Table 7.
Sleep variables from real-time models, compared to each other
| Healthy volunteers test set* | ||
|---|---|---|
| Variable [units]† | Overall bias (95% CI) [P-value] | 95% Limits of Agreement |
| TST [min] | ||
| Actigraphy to radar nightstand | −2.4 (−8.1, 3.3) [0.40] | [−46.5, 41.6] |
| Actigraph to radar ceiling | −0.8 (−10.1, 8.6) [0.87] | [−69.6, 68.1] |
| Radars ceiling to nightstand | −1.0 (−6.8, 4.7) [0.72] | [−43.4, 41.3] |
| SOL [min] | ||
| Actigraphy to radar nightstand | −2.8 (−6.2, 0.7) [0.11] | [−29.4, 23.8] |
| Actigraphy to radar ceiling | 0.1 (−2.5, 2.7) [0.95] | [−18.9, 19.1] |
| Radars ceiling to nightstand | −3.2 (−6.0, −0.3) [0.03] | [−24.3, 18.0] |
| WASO [min] | ||
| Actigraphy to radar nightstand | 16.8 (10.2, 23.4) [<0.001] | [−34.2, 67.9] |
| Actigraphy to radar ceiling | 11.5 (1.5, 21.4) [0.03] | [−62.1, 85.0] |
| Radars ceiling to nightstand | 5.8 (−1.3, 13.0) [0.11] | [−47.2, 58.9] |
| SE [%] | ||
| Actigraphy to radar nightstand | −2.6 (−4.0, −1.2) [<0.001] | [−13.6, 8.4] |
| Actigraphy to radar ceiling | −2.1 (−4.3, 0.1) [0.06] | [−18.2, 14.0] |
| Radars ceiling to nightstand | −0.5 (−1.8, 0.8) [0.43] | [−10.1, 9.1] |
| NW [num] | ||
| Actigraphy to radar nightstand | 6.1 (4.1, 8.1) [<0.001] | [−9.3, 21.5] |
| Actigraphy to radar ceiling | 8.3 (6.3, 10.4) [<0.001] | [−6.7, 23.3] |
| Radars ceiling to nightstand | −2.8 (−4.5, −1.1) [0.01] | [−15.2, 9.6] |
Sleep parameters extracted from the sleep/wake state sequences resulting from the application of a real-time classification model that includes information from five past and zero future epochs, for all nights in the DS1-test (including the ones without consecutive PSG‡ recordings). Since no user marker was available for the nights without PSG, sleep onset latency was calculated from the set bedtime at 23:00 h.
Positive values of bias indicate underestimation by the second sensor relative to the first (p-values from Student's t-test for the null-hypothesis that the bias is zero). Limits of Agreement given as [bias −1.96*SD, bias +1.96*SD]; SD, standard deviation for paired differences.
*n = 12, mean age ± SD: 23.0 ± 3.1 years, 5 male, 11 nights of actigraphy + two radars per participant. The participants were randomly assigned into a training set for model development (n = 63/59 for nightstand/ceiling), and a testing set for validation (n = 63/58 for nightstand/ceiling).
†SOL, Sleep Onset Latency; TST, Total Sleep Time; WASO, Wake After Sleep Onset; SE, Sleep Efficiency; NW, Number of awakenings.
‡PSG, polysomnography.