Scripps Clinic Sleep Center, La Jolla, CA 92037, USA.
J Sleep Res. 2010 Dec;19(4):612-9. doi: 10.1111/j.1365-2869.2010.00835.x.
Wrist actigraphy is employed increasingly in sleep research and clinical sleep medicine. Critical evaluation of the performance of new actigraphs and software is needed. Actigraphic sleep-wake estimation was compared with polysomnographic (PSG) scoring as the standard in a clinical sleep laboratory. A convenience sample of 116 patients undergoing clinical sleep recordings volunteered to participate. Actiwatch-L recordings were obtained from 98 participants, along with 18 recordings using the newer Spectrum model (Philips Electronics), but some of the actigraphic recordings could not be adequately aligned with the simultaneous PSGs. Of satisfactory alignments, 40 Actiwatch recordings were used as a training set to empirically develop a new Scripps Clinic algorithm for sleep-wake scoring. The Scripps Clinic algorithm was then prospectively evaluated in 39 Actiwatch recordings and 16 Spectrum recordings, producing epoch-by-epoch sleep-wake agreements of 85-87% and kappa statistics averaging 0.52 (indicating moderate agreement). Wake was underestimated by the scoring algorithm. The correlations of PSG versus actigraphic wake percentage estimates were r = 0.6690 for the Actiwatch and r = 0.2197 for the Spectrum. In general, using a different weighting of activity counts from previous and subsequent epochs, the Scripps Clinic algorithm discriminated sleep-wake more successfully than the manufacturer's Actiware algorithms. Neither algorithm had fully satisfactory agreement with PSG. Further evaluations of algorithms for these actigraphs are needed, along with controlled comparisons of different actigraphic designs and software.
腕动描记法在睡眠研究和临床睡眠医学中应用越来越广泛。需要对新的动作描记器和软件的性能进行批判性评估。在临床睡眠实验室中，将动作描记器的睡眠-觉醒估计与多导睡眠图（PSG）评分进行比较。一项临床睡眠记录的便利样本中，有 116 名患者自愿参与。从 98 名参与者中获得了 Actiwatch-L 记录，以及 18 名参与者使用较新的 Spectrum 模型（飞利浦电子）的记录，但有些动作描记器记录无法与同时进行的 PSG 充分对齐。在可接受的对齐中，使用 40 个 Actiwatch 记录作为训练集，以经验开发新的斯克里普斯诊所睡眠-觉醒评分算法。然后，前瞻性地评估了 39 个 Actiwatch 记录和 16 个 Spectrum 记录中的斯克里普斯诊所算法，产生了 85-87%的逐时睡眠-觉醒一致率和平均 0.52 的kappa 统计量（表示中度一致）。评分算法低估了清醒状态。PSG 与动作描记器清醒百分比估计值的相关性分别为 Actiwatch 的 r = 0.6690 和 Spectrum 的 r = 0.2197。总体而言，使用来自先前和随后的时间段的活动计数的不同权重，斯克里普斯诊所算法比制造商的 Actiware 算法更成功地区分了睡眠-觉醒。这两种算法都与 PSG 不完全一致。需要进一步评估这些动作描记器的算法，并对不同的动作描记器设计和软件进行对照比较。
