Skip to main content
NPJ Digital Medicine logoLink to NPJ Digital Medicine
. 2024 Jul 1;7:175. doi: 10.1038/s41746-024-01148-y

Author Correction: Self-supervised learning of accelerometer data provides new insights for sleep and its association with mortality

Hang Yuan 1,2, Tatiana Plekhanova 3, Rosemary Walmsley 1,2, Amy C Reynolds 4, Kathleen J Maddison 5,6, Maja Bucan 7, Philip Gehrman 8, Alex Rowlands 3,9, David W Ray 10,11, Derrick Bennett 1,12, Joanne McVeigh 13, Leon Straker 13, Peter Eastwood 14, Simon D Kyle 15, Aiden Doherty 1,2,
PMCID: PMC11217470  PMID: 38951631

Correction to: npj Digital Medicine 10.1038/s41746-024-01065-0, published online 20 May 2024

In this Article the authors have corrected a data cleaning issue. Specifically, some devices (older version of Actigraph) used to collect the wrist-worn accelerometer data for the study could enter sleep mode when the movement detected was low. Based on this, they filtered out the data affected by the sleep mode, only considering data without discontinuity in the data stream. However, another study in their group found this criterion insufficient because when the device enters sleep mode, values from the last timestamp are recorded for the entire second to the device, making it a continuous data stream. Therefore, all the data with repeated values for over one second were filtered out. The impact of the issue was limited to 205 out of 1448 nights of training data. After retraining on the clean dataset, preliminary results show that it enhances the reported sleep staging performance. These changes do not affect the core findings of the research. The accompanying open-source software package has also been updated to reflect these changes. The original article has been corrected.


Articles from NPJ Digital Medicine are provided here courtesy of Nature Publishing Group

RESOURCES