Utility of the statistical and nonlinear analysis for the actigraphic sleep pattern characterization

D. Martin Martinez, P. Casaseca de la Higuera, C. Alberola Lopez, J. Garmendia Leiza, J. Andrés de Llano, S. Alberola Lopez

Research output: Contribution to journalArticlepeer-review


Actigraphy is a useful tool for the assessment of the sleep pattern being mainly addressed by means of the Sadeh’s algorithm; hence several aspects of sleep, such as regularity, have not been appropriately studied so far. This paper strives for showing the utility of both analysis of the sleep registries to complement the sleep pattern characterization. The discriminant capability of some statistical and nonlinear features has been evaluated over two cohorts (institutionalized and non-institutionalized elderly) in which the features resulting from the Sadeh’s algorithm do not show significant differences.

Materials and methods
Case/control study of elderly patients (65 years older). The case group was 144 institutionalized patients, while the control group were 124 patient home-living. Subjects were monitored with the Actigraph GT3x device, 24 h a day from Monday to Thursday, using 1 s epochs. Statistical features are composed by the mean, median, standard deviation, the interquartile range and the variation coefficient (VC). Nonlinear features are formed by those extracted through the analysis with the central tendency measure (CTM) and symbolic dynamics (SD). CTM evaluates the regularity at the ro scale (ro = 0, typically), while the SD (3 symbols long alphabet; 2 symbols/word) provides a set of word appearance probabilities that indicates either regularity or variability; Besides, the Shannon’s entropy (ES) has been also included as complexity measure. All these features have been analysed by means of the U-test of Mann–Whitney to determine the existence of significant differences between the cohorts.


As for the statistical features, only the VC shows significant differences, being higher in the control group (p < 0.05). Regarding the nonlinear features, both CTM and SD give out discriminant features; specifically, the control group shows higher values of CTM (p < 0.04), P02, P20 and P22 (p < 0.02), whereas higher values of ES are achieved in the case group (p < 0.05).

Both the VC, the CTM and the SD are useful to complement the characterization of the sleep pattern. In the current study, these features allow for the assessment of the regularity and the intensity of activity during sleep. Results of both the CTM and the ES point out that the activity of institutionalized elderly is less regular than the activity of those who live at home, which is in line with the results of P22.
Original languageEnglish
Pages (from-to)e181-e181
Number of pages1
JournalSleep Medicine
Issue numberS1
Publication statusPublished - 1 Dec 2013
Externally publishedYes
Event5th World Congress on Sleep Medicine - Palacio de Congresos de Valencia, Valencia, Spain
Duration: 28 Sep 20132 Oct 2013


Dive into the research topics of 'Utility of the statistical and nonlinear analysis for the actigraphic sleep pattern characterization'. Together they form a unique fingerprint.

Cite this