Abstract
Introduction:
Activity monitors are increasingly used in clinical trials to measure physical activity or sedentary behaviour to inform participant outcomes and/or intervention fidelity.
The RECREATE programme seeks to develop and evaluate strategies for reducing sedentary behaviour to improve outcomes after stroke. In the feasibility study, we summarised stroke survivors’ outcomes as a proportion of waking time, so we needed to separate sleep time from waking time for participants.
Gold standards of calculating sleep times using participant observations or plethsmography are impractical for large-scale trials. Automated algorithms for the activPAL monitor can be used to identify sleep times based on long sedentary periods, however these may not be appropriate for less active populations. Alternatives include sleep diaries or assuming a set waking time across all participants.
Methods:
In this feasibility study, participants were asked to wear the activPAL monitor for 7 days at baseline and again at follow-up. We compared sleep times derived using activPAL algorithm with self-reported sleep times from participant diaries using paired mean differences and 95% confidence intervals (CI).
Results:
activPAL data was available for 26 participants at baseline and 13 participants at 4 months. Sleep times calculated from the activPAL algorithm were 2.2 hours longer per day (95% CI 1.3 to 3.2, n=115 days) at baseline and 0.8 hours longer (95% CI 0.0 to 1.6, n=75 days) at 4 months compared to the sleep diary. At baseline, daily sleep times ranged from 3.5 to 20.8 hours from the diary and 1.3 to 23.5 hours from the activPAL algorithm. At 4 months, sleep times ranged from 4.3 to 13.3 hours from the diary and 0.3 to 24 hours from the activPAL algorithm.
In the activPAL algorithm, periods of sedentary behaviour immediately before or after sleep could be incorrectly classified as sleep time, thus under-estimating the amount of sedentary time during waking hours.
Conclusion:
Although it is feasible to collect activity monitor data in stroke survivors, existing automated algorithms to summarise activity outcomes may be inappropriate due to incorrectly classifying sedentary periods as sleep or incorrectly indicating that the activity monitor has been removed. The two proxy measures of automated algorithms and sleep diaries provided different results.
Further research is needed to adapt activity monitor algorithms for use in less active populations, such as stroke survivors. We plan to test this further on a larger sample size in a randomised trial as part of the RECREATE programme.
Authors: Bethan Copsey 1, Jennifer Airlie 2, Ivana Holloway 1,3, Florence Day 1, Coralie English 4, Alison Fergusson 1, Claire Fitzsimons 4, Gillian Mead 6, Lauren Moreau 1, Seline Ozer 2, Amanda Farrin 1, Anne Forster 2,7.
Activity monitors are increasingly used in clinical trials to measure physical activity or sedentary behaviour to inform participant outcomes and/or intervention fidelity.
The RECREATE programme seeks to develop and evaluate strategies for reducing sedentary behaviour to improve outcomes after stroke. In the feasibility study, we summarised stroke survivors’ outcomes as a proportion of waking time, so we needed to separate sleep time from waking time for participants.
Gold standards of calculating sleep times using participant observations or plethsmography are impractical for large-scale trials. Automated algorithms for the activPAL monitor can be used to identify sleep times based on long sedentary periods, however these may not be appropriate for less active populations. Alternatives include sleep diaries or assuming a set waking time across all participants.
Methods:
In this feasibility study, participants were asked to wear the activPAL monitor for 7 days at baseline and again at follow-up. We compared sleep times derived using activPAL algorithm with self-reported sleep times from participant diaries using paired mean differences and 95% confidence intervals (CI).
Results:
activPAL data was available for 26 participants at baseline and 13 participants at 4 months. Sleep times calculated from the activPAL algorithm were 2.2 hours longer per day (95% CI 1.3 to 3.2, n=115 days) at baseline and 0.8 hours longer (95% CI 0.0 to 1.6, n=75 days) at 4 months compared to the sleep diary. At baseline, daily sleep times ranged from 3.5 to 20.8 hours from the diary and 1.3 to 23.5 hours from the activPAL algorithm. At 4 months, sleep times ranged from 4.3 to 13.3 hours from the diary and 0.3 to 24 hours from the activPAL algorithm.
In the activPAL algorithm, periods of sedentary behaviour immediately before or after sleep could be incorrectly classified as sleep time, thus under-estimating the amount of sedentary time during waking hours.
Conclusion:
Although it is feasible to collect activity monitor data in stroke survivors, existing automated algorithms to summarise activity outcomes may be inappropriate due to incorrectly classifying sedentary periods as sleep or incorrectly indicating that the activity monitor has been removed. The two proxy measures of automated algorithms and sleep diaries provided different results.
Further research is needed to adapt activity monitor algorithms for use in less active populations, such as stroke survivors. We plan to test this further on a larger sample size in a randomised trial as part of the RECREATE programme.
Authors: Bethan Copsey 1, Jennifer Airlie 2, Ivana Holloway 1,3, Florence Day 1, Coralie English 4, Alison Fergusson 1, Claire Fitzsimons 4, Gillian Mead 6, Lauren Moreau 1, Seline Ozer 2, Amanda Farrin 1, Anne Forster 2,7.
Original language | English |
---|---|
Publication status | E-pub ahead of print - 6 Oct 2022 |
Event | 6th International Clinical Trials Methodology Conference 2022 - Harrogate, United Kingdom Duration: 3 Oct 2022 → 6 Oct 2022 https://ictmc.org/programme/ |
Conference
Conference | 6th International Clinical Trials Methodology Conference 2022 |
---|---|
Abbreviated title | ICTMC 2022 |
Country/Territory | United Kingdom |
City | Harrogate |
Period | 3/10/22 → 6/10/22 |
Internet address |