Table 3.
Device | Sensitivity | Specificity | Accuracy | PPV | NPV | PABAK | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
All Epochs | Between Nights | All Epochs | Between Nights | All Epochs | Between Nights | All Epochs | Between Nights | All Epochs | Between Nights | All Epochs | Between Nights | |
Actiwatch | 0.95 | 0.95 (0.02) [0.95, 0.96] | 0.35 | 0.37 (0.15) [0.34, 0.4] | 0.90 | 0.90 (0.06) [0.89, 0.91] | 0.94 | 0.94 (0.06) [0.93, 0.95] | 0.38 | 0.39 (0.16) [0.36, 0.42] | 0.80 | 0.80 (0.11) [0.78, 0.82] |
Fatigue Science Readiband | 0.94 | 0.94 (0.07) [0.92, 0.95] | 0.40 | 0.43 (0.20) [0.39, 0.46] | 0.90 | 0.89 (0.07) [0.88, 0.90] | 0.95 | 0.94 (0.06) [0.93, 0.95] | 0.37 | 0.45 (0.22) [0.41, 0.49] | 0.79 | 0.78 (0.15) [0.75, 0.81] |
Fitbit Inspire HR | 0.93 | 0.93 (0.03) [0.93, 0.94] | 0.45 | 0.50 (0.17) [0.47, 0.54] | 0.89 | 0.89 (0.05) [0.88, 0.90] | 0.95 | 0.95 (0.06) [0.94, 0.96] | 0.36 | 0.36 (0.15) [0.33, 0.39] | 0.79 | 0.78 (0.11) [0.76, 0.81] |
Oura Ring | 0.94 | 0.94 (0.03) [0.94, 0.95] | 0.41 | 0.44 (0.17) [0.41, 0.47] | 0.90 | 0.90 (0.06) [0.89, 0.91] | 0.95 | 0.95 (0.06) [0.94, 0.96] | 0.37 | 0.39 (0.16) [0.36, 0.42] | 0.81 | 0.80 (0.11) [0.78, 0.82] |
Polar Vantage V Titan | 0.96 | 0.96 (0.02) [0.95, 0.96] | 0.35 | 0.39 (0.16) [0.36, 0.42] | 0.92 | 0.92 (0.04) [0.91, 0.93] | 0.96 | 0.96 (0.04) [0.95, 0.97] | 0.36 | 0.36 (0.15) [0.33, 0.38] | 0.84 | 0.84 (0.08) [0.82, 0.85] |
Notes: Proportions for EBE agreement metrics are shown for all sleep epochs (versus wake epochs) for all combined epochs as well as between all nights (irrespective of participant) for the devices compared with the corresponding epochs from the reference Dreem headband. Values for between-night analyses are presented as mean (SD) [95% confidence interval]. Higher values (closer to 1.0) indicate better performance on that metric.
Abbreviations: EBE, epoch by epoch; NPV, negative predictive value; PABAK, prevalence and bias adjusted kappa; PPV, positive predictive value; SD, standard deviation.