Journal of the Formosan Medical Association (2013) 112, 545e553

Available online at www.sciencedirect.com

journal homepage: www.jfma-onl ine.com

ORIGINAL ARTICLE

Trajectories of sleep changes during theacute phase of traumatic brain injury: A7-day actigraphy study

Hsiao-Yean Chiu a, Pin-Yuan Chen b,c, Ning-Hung Chen d,e,Li-Pang Chuang c,d,e, Pei-Shan Tsai a,f,g,*

aGraduate Institute of Nursing, College of Nursing, Taipei Medical University, Taipei, TaiwanbDepartment of Neurosurgery, Chang Gung Memorial Hospital, Taoyuan, TaiwancGraduate Institute of Clinical Medical Sciences, College of Medicine, Chang-Gung University,Taoyuan, Taiwand Sleep Center, Chang Gung Memorial Hospital, Taoyuan, TaiwaneDepartment of Pulmonary and Critical Care Medicine, Chang Gung Memorial Hospital,Taoyuan, Taiwanf Sleep Science Center, Taipei Medical University Hospital, Taipei, TaiwangDepartment of Nursing, Wan Fang Hospital, Taipei Medical University, Taipei, Taiwan

Received 17 October 2012; received in revised form 4 June 2013; accepted 6 June 2013

KEYWORDShierarchical linearmodel;

sleep;symptom trajectory;traumatic brain injury

* Corresponding author. Graduate InE-mail address: ptsai@tmu.edu.tw

0929-6646/$ - see front matter Copyrhttp://dx.doi.org/10.1016/j.jfma.201

Background/Purpose: To examine trajectories of change in sleep during the acute phaseof traumatic brain injury (TBI), and whether specific demographic and disease character-istics predicted the initial levels of sleep and the trajectories of change in sleepparameters.Methods: This was a prospective observational study. Fifty-two patients with first-ever TBIwere enrolled. Sleep parameters were measured using actigraphy for 7 consecutive daysafter admission. Hierarchical linear modeling was used for data analyses in 52 TBI patientsand in a subgroup of mild TBI patients (n Z 31).Results: Participants had significant lower sleep efficiency, longer wake time after sleeponset, and longer 24-hour total sleep time (TST) than the normative data (all p < 0.05).Seventy-two percent of participants experienced prolonged 24-hour TST. Both daytimeand 24-hour TST showed a significant downward trend across the study period. An initialGlasgow Coma Scale score < 11 significantly predicted the slope of change of daytimeTST over time. Without initial loss of consciousness and age < 40 years were independentpredictors of the change pattern of 24-hour TST over time. In the mild TBI subgroup, 24-

stitute of Nursing, Taipei Medical University, 250 Wu-Hsing Street, Taipei 110, Taiwan.(P.-S. Tsai).

ight ª 2013, Elsevier Taiwan LLC & Formosan Medical Association. All rights reserved.3.06.007

546 H.-Y. Chiu et al.

hour TST significantly and gradually declined over time. Gender significantly predicted thetrajectory of 24-hour sleep duration.Conclusion: Poor sleep efficiency and longer sleep duration are common symptoms in acuteTBI patients. Both head injury severity and age predicted the trajectories of daytime and24-hour sleep duration during the acute phase of TBI, whereas gender predicted the tra-jectories of 24-hour sleep duration in the mild TBI subgroup.Copyright ª 2013, Elsevier Taiwan LLC & Formosan Medical Association. All rights reserved.

Introduction

Traumatic brain injury (TBI) is an important public healthproblem, and the global incidence rate of TBI is estimatedto be 200/100,000/year.1 Survivors of TBI often suffer fromlong-term physical and psychological problems which inturn may affect a person’s ability to return to work andresumption of daily activities.2,3 Posttraumatic hyper-somnia (PTH) and insomnia are common sleep complaintsreported by chronic and post-acute TBI survivors, withprevalence rates of 20-90% and 30-70%.4�8 Sleep complaintsreported by TBI survivors have been associated with pro-longed rehabilitation unit stay,9 heightened depression andanxiety symptoms,4 and reduced health-related quality oflife.8

Most of the available research has focused on sleepcomplaints of TBI patients during the chronic rehabilitationstage.4�8 A small number of studies reported that disturbednighttime sleep and excessive daytime sleepiness (EDS)may occur in the subacute stage as well.9,10 However, littleis known regarding the change in sleep patterns during theperiod immediately following TBI. Most importantly, paststudies mainly used a cross-sectional approach to investi-gate this matter. The trajectories of TBI-associated changein sleep patterns during the period immediately followingTBI remain uninvestigated. Therefore, the current studywas designed to examine trajectories of change in sleepparameters in individuals with TBI during the periodimmediately following trauma. Age, body mass index (BMI),pain at night, severity of head injury, depression, andanxiety symptoms have been identified as significant pre-dictors of sleep disturbance in TBI survivors during thesubacute or the rehabilitation period in earlier stud-ies.5�7,10�13 Nonetheless, predictors of the trajectories ofchange in sleep patterns during the early phase of TBI havenot yet been explored.

Sleep plays an essential role in the recovery of physicalfunctioning during the period immediately after TBI.Knowledge of the trajectories of change in sleep after TBIand their predictors may provide insights into the devel-opment of effective interventions for this patient popula-tion. Therefore, the purpose of this study was to examinethe following: (1) different trajectories of change in sleepparameters during the acute phase of TBI, and (2) whetherspecific demographic and disease characteristics predictedthe initial levels of sleep and the trajectories of change insleep parameters. We hypothesized that the manifestationsof change in sleep are diverse and are affected by theseverity of head injury, and may decline over time duringthe acute phase after TBI.

Methods

Participants

This study was a prospective observational study. Partici-pants were recruited from three neurological wards of a3000-bed hospital located in northern Taiwan. Patientswere eligible to participate if they were between 18 yearsand 65 years of age, newly diagnosed with a computedtomography (CT)-proven first-time closed brain injury, andadmitted to the neurosurgical ward. Patients wereexcluded if they were shift workers, had a previous TBI, ahistory of psychiatric disease, sleep disturbance, or alcoholabuse prior to TBI, suffered from traumatic injuries to otherparts in addition to TBI, or were admitted to the intensivecare unit (ICU) immediately following TBI.

Measurements

Objective data of sleep parameters were obtained by actig-raphy using the quantitative ActiGraph (ActiGraph, Pensa-cola, FL, USA), a watch-like accelerometer that candifferentiatebetween sleepandwakingbasedon theamountof movement. In brief, for each 60-second epoch interval,data samples taken from theaccelerometer inside thedeviceat a rate of 30Hzwerefirst filtered then accumulated prior tobeing stored in memory. Data were analyzed using ActiLifesoftware (ActiGraph, version 5.3.0). Automatic scoring ofsleep was performed on actigraphical data from each epoch,using the Cole-Kripke algorithm14 to determine minute-by-minute asleep/awake status. Actigraphic measurements ofsleep are comparable to those of PSG,15,16 and are suitable tostudy the sleep patterns of TBI patients.17,18 In order todetermine daytime and nighttime sleep duration, a 7-daysleep diary was used to facilitate actigraphic data analysis.

Sleep variables

Nighttime sleep variables including nighttime total sleeptime (TST), sleep onset latency (SOL), sleep efficiency (SE),and wake time after sleep onset (WASO) were investigatedin the current study. Nighttime TST referred to the amountof actual sleep time that occurred in the nighttime perioddefined by actigraphy and the sleep diary. SE was the ratioof TST to the total time spent in bed. The total time spentin bed was estimated by using the bedtime and wake timerecorded by primary caregivers or participants. SOL wasdefined as the period from bedtime indicated by the pri-mary caregivers or participants to the beginning of sleep.

Sleep trajectories in acute traumatic brain injury 547

WASO quantified the time spent awake after falling into apersistent sleep state. The normal range of nighttime TST,SE, SOL, and WASO for healthy individuals (aged 18-65years) are 375-450 minutes, 83-95%, 16-18 minutes, and 17-50 minutes, respectively.19 Daytime TST referred to theamount of actual sleep time in a sleep period that occurredin the daytime period defined by actigraphy and the sleepdiary. Post traumatic hypersomnia (PTH) is a unique andcommon symptom among TBI patients and referred to anincrease in sleep duration of more than 2 hours per 24 hourswhen compared to that prior to TBI.8,20 It could be char-acterized by an extended nighttime sleep period and day-time sleepiness. Therefore, PTH was operationally definedas an average 24-hour TST (sum of nighttime TST and day-time TST) of longer than 570 minutes in this study.

Procedure

This study was approved by the Research Ethics ReviewCommittee (No. 98-3969B) of the study site. Informed consentwas obtained from participants or from legally authorizedrepresentatives if participants’ Rancho Los Amigos (RLA)Levels of Cognitive Functioning 21,22 was classified as level oneto nine by a trained research nurse. As Glasgow Coma Scale(GCS) is an important predictor of functional outcome 23 andcorrelates with cranial CT imaging,24,25 we classified theseverity of the TBI using initial GCS (IGCS)26 and initial loss ofconsciousness (ILOC), which were derived from medical re-cordsat theemergencydepartment.Objective sleepdatawasmeasured using an actigraphy watch secured to the partici-pant’s nonparalyzed/nontraumatizedwrist within 24 hours ofadmission to the neurosurgical ward and was recorded for 7consecutive days. A 7-day sleep diary was completed by par-ticipants or their caregivers to facilitate data analysis. Atrained research assistant provided verbal instructionsregarding use and care of the actigraph and how to completethe sleep diary. The diary recorded bedtime (time to bed),waketime(timeoutofbed), andnaps.Bedtimewasdefinedasthe timewhen a participant went to bed with the intention tosleep.Waketimewasdefinedas thetimewhentheparticipantwokeupandstartedhis/her day. The timeswereused toguidevisualization of the sleep times. Once all the times wereentered into the program, the ActiLife software (ActiGraph,version 5.3.0) provided the automated scoring of TST, WASO,SE, and SOL. The output was analyzed and verified by one ofthe investigators. All participants completed 7 days of sleeprecording.

Statistical analysis

Statistical analyses were performed by SPSS version 17.0 forWindows (SPSS Statistics, IBM, Armonk, NY, USA) and HLMversion 7.0 (Student version; Scientific Software Interna-tional, Inc., Lincolnwood, IL, USA).27 The significant levelwas set at p < 0.05. Descriptive analyses and frequencydistributions were used to describe the distributions of thedemographic characteristics, disease characteristics, andeach sleep parameters.

Hierarchical linear modeling (HLM) was a commonmethodto estimate the individual growth trajectory. 28,29 Separatetwo-level models were constructed by HLM analysis. Level 1

model was the within-patient equation analysis for each TBIpatient’s trajectory, and Level 2 model was the between-patient equation for cross-sectional differences. In light ofthe heterogeneity of our study sample, statistical analyseswere repeated in a subgroup of study sample comprising 31mild TBI patients.

In order to improve the reliability of the constructs of themodel, preliminary analyses were carried out using correla-tional analyses and the Student t test to select variables forthe model. Potential factors were derived from a review ofthe literatureof sleep disturbance in subacuteandchronicTBIpatients. Preliminary analyses revealed that only IGCS < 11(p < 0.05) was significantly associated with daytime TST.Without ILOC and age> 41 years were significantly associatedwith 24-hour TST (all p < 0.05). Moreover, in the analyses ofthe mild TBI subgroup, preliminary analyses revealed thatonly gender was significantly associated with daytime and 24-hour TST (both p < 0.05).

Results

Demographic and disease characteristics ofparticipants

From March 2009 to February 2010, 60 patients meeting ourcriteria were enrolled in this study. Eight patients droppedout of the study. Five patients withdrew from the studybecause their caregivers experienced difficulties with thestudy procedure or were unwilling to cooperate with theresearch assistant. One patient was transferred to the ICUbecause of worsened conditions. Two patients encounteredactigraphy malfunctions. As a result, data of 52 remainingparticipants were analyzed. Participants aged from 20 yearsto 65 years with a mean age of 34.98 years (Table 1). Themajority of the participants had mild TBI and did not havecomorbid conditions. None of the TBI patients used sedativesduring the observation period. All individuals underwentcranial CT in the emergency department. Eighty-twopercent of the participants used analgesics. Thirty-fivepercent (n Z 18) of the participants were diagnosed ashaving subarachnoid hemorrhage (SAH), 17.3% (nZ 9) of theparticipants had intracerebral hemorrhage (ICH), and one-third of the participants had more than one type of hemor-rhage. Details of study participants are shown in Appendix 1.

Of the 52 participants included, 35 were categorized ashaving mild TBI based on their IGCS (i.e., score ofIGCS Z 13 to 15). Four patients with mild TBI did not cometo the hospital for medical treatment until a few days afterTBI. Therefore, we performed a subgroup analysis on theremaining 31 patients with mild TBI.

Distributions of sleep parameters among patientswith TBI

The average SE, SOL, WASO, nighttime TST, daytime TST, and24-hour TST were 68.7%, 9.8 minutes, 174.0 minutes, 396.1minutes, 307.2minutes, and 703.3minutes, respectively. Theaverage SE and SOL were significantly lower and the averageWASOwere significantly higher in the TBI patients of this studythan the established normative published data (t Z e7.5,p < 0.0001, t Z e2.1, p Z 0.04, and t Z e9.9, p < 0.0001,

Table 1 Distribution of demographic and clinical charac-teristics of the study sample (n Z 52).

Characteristics

Age (y) 34.98 � 14.73BMI 22.74 � 4.01Day after injury 3.94 � 2.68Initial GCS scoreMild 35 (67.3)Moderate 7 (13.5)Severe 10 (19.2)

GenderMale 34 (65.4)Female 18 (34.6)

Cause of injuryAuto/vehicle 41 (78.8)Fall 8 (15.4)Hit by falling object 3 (5.8)

ILOCNo 25 (48.1)Yes 27 (51.9)

ComorbiditiesNo 44 (84.6)Yes 8 (15.4)

Use of medicationsGlycerol/mannitol 17 (36.7)Analgesics 43 (82.6)Sedatives 0 (0)

Data are presented as mean � SD or n (%).BMI Z body mass index; GCS Z Glasgow Coma Scale;ILOC Z initial loss of consciousness; IQR Z interquartile range;SD Z standard deviation.

Table 2 Hierarchical linear models of daytime and 24-hour TST for TBI individuals (n Z 52).a

Intercept, B0 Linear slope, B1

Coefficient (SE) Coefficient (SE)

Daytime TST

Level 1 modelFixed effectIntercept 439.763 (22.665)* �49.143 (12.903)**

Random effectIntercept

(Variance) 13,716.363* 4387.506*Level 2 modelFixed effectIntercept 439.764 (22.173)* �49.143 (12.334)*GCS score < 11 �73.350 (41.907) �59.215 (25.714)**

Random effectIntercept

(Variance) 13,069.153* 3787.099*24-hour TST

Level 1 modelFixed effectIntercept 874.363 (33.439)* �76.143 (15.992)*

Random effectIntercept

(Variance) 31,998.876* 3838.668*Level 2 modelFixed effectIntercept 874.363 (33.410)* �76.143 (14.673)*Without ILOC �15.977 (66.566) 59.475 (29.625)**Age > 41 y �11.092 (68.258) 66.957 (30.325)**

Random effectIntercept

(Variance) 29,685.999* 2159.448*

*p < 0.001.**p < 0.05.GCS Z Glasgow Coma Scale; HLM Z hierarchical linearmodeling; ILOC Z initial loss of consciousness; SE Z standarderror; TBI Z traumatic brain injury; TST Z total sleep time.a Final estimation of fixed effects is reported with robust

standard errors by HLM analysis.

548 H.-Y. Chiu et al.

respectively). Seventy-two percent of the participantsexperienced prolonged 24-hour TST, which is longer than 570minutes, a cutoff defining PTH.

The trajectories of change in sleep parameters

Daytime and 24-hour TST gradually declined over time;however, this pattern of change was not seen with SE, SL,WASO, and nighttime TST. The models listed in Table 2 showthat intercept (B0), linear (B1), and quadric slope (B2) arestatistically significant (all p < 0.05) for daytime and 24-hourTST, indicating that both variables significantly and graduallydecreased during the observation period among TBI patientsin our sample. Although the findings reflected a sample-widedecline for daytime and 24-hour TST, these did not imply thatall patients displayed an identical trajectory. The variancecomponents are calculated byusingmodels and are presentedin Table 2. Substantial differences existed in the trajectoriesof daytime TST and 24-hour TST among the participants,which are illustrated in Fig. 1A and B. These findings indicatedthat further inspection of within-patient differences in theindividual change parameters was warranted.

Predictors of the trajectories of change in sleepparameters

Both linear and quadratic models of changes in daytime TSTover time were tested. The linear model was chosen in our

final analysis because of the statistically significant linearslope and a better fit (based on the goodness-of-fit test:c2 Z 5.312, df Z 2, p Z 0.068). As given in Table 2, theresult of linear slope shows that an IGCS score < 11 is not apredictor of the intercept; however, an IGCS score < 11significantly predicts the between-patient differences inthe slope parameter, indicating that TBI patients with moresevere head injury do not have lower daytime TST at thebeginning of the observational period but they experience asignificantly slower rate of reduction in daytime TST rela-tive to their counterparts over time. To illustrate the ef-fects of IGCS score on the trajectory of daytime TST, theadjusted change curves are displayed in Fig. 2A. An IGCSscore < 11 explained 13.7% of the variance in daytime TSTbetween TBI patients across the observation period [pro-portion of additional variance explained by the model withcovariates Z (4387.506 e 3787.009) / 4387.506 Z 0.137].

Both linear and quadratic models of changes in 24-hourTSTover timewere testedbut only the linearmodel fitted the

Figure 1 Spaghetti plots of the 52 patients’ individual (A)daytime TST and (B) 24-hour TST, and 31 mild TBI patients’individual (C) 24-hour TST trajectories over time.TST Z total sleep time.

Sleep trajectories in acute traumatic brain injury 549

data (p< 0.05). As shown in Table 2, without ILOC and age>41 years are not significant predictors of the intercept;nonetheless, both variables independently and significantlypredict the between-patient differences in the slopeparameter, indicating that patients who are older and thosewithout ILOC do not exhibit shorter 24-hour TST at the

beginning of the observational period but have a faster rateof decline in 24-hour TST relative to their counterparts overtime. As shown in Fig. 2B and C, themanifestation of initiallyprolonged 24-hour TST but shortened 24-hour TST at the endof the observation day is evident in TBI patients with olderage and those without ILOC. The two predictors togetherexplained 43.7% of the variance in 24-hour TST between TBIpatients across the observational period [proportion ofadditional variance explained by the model withcovariates Z (3838.668 e 2159.448) /3838.668 Z 0.437].

Subgroup analyses of sleep parameters in mild TBIpatients

The average SE, SOL, WASO, nighttime TST, daytime TST,and 24-hour TST in mild TBI patients (n Z 31) were 69.6%,10.1 minutes, 165.3 minutes, 250.8 minutes, 384.1 minutes,and 634.9 minutes, respectively. The average SE and SOLwere significantly lower and the average WASO was signifi-cantly higher in these mild TBI patients than the establishednormative published data (t Z e12.90, p < 0.0001,t Z e6.44, p < 0.0001, and t Z 18.68, p < 0.0001,respectively). Sixty-nine percent of the mild TBI participantsexperienced prolonged 24-hour TST, which is longer than 570minutes, a cutoff defining PTH. The mild TBI subgroup hadsimilar manifestations in poor SE, shorter SOL, higher WASO,and prolonged sleep duration as those seen in 52 patients.

In the mild TBI subgroup, as shown in Fig. 1C, only 24-hour TST gradually declines over time (both p < 0.0001);however, this pattern of change was not seen with SE, SL,WASO, daytime, and nighttime TST. The linear model waschosen in our final analysis because of the statisticallysignificant linear slope and a better fit (p < 0.05). As givenin Table 3, the result of a linear slope shows that gender isnot a predictor of the intercept but significantly predictsthe between-patient differences in the slope parameter,indicating that mild TBI patients who are females do notexhibit shorter 24-hour TST at the beginning of the obser-vational period but have a slower rate of decline in 24-hourTST relative to their counterparts over time. To illustratethe effects of gender on the trajectory of 24-hr TST, theadjusted change curves are displayed in Fig. 2D.

Discussion

In comparison to the normative data from the healthypopulation,19 poor sleep efficiency, and frequent awak-ening were observed in the study sample of TBI and themild TBI subgroup. This could reflect trauma to sleep-regulating centers such as the reticular activating system,basal forebrain, or the anterior hypothalamus and/or theirconnections.30,31 Head trauma caused by rapid accelera-tion/deceleration can disrupt the homeostasis of sleeppatterns, resulting in sleep disturbance.32 Our data areconsistent with previous studies evaluating chronic TBIpatients13 and indicate that sleep change can occur notonly during chronic or subacute TBI, but also in the imme-diately post-injury period.

In this study, we found that in the study sample as awhole and in the mild TBI subgroup, sleep duration wasprolonged immediately following trauma. The observed

Figure 2 Trajectories of (A) daytime TST by GCS score, and 24-hour TST according to (B) ILOC and (C) age. (D) The trajectory of24-hour TST by gender in mild TBI subgroup. GCS Z Glasgow Coma Scale; ILOC Z initial loss of consciousness; TST Z total sleeptime.

Table 3 Hierarchical linear models of 24-hour TST forsubgroup analysis of mild TBI (n Z 31).

Intercept, B0 Linear slope, B1

Coefficient (SE) Coefficient (SE)

Level 1 modelFixed effectIntercept 751.45 (29.19)* �29.13 (4.27)*

Random effectIntercept

(Variance) 15,133.74* 15,790.07*Level 2 modelFixed effectIntercept 714.07 (85.15)* �55.26 (12.86)*Female 26.34 (56.67) 18.41 (8.56)**

Random effectIntercept

(Variance) 13,175.50** 15,486.42*

Final estimation of fixed effects is reported with robust stan-dard errors by HLM analysis.*p < 0.001.**p < 0.05.HLM Z hierarchical linear modeling; SE Z standard error;TBI Z traumatic brain injury; TST Z total sleep time.

550 H.-Y. Chiu et al.

trajectories of sleep duration over time are consistent withclinical observations. One potential explanation for thetrajectories is the interaction of brain injury and hypotha-lamic hypocretin system. PTH was reported to be related tohypocretin-1 deficiency in cerebrospinal fluid (CSF) in pa-tients with TBI.33 More recently, a study of 27 TBI patientsfound that most patients in the acute phase of TBI dis-played low or undetectable levels of hypocretin-1 in theCSF, and that the levels increased to the low or normalrange 6 months later.8 It is possible that hypocretinneuronal activity declines in response to injury and re-covers over time even though the cells and/or their axonsare uninjured. This might provide a physiological basis forthe observed prolonged sleep duration in TBI patients andthe trajectories of gradual decrease in daytime and 24-hourTST seen over time in the current study.

By contrast, it should be noted that the observed sleepduration is shorter than the normal range34 toward the endof the observation period in TBI patients with minor headtrauma as they experienced a rapid reduction in 24-hourTST (Fig. 2). Minor severity of head injury is a significantfactor for insomnia in postacute and chronic TBI patients1,8

In this regard, the aforementioned result may provide apossible clue to connect the exhibition of insomnia in theacute and chronic TBI phase, that is, the symptom ofinsomnia may not only be present in the postacute stageand the chronic stage but also exhibits in the acute stage in

Appendix 1. Findings of cranial CT among TBI patients(n Z 52).

Participants Findings ofcranial CT

Patients Findings ofcranial CT

1 SAH 27 EDH and SAH2 SAH 28 SAH and ICH3 ICH 29 ICH4 SAH 30 SAH and ICH5 ICH and IVH 31 ICH6 EDH and ICH 32 SAH and ICH7 SDH and SAH 33 SAH and SDH8 SAH and DAI 34 EDH9 SAH 35 SDH10 SAH 36 SDH11 EDH, ICH, and

SAH37 EDH

12 ICH 38 SAH13 ICH and EDH 39 SAH14 EDH and SDH 40 EDH15 SDH and SAH 41 SAH16 SAH 42 SAH17 ICH, DAI, and

SAH43 ICH

18 ICH 44 SAH19 SDH and SAH 45 SAH20 EDH and SDH 46 SDH and ICH21 SAH 47 ICH22 SAH 48 ICH23 SAH 49 SAH and SDH24 SAH 50 SAH and ICH25 SAH 51 ICH26 ICH 52 SAH

CT Z computed tomography; DAI Z diffuse axonal injury;EDH Z epidural hemorrhage; ICH Z intracerebral hemorrhage;IVH Z intraventricular hemorrhage; SAH Z subarachnoidhemorrhage; SDH Z subdural hemorrhage.

Sleep trajectories in acute traumatic brain injury 551

TBI patients with minor head injuries. Nevertheless, theserelationships and mechanisms need to be elucidated infuture studies.

The result of this study suggested that age significantlypredicted the trajectory of change in 24-hour sleep dura-tion as we found that older TBI patients exhibited a fasterrate of decline in 24-hour TST. This finding is consistentwith a previous report of sleep disturbance in chronic TBIpatients in that older TBI patients exhibited more symp-toms of sleep disturbance during the chronic TBI stage.9 Asthe relationship between age and sleep disturbance and theprevalence of sleep disturbance increases with age asdemonstrated by previous studies,35,36 this notion may beextended to the trajectory of change in sleep duration inTBI patients of older age. Clearly, future research needs toidentify the exact reason why older TBI patients have thispattern of change in 24-hour TST.

Furthermore, our data showed that younger partici-pants (� 40 years) had a slower rate of decline in 24-hourTST and exhibited prolonged sleep duration over time.Forty years of age is an important predictor of the prog-nosis after TBI.37,38 Specifically, it has been found that TBIpatients who were younger than 40 years had significantlybetter prognosis and functional recovery than those ofolder TBI patients. Of note, accumulating evidence sug-gests that sleep enhances plasticity effects after braindamage.39,40 It is therefore reasonable to infer thatobserved prolonged 24-hour sleep duration seen withyounger TBI patients might have beneficial effects onfunctional recovery. Thus, the relationship between ageand 24-hour sleep duration in patients with acute TBIwarrants further investigation.

Interestingly, data from a subgroup analysis of patientswith mild TBI demonstrated that gender significantly pre-dicted the trajectory of change in 24-hour sleep duration aswe found that females exhibited a slower rate of decline in24-hour TST. Although several possible reasons includingsex-specific circadian rhythms in core body temperature,41

female sex hormones,42 or psychological characteristicsmay explain this gender difference,43 the exact underlyingmechanism remain to be determined.

This preliminary study has several limitations. First, therelatively small sample size and recruitment of partici-pants from a single site limit its ability to make moregeneral predictions, especially with regard to participantswith moderate and severe TBI. Second, the study did notinclude patients staying in an intensive care unit aftertrauma. That is, the patients enrolled in this study hadrelatively good prognoses. Third, some of the potentialpredictors of sleep disturbances reported by previousstudies, including pain at night, anxiety, and depressionlevel, were not available for analysis in this study. Fourth,the time after injury is not homogeneous in our partici-pants. Lastly, sleep parameters were not measured bypolysomnography, the gold standard for evaluating sleepdisorders. Although polysomnography can record anddifferentiate sleep stages precisely, there were significantreasons to use actigraphy in this study. For example, acuteTBI patients can be irritable and agitated, which can makethem less likely to co-operate with polysomnography.Furthermore, polysomnography cannot trace changes incircadian rhythms. By contrast, actigraphy is convenient,

easy to use, and suited to evaluate changing sleep patternsin TBI patients.

In conclusion, awaking frequently during sleep, poorsleep efficiency, and prolonged sleep duration arefrequently experienced by TBI patients during the acutephase of brain trauma. Additionally, this prospective studysuggested that the daytime and 24-hour TST significantlyand gradually declined over time, and that age and severityof head injury were significant predictors of the trajec-tories of daytime and 24-hour sleep duration during acutephase following TBI. Mild TBI patients also exhibitedfrequent awakenings during sleep, poor sleep efficiency,and prolonged 24-hour sleep duration after acute headtrauma. Moreover, in the period immediately following TBI,24-hour TST significantly and gradually declined over timein mild TBI patients; gender significantly predicted thetrajectory of 24-hour sleep duration in mild TBI patients.Future studies should explore the mechanisms underlyingsleep patterns and the effects of sleep duration on func-tional outcomes (i.e., cognitive function) among acute TBIpatients.

552 H.-Y. Chiu et al.

References

1. Bryan-Hancock C, Harrison J. The global burden of traumaticbrain injury: preliminary results from the Global Burden ofDisease Project. Inj Prev 2010;16(Suppl. 1):A1e289.

2. McAllister TW, Sparling MB, Flashman LA, Guerin SJ,Mamourian AC, Saykin AJ. Differential working memory loadeffects after mild traumatic brain injury. Neuroimage 2001;14:1004e12.

3. Englander J, Hall K, Stimpson T, Chaffin S. Mild traumatic braininjury in an insured population: subjective complaints and re-turn to employment. Brain Inj 1992;6:161e6.

4. Verma A, Anand V, Verma NP. Sleep disorders in chronic trau-matic brain injury. J Clin Sleep Med 2007;3:357e62.

5. Fichtenberg NL, Millis SR, Mann NR, Zafonte RD, Millard AE.Factors associated with insomnia among post-acute traumaticbrain injury survivors. Brain Inj 2000;14:659e67.

6. Watson NF, Dikmen S, Machamer J, Doherty M, Temkin N.Hypersomnia following traumatic brain injury. J Clin Sleep Med2007;3:363e8.

7. Guilleminault C, Yuen KM, Gulevich MG, Karadeniz D, Leger D,Philip P. Hypersomnia after head-neck trauma: A medicolegaldilemma. Neurology 2000;54:653e9.

8. Baumann CR, Werth E, Stocker R, Ludwig S, Bassetti CL. Sleep-wake disturbances 6 months after traumatic brain injury: aprospective study. Brain 2007;130:1873e83.

9. Makley MJ, English JB, Drubach DA, Kreuz AJ, Celnik PA,Tarwater PM. Prevalence of sleep disturbance in closed headinjury patients in a rehabilitation unit. Neurorehabil NeuralRepair 2008;22:341e7.

10. Rao V, Spiro J, Vaishnavi S, Rastogi P, Mielke M, Noll K, et al.Prevalence and types of sleep disturbances acutely aftertraumatic brain injury. Brain Inj 2008;22:381e6.

11. Ouellet MC, Beaulieu-Bonneau S, Morin CM. Insomnia in pa-tients with traumatic brain injury: frequency, characteristics,and risk factors. J Head Trauma Rehabil 2006;21:199e212.

12. Castriotta RJ, Wilde MC, Lai JM, Atanasov S, Masel BE, Kuna ST.Prevalence and consequences of sleep disorders in traumaticbrain injury. J Clin Sleep Med 2007;3:349e56.

13. Parcell DL, Ponsford JL, Rajaratnam SM, Redman JR. Self-re-ported changes to nighttime sleep after traumatic brain injury.Arch Phys Med Rehabil 2006;87:278e85.

14. Sadeh A, Sharkey KM, Carskadon MA. Activity-based sleep-wake identification: an empirical test of methodological is-sues. Sleep 1994;17:201e7.

15. Ancoli-Israel S, Cole R, Alessi C, Chambers M, Moorcroft W,Pollak CP. The role of actigraphy in the study of sleep andcircadian rhythms. Sleep 2003;26:342e92.

16. Littner M, Kushida CA, Anderson WM, Bailey D, Berry RB,Davila DG, et al. Practice parameters for the role of actigraphyin the study of sleep and circadian rhythms: an update for2002. Sleep 2003;26:337e41.

17. Makley MJ, Johnson-Greene L, Tarwater PM, Kreuz AJ, Spiro J,Rao V, et al. Return of memory and sleep efficiency followingmoderate to severe closed head injury. Neurorehabil NeuralRepair 2009;23:320e6.

18. Zollman FS, Cyborski C, Duraski SA. Actigraphy for assessmentof sleep in traumatic brain injury: case series, review of theliterature and proposed criteria for use. Brain Inj 2010;24:748e54.

19. Ohayon MM, Carskadon MA, Guilleminault C, Vitiello MV. Meta-analysis of quantitative sleep parameters from childhood to oldage in healthy individuals: developing normative sleep valuesacross the human lifespan. Sleep 2004;27:1255e73.

20. Kempf J, Werth E, Kaiser PR, Bassetti CL, Baumann CR. Sleep-wake disturbances 3 years after traumatic brain injury. JNeurol Neurosurg Psychiatr 2010;81:1402e5.

21. Hagen C, Malkmus D, Durham P. Levels of cognitive func-tioning. Rehabilitation of the head injured adult: compre-hensive physical management. In: Dowey CA, editor.Professional Staff Association of the Rancho Los Amigos Hos-pital, Inc.; 1979.

22. Hagen C. Levels of cognitive functioning. Rehabilitation of thehead injured adult: comprehensive physical management. In:Dowey CA, editor. 3rd ed. Professional Staff Association of theRancho Los Amigos Hospital, Inc; 1998.

23. Balestreri M, Czosnyka M, Chatfield DA, Steiner LA, Schmidt EA,Smielewski P, et al. Predictive value of Glasgow Coma Scaleafter brain trauma: change in trend over the past ten years. JNeurol Neurosurg Psychiatr 2004;75:161e2.

24. Lee TT, Aldana PR, Kirton OC, Green BA. Follow-up comput-erized tomography (CT) scans in moderate and severe headinjuries: correlation with Glasgow Coma Scores (GCS), andcomplication rate. Acta Neurochir (Wien) 1997;139:1042e7.discussion 7e8.

25. Morgado F, Rossi L. Correlation between the Glasgow ComaScale and computed tomography imaging findings in patientswith traumatic brain injury. Radiologia Brasileira 2011;44:35e41.

26. Campbell CG, Kuehn SM, Richards PM, Ventureyra E,Hutchison JS. Medical and cognitive outcome in children withtraumatic brain injury. Can J Neurol Sci 2004;31:213e9.

27. Raudenbush SW, Bryk AS, Cheong Y, Congdon R, Toit M. HLM7:hierarchical linear and nonlinear modeling. Chicago: ScientificSoftware International; 2010.

28. Raudenbush SW. Comparing personal trajectories and drawingcausal inferences from longitudinal data. Annu Rev Psychol2001;52:501e25.

29. Raudenbush SW, Bryk AS. Hierarchical linear models: appli-cations and data analysis methods. 2nd ed. Thousand Oaks,CA: Sage Publications; 2002.

30. Saper CB, Cano G, Scammell TE. Homeostatic, circadian, andemotional regulation of sleep. J Comp Neurol 2005;493:92e8.

31. Saper CB, Scammell TE, Lu J. Hypothalamic regulation of sleepand circadian rhythms. Nature 2005;437:1257e63.

32. Graham DI, Lawrence AE, Adams JH, Doyle D, McLellan DR.Brain damage in fatal non-missile head injury without highintracranial pressure. J Clin Pathol 1988;41:34e7.

33. Baumann CR, Stocker R, Imhof HG, Trentz O, Hersberger M,Mignot E, et al. Hypocretin-1 (orexin A) deficiency in acutetraumatic brain injury. Neurology 2005;65:147e9.

34. Ohayon MM, Vecchierini MF. Normative sleep data, cognitivefunction and daily living activities in older adults in the com-munity. Sleep 2005;28:981e9.

35. Ancoli-Israel S. Sleep and its disorders in aging populations.Sleep Med 2009;10(Suppl. 1):S7e11.

36. Wolkove N, Elkholy O, Baltzan M, Palayew M. Sleep and aging:1. Sleep disorders commonly found in older people. CMAJ2007;176:1299e304.

37. Kuo JR, Lo CJ, Lu CL, Chio CC, Wang CC, Lin KC. Prognosticpredictors of outcome in an operative series in traumatic braininjury patients. J Formos Med Assoc 2011;110:258e64.

38. Katz DI, Alexander MP. Traumatic brain injury. predictingcourse of recovery and outcome for patients admitted torehabilitation. Arch Neurol 1994;51:661e70.

39. Carmichael ST, Chesselet MF. Synchronous neuronal activity isa signal for axonal sprouting after cortical lesions in the adult.J Neurosci 2002;22:6062e70.

40. Stickgold R, Hobson JA, Fosse R, Fosse M. Sleep, learning, anddreams: off-line memory reprocessing. Science 2001;294:1052e7.

41. Campbell SS, Gillin JC, Kripke DF, Erikson P, Clopton P. Genderdifferences in the circadian temperature rhythms of healthyelderly subjects: relationships to sleep quality. Sleep 1989;12:529e36.

Sleep trajectories in acute traumatic brain injury 553

42. Sowers MF, Zheng H, Kravitz HM, Matthews K, Bromberger JT,Gold EB, et al. Sex steroid hormone profiles are related tosleep measures from polysomnography and the PittsburghSleep Quality Index. Sleep 2008;31:1339e49.

43. Lindberg E, Janson C, Gislason T, Bjornsson E, Hetta J,Boman G. Sleep disturbances in a young adult population: cangender differences be explained by differences in psychologi-cal status? Sleep 1997;20:381e7.