Title of Dataset: Using commercial wearables to estimate activity levels of nursing home residents with dementia
Name: Dr. Adriana Petronella Anna van Beek ORCID: https://orcid.org/0000-0001-08077-4855 Institution: NIVEL, Netherlands Email: [email protected]
Name: Dr. Matthijs van Leeuwen ORCID: https://orcid.org/0000-0002-0510-3549 Institution: LIACS, Leiden Email: [email protected]
Name: Daniela Gawehns ORCID: https://orcid.org/0009-0002-9678-9012 Institution: LIACS, Leiden Email: [email protected]
Name: Suzanne Portegijs ORCID: https://orcid.org/0009-0002-4367-3235 Institution: NIVEL, Utrecht Email: [email protected]
May 2021 (not the exact dates to preserve privacy of participants)
Kaatsheuvel, the Netherlands
ZonMw funded research. Projectnumber: 733050846
The data of this project comes with one datafile called "linkedData21.csv" this file includes all data needed to recreate the correlation matrix in the manuscript. It containes only complete sets (i.e., only people with both, wearables and observational data).
- linkedData.csv : Contains behavioral observations, normalized 24 hrs data and MAD data information.
The MAD data is calculated over 5sec intervals and this forms the basis of the following derivatives: SD Median Qant Frag Rel
The 24hrs data was cleaned of time intervals that the watch was not in use (not on patient wrist but recording while being charged or being with the research team).
The 24 hrs data and the behavioral observations are normalized to 30 min. See more details in attributes list.
| Attribute name | Description |
|---|---|
| experiment_id | identifier of the experiment |
| dataset_label | dataset label of the experiment |
| plate_number | the identifier of the plate |
| plate_content | the description of the plate content |
Hardware used: Samsung Wearable, Gear Fit 2 Pro Software used: WEARDA https://research-software-directory.org/software/wearda Note on wearable data collection: we have different timelines per patient, some participated 4 days for 8 hrs each day, others missed hours due to changing sleep cycles or leaving the building to visit family or a doctor's appointment. The data got was normalized to 30 min intervals to make up for those difference. The MEDLO observation schema was used for the behavioral observations.