Effect of Severity of Post-Traumatic Confusion and Its Constituent Symptoms on Outcome After Traumatic Brain Injury

Study Population 

The study population for the present study consisted of all qualified persons with TBI who were admitted to an inpatient brain injury neurorehabilitation program within a free-standing rehabilitation hospital during the study period, April 2003 through February 2006. Data collection for this investigation was approved by the appropriate institutional review board. Participants were recruited as part of the National Institute on Disability and Rehabilitation Research Traumatic Brain Injury Model Systems (TBIMS) program. Inclusion criteria for the TBIMS program12 include: medically documented TBI; treatment at an affiliated level 1 trauma center within 24 hours of injury; receipt of inpatient rehabilitation at the TBIMS-affiliated rehabilitation unit; admission to inpatient rehabilitation within 72 hours of discharge from acute care; age of 16 years or older at the time of injury; and provision of informed consent by the person with injury or an appropriate proxy. Additional inclusion criteria for the present investigation were: (1) recovery to a responsive state prior to rehabilitation discharge (vegetative or minimally conscious patients were excluded) and (2) clinical presentation and reason for rehabilitation primarily reflective of TBI. There was no upper limit on participant age.

Data Collection 

We collected demographic data (age, sex, years of education) and injury severity data (Glasgow Coma Scale [GCS] score, time to follow commands [TFC], acute care length of stay [LOS], rehabilitation LOS) through review of medical records. GCS scores were obtained at admission to the emergency department. TFC was defined as the interval in days from the date of injury to the date that the participants followed simple motor commands on 2 consecutive assessments. Data on severity and symptoms of confusion were collected using the CAP. CAP data were collected serially on all participants, but assessments closest to 21 days postinjury were used for this investigation. This time was selected because the majority of potential participants had evaluations available close to this time. The number of symptoms of confusion (range, 0–7) at the 21-day postinjury CAP was the measure on severity of confusion used in this investigation. We included participants who did not meet our criteria for PTC because our previous investigations have shown that persons in early recovery from TBI continue to have some symptoms of confusion even after they no longer meet full criteria for PTC.10 These ratings were collected by neuropsychologists, neuropsychology fellows, and neuropsychology technicians, all with extensive training in the use of this instrument. One study outcome, employability, was collected using the Disability Rating Scale (DRS). These data were based on a consensus rating of the treatment team at discharge from inpatient rehabilitation. Most members of the treatment team were not aware of the goals of the present investigation. The other study outcome, productivity, was collected at 1 year postinjury by a research assistant who was unaware of the goals of this investigation. Persons who were employed earning at least minimum wage or who were in school (high school or college) making academic progress toward a degree were coded as productive. All others were coded as nonproductive.

Measures 

Data Analysis 

We presented descriptive data for demographic characteristics (age, years of education, sex), injury severity (GCS, TFC, acute LOS, rehabilitation LOS), confusion severity and symptoms (days from injury to CAP evaluation [CAP days], total number of CAP symptoms [total CAP], presence of each of the 7 symptoms [disorientation, cognitive impairment, fluctuation, agitation, nighttime sleep disturbance, decreased daytime arousal, psychotic type symptoms]), and study outcomes (employability, productivity) as quartiles for continuous variables and frequency and proportions for nominal variables. Participants classified as confused on the 21-day postinjury CAP were compared with participants classified as nonconfused on a variety of study measures using chi-square tests for dichotomous variables and Wilcoxon and Mann-Whitney tests for continuous variables.

The primary study hypotheses regarding prediction of employability and productivity outcomes were examined using multivariable logistic regression models with age, years of education, GCS score, TFC, and total CAP as candidate predictors and employability and productivity as separate outcomes. Missing values of predictor variables were imputed. The TFC interval was censored at 45 days to prevent undue effects of outliers. Effects for all predictors are odds ratios (ORs). For all continuous predictors, the ORs indicate the change in the odds of the outcome (employability, productivity) as the value of the predictor variable changes from the 25th percentile to the 75th percentile and are called interquartile range coefficients.

We examined the secondary, exploratory hypothesis regarding the prognostic value of presence (or absence) of each symptom of confusion for predicting employability at rehabilitation discharge and productivity at 1 year postinjury using simple, binary logistic regression analyses. Separate predictive models were calculated for each of the 7 symptoms for each of the 2 outcomes. For these analyses, presence of a symptom was coded 0 and absence of the symptoms was coded 1.

Study Population 

During the study period, 195 potential participants were admitted to the study site. Of these, 16 did not recover to a responsive state and 3 had clinical presentations that were substantially influenced by non-TBI disorders (spinal cord injury, dementia, deafness). An additional 5 potential participants declined to give consent and 3 participants who were enrolled were inadvertently missed for collection of key study measures. The resulting study sample comprised 168 participants. This study sample included 38 participants (23% of the current study sample) who were reported on in our previous investigation of confusion and outcome after TBI.11 The productivity outcome was collected at median time of 382.0 days postinjury (25th percentile, 339.5; 75th percentile, 414.0). At follow-up, 1 participant was deceased, 9 were retired, 5 could not be contacted, and 21 were not due for follow-up at the time of analysis. These 36 participants were excluded from the productivity analyses, but the remaining 132 participants were included in these analyses. All other data were complete for all participants except for 2 missing values for TFC. These values were predicted using GCS scores and these imputed values were used in all subsequent analyses.

Table 1 shows descriptive data for demographic, injury severity, CAP days, total CAP, and study outcomes for all participants, nonconfused participants, and confused participants. Participants were classified as nonconfused or confused based on the CAP evaluation obtained closest to 21 days postinjury. Figure 1 shows the proportions of all, nonconfused, and confused participants rated positive for each symptom of confusion on the 21-day postinjury CAP. Mann-Whitney U analyses revealed that nonconfused patients differed significantly from confused patients on years of education, GCS, TFC, acute LOS, rehabilitation LOS, CAP days, and total CAP. As expected and consistent with our previous findings, confused patients sustained more severe injuries as indicated by GCS, TFC, and both LOS variables. The longer interval from injury to CAP evaluation was due to confused patients being more likely to be admitted to rehabilitation at greater than 21 days postinjury. By definition, confused patients had greater severity of confusion (total CAP) than nonconfused patients. The significance of the difference in years of education is unclear but it could indicate that patients with a lower level of functioning prior to TBI were at greater risk for confusion.

Table 1. Description of the Study Sample on Demographic, Injury Severity, Confusion, and Outcome Variables (N=168)

		All (N=168)	Nonconfused (n=85)	Confused (n=83)
	Descriptors	Median (25th, 75th Percentiles)	Median (25th, 75th Percentiles)	Median (25th, 75th Percentiles)
	Continuous
	Age (y)	27.0(20.0,45.5)	24.4(18.6,44.1)	28.6(20.7,48.6)
	Education⁎ (y)	12.0(10.0,12.2)	12.0(10.0,13.5)	11.0(10.0,12.0)
	GCS score†	7.0(5.0,11.2)	9.0(6.0,13.0)	7.0(4.0,10.0)
	TFC‡ (d)	4.0(1.0,11.2)	2.0(1.0,5.0)	7.0(2.0,17.0)
	Acute LOS‡ (d)	19.0(12.0,27.0)	13.0(10.0,19.5)	23.0(18.0,30.0)
	Rehabilitation LOS‡ (d)	16.0(13.0,22.8)	14.0(11.0,17.5)	20.0(15.0,25.0)
	CAP days‡	21.0(19.0,29.0)	19.0(16.0,22.5)	24.0(21.0,31.0)
	Total CAP‡	3.0(2.0,5.0)	2.0(0.0,2.0)	5.0(4.0,6.0)

	Categorical	n(%)	n(%)	n(%)
	Sex
	Male	113(67)	57(67)	56(68)
	Female	55(33)	28(33)	27(32)
	GCS score
	Very severe (3–5)	44(26)	18(21)	27(33)
	Severe (6–8)	52(31)	24(28)	27(33)
	Moderate (9–12)	34(20)	17(20)	17(20)
	Mild (13–15)	38(23)	26(31)	12(14)
	Outcomes
	Employability‡
	Yes	69(41)	53(62)	16(19)
	No	99(59)	32(38)	67(81)
	Productivity† (n=132)
	Yes	48(36)	34(49)	14(23)
	No	84(64)	36(51)	48(77)

⁎ P<.05; † P<.02; ‡ P<.001 (differences between nonconfused and confused participants).

View Large Image Download to PowerPoint

Fig 1. Percentages of all, nonconfused, and confused patients showing confusion symptoms (N=168). Abbreviations: Arouse, decreased daytime arousal; Cog imp, cognitive impairment; Disorient, disorientation; Fluct, fluctuation in symptom presentation; Psychotic, psychotic-type symptoms; Sleep dist, sleep disturbance.

Chi-square analyses revealed that confused participants were more likely to show each of the 7 confusion symptoms. The confused group did not differ from the nonconfused group in sex composition. Consistent with prior investigations, confused patients had poorer outcomes for both the employability and productivity outcomes.

Multivariable Logistic Regression Analyses 

Table 2 shows the multivariable logistic regression model for predicting employability at discharge from inpatient rehabilitation. After adjustment for all other predictors, TFC and total CAP made unique contributions to prediction of this outcome. Participants with TFC values at the 75th percentile (indicating more severe injuries) had only .54 times the odds of being rated as employable at rehabilitation discharge as compared with participants at the 25th percentile. Stated another way, participants at the 25th percentile of TFC had almost twice the odds of being employable as those at the 75th percentile. Participants with total CAP values at the 75th percentile (more confused) had only .34 times the odds of being employable as compared with those at the 25th percentile. Stated another way, those at the 25th percentile had almost 3 times the odds of being employable as those at the 75th percentile. There was a trend for an effect for years of education with higher levels of education tending to be associated with a greater likelihood of being employable.

Table 2. Multivariable Logistic Regression Model for Predicting Employability at Discharge From Inpatient Rehabilitation (N=168)

	Predictor	Comparison	OR	95% CI	P
	Age	20.0,45.5	0.71	0.38–1.31	.27
	Education	10.0,12.2	1.39	0.96–2.02	.08
	GCS	5.0,11.2	0.78	0.40–1.49	.44
	TFC	1.0,11.2	0.54	0.32–0.93	.03
	Total CAP	2.0,5.0	0.34	0.19–0.61	.001

Abbreviation: CI, confidence interval.

Table 3 shows the multivariable logistic regression model for predicting productivity at 1 year follow-up. After adjustment for all other predictors, age and total CAP made unique contributions to prediction of this outcome. Participants at the 75th percentile of age had only .39 times the odds of being productive as compared with younger participants at the 25th percentile of age. Stated another way, participants at the 25th percentile of age had 2.56 times the odds of being productive as those at the 75th percentile. Participants at the 75th percentile of confusion severity had only .50 times the odds of being productive as those at the 25th percentile. Stated another way, less confused participants were twice as likely to be productive at 1 year follow-up as participants who were more confused on the 21-day postinjury CAP evaluation. Again there was a trend for those with higher preinjury levels of education to have more favorable outcomes.

Table 3. Multivariable Logistic Regression Model for Predicting Productivity at 1 Year Postinjury (n=132)

	Predictor	Comparison	OR	95% CI	P
	Age	20.2,43.9	0.39	0.18–0.87	.02
	Education	10.0,13.0	1.65	0.92–2.97	.09
	GCS	5.0,10.2	1.40	0.73–2.68	.31
	TFC	1.0,11.5	0.71	0.43–1.18	.18
	Total CAP	2.0,5.0	0.50	0.27–0.93	.03

Simple Binary Logistic Regression Analyses 

Results of simple binary logistic regression analyses examining prediction of employability and productivity outcomes from the presence or absence of each of the 7 symptoms of confusion at the 21-day postinjury CAP evaluation are presented in table 4. Each symptom showed a simple (unadjusted) effect on employability at rehabilitation discharge. ORs ranged from 1.85 to 7.14. Participants who were not rated as having nighttime sleep disturbance on the 21-day postinjury CAP were 1.85 times as likely to be employable as those who did have nighttime sleep disturbance. Participants who were not disoriented on the 21-day postinjury CAP were 7.14 times as likely to be employable as those who were disoriented. ORs for other symptoms ranged from 3.57 to 4.54.

Table 4. Simple Binary Logistic Regression Analyses Examining Prediction of Employability (N=168) and Productivity (n=132) from the 7 Symptoms of Confusion

	Symptoms	Employability			Productivity
		OR	95% CI	P	OR	95% CI	P
	Disorientation	7.14	3.33–14.28	.001	3.12	1.45–6.67	.004
	Cognitive impairment	3.57	1.89–7.14	.001	4.54	2.12–10.00	.001
	Fluctuation	4.54	2.22–9.09	.001	3.33	1.49–7.14	.003
	Agitation	3.85	1.96–7.69	.001	3.12	1.45–6.67	.004
	Nighttime sleep disturbance	1.85	1.00–3.44	.050	1.37	0.67–2.78	.393
	Decreased daytime arousal	3.57	1.61–7.69	.002	1.69	0.73–4.00	.212
	Psychotic-type symptoms	4.17	1.78–10.00	.001	14.28	3.03–50.00	.001

Presence of all symptoms except nighttime sleep disturbance or daytime decreased arousal had effects on productivity at 1 year follow-up. These ORs ranged from 3.12 to 14.28. Participants who were not disoriented at the 21-day postinjury CAP were 3.12 times as likely to be productive at 1 year follow-up as those who were disoriented. The effect for agitation was of the same magnitude with those who were not agitated having 3.12 times the odds of being productive as compared with those who were agitated. Participants who did not have psychotic-type symptoms at the 21-day postinjury CAP were 14.28 times as likely to be productive at 1 year follow-up as those who did have psychotic-type symptoms. ORs for cognitive impairment and fluctuation were 4.54 and 3.33, respectively.

Study Limitations 

We regard the current findings as preliminary and requiring replication in new samples. For the current study sample, detailed histories of premorbid psychiatric history or family psychiatric history were not obtained. It is possible that these factors may influence the patterns of symptoms of confusion seen in these patients. Also, the severity of confusion at the time of the CAP evaluation was confounded with the time interval from injury until the CAP evaluation. As noted above, this was due to more confused patients having longer acute care hospital LOSs. For this investigation, we included participants not meeting PTC criteria because, as shown in figure 1, these persons often have some symptoms of confusion. A later study could investigate the impact of confusion severity on outcome among only those patients who meet criteria for PTC.