More than a million people in the United States are affected by traumatic brain injury (TBI) annually [
Pittsburgh Sleep Quality Index (PSQI), which is divided into 7 components, is a questionnaire frequently used for the evaluation of sleep problems in clinical and healthy populations [
Therefore, this study aimed to determine the patterns of sleep problem associated with the mTBI by use of the PSQI. Specifically, we performed an analysis by the propensity score model to describe the characteristics of sleep problems among the patients following acute mTBI.
All of the mTBI patients aged ≥17 years who were admitted to any of the 3 affiliated hospitals of Taipei Medical University (TMU) between March 2010 and February 2013 were recruited. The definition of mTBI was based on the diagnostic criteria established by the American Congress of Rehabilitation Medicine, which consist of a Glasgow Coma Scale (GCS) score of 13–15 at presentation, loss of consciousness for <30 min, and normal head computed tomography findings. Patients who had a history of cerebrovascular disease, psychiatric comorbidities, epilepsy, alcohol abuse, sleep-wake modifying treatment, previous TBI, or severe systemic medical illness were excluded. In addition, volunteers who were older than 17 years old and did have no brain injury were recruited into the control group. The exclusion criteria for the control participants were the same as those for the mTBI patients. All patients were initially contacted by telephone and 675 mTBI patients were recruited. Among the mTBI patients, 171 (25.33%) provided informed consent and completed a baseline assessment during an initial evaluation within 1 month of experiencing an mTBI. The study protocol was approved by the Joint Institution Review Board at TMU.
The number of participants for each PSQI component was calculated, and differences in trends between the mTBI and control groups were compared using the Cochran-Armitage test. Also, the association between scales and the other confounders was assessed via Spearman’s correlations. In this study, the participants in the control group were assumed to represent the general population. The control participant recruited without matching the age and gender of the mTBI group. In order to generate a quasirandomized design, the propensity score method was used to account for selection biases and potential confounding factors. The propensity scores were calculated by the logistic regression to estimate the probability of each patient on the basis of age, sex, and questionnaires. The best model was selected according to AIC stepwise algorithm. The effects for each component were assessed via cumulative logit regression. In all statistical tests, a
In this study, we recruited 675 mTBI patients and 186 control participants, of whom 171 and 145 subjects, respectively, completed the PSQI and other questionnaires and signed the informed consent. The participants’ demographic information is shown in Table
Demographic data of the mTBI and control groups.
Variables | mTBI | Control |
|
---|---|---|---|
|
171 | 145 | |
Glasgow outcome score | 14.98 (0.12) | 15 | 0.99 |
Age at injury, mean (SD) | 38.57 (15.09) | 32.18 (10.63) | <0.001 |
Men, |
56 (32.74%) | 38 (26.21%) | 0.06 |
Years of education (SD) | 15.33 (1.95) | 14.93 (2.23) | 0.88 |
Married, |
59 (34.50%) | 44 (30.34%) | 0.27 |
Employed, |
83 (48.53%) | 70 (48.28%) | 0.47 |
Headache, |
107 (62.57%) | 30 (20.69%) | <0.001 |
Mechanism of injury, |
|||
Transportation accident | 92 (53.80) | — | |
Fall | 57 (33.33) | — | |
Other | 22 (12.87) | — | |
Psychological evaluations | |||
BAI | 9.30 (10.14) | 2.62 (3.62) | <0.001 |
BDI | 8.75 (8.13) | 5.18 (7.24) | <0.001 |
ESS | 7.33 (4.28) | 6.49 (3.51) | 0.12 |
PSQI | 7.23 (3.92) | 5.65 (3.45) | <0.001 |
SD: standard deviation.
The stratified numbers of participants in each category of PSQI components are shown in Table
Numbers (percentages) of participants with each PSQI subscore in the mTBI and control groups.
PSQI | Number (%) |
|
||
---|---|---|---|---|
mTBI |
Control |
|||
Sleep duration | >7 h | 92 (54) | 74 (51) | 0.44 |
6-7 h | 34 (20) | 47 (32) | ||
5-6 h | 28 (16) | 12 (8) | ||
<5 h | 17 (10) | 12 (8) | ||
|
||||
Sleep disturbance | None | 6 (4) | 8 (6) | <0.01 |
1–9 | 99 (58) | 111 (77) | ||
10–18 | 57 (33) | 23 (16) | ||
19–27 | 9 (5) | 3 (2) | ||
|
||||
Sleep latency | <15 min and not during the previous month | 43 (25) | 35 (24) | 0.10 |
16–30 min and less than once/wk | 59 (35) | 67 (46) | ||
31–60 min and once or twice/wk | 45 (26) | 34 (23) | ||
>60 min and >3 times/wk | 24 (14) | 9 (6) | ||
|
||||
Daytime dysfunction | No problems | 43 (25) | 54 (37) | 0.07 |
Minor problems | 94 (55) | 65 (45) | ||
Considerable problems | 26 (15) | 22 (15) | ||
Major problems | 8 (5) | 4 (3) | ||
|
||||
Habitual sleep efficiency | ≥85% | 106 (62) | 110 (76) | 0.01 |
75%–84% | 33 (19) | 20 (14) | ||
65%–74% | 11 (6) | 4 (3) | ||
<65% | 21 (12) | 11 (8) | ||
|
||||
Subjective sleep quality | Very good | 11 (6) | 27 (19) | <0.01 |
Relatively good | 55 (32) | 73 (50) | ||
Relatively poor | 78 (46) | 34 (23) | ||
Very poor | 27 (16) | 11 (8) | ||
|
||||
Use of sleep medication | Never during the previous month | 148 (87) | 130 (90) | 0.14 |
Less than once/wk | 6 (4) | 7 (5) | ||
Once or twice/wk | 2 (1) | 3 (2) | ||
≥3 times/wk | 15 (9) | 5 (3) |
Percentage of each score for seven sleep components.
The
Group | Age | BAI | BDI | ESS | |
---|---|---|---|---|---|
Sleep duration | mTBI | 0.01 | <0.01 | 0.02 | 0.6 |
Control | 0.16 | 0.22 | 0.65 | 0.14 | |
|
|||||
Sleep disturbances | mTBI | 0.17 | <0.01 | <0.01 | 0.04 |
Control | 0.58 | <0.01 | 0.01 | 0.04 | |
|
|||||
Sleep latency | mTBI | 0.67 | <0.01 | <0.01 | 0.55 |
Control | 0.15 | <0.01 | <0.01 | 0.16 | |
|
|||||
Daytime dysfunction | mTBI | 0.09 | <0.01 | <0.01 | <0.01 |
Control | 0.02 | <0.01 | <0.01 | <0.01 | |
|
|||||
Habitual sleep efficiency | mTBI | 0.29 | 0.01 | <0.01 | 0.62 |
Control | 0.13 | 0.03 | 0.03 | 0.10 | |
|
|||||
Subjective sleep quality | mTBI | 0.92 | <0.01 | <0.01 | 0.06 |
Control | 0.89 | <0.01 | <0.01 | <0.01 | |
|
|||||
Use of sleep medication | mTBI | 0.04 | <0.01 | 0.01 | 0.44 |
Control | 0.04 | 0.16 | 0.15 | 0.73 |
The confounders were assessed for both groups via the logistic regression. After stepwise, the final model included four covariates, age, sex, BAI, and headache (as shown in Table
Results of propensity score model.
Estimate |
|
|
---|---|---|
Age | 0.05 | <0.01 |
Sex | −1.28 | <0.01 |
BAI | 0.19 | <0.01 |
Headache | 1.74 | <0.01 |
BDI | −0.02 | 0.38 |
ESS | −0.02 | 0.58 |
|
||
After stepwise | ||
Estimate |
|
|
|
||
Age | 0.05 | <0.01 |
Sex | −1.32 | <0.01 |
BAI | 0.17 | <0.01 |
Headache | 1.71 | <0.01 |
Results of mTBI effect via cumulative logit model with or without propensity score as a covariate.
Model | Estimate | sd |
|
|
---|---|---|---|---|
Sleep duration | ||||
Crude | 0.0694 | 0.2136 | 0.3249 | 0.75 |
Propensity score | −0.7122 | 0.2685 | −2.6525 | <0.01 |
Sleep disturbance | ||||
Crude | 0.9691 | 0.2496 | 3.8827 | <0.01 |
Propensity score | −0.0498 | 0.3087 | −0.1613 | 0.87 |
Sleep latency | ||||
Crude | 0.2998 | 0.2496 | 3.8827 | 0.14 |
Propensity score | −0.3154 | 0.2477 | −1.2734 | 0.20 |
Daytime dysfunction | ||||
Crude | 0.4126 | 0.2153 | 1.9166 | 0.06 |
Propensity score | −0.1134 | 0.2617 | −0.4332 | 0.66 |
Habitual sleep efficiency | ||||
Crude | 0.6535 | 0.2459 | 2.6572 | <0.01 |
Propensity score | 0.0722 | 0.3016 | 0.2394 | 0.81 |
Subjective sleep quality | ||||
Crude | 1.1907 | 0.2201 | 5.4096 | <0.01 |
Propensity score | 0.7300 | 0.2532 | 2.8831 | <0.01 |
Use of sleep medication | ||||
Crude | 0.3366 | 0.3522 | 0.9557 | 0.34 |
Propensity score | −0.5840 | 0.4656 | −1.2544 | 0.21 |
Results of the proportional-odds cumulative logit model.
Model | Estimate | sd | 95% CI | Odds ratio | |
---|---|---|---|---|---|
Sleep duration | |||||
Intercept | log( |
1.2207 | 0.2641 | (0.70, 1.74) | |
Intercept | log( |
2.4835 | 0.2926 | (1.91, 3.06) | |
Intercept | ( |
3.5940 | 0.3403 | (2.93, 4.26) | |
mTBI | −0.7122 | 0.2685 | (−1.24, −0.19) | 0.491 | |
|
|||||
Subjective sleep quality | |||||
log( |
−0.8032 | −0.2837 | (−1.32, −0.28) | ||
log( |
1.1130 | 2.1525 | (1.11, 2.15) | ||
log( |
3.2958 | 4.6749 | (3.30, 4.67) | ||
0: very good; 1: relatively good; 2: relative poor; 3: poor | |||||
mTBI | 0.7300 | 0.2532 | (0.23, 1.23) | 2.075 |
Propensity score density for mTBI (solid line) and control (dotted line) participants.
All of the mTBI patients in this study had suffered nonblast injuries [
One finding of particular interest was that sleep problems following mild traumatic brain injury can be associated with anxiety and depression. Therefore, the group effect was evaluated after adjusting by the propensity score model including anxiety and depression scores in order to achieve the quasirandomized observation study. In the mTBI group, we observed that the PSQI subscore for sleep disturbance was moderately to strongly associate with the BAI score. We suggest that some features of anxiety, such as increased arousal at night, can lead to sleep disruption and then result in changes of sleep duration and subjective sleep quality [
There are a few limitations in the study. First, the subjective data were collected instead of objective data, such as adrenocorticotropic hormone (ACTH) and cortisol which were associated with sleep quality [
In conclusion, our study results indicate the characteristics of sleep problems following mTBI. The patients with mTBI had significantly different sleep duration and sleep quality after adjusting all other confounders. These findings could potentially increase physicians’ understanding of the consequences of the mTBI. Studies with longer-term followup and analyses of biomarkers such as ACTH and cortisol are recommended to facilitate the raising an optimal management of sleep disturbance in mTBI patients.
The authors declare that there is no conflict of interests regarding the publication of this paper.
This study was supported by Grants NSC 98-2321-B-038-005-MY3, NSC 101-2321-B-038-005, 103TMU-SHH-24, and DOH 101-TD-B-111-003.