Clinical PsychologyReview,Vol. 16, No. 5, pp. 19%413.1996 Copyright 0 1996Fkvier Science Ltd Printed in the USA. AU rights resewed 027%7x%/96 $15.00+ .90
NOCTURNAL ACTIVITY AND SLEEP ASSESSMENT Warren W. Tvon Fordham
ABSTRACT. Sl.qb occupies aj#nvxima~y one third ofour lives yetitsmodern study has lag0 lj been w&c&d to the use of the polygraph and to sCf_ via sleep logs. Actigmphy is a method for obtaining objectivebehavioral samples of noctumal as well as diurnal activity. Activity is continuously measuwd and rewnkd in memory ai user +cijicd in&n&s, usual.& 1 minuk, 24-bow-s a day for days and we&s if desired This mvicw wncentrates on studies validatingactigraphy for s&b assessment and briefly addresses applications to DSM-IVsle+ disorders. Sleep onset is a continuous rather than discz&e event. [email protected]
+se Sbg-onrct Spectrum is discussed and actigmphy is validated within this contact.
WE ALL SLEEP and many of us spend approximately one-third of our lives doing so. DSM-IV (APA, 1994) defines 10 sleep disorders open to behavioral evaluation. Many other DSM-IV disorders entail similar sleep disturbances. Sleep is therefore of broad psychological interest. Nearly all that is widely published about sleep derives from sleep laboratories by polysomnographers and other EEG specialists, not psychologists. Evaluation of sleep disorders is rarely conducted by psychologists, and then their inquiry is often limited to self-reports entered retrospectively the next morning into a sleep log. Exceptions include Birrell (1983)) Bonato and Ogilvie (1989), Espie, Lindsay, and Espie (1989), Franklin (1981), Granada and Hammack (1961), Kelley and Lichstein (1980), Perry and Goldwater (1987), Sack, Blood, Percy and Pen (1995), Stickgold and Hobson (1994), and Viens, DeKonick, Van den Bergen, Audt, and Christ (1988), who used some form of behavioral response. Activity measurement is an alternative approach to measuring sleep. Kryger, Roth, and Dement’s (1994) recent 95 chapter second edition of their preeminent text entitled “Principles and Practice of Sleep Medicine” does not mention the term actigraphy in the index. Previous sleep-related reviews published in major psychology journals have mentioned but not adequately discussed actigraphy (Lacks & Morin, 1992; Lichstein & Riedel, 1994). Parts I and II of Ogilvie and Harsh (1994) discuss Correspondence should be addressed to Warren W. Tryon, Department of Psychology, Fordham University, Bronx, NY 104585198. I97
behavioral measures of sleep but not actigraphy. The primary purpose of this review is to evaluate the validity of actigraphy for sleep assessment and to briefly mention DSM-IV applications. Our knowledge of sleep presentlyrestsmainly on one- or two-nightbehavioralsarnples because the cost of obtaining larger samples from extended sleep laboratory study is prohibitive. Little is known about normal and pathological sleep under natural conditions because of the difficultiesand expense associatedwith lab and home polysomnography. Good information about variabilityof normal and pathological sleep is lacking, partly because data collection has been restrictedalmost exclusively to polysomnography (PSG). Behavioral response (see above) has sometimes been used to study sleep, but these methods require activeparticipationby the subject and this raises the question of how consistentlymotivated subjects are to comply with all instructions.It is especially difficult to verify reported motivation. Response to auditory stimuli presumes normal hearing. Actigraphy requires only that the subject wear a small light weight device thereby minimizing compliance problems. Compliance can be verifiedby the recorded data. Issuesof auditoryacuityare avoided.Technological advances allow psychologists to obtain quantitative measurements of activity every minute of the day and night for up to 22 consecutivedays prior to down loading the data to a PC through a serial interface (Tryon, 1991a, pp. 23-60, Tryon & Williams, in press). These data also allow psychologiststo comprehensivelyquantifysleepwake cycles over days, weeks, or months under natural conditions; neither polysomnography or behavioral response devices provide circadian data. Professionalrecognition of the clinical applicabilityof actigmphy to the evaluation of sleep and its disorders has recently been given by the American Sleep Disorders Association through their publication of “Practice Parameters for the Use of Actigmphy in the Clinical Assessment of Sleep Disorders” (Sadeh, Hauri, Rripke, & Lavie, 1995; Standardsof Practice Committee, 1995). Questions about the validityof actigraphyare generallybased on two assumptions: (a) the brain goes to sleep as a dii Creteevent, and (b) one measure of this event is superior to all others and deserves the statusof a “gold standard” by which all other measuresare validated.We question both assumptionson empirical and theoretical grounds. The first section of this article reviewsevidence for sleep onset as a specific sequence of events which we call the SleepOnset Spectrum (SOS). The second section evaluatesthe validityof actigraphy in terms of the SOS. The third and final section briefly considers clinical (DSM-IV) applicationsof actigraphy. SLEEP-ONSET SPECTRUM The theoreticaland empirical data reviewedbelow regardingsleep-onset (SO) strongly indicate that SO is not a discrete event, but entails a series of events occurring in a predictable order (Rechtschaffen, 1994). We are not fully awake one instant and soundly asleep the next. It makes evolutionarysense that people can rest their limbs through relaxation and immobility, and rest their eyes through lid closure, while maintaining auditory contact with the environment. Ogilvie and Wilkinson (1988) refer to a Sleep-Onset Period (SOP) in contrast to sleep-onset as a discrete incident, Tryon (1991a) used the term Sleep-Onset Spectrum (SOS) to emphasize the orderly sequence of SO events, the empirical measures of which are here called mo&tx. Consequently, Sleep-Onset Latency (SOL) varies systematicallywith the selected SO marker. Waking up is much more rapid for most people than is going to sleep. The SOS consists of the following five phases.
Nocturnal Activity and [email protected]
Phase 1: Immobility Transition from normal activity to a period of quiescence and immobility marks the first SOS phase. Activity diminishes from a more active to a less active level prior to going to bed. Activity decreases further once in bed, resulting in a period of protracted immo bility. Wrist and/or waist-worn computerized activity monitors with onboard clocks, known as actigmphs, can conveniently measure and record motility over user defined epochs: often 1 minute but sometimes 30 seconds when measuring only sleep. Table 1 summarizes six studies demonstrating that subjects become inactive prior to evidencing EEG Stage 1 sleep. The only discrepant data, reported by Ham-i and Wisbey (1992)) concern 15 of 36 subjects who satisfied EEG Stage 1 sleep criteria before satisfying actigraphic sleep criteria. The most discrepant subject had severe Periodic Limb Movement disorder and therefore should not have been sleep scored. Eight of the remaining 14 discrepant subjects were diagnosed as Sleep State Misperception. Apparently, thii dis order entails limb movement that continues after EEG Stage 1 sleep begins.
TABLE 1. Studies Reporting that Actigmphic SleeponSet Latency (SOL) I’reads PolyEomnoglaphicSOL Study
Mullaney,Rripke, & Messin (1980)
1. Scoring “errors”largelyentail subjectslying awake motionless. 2. Actigtaphic scored sleep and EEG scored wake occurred three times more often than vice yersa. 3. Act&mph Total Sleep Time (TST) is, on average, 15 minutes greater than EBG TST.
Webster,Rripke, Messin, Mullaney,& Wybomey (1982)
1. EEG sleep onset occurs after all activityceases. 2. The conditional probabilityof misscoringwake as sleep is .062 compared to the conditional probabiliityof misscoringsleep as wake is .039.
Stampi & Broughton (1989)
1. Actigraphsoverestimatesleep time due to quiet wakefulness.
Cole & Rripke (1989)
1. Used Webster et al. (1982) methods for restoring sleep, actuallyquiet wakefulness,as wake.
Cole, Rripke, Gruen, Mullaney,8~ Gillin (1992)
1. Subject’s become inactive a few minutes before
Ham-i & Wisbey (1992)
EEG Sleep Stage 1. 2. EEG wake is misscored as actigraphicsleep 3.5 times as often as EEG sleep is misscoredas actigraphic wake. 3. Restoring rules can reduce the “false sleep” to ‘false wake” ratio from 3.5 to 1 to 2.5 to 1. 1. Actig-raphyover estimatessleep in patientswith insomnia but not Sleep State Misperception (SSM) from 9 to 105 minutes (n = 20, M= 41.4, SD= 23.6 minutes). 2. Actigraphyunder estimatessleep primarilyin patientswith SSM from 1 to 21’7 minutes (a = 15, M = 63.0, SD = 58.5 minutes).
Phase 2: Decreased Muscle Tone The transitionfrom normal to markedlyreduced muscle tone marks the second SOS phase (Carskadon & Dement, 1994; Chase & Morales, 1994), resulting in the drop ping of hand-held objects. This phenomenon has been known for over a half century and served as a “gold standard” for validatingthe emergence of EEG alpha waves as a marker for the onset of Stage 1 sleep (Blake, Gerard, & Rleitman, 1939; Loomis, Harvey, & Hobart, 1937; Perry & Goldwater, 1987; Snyder & Scott, 1972). Ogilvie, Wilkinson, and Allison (1989) measure the drop-point using a hand held ‘deadman” switchrequiring90 grams of pressureto maintainclosure. When muscle tone decreases below this level, the switch opens, marking the droppoint. Perry and Goldwater (1987) accomplish the same end by instructingsubjects to continuously close a telegraph key by maintaining constant extensor tension using the third finger of the preferred hand. Decreased muscle tension associatedwith sleep onset causes the finger to drop and the switch to open. Franklin (1981), Tiyon, Gruen, and Reitman (1995), and Viens et al. (1988) describe variationsof the Ogilvie et al. apparatussuitable for home use. Viens et al. (1988) report average SOL for Stage 1, Stage 2, and a SOL device as 29.0,36.0, and 38.2 minutes, respectively.That the SOL device is more closely associatedwith EEG Stage 2 than Stage 1 sleep may be explained by the pressure required to keep the switch closed. Ogilvie et al. (1989) required 90 grams of force. Neither Franklin nor Viens et al. reported required switchclosure forces. The device described by Tryon et al. (1995) can be calibrated to 90 grams, or small variations around this figure and, therefore, can be tailored for the individualsleeper if desired. Requiringless than 90 grams of pressuremeans that switchclosure can be maintained longer, perhaps until Stage 2 sleep begins. The 9Ogram value is recommended by the fact that EEG Stage 1 sleep wasvalidatedagainstdropping light hand held objects and the 90 gram level is associatedwith the onset of EEG Stage 1 sleep A lesser,and yet to be determined pressure, seems capable of marking Stage 2 sleep-onset. Phase 3: EEG Sleep Stage 1 EEG changes defining Stage 1 sleep (Rechtschaffen 8c Kales, 1968) mark the third phase of the SOS. Birrell (1983) provide a “liberal”and %onServative”EEG definition of sleep onset. The former allows transientawakening to occur after the first occurrence of Stage 1 sleep has been scored and the latter position does not. Hauri and Olmstead (1983) describe three additional EEG SOL criteria: (a) the first epoch scored Stage 2 sleep, (b) the beginning of the first 15 minutes of Stage 2 sleep not interrupted by epochs scored as Stage 1 or as Wake, and (c) the beginning of the first 30 minutes of Stage 2 sleep not interrupted by epochs scored as Stage 1 or as Wake. These criteria for sleep onset span a sufficientlylarge portion of the SOS that the more conservativeones coincide with Phase 4 of the SOS described next. They currentlyserve as the “gold standard” for sleep onset. Table 1 shows that actigraphicsleep onset occurs before EEG Stage 1 sleep onset and, therefore, before Stage 2 and more conservativeEEG SOL measures. Phase 4: Auditory Threshold increase Auditory threshold increasesmark the fourth SOS phase. Birrell (1983) reported that auditory response SOL occurred, on average, 15.6 minutes after “liberal” EEG Stage 1 and 14.9 minutes after “conservative”EEG Stage 1, and 6.3 minutes after EEG Stage 2 sleep onset. During EEG Stage 1 sleep, subjectsrespond when their name is spoken
Nocturnal Activity and [email protected]
softly or if they are touched (Carskadon 8c Dement, 1994). Bonnet and Moore (1982) report that auditory threshold rises rapidly within 1 minute of the first EEG sleep spindle. Auditory threshold increases take place mainly during EEG Stage 2 sleep (Bonato & Ogilvie, 1989). Subjects no longer respond to their name when spoken softly or to a light touch. Subjects no longer respond to normal external stimuli (Lindsley, 1957; Ogilvie & Wilkinson, 1984). Sack et al. (1995) simultaneously compared three SOL markers in three female subjects sleeping at home for two nights: wrist actigraphy, polysomnography, and behavioral-response monitoring (BKM) requiring subjects to close a microswitch within 5 seconds of a tone or light which was presented about once a minute. Each subject underwent five sleep latency trials per night. The graphic results for both conditions showed, with minor exception, that actigraph SOL occurred before PSG SOL which occurred before BKM SOL. Table 2 summarizes five studies reporting auditory threshold and self-report SOL values on the same subjects. These studies consistently demonstrate that auditory threshold increases prior to self-reported SOL. This makes sense because subjects probably base their perception of sleep onset on not being able to hear given that their eyes are shut and lights are off. The average difference of 10.17 minutes is statistically significant (C(9) = 3.53, p < .Ol). The 95% confidence interval for this mean range from 3.66 minutes to 16.68 minutes. Phase
5: Perceived Sleep
Table 2 demonstrates that in all but two instances, self-reported SO occurs after auditory threshold determined SO. One of these two discrepant cases (-0.2 min) is essentially a tie. These data support the conclusion that self-reported SO comes after auditory threshold increase in the SOS. Hauri and Olmstead (1983) cite five studies documenting that insomniacs seriously overestimate SOL when appearance of the first sleep spindle, K-complex, is used to define SOL (Baekeland & Hoy, 1971; Bier, Kales, Leo, & Slye, 1973; Carskadon et al., 1976; Frankel, Coursey, Buchbinder, & Snyder, 1976; Johns, 1975). Ham-i and Olmstead present data on 10 good sleepers (five female, five male, average age = 42 years) showing that subjective SOL occurs within 2 minutes of the first Kcomplex. Data on 56 insomniacs divided into four groups show subjective SOL ranging from 10.4 to 85.4 minutes after the first Kcomplex. These data clearly demonstrate that subjective SOL occurs after the first K-complex in insomniacs. Chambers (1994) calls attention to a lO-minute average difference between actigraphic SOL and sleep log SOL in data collected by Ham-i and Wisbey (1992). Subjective SOL occurred after actigraphic SOL for every one of the 36 insomniacs studied by Harui and Wisbey (1992). Carskadon et al. (1976) compared subjective and EEG SOL in 122 subjects. Fiftynine subjects were patients referred for chronic insomnia. Sixty-three were volunteers, 48 of whom had consulted physicians for sleep problems. The average EEG defined SOL was 26.2 minutes compared to an average self-reported SOL of 61.7 minutes. Sixty subjects overestimated their the EEG SOL by 15 minutes and 15 subjects overestimated the EEG SOL by 1 hour. That relative rank order was maintained despite varying SOS durations is evidenced by a SOL correlation of rho (61) = 65, p < .OOl for men, and rho (57) = 60, p < -001 for women. Other data published by Birrell (1983) confirm the above described SOS. “Liberal” EEG Stage 1 SOL averaged 4.6 minutes, “conservative” EEG Stage 1 SOL averaged
TABLEI 2. SleepOnset
Latency as Determined Self-Report
Auditory SelfNights Threshold Report Difference
F5onato & Ogilvie (1989)
18-39 Mean = 21
1 2 3
27.8 18.3 17.1
44.2 25.6. 22.9
16.4 7.3 5.8
Mean = 43.2
24.3 33.66 23.49
20.7 42.23 28.14
-3.6 10.17 9.11
Espie, Lindsay, &E&pie (1989)
Lichstein, Nickel, Hoelscher, & Kelley (1982)
Lichstein, Hoelscher, E&in, & Nickel (1983)
20-52 Median = 33
11.0 minutes, EEG Stage 2 SOL averaged 19.6 minutes, the auditory sleep monitor SOL averaged 20.2 minutes, and subjective SOL averaged 25.9 minutes. Notice the similarity between EEG Stage 2 and the auditory sleep monitor SOL. SOS FEATURES The following SOS characteristics create the context in which actigmphy will be evaluated. 1. Actigraphic measured immobility usually precedes EEG Stage 1 sleep, which usually precedes markers of auditory threshold increases, which usually precede perceived sleep-onset. 2. The temporal distribution characterizing the SOS makes it inappropriate to expect exact temporal agreement between pairs of SOS markers unless they are equivalent markers of the same theoretical events. Markers of different SOS phases necessarily occur at different times. 3. Sleep-onset markers are closely spaced in time when sleep-onset is rapid and are more widely spaced in time when sleep-onset is slow. SOS duration is defined as the time from the first to last SOS marker.
Nocturnal Activity and Slag
4. SOS tion Less SOS SOS
markers are positively correlated within subjects over time on the assump that they conserve their relative positions within the sleep-onset process. positive correlations may be found across subjects due partly to different lengths. The magnitude of all these correlations is directly proportional to marker variability. Range restriction attenuates correlation. VALIDATION
OF WRIST ACTICRAPHY
Sleep-Onset Latency The literature on the validity of wrist actigraphy developed in the absence of an understanding of the SOS. Consequently, actigtaphy has been primarily evaluated on the basis that it should duplicate EEG Sleep Stage 1 timing. This expectation is inap propriate in light of the SOS and consequently previous reviews have inappropriately evaluated actigraphy against EEG SOL. The primary purpose of this review is to reevaluate actigraphy based on the SOS. Five of the six studies presented in Table 1 consistently demonstrate that motoric immobility precedes EEG Stage 1 sleep consistent with SOS features 1 and 2. Ham-i and Wisbey (1992) indicate that patients with Periodic Limb Movements during Sleep and Sleep State Misperception are exceptions to this rule. The case of PLMS is more understandable in that actigraphy cannot discriminate involuntary from voluntary movements. Suadling, Warley, and Sharpley (1987) reported actigmphic sleep onset to be 0.9 minutes (m = 9) after EEG sleep onset in 12 normals and two sleep disorder patients. Other Sleep Mewres Table 3 identifies 14 pertinent studies and describes their samples in terms of sex, age, and diagnosis. Table 4 reports correlations between polysomnography and actigraphy. Appreciation of this Table is facilitated by a brief reminder that validity is typically established in psychology using correlation coefficients. The data may be considered in light of the validity data employed in evaluating IQ measures. IQ testing is generally recognized as one of psychology’s most valid and widely administered tests. Kaufman (1999) reports WAISR correlations with validity criteria ranging from .45 to .70, averaging in the low 66s. The correlations in Table 4 range from .49 to .98. That polysomnography and actigraphy continuously collect data throughout the night enables one to move beyond the usual correlational basis for demonstrating validity. Table 5 summarizes percent agreement from studies where simultaneous actigraphy and polysomnography data have been collected. The 24 entries for sleep average 92.2% (SD= 5.1%). The 12 entries for wake average 70.6% (SD= 9.8%). The 28 entries for sleepwake discrimination average 87.3% (SD = 11.4%). These data must be evaluated against two important factors. First, actigmphy and EEG are different SOS markers occurring at different points in the SOS process and therefore should not agree 100% of the time. Greater agreement is expected in better sleepers, where SOS is shorter than in problematic sleepers. Second, actigraphy cannot agree more with EEG scoring than EEG scoring agrees with itself. Reliability sets an upper bound on validity. Validity coefficients cannot consistently exceed the square root of reliability coefficients (Gulliksen, 1950, pp. 95-97). PSG Reliability Tryon (1991a, pp. 167-168) disc usses uncertainty associated with EEG scoring. Mullaney, Rripke, and Messin (1986) reported EEG reliability across 10 subjects to be 96.5% on
21-59 Mean = 42 43-72 Mean= 52 20 20-76 Mean = 43 3-13 Mean = 9.1
N= 17 N=7 N-2 N=5 N=4 N=2 N=l N=l M=S M=l,F=3
F=l N= 13
Webster, Rripke, Messin, Mullaney, & Wybomey (1982)
Levine, Moyles, Roehrs, Fortier, & Roth (1986)
Zomer et al. (1987)
Newman, Stampi, Dunham, & Broughton (1988)
Stampi & Broughton (1989)
Sadeh, Alster, Urbach, & Lavie (1989)
N= 16 N= 25
18-26 Mean = 33.6
Mullaney, Rripke, & Messin (1980)
Rripke, Mullaney, Messin, & Wybomey (1978)
Patients Insomnia Sleep apnea
Normals Sleep apnea Insomnia Sleep erection PLMS Paroxysmal awakening
39 Normalsa 63 Patients
TABLE 3. Identification Number, Reference, and Demographic Information for Studies ValidatingWrist ActigraphyAgainst
M=lS,F=23 Mean = 45 N= 13
Hauri & Wisbey (1992)
Sadeh, Sharkey, & Carskadon (1994)
Note.PLMS = periodic limb movements during sleep and DIMS = disorden in maintaining sleep. Yiixteen subjects were recorded twice and one was recorded three times. bActigraphs were placed on the leg rather than the wrist because subjects were infants.
Cole, Eripke, Gruen, Mullaney, & Gillin (1992)
20-25 Mean = 23 lo-16 Mean = 14
Insomnia with mental disorder Psysiologic insomnia Sleep state Misperception Ideopathic insomnia PLMS
15 Normal 3 Elderly normal 12 Psychiatric 4 Sleep apnea 3 DIMS 3 Bereaved 1 Back pain
Allen, Eripke, Poceta, Erman, 8cMitler (1992)
6 Normals 4 Sleep disorder 6 Depressed or schizophrenic 4 Bereaved widows
Sadeh, Lavie, Scher, Tirosh, &Epstein (1991)
Cole & Eripke ( 1989)
TABLE 4. Correlations Between Polysomnographyand Actipphy Derived Estimates of Total Sleep Tie (TST), Sleep, Percent Sleep (96 Sleep), Sleep lZffkiency, and Wake After Sleep Onset (WASO) lD#f 1 2a
r(3) = .98 r(160) = .89 r(37) = .81 r(61) = .97 r(H) = .82 r(12) = .94b
WAS0 r(3) = .85 r(1cM-l) = .70 r(37) = .56 r(67) = .87
r(2) = -96
411) = .72 r(18) = .91 r(19) = .77
r(l8) = .89 r(19) = .82
r(ll) = .91 r(ll) = .81 1114)= .79 r(23) = .63 r(ll) = .56 r(18) = .85 r(19) = .71
r(18) = .49 r(19) = -63
rIhe first entry is for all 102 recordings, the second entry is for 39 recordings, and the third entry is for 63 nonpatient recordings. bOne subject lay motionless in bed for SO minutes.Excluding this subject increases the Total Sleep Time correlation between PSG and Actigraphy to that reported. CEnuies are for normals, children, insomnia, and apnea dEntrles are for training and validation samples.
average. Reliability coefficients differed as a function of FXG parameter. They ranged from a high of r(S) = .999 for total sleep period to a low of r(8) = .899 for number of midsleep awakenings. OgiIvie and Wilkinson (1988) reported that interrater epoch-tmpoch EEG sleep scoring agreement values range from 80% to 98%. Cole, Kripke, Gruen, Muhaney, and Gillin (1992) reported PSG reliability of 94.19%. Spiegel (1981, p. 62) reported that Stage 1 sleep is scored with only 60% reliibility whereas Stage 2 sleep is scored with 90% reliability. Percent agreement can be converted into phi, a special case of the Pearson correlation coefficient r. The 80-98% PSG agreement implies reliabilityphi coefficients of from 64 to .96 and the square root of these values implies maximum validitycoefficients of from .80 to .98. The averageTable 4 validitycoefficient of .79 essentiallyties the lower bound of this interval. Spiegel’s (1981, p. 62) lower PSG estimate of 60% agreement implies a reliabilityphi coefficient of 36 and a lower estimate of the maximum validitycoefficient of 60. The entire .49 to .98 range of Table 4 validitycoefficients lies above this lower validity bound. Taken together, actigraphy corresponds about as closely with polysomnogmphy as psychometricprinciples allow.
Absolute Comparisons The greatest level of scrutiny entails comparing numerical estimates of PSG and actigraphic estimated total sleep time, percent sleep, sleep efftciency, wake after sleep onset, and minutes of sleep. Table 6 presents these results. Again, it is important to remember that absolute agreement should not occur given that we are comparing two different SOS markers. In good sleepers, these markers can, and often are, only a few
TABLE 5. Percent Agreement Between Polysomnogmpb=dANV=PlV Derived Estimatea of Total Sleep Time (T!3T), Sleep, Wake, and SleepWake S--h3 ID#
All 3s = 94.5 NonPts = 96.3 Pts= 91.6 96.0 93.0
3 4 5 7a
Exp2=94.5 Exp 3 = 93.9 92.6 90.0 99.7 78.8 96.1 95.5 88.3 92.9 95.4 92.1 91.7 88.9 90.7 79.6 88.5 87.7
63.5 76.2 66.0 48.5 56.5
96.4 97.9 94.6 95.8 96.4 94.8
78.6 74.3 74.5 79.8 75.4 76.5
85.3 91.9 88.0 88.3 85.3 89.8 89.2 84.0 83.9 83.1 89.7 59.2 91.8 87.9 88.3 41.3to 96.8 Mean = 82.1 92.8 92.6 91.2 92.5 92.5 91.4
aEntriesare for baseline, ultrashortsleep periods, and recoverysleep. bEntries are for calibration normals, validation normals, child patients,insomnia patients, and apnea patients. ‘Xntries are for normal, sleep disorder, psychiatric,bereaved, and all subjects. dEntriesare for normal, elderly normal, psychiatric,sleep apnea DIMS (diiculty in maintaining sleep), back pain and all diagnoses. The first entry is for the training sample, first half of the subjects,and the second entry is for the validation sample, second half of the subjects in each categoty. c41.3% agreement is for a patient with severe PLMS, 96.8% is for a patientwith physiologic insomnia. fEntriesare for 10 calibrationadults. 10validationadults, and 16validationadolescentsfor the nondominant wristfollowed by the dominant wrist.
TABLE 6. Absolute Values for Sleep Time (ii Minutes), Percent Sleep (96 Sleep), Sleep Efikiency, and Wake After Sleep Onset (W-O), for Polysomnography (PSG) and Actigraphy (Act) for Studies idemtifiedin Table 4 (Iw Sleep (min) PSG
434 348 442 462 502
410 345 439 453 473
396 383 370
399 486 385
% Sleep PSG
Sleep Eff PSG
PSG ACT 9454 4 10 2
87.1 88.9 63.3 82.2
loo 46 1 55 23
85.7 86.5 78.6 83.5
313 339 373 369 325 344c
341 364 337 278 238 3llC
aEntriesare for baseline, ultrashort sleep, and recovery sleep. bEntries are for normals, children, insomnia, and apnea. %I1 36 subjects.
apart. Considerably greater differences are found in poor sleepen. Mullaney et al. (1986) indicate that an average difference of 15 minutes is not unusual. None of the differences in Table 6 appear to be detrimental for either clinical or research applications. None of these differences is enough to mistake a poor sleeper for a good one or vice versa. The willingness of clinicians to accept self-reported information and to use brief screening tests with low reliability and validity coefficients indicates a tolerance for considerably greater measurement imprecision than that reflected in Table 6. Research applications of tests and self-report instruments are often based on correlation coeffkients of approximately 3. Given a reliability coefficient of .8 and a validity coeffkient of 3, valid variance is but .09/.64 = 14% of what the test reliably measures. Research investigators rarely express concern that the other 86% of what minutes
Noctumal Activity and Slqb
their test reliably measures may interact with, and perhaps confound, measurement of valid variance. Such measurement imprecision is considerably greater than that shown in Table 6. PROFESSIONAL
The Standards of Practice Committee (1995) of the American Sleep Disorders Association recommends actigraphy as a [email protected]
&&nct for the diagnosis and treatment of: (a) insomnia, (b) circadian+hythm disorders, and (c) excessive sleepiness under specific conditions set forth on pages 286-287. A minimum behavioral sample consisting of at least three consecutive 24hour periods was recommended. These conclusions and recommendations, while positive, were not informed by the SOS, but presumed sleep onset to be discrete and best measured by EEG. Consequently, they are unnecessadly restrictive. APPLICATION
TO DSM-IV SLEEP DISORDERS
The ability of actigraphy to detect awakening has never been questionned. This ability coupled with its sleep-onset perspective makes actigraphy more informative regarding DSM-IV sleep disorders than has previously been recognized. Tiyon (in press) covers these applications in greater detail than space permits here. Some of the more pertinent points are summarized next. Dyssomnias include Primary Insomnia (307.42)) which entails difficulty initiating or maintaining sleep. Actigraphy can document the time of each awakening and therefore allows one to calculate the interval between awakenings, as well as determine their fiequency and distribution throughout the sleep period. Primary Hypetsomnia (307.44) entails difficulty awakening and daytime naps of an hour or more. Actigraphy can document the duration of the major nocturnal sleep period plus extended daytime nap ping. Narcolepsy (347) is characterized by recurrent lo-66 minute sleep attacks every several hours. Actigmphy data can be computer sleep scored every minute of each 24 hour period to identify the frequency and duration of daytime sleep periods. Diary entries and clinical interviews can distinguish voluntary from involuntary sleep periods. Sleep-Wake Circadian Rhythms, also called Circadian Rhythm Sleep Disorder (CRSD) (307.45; p. 578)) require documenting sleepwake patterns. Actigraphy excels in its ability to continuously track sleepwake periods over days and weeks (Gruen, 1987), thereby making it arguably the preferred method of documenting CRSD. Activity is routinely monitored in the labomtory to study circadian rhythm because it is theoretically second only to core body temperature for this purpose (Hanington, Rusak, & Mistlberger, 1994; Kryger, Roth, & Carskadon, 1994; Richardson, 1994). Tryon (1991a, pp. 68-71) reviews earlier actigraphic studies of circadian rhythm. Recent actigraph studies include 50 healthy infants aged 9-24 months (Sadeh, 1994), 11 healthy infants aged 12-48 months and 63 sleep disturbed infants from 9-24 months (Sadeh, Lavie, Scher, Tirosh, 8c Epstein, 1991), 40 hospitalized children aged 3-8 years (Williams, White, Powell, Alexander, & Conlon, 1988), 23 normal adults aged 22-54 years (Brown, Smolensky, D’Alonzo, & Redman, 1990), 11 normal adults aged 60-83 years (Mason 8c Tapp, 1992), 43 healthy adults aged 21-83 years (Renfrew, Pettigrew, & Rapoport, 1987), 29 elderly insomniacs and 22 elderly controls (Polk&, Perlick, & Linsner, 1992), 12 persons with Alzheimer’s Disease aged 71-86 (Witting, Kwa, Eikelenboom, Mirmimn, & Swaab, 1990), 26 depressed inpatients aged 44 years (Raoux et al., 1994), and 25 post coronary artery bypass surgery patients 64 years of age (Redeker, Mason, Wykpisz, Glica, & Miner, 1994). The point of these citations is to document that actigraphy is rather widely used to study human circadian rhythms.
Parasomnias include Nightmare Disorder (S07.47) and Sleep Terror Disorder (397.48)) both of which entail sudden waking. Actigraphy can clearly document when sudden waking occurs. Sleepwalking Disorder (S07.46) entails the performance of motor behavior during sleep, including rising from bed and walking about. Wrist and waist actigraphy can be very informative regarding such behaviors. Other DSM-l?r Sleep Disorders include Insomnia (887.42) and Hypersomnia (807.44) “Belated to Another Mental Disorder” other than Primary Insomnia or Hypersomnia and sleep disorders due to a General Medical Condition (780.~~). Actigraphy therefore informs us about Major Depressive Episodes, Manic Episodes, Hypomanic Episodes, Mixed Episodes, Bipolar I and II, Melancholia, ADHD, Generalized Anxiety Disorder, PTSD, Catatonic Schizophrenia, and Schhoaffective Disorder (Tryon, 1986, 1991a, 1991b, in press). Substance-induced sleep disorders can also be detected using actigraphy (e.g., Tryon, 1991a, pp. 197-20’7). Actigmphy can evalute several aspects of sleep hygiene (Ham-i, 1991; Zamone, 1994). First, it can determine the regularity with which clients attempt sleep. The time at which they begin quiet repose prior to sleep-onset can easily be determined with a temporal accuracy of 1 minute from standard act&mph data Second, actigraphy can identify the times clients awake. Third actigraphy can determine the number of minutes spent napping during the morning, afternoon, and/or evening when the person is expected to be awake. Compliance with wearing actigraphs can be checked by examining the record for consecutive epochs of zero activity. A&graph removal leaves a very obvious signature in that many consecutive epochs of zero wrist activity are very rare in persons who are awake but always present when actigraphs are removed. A&graph removal can easily be spotted from activity plots. Exercise may sometimes be prescribed to facilitate sleep at night and/or to improve health generally. Actigraphy can document the duration and extent to which activity is taken and the time of day activity begins and ends. CONCLUSIONS Sleep onset is a gradual process characterized by at least five phases. Because actigraphy marks Phase 1 and se&eported sleep onset marks Phase 5, these two convenient indices can be used to estimate SOS duration. Good sleepers are known to have a short SOS duration and poor sleepers a long SOS duration. Future research and clinical study should be directed toward methods for reducing SOS duration in poor sleepers. The ability of actigraphy to document waking as well as sleeping makes it informative when evaluating a variety of DSM-IV sleep disorders. Actigraphy is especially well suited for the quantification and assessment of Circadian Rhythm Sleep Disorders. Sadeh, Sharkey, and Carskadon (1994) identify between device variability and the possibility of introducing breathing artifact when the hand rests on the chest while asleep. These problems are minimized by the Actillume (Ambulatory Monitoring, Inc., Ardsley, NY) and a new actigraph described by Tryon and Williams (in press). Both devices quantity activity into 128 levels thereby insuring that small readings are always associated with minor movements such as respiration. Dynamic spinning calibration of the new a&graph makes them highly comparable and therefore interchangeable. REFERENCES Allen, T. G., Kripke,D. F.. Pocera. J. S., Em-tan, hf. K, SCMider, M. hi. (1992). Actillume experience in the sleep disorders clinic. Sbcp &scat& 21. 330. American Psychiatric Association. (1994). LXagmshc and sfatisticat manual of men&d diudm (4th ed.). Washington DC: Author.
Nocturnal Activity and Slqb
Baekeland, F.,& Hoy, P (1971). Reported vs. recorded skp ch~teristi~. Atchiw.~ of cnnol Alclrioh): 24, 548-551. Bimll, P. C. (1983). Behavioral, subjective,and electroencephalograph indices of sleep onset latency and sleep duration.Jownal ofBchat&ml Aswsment,5,179-196. Bixler, E. O., Kales, A, Leo, L A., & Slye, T. (1973). A comparison of subjective estimatesand objective sleep laboratory finding in insomniac patients.sbcp w 2.143. Blake, H., Gerard, R. W., SeRleitman. N. (1939). Factorsinfluencing brain potentialsduring sleep.Jound
Bonato, R A, & ogilvie, R. D. (1989). A home evaluationof a behavioral response measure of sleepwake fulness. Pcnxptualand Motor&l& 68.87-96. Bonnet, M. H.. & Moore, S. E. (1982). The threshold of sleep: Perception of sleep as a function of time asleep and auditory thresbold. S&+, 5,267-276. Brown, A. C., Smolensky,M. H.. D’Alonzo, G. E., St Redman, D. P. (1996). ActigtaphyzA means of assessing circadian patterns in human activity.awn&u& Inbmotiond. 7,125-133. Car&don, M. A., SeDement, W. C. (1994). Normal human sleep:An overview,ln hi. H. Rq~er, T. Roth, & W. C Dement (Eds.), Aincjplcsandpm&c ofsir+ me&c& (2nd ed.. pp. 1625). Dmdon: W. B. Smmden Co. Camkadon, M. A, Dement, W. C., Mitler, hf. hi., GuiBeminault,C., Zarcone, I? P., SCSpiegel, R (1976). Selfreports versus sleep laboratory findings in 122 drug-free subjects with complaints of chronic insomnia. Anlaian~ lq+chiany, 12 1382-1388. Chambeta, M. J. (1994). Actigtaphy and insomnia: A closer look Part 1. sbcp 17,465-468. Chase. M. H.. & Morales, F. R (1994). Tbe control of motoneurons during deep. In M. H. Rryger.T. Roth, & W. C. Dement (Eds.), Z%inci+ and pmcticcof sbcp medick (2nd cd.. pp. 163-175). London: W. B. 8aunders Co. Cole, R J.. SCRtipke. D. F. (1989). F%gress in automaticsleepwakescoring by wrist a&graph. Slqb Racamch, IS. 331. Cole, R J., Rtipke. D. F., Gruen. W., Mullaney, D. J., & Gillin, J. C. (1992). Automatic sleepwake identifica tion from wrist activity.sbp Z5,461-469. Espie. C. A., Lindsay,W. R, Espie, L C (1989). Use of the sleep assessmentdevice (Rell9 and Lichstebr, 1986) tovalidate insomniacs self+eport of sleep pattem.&mbal o~+hoF&o& and B&wiomlAuannny II. 71-79. Ftankel, B. L, Coursey, R I).. Buchbinder, R. SCSnyder.F. (1976). Recorded and reported sleep in chronic primary inwmnia. Ati ofGnanlI%y&any, 33,615-623_ Franklin.J. (1981). The measurement of sleep onset latency in insomnia. lk&mimu &rat& and 7&m&, 19.547-549. Granada,kM.,&H ammack,J. T. (1961). Operant behavior during sleep. S&ncc, 133,1485-1486. Gnren, W. (1987). Wrist-a&graph measurements of sleepwake pattema during a 28&y Westerly trip around world. Fij?hIntanroriad Congns of Siyb lbntrh (Poster 496), p. 691. Gulliksen, H. (1956). Theay ofmcnrolt&s. New York Wi19. Hanington, M. E., Rusak, B., SCMistlberger, R. E. (1994). Anatomy and physiologyof the mammalian cir~system.InM.H.Kryger,T.Roth.&W.C.Dement(Eds.),~andpmdiacof~~(2nd ed.. pp. 286-366). London: W. B. Saunders Co. Ham-i, P.J. (1991). sleep hygiene, relaxation therapy,and cognitive interwntions. In P.J. Hauri (Ed.), Case studiain insomnia(pp. 66-84). New Yorlr: Plenum. Hauri. P., & Ohnstead, E. (1983). What ia the moment of sleep onset for insomniacs?St&p,6,16-15. Hauri, P.J.. & Wubey, J. (1992). Wrist actigtaphy in insomnia. Sk+, 15.293-361. Johns, M. W. (1975). Preliminary communication: Factor analysisof objective and subjectivechamcteristics of a night’s sleep. [email protected]
MulGnc, 5,413-418. ffiufinan, A. s. (1996). Assess& adokxcnf and a&t indigmz (pp. l-29). Boston: ARyn and Bacon, Inc. Rell9, J. E., SCLichstein, R. L. (1986). A sleep assessmentdevice. B&uioml~ 2,135-146. Eripke, D. F., Mulian9, D. J., Messin. 8, & Wybomey, V. G. (1978). Wrist actigraphic measuresof sleep and rbytbms. E&bwnc+~~ and Uinicol Newybh~, 44,674-676. Rryger, M. H., Roth, T.. & Carskadon, M. (1994). Circadian rhythms in humans: An overview. In M. H. Rryger. T. Roth, SCW. C. Dement (Eds.), &%a& and @o&ice O/SIC+wd&te (2nd ed.. pp. 361-368). London: W. B. Saunders Co. Eryger,M. H., Roth, T., & Dement, W. C. (1994). Efinciplcr undpncticcofsbepWI&% (2nd ed.). Philadelphia: W. B. Saunders. Lacks, P., & Morin, C. M. (1992). Recent advances in the assessmentand treatmentof insomnia.Jmrnd of Gmwlting and ClinicalRychologr,60.586-594. Levine, B., Moyies, T., Roehrs, T., Fortier,J.. & Roth, R. (1986). Actigraphic monitoring and polygraphic recording in determination of sleep and wake. S&J Rwarrh, 15,247.
212 Lichstein, K L., Hoehcher.
U!U!T~ T. J., Eakin, T. L, & Nickel, R (1983). Empirical sleep assessment
A convenient, inexpensive approach. Joumal of lkhmkd A.swsm& 5,l l l-l 18. Lichstein, K L., Nickel, R, Hoelscher. T. J.. & Eelley, J. E. (1982). Clinical validation
in the home:
of a sleep assessment
device. B&ar&ur Rewwrh and Thou& 20,292-298. Lichstein, K L, & l&&1, B. W. (1994). Behavioral assessment and treatment of insomnia: A review with an emphasis on clinical application. S&zuior [email protected]
, 2% 659-688. Lindsley, 0. R (1957). Operant behavior during sleep: A measure of depth of sleep. S&rue, 126,1290-1291. Loomis, A. L, Harvey, E. N., & Hobart, G. A. (1937). Cerebral states during sleep as studied by human brain potentiab.&wal ofE+im&ul Pqcho&y, 21,127-144. Mason, D. J., & Tapp. W. (1992). Measuring circadian rhythms: Actigraph versus activation checkliit. Wairm Joumal ofNu&ngResauch, 14, S58-S79. Mullaney, D. J., Kripke, D. F., & Messin, S. (1980). Wtistactigraphic estimation of sleep time. s&p. 3,8S-92. Newman, J., Stampi, C.. Dunham, D. W., & Broughton. R (1988). Does wrist-actigraphy approximate tradetional polysomnogmphic detection of sleep and wakefulness in narcolepsycatapelxy? sbap Resee&, 17,343. Ogilvie, R D., & Harsh, J. R (1994). sbcp ens& Nonaal and [email protected]
Washington, DC: American Psychological Association. Ogiltie, R D., & Wilkinson, R T. (1984). The detection of sleep onset: Behavioral and physiological convergence. w, 21.510-520. Ogilvie, R D.. & Wilkinson, R T (1988). Behavioral versus EEGbased monitoring of all-night sleep-wake patterns. srcrp II. 139-155. Ogihrie, R D., Wdkinson, R T., & Allison, S. (1989). The detection of sleeponset: Behavioral, physiological, and subjective convergence. Sf+, 12458-474. Perry, T. J., & Goldwater, B. C (1987). A passive behavioral measure of sleep onset in high-alpha and lowalpha subjects. Pqcm, 24.657-665. Pollak, C. P., Perlick, D., & Linsner. J. P. (1992). Daily sleep reports and circadian rest-activity cycles of elderly community residents with insomnia. Bioiqkaf Rychhy, 32,1019-1027. Raoux. N.. Benoit, 0.. Dantchev, N., Denise, P., Franc, B., Alliire, J. F., SCWidlocher, D. (1994). Circadian pattern of motor activity in major depressed patients undergoing antidepressan t therapy. Belationship &runt%, 52 85-98. between actigraphic meamres and clinical course. eu Rechtschaffen, A (1994). Sleep onset: Conceptual issues. In R D. Ogilvie &J. R Harsh &is.), .%$ our& plwava (pp. 8-17). Washington, DC: American Psychological Associion. Nouaaiandabnomml Rechtschatfen, A, & Eales. A. (1968). A man& of standanf temtinolqg, techniques and swing spemjiw slap states of human subjsts (NIH Pub #204). Washington, DC Superintendent of Documents, Book 162 or Los Angeles: UCLA, Brain Information ServiceBrain Research Institute, University of Caliiomia at Los Angeles. Redeker, N. S.. Mason, D. J., Wykpisz, E., Glica, B., & Miner, C. (1994). First postoperative week activity patterns and recovery in women after coronary artery bypass surgery. Nun&q &war+ 43,168-l 73. Renfiew, J. W., Pettigrew, K D.. & Bapoport, S. 1. (1987). Motor activity and sleep duration as a function of age in healthy men. Physiokgy and Behavk, 41,627-634. Richardson, G. (1994). Circadian rhythms in mammals: Formal properties and environmental influences. In M. H. Etyger, T. Roth, 8c W. C. Dement (Eds.), Efinn’pla and pnrdicc of sbcp snedicinc (2nd cd.. pp. 277-285). London: W. B. Saunden Co. Sack, R L., Blood, M. L.. Percy, D. L., & Pen, J. C. (1995). A comparison of sleep onset latency measured by PSC, actigraphy and behavioral response. Slqb Rcruw& 24,540. Sadeh, A. (1994). Assessment of intervention for infant night wakening: Parental reports and activitybased home monitoring. Joumal of Gmsulting and Clinical Psychology, 62,6!3-66. Sadeh, A., Ahter, J.. Urbach, D.. & Iavie, P (1989). Actigraphically based automatic bedtime sleepwake scoring Validity and clinical applications. Journal of Anhuiaby Monitoring 2,209-216. Sadeh, A., Hauti, P J.. Kipke, D. F.. & Lavie, P (1995). The role of actigraphy in the evaluation of sleep disorders. Sk+, IS, 288-802. Sadeh, A., L&e. P., Scher, A., Tirosh, E., & Epstein, R. (1991). Actigraphic home-monitoring sleep&sturbed and control infants and young children: A new method for pediatric assessment of sleep-wake patterns. Pediahiq 87.494-499. Sadeh, A.; Sharkey, K, & Car&don, M. A. (1994). Activity-based sleepwake identification: An empirical test of methodological issues. [email protected]
, 17.201-207. Snyder, F., & Scott, J. (1972). The psychology of sleep. In N. S. Greenfield & R. A Stemback (Eds.), Handbook of p+op+iofqy (pp. 645-708). Toronto: Halt, Rinehart and Winston. Spiegel, R. (1981). slcg and s~ness in advanced age New York: SP Medical & Scientific Books. Stampi, C.. & Broughton, R. (1989). Ultrashort sleepwake schedule: Detection of sleep state through wrist actigraph measures. Slqb Rescatrh, 18,100.
Nocturnal Activity and Slqb
Standards of Practice Committee. (1995). Practice parameters for the use of actigtaphy in the clinical assessment of sleep disorder. sleep 18.285-287. Stickgold, R., & Hobson, J. A (1994). Home monitoring of sleep onset and sleeponset mentation using the Nightcap (tm). In IL D. Ogilvie &J. R Harsh (Fds.), slog onset:Normal end a6nem&jmxxscs (pp. 141-160). Washington, DC American Psychological Association. Stradling, J., WarIey, A., & Sharpley, A. (1987). Wrist actigraphic assessment of sleep. &?/I Intemational GmgmssofSle$ Restmrh Abstracts, (Poster 521), p. 672. Tryon, W. W. (1986). Motor activity measurements and DSM-III. In M. Hemen, R M. EisIer, & P. Miller (Eds.), Ifognss in &enicr rnodij&iou (Vol. 20, pp. 35-66). San Diego, CA: Academic Publishers. Tryon, W. W. (1991a). Actiuify mearumnmt in @ychdogl and m&&zinc.New York: Plenum. Tryon, W. W. (1991b). Motoric assessment and DSM-III-R In M. Hemen & S. M. Turner (Fds.), AduUpy &pathology and diugnh (pp. 413-440). New York: Wiley. Ttyon, W. W. (in press). Motor activity and DSM-IV. In S. Turner 8c M. Hetsen (Eds.), Adub fa$wpt~ and diqnask (3rd ed., pp. 413-440). New York Wiley. Tryon, W. W., Gruen, W., & Reitman, M. (1995). A simple handheld device for measuring sleeponset Iatency. Manuscript in preparation. Tryon, W. W., & Williams. R (in press). Fully proportional actigmphy: A new instrument. E&avior Rcrmnh MiiholbImhumentsandcompurcrs. views, M., DeEoninck, J., Van den Bergen, R, Audet, R, & Christ, G. (1988). A refined switch+tctivated time monitor for the measurement of sleeponset latency. Ii&uiour Rcsacrnhand Tw, 24 Ml-273 Webster, J. B., Bripke. D. F., Messin, S., MuIlaney, D. J.. & Wybomey, G. (1982). Au activitybased sleep monitor system for ambulatory use...%$I, 5,389-399. Wdliams, P. D.. White, M. A., Powell, G. M., Alexander, D. J.. & Conlon, M. (1986). Activity level in hospitalized children during sleep onset latency. Cosrpulcnia Nuking, 6,70-76. Witting, W., Kwa, I. H., Eikelenboom, P., Mh-mitan, M., & Swaab, D. F. (1990). Alterations in the circadian restsctivity rhythm in aging and Alzheimer’s Disease. [email protected]
Rychiatry, 27,563-572. Zarcone. V. P., Jr. (1994). Sleep hygiene. In M. H. Eryger, T. Roth, & W. C. Dement (Eds.), Ainciplcs und jnmtiu of slqb atedick (2nd ed., pp. 542-546). London: W. B. Saunders Co. Zomer, J., Pollack, I., Tzischinsky, 0.. Epstein, R. Ulster,J., LkLavie. P. (1987). Computerized assessment of sleep time by wrist actigtaph. [email protected]
Inten& Ckmgmss of Sl+ Resmmh (Poster 528), p. 528.