023. Version: 1.1.0 <View latest version>
When using this resource, please cite:
(show more options)
Cho, P., Kim, J., Bent, B., & Dunn, J. (2023). BIG IDEAs Lab Glycemic Variability and Wearable Device Data (version 1.1.0). PhysioNet. https://doi.org/10.13026/e827-h193.
Please include the standard citation for PhysioNet:
(show more options)
Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.
Abstract
This study aimed to determine the feasibility and effectiveness of wearable devices in detecting early physiological changes prior to the development of prediabetes [1-3]. The study generated digital biomarkers for remote, mHealth-based prediabetes and hyperglycemia risk to classify which individuals should undergo further clinical testing. The primary inclusion criteria were subjects aged 35-65 years, inclusive, including only post-menopausal females, with a point of care A1C measurement between 5.2-6.4%, inclusive. Blood was collected during the study for measurement of glucose, hemoglobin A1C, lipoproteins, and triglycerides. Participants wore a Dexcom 6 continuous glucose monitor (CGM) and an Empatica E4 wristband for 10 days while receiving a standardized breakfast meal every other day. At the end of the 10 days, the participant returned to the clinic for an oral glucose tolerance test (OGTT). Research data collected includes physiological measurements from wearable devices such as heart rate, accelerometry, and electrodermal conductance.
Background
An estimated 90% of the 86 million US adults with prediabetes are undiagnosed. Prediabetes can lead to severe health consequences including type 2 diabetes and cardiovascular disease. Elevated fasting glucose levels even below prediabetes thresholds can contribute to increased cardiovascular risk, and the risk of cardiovascular death is 1.7 times higher for people with type 2 diabetes. Prediabetes is a complex and highly heterogeneous condition, which has resulted in ill-defined criteria for clinical classification and diagnosis. In-person clinic visits and invasive blood testing are needed for screening which further reduces detection rates. Recent dramatic improvements in mobile health (mHealth) technologies provide an unprecedented opportunity to address these gaps to better define prediabetes and revolutionize its detection.
Methods
Empatica E4 desktop computer application - this desktop application is developed by Empatica and downloaded from the Empatica website. The data from the wristband device (photoplethysmography, galvanic skin response, skin temperature, and accelerometry) was uploaded to the application by the user via USB. The data was stored in secure cloud storage managed by Empatica, accessible by the subject and research team only.
Empatica E4 real-time mobile (iOS application) application is developed by Empatica to display parameters generated by the Empatica E4 wristband (photoplethysmography, galvanic skin response, skin temperature, and accelerometry). The data was synced through Bluetooth BLE and data was stored in secure cloud storage managed by Empatica (same as desktop computer application), accessible by the subject and the research team only. The subject downloaded the application onto their mobile device and we provided a key for them to access the application.
The Dexcom Clarity mobile app was used to monitor the data produced by the Dexcom G6 continuous glucose sensor that the participant was wearing. This app was downloaded from the internet. The participant was needed to initiate sharing for the study team to gain access to this data. The participant, study coordinators, investigators, and analysts had access to this data. The Dexcom clarity app was downloaded to the participant's mobile phone. The PHI collected by the Dexcom Clarity app was the interstitial glucose levels of the participant. The data was stored on a private drive of the Endocrinology server. Data was deidentified by assigning each participant a unique ID and time shifting the observations. The de-identified data was shared with analysts from the biomedical engineering department.
The Dexcom Clarity web application allowed subjects to upload glucose data from their Dexcom CGM receiver or smartphone, view the data, and share the data with the research team. Data was uploaded from the CGM sensor via USB or Bluetooth.
Empatica E4 Manager is a desktop application used to upload the data stored in the E4 (wearable wristband) to Empatica secure cloud storage and access with the E4 connect web platform. The desktop application was downloaded on the subject's computer(s) and the subject was able to upload the data from the wristband via USB.
We accessed Duke patient populations and sought Duke patients with A1C levels in the high normal and prediabetic range. We screened for participants who were between 35-65 years of age. The participants were screened without regard for socioeconomics, ethnicity, skin color, or religion. We accessed MaestroCare to screen for patients that had not opted out of clinical research opportunities. The subjects were excluded if they have a history of chronic obstructive pulmonary disease (COPD), cardiovascular disease, cancer history, or chronic kidney disease. The subjects were excluded if they do not have access to a laptop or computer (required to be able to upload data daily from the wearable device). The study coordinator contacted patients using an IRB-approved phone script. Subjects received parking validations for the screening visit and Visits 1 and 2. Subjects were compensated $75 for Visits 1 and 2. The total compensation per subject for the entire study was $150 dollars plus the cost of parking validations for all three visits.
During the duration of the study, participants were provided standard breakfast meals to consume every other day. Participants were provided a food log and were advised to log all food items consumed during the study. Date shifting was performed in the same manner as the wearable device data to ensure the de-identification of participants.
Data Description
Study participants (n = 16) with elevated blood glucose in the normal range were monitored with the Dexcom G6 continuous glucose monitors and Empatica E4 wrist-worn wearable devices for 8-10 days. Note that all data is time-shifted (by date) to prevent reidentification.
The Dexcom G6 measures interstitial glucose concentration (mg/dL) every 5 min and the Empatica E4 measures photoplethysmography (PPG), electrodermal activity (EDA), skin temperature, and tri-axial accelerometry, resulting in a total of 7 features. PPG was sampled at 64 Hz, providing heart rate (HR) values every second along with a blood volume pulse (BVP) signal from which interbeat interval (IBI) data was computed. EDA and skin temperature were sampled at 4 Hz and accelerometry was sampled at 32 Hz.
Datasets
All data for each participant is contained in a folder numbered 001 through 016, with a separate CSV file for each feature. The seven features are:
- tri-axial accelerometry (ACC.csv),
- blood volume pulse (BVP.csv),
- interstitial glucose concentration (Dexcom.csv),
- EDA (EDA.csv),
- HR (HR.csv),
- IBI (IBI.csv), and
- skin temperature (TEMP.csv).
- Food Log (Food_Log_"ID#".csv)
Gender and HbA1c information for each participant can be found in file “Demographics.csv” in the root folder.
Columns
A summary of the different data files is provided below. The raw output of Empatica E4 data has been modified from its original format for readability [2].
- ACC: data will include the “Timestamp” as a datetime value and accelerometer data will be given for the "X”, “Y”, “Z” orientation.
- BVP: refers to blood volume pulse; data will include “Timestamp” as a datetime value and “Value” as the measurement recorded at the time.
- Dexcom: refers to interstitial glucose concentration; data will include “Timestamp” as a datetime value and “Value” as the measurement recorded at the time.
- EDA: refers to electrodermal activity; data will include “Timestamp” as a datetime value and “Value” as the measurement recorded at the time.
- TEMP: refers to skin temperature; data will include “Timestamp” as a datetime value and “Value” as the measurement recorded at the time.
- IBI: refers to interbeat interval; data will include “Timestamp” as a datetime value and “Value” as the measurement recorded at the time.
- HR: refers to heart rate; data will include “Timestamp” as a datetime value and “Value” as the measurement recorded at the time.
- Food Log: refers to food items consumed by that participant throughout the study; data will include "date" as a date value, "time_of_day" as a time value, "time_begin" as a datetime value, "time_end" as a datetime value, "logged_food" as a string value, "amount" as a numeric value, "unit" as a string value, "searched_food" as a string value, "calorie" as a numeric value, "total_carb " as a numeric value, "dietary_fiber", "sugar" as a numeric value, "protein" as a numeric value, and "total_fat" as numeric value
Usage Notes
This data has been used in the publication “Engineering digital biomarkers of interstitial glucose from noninvasive smartwatches” and “Non-invasive wearables for remote monitoring of HbA1c and glucose variability: Proof of concept”. The data can be reused to correlate glycemic variability with measurements from a research-grade wrist-worn wearable device.
There are no known limitations with this dataset. OGTT data were not collected because the files were corrupted. All variables did not suffer from high missingness as participants were relatively adherent to wearing the devices. Complementary code is available on the DBDP website and complementary datasets collected by the BIG IDEAs lab are available on DHDR [4, 5].
Release Notes
Initial release version 1.0
Included food logs for participants in version 1.1.0
Ethics
Data was collected through the Duke University Health Side system under the Biomedical Engineering department. Subjects provided written consent and subjects unable to provide legally effective consent were not approached for participation in the study. The consent and study procedures were held in a clinical room in the Diabetes Research Clinic at the Duke Hospital. Subjects were provided as much time as they need to review the document and were allowed to ask questions. They were also informed that they were under no obligation to sign the consent.
Acknowledgements
Duke MEDx supported this work. Brinnae Bent was a Duke Forge predoctoral fellow. Jessilyn Dunn is a MEDx Investigator and a Whitehead Scholar. This work was also guided by Dr. Mark Feinglos, our collaborator, mentor, and friend. His insights and ideas were integral to this study and we are grateful for the time we had to learn and work together.
Conflicts of Interest
The author(s) have no conflicts of interest to declare.
References
- Bent, B., Cho, P. J., Henriquez, M., Wittmann, A., Thacker, C., Feinglos, M., Crowley, M. J., & Dunn, J. P. (2021, June 2). Engineering digital biomarkers of interstitial glucose from noninvasive smartwatches. Nature News. Retrieved June 14, 2022, from https://www.nature.com/articles/s41746-021-00465-w#article-info
- Bent, B., Cho, P. J., Wittmann, A., Thacker, C., Muppidi, S., Snyder, M., Crowley, M. J., Feinglos, M., & Dunn, J. P. (2021, June 1). Non-invasive wearables for remote monitoring of HbA1c and glucose variability: Proof of concept. BMJ Open Diabetes Research & Care. Retrieved June 14, 2022, from https://drc.bmj.com/content/9/1/e002027
- Cho PJ, Bent B, Wittmann A, Merwin RM, Thacker CR, Feinglos MN, et al. 73-LB: Expanding the Definition of Intraday Glucose Variability. Diabetes. 2020 Jun 1;69(Supplement_1):73-LB.
- DBDP. (n.d.). Retrieved September 1, 2022, from https://dbdp.org/
- Digital Health Data Repository (DHDR). (2022). DBDP. https://github.com/DigitalBiomarkerDiscoveryPipeline/Digital_Health_Data_Repository (Original work published 2021)
Access
Access Policy:
Anyone can access the files, as long as they conform to the terms of the specified license.
License (for files):
Open Data Commons Attribution License v1.0
Discovery
DOI (version 1.1.0):
https://doi.org/10.13026/e827-h193
DOI (latest version):
https://doi.org/10.13026/w591-tp72
Topics:
biomedical engineering
pre-diabetes
biomarkers
Project Website:
https://github.com/DigitalBiomarkerDiscoveryPipeline/cgmquantify.
Corresponding Author
Files
Total uncompressed size: 34.1 GB.
Access the files
- Download the ZIP file (4.7 GB)
-
Download the files using your terminal:
wget -r -N -c -np https://physionet.org/files/big-ideas-glycemic-wearable/1.1.0/
-
Download the files using AWS command line tools:
aws s3 sync --no-sign-request s3://physionet-open/big-ideas-glycemic-wearable/1.1.0/ DESTINATION
Name | Size | Modified |
---|---|---|
Parent Directory | ||
ACC_003.csv (download) | 560.0 MB | 2022-07-31 |
BVP_003.csv (download) | 876.6 MB | 2022-07-31 |
Dexcom_003.csv (download) | 135.7 KB | 2022-07-31 |
EDA_003.csv (download) | 58.1 MB | 2022-07-31 |
Food_Log_003.csv (download) | 5.9 KB | 2023-02-21 |
HR_003.csv (download) | 10.7 MB | 2022-07-31 |
IBI_003.csv (download) | 6.3 MB | 2022-07-31 |
TEMP_003.csv (download) | 49.3 MB | 2022-07-31 |