Predicting health-related quality of life 2 years after moderate-to-severe traumatic brain injury


N. Andelic, Department of Physical Medicine and Rehabilitation, Oslo University Hospital, PB 4950 Nydalen, 0424 Oslo, Norway

Tel.: +47 22118687

Fax: +47 22115022




To describe health-related quality of life (HRQL) 2 years after moderate-to-severe traumatic brain injury (TBI) and to assess predictors of HRQL.

Materials and methods

A prospective cohort study of 91 patients, aged 16–55 years, admitted with moderate-to-severe TBI to a trauma referral centre between 2005 and 2007, with follow-up at 1 and 2 years. Mean age was 31.1 (SD = 11.3) years, and 77% were men. Injury severity was evaluated by the Glasgow Coma Scale (GCS), head CT scan (using a modified Marshall Classification), Injury Severity Score (ISS) and post-traumatic amnesia (PTA). The Functional Independence Measure (FIM), Community Integration Questionnaire (CIQ), Beck Depression Inventory (BDI) and Medical Outcomes 36-item Short Form Health Survey (SF-36) were administered at follow-up visits. The main outcome measures were the Physical Component Summary (PCS) and Mental Component Summary (MCS) of the SF-36.


HRQL appears to be relatively stable between 1 and 2 years after injury. In the multivariate linear regression, younger age (β = −0.20, P = 0.032), more severe TBI (β = 0.28, P = 0.016), more severe overall trauma (β = 0.22, P = 0.026), higher levels of community integration (β = 0.36, P = 0.019) and higher positive change in PCS scores from 1 to 2 years (β = 0.41, P < 0.001) predicted better self-reported physical health 2 years post-TBI. Lower scores for depression (β = −0.70, P < 0.001) and a higher positive change in MCS scores (β = 0.62, P < 0.001) predicted better self-reported mental health.


Future interventions should focus on aspects related to HRQL that are more easily modified, such as physical functioning, home and social integration, productivity, and mental and emotional status.


Traumatic brain injury (TBI) is common in young and otherwise healthy adults, resulting in a high cost to society due to the years of life lost to disability or death [1]. In the case of TBI, a single traumatic incident may result in long-lasting or permanent physical, mental and social impairments and diminished quality of life [2].

Over the past decade, patients’ subjective perception of their physical, emotional and social well-being (i.e. health-related quality of life, or HRQL) has been recognized as an important outcome of TBI rehabilitation [3]. Several studies have found that TBI patients report a lower HRQL compared with the general population [3-6]. HRQL may change over time because recovery from TBI is a long and complex process. The course of recovery may vary due to a number of factors, such as severity of the injury, time since the injury and the domain of functioning that is being assessed [5]. However, some long-term follow-up studies found little improvement in HRQL over time, and factors related to HRQL in the long term remain similar to those found in more acute studies [7].

The literature reveals conflicting results concerning the relationship between injury severity and HRQL [3, 8-11]. However, there is an agreement that improved physical functioning, perceived mental health, participation in productive activities and social supports improve HRQL [12, 13].

In Scandinavia, there have been several studies that address HRQL related to long-term follow-up after TBI [3, 4], but further studies are warranted to examine HRQL profiles in a shorter time perspective, such as 1–2 years after moderate-to-severe TBI.

In this study, we aimed to (i) describe HRQL 2 years after moderate-to-severe TBI, (ii) explore the change in HRQL from 1 to 2 years, (iii) assess preinjury and injury-related factors and post-injury functioning as predictors of physical and mental health at 2 years and (iiii) compare HRQL in our study sample with another TBI population in the USA [6] and the general population of Norway [14].

Materials and methods


Two-year follow-up of a prospective cohort of patients with acute TBI admitted to the Oslo University Hospital, a trauma referral centre for the south-east region of Norway from May 2005 to May 2007.

The inclusion criteria were (i) age 16–55 years, (ii) residence in eastern Norway, (iii) admission with ICD-10 diagnosis S06.0–S06.9 within 24 h of injury and (iv) moderate-to-severe TBI with a Glasgow Coma Scale (GCS) score of 3–12 before intubation [15].

The exclusion criteria were (i) previous neurological disorders, (ii) associated spinal cord injuries, (iii) severe psychiatric or substance abuse disorders and (iv) unknown address or incarceration.

A total of 160 patients met the inclusion criteria. Of these, 27 (17%) refused to participate, thus 133 patients remained. Twenty-three (17%) patients died in acute or post-acute care and nine (7%) patients dropped out of the study before the 2-year follow-up. Ten (8%) patients were excluded because they were unable to follow commands, comprehend or answer questions properly resulting in a total study sample of 91 (68%) patients.


HRQL was measured by the Medical Outcomes 36-Item Short Form Health Survey (SF-36) [16]. In this study, SF-36 was assessed at both one and 2 years. The SF-36 measures HRQL along eight subscales: physical function (PF), role limitations due to physical health (RP), bodily pain (BP), general health (GH), vitality (VT), social function (SF), role limitations due to emotional health (RE) and mental health (MH). In addition, a single item reports the changes in overall health over the past year. Raw scores were transformed into a scale score from 0 to 100 (worst to best). The subscales were weighted and calculated into the Physical Component Summary (PCS), consisting of the first four SF-36 subscales (PF, RP, BP and GH) and the Mental Component Summary (MCS), consisting of the latter four SF-36 subscales (VT, SF, RE and MH). The results were converted into standardized T-scores, with a mean value of 50 ± 10. We computed these scores at the SF-36 website (, which is based on Norwegian normative data. The internal consistency of the PCS subscales and MCS subscales in the present study was measured with Cronbach's alpha and found satisfied (α = 0.74 and α = 0.80, respectively). Physical domains subscales correlated strongly (r > 0.6) with PCS and weakly (r < 0.3) with MCS, while mental domains subscales displayed the opposite relationship, except for RP and GH (r = 0.34 and r = 0.40, respectively). The correlation between PCS and MCS approached zero (r = −0.04), indicating the validity of the construction of the summary scores [16].

Preinjury factors

Age, gender, marital status, education, employment and substance use (e.g. alcohol or/and drugs) at the site of injury (revealed by clinical and/or laboratory evaluation) were recorded in the acute phase.

Injury-related variables

Recorded variables related to injury included Glasgow Coma Scale (GCS) score, computed tomography (CT) head scan (using modified Marshall Classification), Injury Severity Score (ISS) and length of post-traumatic amnesia (PTA).

A modified Marshall Classification [17] was used to classify TBI severity into less severe (no visible injury or small intracranial injury, score < 3) and more severe (significant intracranial abnormalities, score ≥3) as shown on a CT scan. The ‘worst’ scan within the first 24 h was used for severity classification in this study.

ISS [18] assesses patients’ overall trauma score, taking into account multiple injuries. The ISS range from 1 to 75 (best to worst), and an ISS score >15 indicates a patient with major trauma.

The duration of PTA was defined as the number of days between the injury and reaching a total score ≥76 on the Galveston Orientation and Amnesia Test (GOAT) [19].

Post-injury functional level at 1 year

Functional status 1-year post-TBI was measured by Beck Depression Inventory (BDI), Functional Independence Measure (FIM) and Community Integration Questionnaire (CIQ), all three frequently used in TBI research [20-22].

Beck depression inventory [23] is a 21-item self-report instrument designed to screen for depression, primarily cognitive and affective symptoms. BDI scores range from 0 to 63; scores >12 indicate depression [24].

FIM [25] is an 18-item scale that evaluates specific activities of daily living (ADL), with a total score range from 18 to 126. Higher scores indicate greater independence in ADL; scores ≤108 indicate the need for assistance.

CIQ [21] is a 15-item scale with a total score of 29 within three domains: home integration (score range 0–10), social integration (0–12) and productive activities (0–7). Higher scores indicate greater integration and fewer participation restrictions.


Demographic variables (age, gender, education, marital status, preinjury employment and substance use), injury-related characteristics (cause of injury, GCS, ISS, CT head scans, PTA) and length of hospital stays were collected during admissions for acute TBI. At the follow-up visits, the physiatrist at the outpatient department assessed patients. The FIM, CIQ, BDI and SF-36 were administered.

The study was approved by the Regional Committee for Medical Research Ethics, East Norway. We received written informed consent from all participants in the study.


The Statistical Package for the Social Sciences (SPSS) version 18 was used. Statistical significance was set at P = 0.05. Descriptive data are shown by proportion, mean values with standard deviations (SD) and median with interquartile range (IQR). Spearman's Correlation, t-tests and Mann–Whitney U-tests were conducted for continuous variables, and chi-squared tests were used for categorical variables to examine HRQL differences in demographic and injury characteristics. Paired t-tests were used to compare SF-36 subscales, PCS and MCS from 1 and 2 years of follow-up.

The dependent variables in the two regression models were PCS and MCS at 2 years. Of the independent variables (Table 1), those with P < 0.10 from the simple regression were entered into a multiple linear regression model to quantify their predictive impact on PCS and MCS (Table 2). CT scores and ISS did not reach the probability level of < 0.10, but they were entered into multiple regression analysis as indicators of injury severity. Hierarchical linear regression models were performed, and results are presented as R2, R2 change and standardized beta values (Table 2). Before conducting the multiple regression analysis, possible multicollinearity was examined using the variance inflation factor (VIF). Distribution of the residuals was examined for normality, and influential data points were examined using Cook's distance. The PCS and MCS, CIQ and BDI were normally distributed while FIM was skewed and therefore log-transformed when performing regression analyses. None of the variables showed correlations between each other at r > 0.7 (data not shown).

Table 1. Demographics, injury characteristics and functional status 1-year post-TBI (n = 91)
  1. *PTA was reported for 87 patients. BDI score was reported for 85 patients.

Sex, n (%)
Male70 (76.9)
Female21 (23.1)
Age (time of injury), years
Mean (SD)31.1 (11.3)
Living arrangements (time of injury), n (%)
Married, cohabitant, living with family54 (59.3)
Divorced, living alone37 (40.7)
Education (time of injury), n (%)
≤12 years52 (57.1)
>12 years39 (42.9)
Employment (time of injury), n (%)
Employed77 (84.6)
Unemployed14 (15.4)
Substance use (at the site of injury), n (%)
Yes40 (44.0)
No51 (56.0)
Cause of injury, n (%)
Traffic accidents51 (56.0)
Falls24 (26.4)
Violence10 (11.0)
Other6 (6.6)
Mean (SD)7.4 (3.1)
CT-Marshall scores, n (%)
< 346 (50.5)
≥345 (49.5)
ISS (acute phase)
Mean (SD)29.5 (13.1)
PTA, days (acute phase)*
Mean (SD)27 (30)
Length of stay in acute care
Median (IQR)22 (22)
Rehabilitation length of stay
Median (IQR)31 (63)
FIM (1-year post-injury)
Mean (SD)120.8 (10.0)
CIQ total (1-year post-injury)
Mean (SD)19.2 (5.2)
BDI score (1-year post-injury)
Mean (SD)10.7 (8.6)
Table 2. Results from the multiple hierarchical regression models of SF-36 component summary scores (n = 91)
Step 1Step 2Step 3Step 4Step 1Step 2Step 3Step 4
  1. Standardized beta coefficients are presented. Significance levels: *0.05; 0.01;<0.001.

  2. Employment status and substance use at the site of injury.

Age at injury−0.145−0.151−0.160−0.201*−0.060−0.0650.0280.089
Substance use0.236*0.2110.1690.136−0.204−0.161−0.119−0.055
CT-Marshall 0.261*0.3360.282* 0.0620.026−0.004
ISS 0.1000.1450.223* 0.067−0.011−0.008
PTA −0.405−0.293−0.242 0.0440.0300.080
FIM  −0.0330.150  −0.239−0.042
CIQ  0.4430.362*  −0.032−0.095
BDI  0.1880.160  −0.570−0.699
PCS change   0.412    
MCS change       0.615
R 2 0.1130.2260.3050.4520.1430.1590.3610.689
Adjusted R20.0800.1650.2200.3770.1100.0930.2830.646
R2 change0.1130.1130.0790.1470.1430.0160.2020.328


Data on the 91 participants are presented in Table 1. The mean age was 31.1 (SD 11.3) years, and 77% were men. According to the GCS, 64.8% of patients suffered severe TBI. On CT head scans, 49.5% sustained severe TBI with significant intracranial abnormalities. According to the ISS, 85% of the patients were determined to have major trauma. The mean PTA was 27 days (range 6 h-128 days).

According to the FIM, only 9% of patients were in need of personal assistance (≤108), while 48.4% were highly independent in ADL (FIM score of 126) at 1 year. The mean CIQ subscale scores were 6.0 (SD 2.8) for home integration, 9.2 (SD 2.2) for social integration and 4.0 (SD 2.1) for productive activities. Of 49 working patients at 1 year, 69% worked more than 20 h per week. The total number of patients working at the 2-year follow-up declined slightly from 49 to 43, whereas the proportion of those working more than 20 h per week was stable. The BDI showed that 28% of patients were depressed at 1 year, slightly increasing to 34% at the 2-year follow-up.

HRQL 2 years after moderate-to-severe TBI

The mean scores of SF-36 subscales are shown in Fig. 1. There were no significant differences between men and women or married and unmarried patients for any of the SF-36 subscales. Patients aged >31 years showed significantly lower scores for RP (P = 0.05) than those aged ≤31 years, while BP approached significance (P = 0.06). Patients with education >12 years showed a significantly better PF (P = 0.05). Patients who were employed at the time of injury showed significantly better HRQL in six SF-36 subscales when compared with unemployed patients.

Figure 1.

Mean scores of SF-36 subscales and T-scores of physical and mental component summary (PCS and MCS) of 91 patients 1 and 2 years after TBI.

There were no significant differences in SF-36 subscales between patients with less severe vs more severe intracranial injuries. Patients with less severe overall trauma had worse BP scores (P = 0.003). Significant differences were found between independent patients (FIM > 108) and those needing assistance (FIM ≤ 108) in PF (P = 0.001), RP (P = 0.03), GH (P = 0.02) and SF (P < 0.001). According to the CIQ, significant differences were found between patients, with a total CIQ score ≤19 compared with >19 in six SF-36 subscales, but not in BP and RE. Depressed patients (BDI > 12) scored significantly poorer in all SF-36 subscales compared with non-depressed patients.

Change in HRQL from 1 to 2 years

There were no significant differences in the mean scores of SF-36 subscales between 1 and 2 years (see Fig. 1). On the SF-36 single item that assesses the patient's changes in HRQL over the past year, 63% reported some improvement or no difference, and only 11% experienced poorer HRQL at 2-year follow-up. Finally, 26% of patients reported better HRQL at 2 years compared with the year before.

Predictors of HRQL at 2 years

In the prediction analysis, the PCS and MCS scores were used as dependent variables. At 2 years, the mean PCS was 42.6 (10.7) and median 44.4 (IQR 16). The mean MCS was 44.7 (11.9) and median 46.1 (IQR 16). The four blocks of independent variables were (I) age, employment and substance use at the site of injury, (II) CT-Marshall scores, ISS and PTA, (III) functioning at 1 year (FIM, CIQ and BDI scores) and (IV) changes in PCS and MCS scores between 1 and 2 years. The results of the hierarchical multiple regression models are presented in Table 2.

The final PCS model explained 38% (adjusted R2 = 0.38) of the variance in PCS, while the MCS model explained 65% (adjusted R2 = 0.65) of the variance in MCS.

HRQL comparison with the TBI population in the USA and the general population of Norway

The study population 2 years post-TBI had consistently lower HRQL compared with the Norwegian general population, but was quite similar to the USA study population at least 1-year post-TBI (see Fig. 2).

Figure 2.

Mean SF-36 subscale score profile; the study population (n = 91) at 2-year post-injury compared to the general population of Norway (n = 2323) and a group of patients with moderate-to-severe TBI from USA (n = 228).


The present study is the first Scandinavian study to investigate HRQL 2 years after moderate-to-severe TBI. The results indicate that HRQL appears to be relatively stable from 1 to 2 years of follow-up. Lippert-Gruner et al. [26] reported significant improvements in SF-36 subscales of severe TBI patients from 6 months to 1 year. A plateau of functional recovery is reported between 1 and 2 years post-injury [2], which may explain our findings of stable HRQL. These findings are in accordance with a previous study from the USA [22], which reported relatively stable SF-36 mean scores over the 5 years after discharge from TBI rehabilitation. However, the present study is in contrast with a published paper on moderate-to-severe TBI from China that reported that HRQL significantly improved from 1 to 2 years after discharge from a trauma centre [27]. Ethnic/cultural differences and different study methodologies may play a role in these contradictory results.

No significant gender differences were found on SF-36 subscales, in contrast with other studies reporting lower HRQL for women suffering TBI [5, 27, 28] and in the general population [14]. Our findings may reflect the low proportion of women in this study and the limited age range (16–55 years). In line with previous results, patients employed at the time of injury reported significantly higher scores on SF-36 subscales, reflecting the importance of a productive lifestyle to HRQL [13]. There were no significant differences in SF-36 subscales between TBI severity groups, in line with previous long-term outcome studies [4]. Patients with fewer limitations in functional ability showed higher mean scores in SF-36 subscales belonging to the physical component, reflecting better physical capacity of these patients. Patients with fewer restrictions in community integration showed higher mean scores in SF-36 subscales, suggesting the importance of reintegration into society to achieving a better HRQL. Depressed patients showed lower mean scores in all SF-36 subscales, reflecting the fact that emotional status influences HRQL [28].

In this study, a moderate proportion of the variance in PCS (38%) at 2 years was predicted by preinjury variables (age, preinjury employment and substance use), injury severity and functional level (functional ability, emotional status and participation) at 1 year. Physical functioning is highly affected by age [14]. As expected, lower age predicted better PCS in this study. The severity of intracranial injury was clearly related to the PCS in all regression steps (i.e. less severe intracranial injury predicted worse PCS), making the results inconsistent with some studies [5, 27], and in line with others [8, 9, 29]. One possible explanation is that individuals who are afflicted with fewer problems overall may be more bothered by physical problems [9]. In addition, individuals with more severe injuries may have an altered awareness of their functional difficulties and may not distinguish their problems in the same way as those who are less severely injured [7]. Another possible explanation is that our patients had a relatively high functional level preinjury, and thus, residual physical limitations may influence their ability to regain their former physical function, resulting in a lower PCS. ISS was related to the PCS in the final regression model, showing that patients with less severe overall trauma experienced poorer physical functioning. The explanations mentioned above may be applicable here as well. However, some studies found that ISS independently predicted the PCS in severely injured patients [30].

Community integration was one of the strongest predictors of PCS results, indicating that patients with a higher capability of being integrated will show higher PCS scores. Previous studies have reported that productivity is a cornerstone to achieving a high HRQL [3] through greater self-fulfilment and increased opportunities [13]. As expected, the gains in physical health from 1 to 2 years were also a strong predictor of higher PCS scores in this study.

A large proportion of the variance in the MCS (65%) was predicted in this study by depression and changes in mental health from 1 to 2 years post-TBI. As expected, fewer depressive symptoms at 1-year post-injury and positive changes in mental health resulted in higher MCS scores. These results are in line with other studies that found that depression is strongly associated with overall HRQL [31, 32]. According to Granger et al. [31], depression accounts for nearly 50% of the variance in subjectively expressed quality of life. Findler et al. [6] reported high correlations between the SF-36 mental health subscales scores and the BDI. Neither injury severity nor community integration predicted MCS scores. One possible explanation for the lack of association may be that the MCS described symptoms associated with depression [33]. However, injury-related cognitive symptoms may influence these results, as we evaluated patients with the BDI, which primarily screens for the cognitive and affective symptoms [23]. Roughly, one-third of patients in this study were considered to be depressed, in accordance with previous TBI studies [34]. A lack of adequate patient evaluation, inadequate medical treatment and low treatment compliance are possible reasons for the persistence of depressive symptoms [34].

In this study, population norms were used as a reference group [14]. Population norms have been used to provide reference values for post-injury HRQL data [35]. As expected, the study sample showed lower HRQL in all SF-36 domains compared with the general population of Norway [14]. This finding is consistent with the international literature that compared the HRQL of TBI patients with normative groups of people [6]. The fact that the HRQL reported in this study is similar to the American SF-36 validation study of the TBI population strengthens our findings [6].

This study has several limitations that should be addressed. The age span and the TBI severity may limit generalization of the results. However, the demographic and injury characteristics were similar to other studies on moderate-to-severe TBI [10, 36]. As the level of consciousness in the early phase of TBI might be obscured because of substance influence, medical sedation, seizures, shock, etc., we cannot rule out a potential bias introduced by using early GCS score as the inclusion criteria [1]. However, as CT findings are not influenced by these factors, we used CT-Marshall score as an indicator of brain injury severity in this study. The small sample size limited the number of explanatory variables used in the regression analyses. Compared with single SF-36 subscales, the PCS and MCS scores have some statistical advantages in modelling as scores are normally distributed and reliable in the general population [16]. Because patients’ self-reports on the SF-36 subscales scores are generally low (see Fig. 2), the components summary scores may be a sensitive measure to detect TBI-related physical and mental health effects. However, the SF-36 as a generic tool did not integrate disease-specific issues of TBI patients such as cognition and self-awareness.

As mentioned above, an altered awareness may influence self-reports in patients with severe TBI. In future TBI studies using SF-36, it may be wise to involve close relatives to check for reliability of patients’ self-reported HRQL.


Our results indicated that HRQL after moderate-to-severe TBI appeared to be relatively stable from 1 to 2 years of follow-up. Younger age, more severe intracranial injury and overall trauma, higher level of community integration at 1 year and positive changes in physical health from 1 to 2 years were significant predictors of better PCS. However, no depression and positive changes in mental health from 1 to 2 years of follow-up were associated with better MCS. Future interventions should focus on targeting more modifiable aspects related to HRQL, such as physical function, home and social integration, productivity and emotional status. In this context, it is important to point out a positive impact of physical exercises in the amelioration of physical and emotional deficits induced by TBI [37].


The authors would like to thank all the subjects for their participation. Special thanks to Morten Hestnes for the extraction of trauma scores from the Trauma Register, and Tone Jerstad, neuroradiologist, for the CT assessments. This study was funded by grants from the Norwegian Health South-East Authority and The Research Council of Norway.

Conflict of interest