SEARCH

SEARCH BY CITATION

Keywords:

  • actigraphy;
  • children;
  • polysomnography;
  • sleep;
  • validation;
  • wake

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Subjects
  6. Results
  7. Discussion
  8. Acknowledgement
  9. References

There have been limited studies of the validation of actigraphy for the determination of sleep and wake in children and in this study we aimed to compare wrist actigraphy with polysomnography (PSG). We studied 45 children (29 M/16 F), aged between 1 and 12 years (5.8 ± 2.7 years, mean ± SD). Actigraphic data were collected during standard overnight PSG. Data from the actiwatch were analysed over four separate activity threshold settings (low, medium, high, auto). Actigraphic data were compared epoch-by-epoch with the matching PSG. Sleep time was not different from PSG values for the low or auto activity thresholds, but was significantly less on the medium and high activity thresholds (P < 0.05). In contrast, the low and auto activity thresholds significantly underestimated wake time (P < 0.05), whilst that recorded on the medium and high activity thresholds were not different to PSG values. Agreement rates across the thresholds were all high ranging from 85.1% to 88.6%. Predictive value for sleep and sensitivity were also high with values ranging from 91.6% to 94.9% and 90.1% to 97.7%, respectively. In contrast, predictive value for wake and specificity were low ranging between 46.7–65.6% and 39.4–68.9%, respectively. There was no effect of subject age, OAHI or PSG arousal index on AR for any of the activity thresholds. We conclude that actigraphy is a reliable method for determining sleep in children when compared against PSG. Actigraphy may be a useful tool in paediatric sleep clinics and research.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Subjects
  6. Results
  7. Discussion
  8. Acknowledgement
  9. References

An alternative to polysomnography (PSG) for assessing sleep/wake patterns is actigraphy, which has gained increasing popularity over the last 30 years. Actigraphs are small movement detectors (accelerometers) generally placed on the wrist which can distinguish sleep from wake using algorithms to quantify the reduced movement associated with sleep. Actigraphy has advantages over conventional PSG for the study of sleep in that the devices are small, can be used over long periods of time, do not require sophisticated recording equipment or specialist staff for their use, thus are low cost and can be used in the home. A number of studies have been carried out in healthy adults to assess the validity of actigraphy for assessing sleep and wake against the gold standard of PSG and have found good agreement (97%) for total sleep time (Jean-Louis et al., 1996), and overall agreement rates (AR) of 91–93% (Jean-Louis et al., 2001; Sadeh et al., 1994). Recently, our group has reported that actigraphy correctly predicts sleep in infants under 6 months of age (So et al., 2005). Both our study and those previously reporting on the use of actigraphy in infants have demonstrated that the threshold sensitivity of the actiwatch recording was important as this differed with age (Sadeh et al., 1995a). To our knowledge, there has been only one validation study comparing actigraphy with PSG in children (n = 13) and this study did not account for actiwatch threshold sensitivity (Sadeh et al., 1989). Thus, the aim of this study was to compare actigraphy at different threshold sensitivities with the PSG determination of sleep and wake in children.

Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Subjects
  6. Results
  7. Discussion
  8. Acknowledgement
  9. References

The Monash Medical Centre Human Ethics Committee granted ethical approval for this project. Subjects were volunteers recruited from children attending the Melbourne Children’s Sleep Unit for routine assessment of sleep-disordered breathing (SDB). Written informed consent was obtained from parents, and no monetary incentive was provided.

Subjects

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Subjects
  6. Results
  7. Discussion
  8. Acknowledgement
  9. References

Forty-five children (29 M/16 F) aged 1–12 years (5.8 ± 2.7 years, mean ± SD) were recruited sequentially between April and June 2005. Children with syndromes affecting limb movement, such as Duchenne muscular dystrophy, were excluded. An obstructive apnoea hypopnoea index (OAHI) defined as the total number of obstructive apnoeas, mixed apnoeas and obstructive hypopnoeas per hour of sleep was calculated. Classification of SDB severity followed our current practice: normal/Primary Snoring (PS; OAHI <1 events per hour and normal ventilation and sleep architecture, n = 31); mild obstructive sleep apnoea (OSA; OAHI between 1 and 5 events per hour, n = 8); moderate OSA (OAHI between 5 and 10 events per hour, n = 3) and severe OSA (OAHI >10 events per hour, n = 3).

Recording methods

Electroencephalograms (C4-A1, O2-A1), left and right electro-oculograms, submental electromyogram and electrocardiogram were attached for routine clinical PSG. Thoracic and abdominal breathing movements (Resp-ez Piezo-electric sensor, EPM Systems, Midlothian, VA, USA), oxygen saturation (Biox 3700e Pulse Oximeter, Ohmeda, Louisville, CO, USA) and transcutaneous carbon dioxide (TINA TCM3, Radiometer, Denmark, Copenhagen), nasal pressure and oronasal airflow were also recorded. Recordings were made on Compumedics Series S Sleep System (Compumedics, Melbourne Australia).

In addition, an actigraph (Actiwatch AW64, Mini Mitter Company Inc., Sunriver, OR, USA), time synchronized to the PSG, was placed on the non-dominant wrist (Sadeh et al., 1994). The actiwatch weighed 16 g, was sensitive to 0.01 g, sampled at a rate of 32 Hz and had a non-volatile memory of 64 kb.

Data analysis

Sleep state and arousal from sleep were scored from the PSG according to standard criteria (ASDA, 1992; Rechtschaffen and Kales, 1968). Previous literature in children has highlighted that approximately 50% of respiratory events are not terminated with an arousal as defined by ASDA criteria; however, they are commonly associated with increased autonomic activity (Katz et al., 2003; McNamara et al., 1996). Therefore, as per our current clinical practice, included in the arousal index were autonomic arousals that did not reach ASDA criteria. Autonomic arousals were scored if there were 2 of an increase in submental electromyogram, heart rate or gross body movement. Data from the actiwatch were coded into sleep and wake in 30 s epochs using commercially available software (Actiware®-Sleep V3.3, Mini Mitter Company Inc., Sunriver, OR, USA). Data were analysed at four separate activity threshold settings: low = 80 counts per epoch; medium = 40 counts per epoch; high = 20 counts per epoch and auto =(mean score in active period × 0.888)/epoch length. Sleep was scored when total activity counts were less than or equal to the activity threshold setting according to the following formula (Actiware®-Sleep V3.3).

  • image

A = total activity counts; E = activity counts during scored epoch; En = activity counts during previous or successive epochs.

For analysis, both PSG and actiwatch data were reduced to binary form (0 = wakefulness and 1 = any sleep state). AR, predictive value for sleep (PVS), and predictive value for wakefulness (PVW) were calculated as described previously (So et al., 2005) and as shown in Table 1. AR was defined as the proportion of observations for which PSG ratings were in accordance with actiwatch ratings for both sleep and awake; PVS was calculated as the probability that the actiwatch prediction was correct by PSG criteria for sleep; and PVW as the probability that the actiwatch prediction was correct by PSG criteria for wake. We also calculated sensitivity, the ability of the actiwatch to predict true sleep and specificity the ability of the actiwatch to predict true wake. Data were compared between each activity threshold with Friedman repeated measures anova on ranks with Dunn’s post hoc analysis. ARs for each activity threshold were compared against age, OAHI, PSG arousal index and PSG sleep efficiency with regression analysis (Sigma Stat 3.01, Systat Software Inc., Richmond, CA, USA). Demographic data are presented as mean ± SD, and sleep/wake data as median and interquartile range (IQR) with significance set at P < 0.05.

Table 1.   Calculation of PVS, PVW, sensitivity, specificity and agreement rate
 SleepPSG
WakefulnessTotal
  1. Predictive value for sleep (PVS) = 100 × 794/834 = 95.2%; predictive value for wakefulness (PVW) = 100 × 107/188 = 56.9%; sensitivity of actigraphy to sleep = 100 × 794/875 = 90.1%; specificity of actigraphy to wakefulness = 100 × 107/147 = 72.8%; agreement rate = 100 × (794 + 107)/1022 = 88.2%.

Actigraphic prediction of sleep79440834
Actigraphic prediction of wakefulness81107188
Total8751471022

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Subjects
  6. Results
  7. Discussion
  8. Acknowledgement
  9. References

A total of 22 396.5 min of sleep and wake were analysed. Median sleep and wake time per study and sleep efficiency recorded from PSG and for each of the four Actiwatch activity thresholds are presented in Table 2. Median recorded sleep time was not different from PSG values for the low or auto activity thresholds, but was significantly less on the medium and high activity thresholds (P < 0.05). In contrast, the low and auto activity thresholds significantly underestimated wake time (P < 0.05), whilst that recorded on the medium and high activity thresholds were not different to PSG values. Sleep efficiencies measured from the PSG were also not different to those measured from the actiwatch on the low and auto activity thresholds.

Table 2.   Median and IQR recorded total sleep and wake times and sleep efficiency for PSG and the four Actiwatch activity settings (n = 45)
 Sleep time (min)P-valueWake time (min)P-valueSleep efficiency (%)P-value
  1. P-values are in comparison to PSG values.

PSG440 (382–465) 62 (37–92) 88.0 (80.8–92.1) 
Low424 (397–453)NS21 (9–38)<0.0585.6 (79.0–90.3)NS
Medium402 (376–433)<0.0540 (20–61)NS82.5 (76.2–87.8)<0.05
High388 (358–417)<0.0559 (35–81)NS79.5 (72.8–85.2)<0.05
Auto426 (404–459)NS20 (10–29)<0.0587.7 (80.3–92.1)NS

Values for AR, PVS, sensitivity, PVW and specificity are presented in Table 3. Overall, median ARs were all high ranging from 85.1% to 88.6%. PVS and sensitivity were also high with values ranging from 91.6% to 94.9% and 90.1% to 97.7%, respectively. In contrast, PVW and specificity were low ranging between 46.7–65.6% and 39.4–68.9%, respectively. AR, sensitivity and PVW recorded on the high activity threshold were all significantly lower (P < 0.05) than that recorded on the other three activity thresholds. In contrast, PVS and specificity were higher (P < 0.05) recorded on the high activity threshold compared with the other three activity thresholds.

Table 3.   Summary of data relating actigraphy to PSG (n = 45): agreement rate (AR), predictive value for sleep (PVS) sensitivity, predictive value for wakefulness (PVW) and specificity for each activity threshold. All values expressed as median and IQR
Activity thresholdAR (%)PVS (%)Sensitivity (%)PVW (%)Specificity (%)
  1. *P < 0.05 significantly different from high activity threshold; P < 0.05 significantly different from medium activity threshold.

Low88.2 (85.2–92.7)*92.5 (85.8–97.0)*96.5 (94.4–98.8)*62.3 (44.0–83.7)*39.4 (15.5–67.3)*
Medium87.3 (83.8–91.7)*94.4 (87.3–97.9)*93.9 (90.9–97.1)*56.9 (37.1–74.5)*59.0 (28.7–82.1)*
High85.1 (81.7–89.4)94.9 (89.5–99.0)90.1 (85.3–94.6)46.7 (30.5–65.3)68.9 (40.6–92.6)
Auto88.6 (85.0–92.4)*91.6 (85.6–95.9)*97.7 (96.2–98.4)*65.6 (47.1–84.1)*39.4 (22.9–53.9)*

There was no effect of subject age, OAHI or PSG arousal index on AR for any of the activity thresholds. However, there was a positive correlation between sleep efficiency and AR for all activity thresholds (low: r = 0.79, P < 0.001; medium: r = 0.70, P < 0.001; high: r = 0.48, P < 0.001; auto: r = 0.83, P < 0.001).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Subjects
  6. Results
  7. Discussion
  8. Acknowledgement
  9. References

Our study has demonstrated for the first time that actigraphy is a reliable method for determining sleep in children between 1 and 12 years of age when compared against PSG. Overall we found high ARs of 85.1–88.6% between actigraphy and PSG. Actigraphy was reliable for determining total sleep time and sleep efficiency recorded on the low and auto activity settings, with no differences between that scored with PSG. However, actigraphy was less reliable for determining wakefulness.

In adults, it has been demonstrated that actigraphy provides an overall AR for recording total sleep time between 88% and 98% (Ancoli-Israel et al., 2003), with one study reporting a discrepancy of only 12 min (Jean-Louis et al., 1996). Our study in children supports these findings, with discrepancies of only −16 min for the low and −14 min for the auto actigraphy activity thresholds when compared with PSG. In addition, sleep efficiencies were also well correlated for the low and auto activity thresholds as has been previously reported in adults (Kushida et al., 2001) with differences of only −2.4% and −0.3%, respectively.

Our finding of high ARs between actigraphy and PSG on an epoch-by-epoch basis is similar to those reported in both normal adults (90.2%) and children (89.9%) (Sadeh et al., 1989, 1995b), although somewhat lower than a study of 16 adolescents and 20 young adults which reported ARs of 91–93% (Sadeh et al., 1994). Our lower ARs may have been due to our sample of children with SDB, as adults with OSA also have lower AR (Sadeh et al., 1994, 1995b) or to the younger age of our sample population. It is important to note that the studies by Sadeh were conducted using a different brand of Actiwatch which used different algorithms for determining sleep and wake.

Although our study found good agreement between PSG and actigraphy for determining sleep 91.6–94.9%, the PVW between the two measures for determining wake was poor (46.7–65.6%). This finding has also been reported in adults (Blood et al., 1997). We speculate that this discrepancy may be because children sleeping in the unfamiliar environment of the sleep laboratory tend to lie awake without moving as suggested previously (Kushida et al., 2001). In support of this, a recent study in healthy adults has also reported low specificity and suggested that this may be due to the small number of wake epochs compared with epochs of sleep recorded (de Souza et al., 2003).

Our finding that the low and auto activity threshold settings were the best for predicting sleep is also supported from adult studies (Kushida et al., 2001). In contrast, the actigraphy was best at predicting wake at the high and medium activity threshold settings with a discrepancy of only −3 and −22 min, respectively, and this has also been reported for adults (Kushida et al., 2001).

Our study supports the suggestion that actigraphy provides a useful tool for the non-invasive measurement of sleep/wake patterns; however, there are limitations to its use. Actigraphy is unable to differentiate between periods of quiet wakefulness and sleep. Without documentation of sleep and waking provided by other means such as sleep diaries, actigraphy may overestimate sleep. In addition, actigraphy is also prone to artefacts such as external motion, thus causing confusion between sleep and wake (Sadeh et al., 1995b). Despite these limitations, actigraphy is simple to use, is relatively inexpensive and provides little disruption to normal sleep/wake patterns. Actigraphy thus presents researchers with a useful tool for continuous prolonged recordings of sleep patterns in the natural environment.

In conclusion, our study has demonstrated that actigraphy can be used as a reliable indicator of sleep in children when used on low or auto settings. This validation study may lead to actigraphy becoming more widely used in paediatric clinical and research settings.

Acknowledgement

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Subjects
  6. Results
  7. Discussion
  8. Acknowledgement
  9. References

We would like to thank the parents and children who participated in this study.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Subjects
  6. Results
  7. Discussion
  8. Acknowledgement
  9. References
  • Ancoli-Israel, S., Cole, R., Alessi, C., Chambers, M., Moorcroft, W. and Pollak, C. P. The role of actigraphy in the study of sleep and circadian rhythms. Sleep, 2003, 26: 342392.
  • ASDA. EEG arousals scoring rules and examples: a preliminary report from the sleep disorders atlas task force of the American sleep disorders association. Sleep, 1992, 15: 173184.
  • Blood, M. L., Sack, R. L., Percy, D. C. and Pen, J. C. A comparison of sleep detection by wrist actigraphy, behavioral response, and polysomnography. Sleep, 1997, 20: 388395.
  • Jean-Louis, G., Von Gizycki, H., Zizi, F., Fookson, J., Spielman, A., Nunes, J., Fullilove, R. and Taub, H. Determination of sleep and wakefulness with the actigraph data analysis software (adas). Sleep, 1996, 19: 739743.
  • Jean-Louis, G., Kripke, D. F., Cole, R. J., Assmus, J. D. and Langer, R. D. Sleep detection with an accelerometer actigraph: comparisons with polysomnography. Physiol. Behav., 2001, 72: 2128.
  • Katz, E. S., Lutz, J., Black, C. and Marcus, C. L. Pulse transit time as a measure of arousal and respiratory effort in children with sleep-disordered breathing. Pediatr. Res., 2003, 53: 580588.
  • Kushida, C. A., Chang, A., Gadkary, C., Guilleminault, C., Carrillo, O. and Dement, W. C. Comparison of actigraphic, polysomnographic, and subjective assessment of sleep parameters in sleep-disordered patients. Sleep Med., 2001, 2: 389396.
  • McNamara, F., Issa, F. G. and Sullivan, C. E. Arousal pattern following central and obstructive breathing abnormalties in infants and children. J. Appl. Physiol., 1996, 81: 26512657.
  • Rechtschaffen, A. and Kales, A. A Manual of Standardized Terminology, Techniques and Scoring Systems for Sleep Stages of Human Subjects. U.S. Government Printing Office, Washington, DC, 1968.
  • Sadeh, A., Alster, J., Urbach, D. and Lavie, D. Actigraphically based automatic bedtime sleep-wake scoring: validity and clinical applications. J. Amb. Monit., 1989, 2: 209216.
  • Sadeh, A., Sharkey, K. M. and Carskadon, M. A. Activity-based sleep-wake identification: an empirical test of methodological issues. Sleep, 1994, 17: 201207.
  • Sadeh, A., Acebo, C., Seifer, R., Aytur, S. and Carskadon, M. A. Activity-based assessment of sleep-wake patterns during the 1st year of life. Infant Behav. Dev., 1995a, 18: 239337.
  • Sadeh, A., Hauri, P. J., Kripke, D. F. and Lavie, P. The role of actigraphy in the evaluation of sleep disorders. Sleep, 1995b, 18: 288302.
  • So, K., Buckley, P., Adamson, T. M. and Horne, R. S. Actigraphy correctly predicts sleep behavior in infants who are younger than six months, when compared with polysomnography. Pediatr. Res., 2005, 58: 761765.
  • De Souza, L., Benedito-Silva, A. A., Pires, M. L., Poyares, D., Tufik, S. and Calil, H. M. Further validation of actigraphy for sleep studies. Sleep, 2003, 26: 8185.