Heterogeneity of health-related quality of life after mild traumatic brain injury with systemic injury: a cluster analytic approach

Abstract

Purpose

Health-related quality of life (HRQoL) is a key component of evaluating outcome after mild traumatic brain injury (mTBI). As outcome is heterogeneous following mTBI, it is relevant to examine individual differences in HRQoL. This study investigated whether multiple homogenous subgroups could be meaningfully identified, 10 weeks after hospitalised mTBI with systemic injury, on the basis of HRQoL profiles.

Methods

Ninety-one adults were assessed for HRQoL, pain, fatigue, sleep quality, psychological distress, cognition and post-concussion symptoms.

Results

Cluster analyses revealed three separate subgroups based on physical, mental, social and energy HRQoL. One group (42%) demonstrated normative levels of HRQoL on all subdomains. The remaining two groups demonstrated significantly reduced HRQoL on all subdomains. These groups had equivalently poor mental, social and energy HRQoL, but the smallest group (27%) had significantly poorer physical HRQoL. Multinomial logistic regression revealed that pain significantly and independently predicted group membership for the particularly poor physical HRQoL group. Fatigue was the only significant independent predictor of group membership for the remaining group with reduced HRQoL.

Conclusion

These findings suggest more than 50% of hospitalised individuals with mTBI and systemic injury, have reduced HRQoL, 10 weeks after mTBI. Pain and fatigue warrant clinical attention in these individuals.

IMPLICATIONS FOR REHABILITATION

• Mild traumatic brain injury is a common event that has been shown to be associated with persistently reduced health-related quality of life in approximately 50% of individuals 6 to 12 months after injury.

• Health-related quality of life likely varies between individuals after injury

• Relative to the “normal” population, most individuals in this cohort of individuals with mTBI and systemic injury had reduced mental, social and energy quality of life 10 weeks after injury.

• Fatigue and pain are important factors in reduced health-related quality of life after mTBI with systemic injury.

• Further research is needed to determine whether these fatigue and pain issues are related to mTBI-factors, such as headache, and/or related to systemic injury factors, which are common in the mTBI population.

Keywords:

• Mild traumatic brain injury
• quality of life
• fatigue
• pain
• post-concussion symptoms
• psychological distress

Mild traumatic brain injury (mTBI) is common in the general population (excluding professional athletes and war veterans), with more than 100–300 hospital-treated cases per 100 000 individuals occurring each year [1]. It has been recognised that health-related quality of life (HRQoL) is important to study after mTBI [2] as it is a key component of evaluating outcome after injury [3]. Health-related quality of life measures the subjective experience of how an illness affects the physical, mental and social components of an individual’s life and provides standardised information on recovery [2–4]. Cross cultural research has indicated that spiritual, functional impacts, economic and environmental dimensions are also relevant aspects of QoL, but physical, mental and social aspects remain the key consensus components of HRQoL [3].

Previous research has reported that HRQoL after mTBI is significantly poorer than age and gender matched normative population data at both 3 and 12 months after injury [5], although others have found HRQoL is consistent with general population levels by 1 year post-injury [6]. A range of factors have been associated with poorer HRQoL after mTBI in adults, including increased levels of depression and PTSD symptomatology [7], presence of post-concussion syndrome (PCS) [8], increased PCS symptoms [5,9], impaired cognition [10], older [6,11] and younger [10] age, and female sex [6,10–12]. Surprisingly, given systemic/non-brain related injury being common after mTBI [13], there has been no consideration of the impact of pain and fatigue/sleep difficulties on HRQoL after mTBI.

At a group level, each of the different components of HRQoL (physical, mental and social) have been found to be reduced after mTBI, at least during the first 6–9 months after injury [7,8,14]. This contrasts the longstanding finding that complete recovery occurs for 80% of adults who experience a mTBI within 6–12 weeks after injury [15]. Given this discrepancy between traditional measures of recovery and subjectively reported quality of life, greater understanding of the factors contributing to HRQoL after mTBI is needed. In particular, improved understanding of HRQoL after mTBI will likely result in identification of “at risk” HRQoL factors and/or profiles, which can influence clinical management and intervention decisions, and facilitate better patient outcomes.

Although previous research has examined factors that are associated with HRQoL, as well as the trajectory of HRQoL changes over time [11], there is a dearth of research that has considered individual differences in how HRQoL is perceived after mTBI. Individual differences may be particularly relevant because HRQoL is a subjective concept. Given that mTBI diagnosis can include heterogeneous injury severities [16,17] and outcomes are also heterogeneous [1,18], it is likely that variations exist between individuals’ HRQoL profiles. Cluster analysis is a useful approach to examine differences between individuals’ profiles as it determines whether latent subgroups exist that have different HRQoL profiles. Specifically, it creates relatively homogenous subgroups from heterogeneous data, based on patterns of performance and similarities across levels [19,20]. This can enable identification of groups that may benefit from different management approaches [19].

The aim of this study was to determine if multiple subgroups could be meaningfully identified in a cohort of premorbidly healthy adults, approximately 10 weeks after being admitted to hospital following systemic injury and mTBI, on the basis of physical, mental and social HRQoL. This timeframe was chosen to examine individuals at the end of the “typical” recovery period (6–12 weeks). A further aim was to validate these clusters with respect to demographic, clinical and cognitive factors. It was hypothesised that separate clusters of individuals with distinct HRQoL profiles would be identified and that there would be significant differences between these clusters with respect to demographic, clinical and cognitive factors that predict group membership.

Method

Ethical approval was obtained from the Human Research Ethics Committees of The Alfred hospital and Royal Melbourne Hospital.

Participants

Participants comprised 91 adults (69 male, 22 female) who had experienced a mTBI between September 2015 and February 2020 and been consecutively admitted to The Alfred hospital or Royal Melbourne Hospital, Melbourne, Australia. Initial hospital admission occurred as a consequence of experiencing any traumatic injury (i.e., primarily systemic and/or head). All individuals who had been admitted with traumatic injury were considered potentially eligible. Determination regarding whether an individual fulfilled criteria for mTBI was based on medical record data together with data obtained from structured clinical interview. Identical recruitment processes were undertaken at each hospital. Assessment was undertaken 6–12 weeks after injury, in the outpatient department of The Alfred hospital.

A mTBI event was defined according to World Health Organisation criteria [21]. Briefly, individuals needed to demonstrate one or more of: (i) confusion or disorientation, loss of consciousness for 30 min or less, post-traumatic amnesia (PTA) less than 24 h, and/or other transient neurological abnormalities not requiring surgery; (ii) Glasgow Coma Scale (GCS) score of 13–15 after 30 min, or later upon presentation for healthcare. Exclusion criteria were: any previous neurological history (other than <3 previous self-reported non-specific “concussion” events, which were not associated with any medical attention, diagnosis or self-reported consequence; 75% of the sample had 0 previous self-reported “concussions”), diabetes, history or current use of significant drug use, history of or current heavy alcohol consumption (>5 standard drinks/day), history of diagnosis or treatment for any significant psychiatric disorder, current/recent (during previous 12 months) diagnosis or treatment of depression and/or anxiety and/or post-traumatic stress disorder, current TBI as a result of physical assault/attack and lack of conversational English fluency.

Measures

Measure of health-related quality of life

The RAND 36-Item Short Form (SF-36) Health Survey (version 1) [22] was used to measure self-reported HRQoL. The SF-36 is one of the measures that are recommended for examining generic HRQoL with the TBI population [2]; it examines physical, mental and social functioning and is valid and reliable in TBI populations [23]. Eight domains of QoL can be derived from the SF-36, with internal reliability ranging from acceptable to excellent. These domains are: Physical Functioning (PhysFxn; α = 0.93), Emotional Well-Being (EmotWellBeing; α = 0.90), Energy/Vitality/Fatigue (Energy; α = 0.86), Social Functioning (SocFxn; α = 0.85), Role limitations due to physical health (α = 0.84), Role limitations due to emotional problems (α = 0.83), Pain (α = 0.78) and General Health (α = 0.78). The current study sample size was sufficient for cluster analyses derived from 4 variables [24]. Consequently, only 4 of the 8 SF-36 domains were analysed. As key components of HRQoL [3], the measures of physical, mental and social QoL from the SF-36 (PhysFxn, EmotWB, SocFxn) were included in data analyses. There is a longstanding literature demonstrating that lack of energy/fatigue comprises both a physical and mental component [25], is prevalent at least 6 months after mTBI and affects QoL [26]. Given this, together with the good internal reliability of the Energy domain, Energy QoL was included in cluster analyses.

Measure of post-concussion symptoms

The Rivermead Post Concussion Symptoms Questionnaire (RPQ) is a widely used measure of PCS. It assesses physical (10 items), psychological (3 items) and cognitive (3 items) symptoms experienced during the past 24 h [27]

Measures of subjective sleep disturbance, fatigue and pain

The Pittsburgh Sleep Quality Index (PSQI) [28] was used to measure self-reported sleep disturbance. The PSQI is an 18-item questionnaire from which a total score, Global PSQI, can be obtained. It has high levels of internal consistency (Cronbach’s alpha = 0.83) [28] and has been validated in mTBI samples [29]. The Multidimensional Fatigue Inventory (MFI) [30] was used to measure fatigue. The MFI is a 20-item questionnaire from which a total score, MFI Total, can be obtained. It has good internal consistency (Cronbach’s alpha > 0.076) and has been used in the mTBI population previously [25]. The Short-Form McGill Pain Questionnaire (SF-MPQ-2) [31] was used to measure pain. It has excellent reliability and validity and has been used with the mTBI population [32]. An overall pain value (Total Pain) can be derived as a unitary measure of pain.

Measures of depression, anxiety and post-traumatic stress

Three widely used, valid and reliable questionnaires of psychological distress were used: The Inventory of Depressive Symptomatology (IDS) is a 30-item measure of severity of overall depression [33]. The Beck Anxiety Inventory (BAI) is a 21-item measure of anxiety symptomatology [34]. The PTSD Checklist for the DSM-5 (PCL-5) [35] is a 20-item measure of the symptoms of PTSD defined by DSM-5. To facilitate data reduction, a single variable of Psychological Distress was created from standardised performances on the IDS, BAI and PCL-5. The IDS and BAI are 4-point scales (range 0–3), whereas the PCL-5 is a 5-point scale (0–4). Consequently, each response on the PCL-5 was multiplied by 0.75, resulting in all measures having item responses on a 4-point scale (range 0–3). Performance on these measures were then summed together, resulting in a single index of psychological distress, with a possible range of 0–213.

Measures of cognition

The Symbol Digit Modality Test – oral version (SDMT) is a measure of processing speed that is sensitive to cognitive impairment after mTBI [36]. On this version of the SDMT, the final score was number of correct items verbalised within 2 min. The Digit Span (DSp) subtest from the Wechsler Adult Intelligence Scale – 4th Edition (WAIS-IV) was used as the measure of attention; it is a valid, reliable and widely-used measure [37]. The Rey Auditory Verbal Learning Test (RAVLT) is a reliable and valid measure of memory [38]. The total number of items learned on the 5 list learning trials (Total) was used to assess memory function. The Trail Making Test Part B (TMTB) was used as a measure of mental flexibility as it has high reliability and validity [39].

Assessment of performance validity

The DSp subtest from the WAIS-IV [37] was used as a measure of effort [40]. Participants were identified as having problematic effort on testing if they failed on the subscales of Age Scaled Score Total (Fail = 5 or less) and Longest Digits Forward (Fail = 4 or less) [40], which have been shown to have a likelihood ratio that successfully identifies poor effort [41].

Procedure

Participants with mTBI were initially recruited into the study on the ward within 1–4 days after injury, during their inpatient stay. Written informed consent to participate was given followed fulfilment of inclusion/exclusion criteria. Following discharge, participants were contacted by phone and attended The Alfred hospital to undertake individual assessment 6–12 weeks after injury. All participants completed the measures in the following order:SDMT, RAVLT, DSp, TMT, RPQ, SF-36, MFI, PSQI, IDS, BAI, PCL-5, SF-MPQ-2.

Data analysis

Statistical assumptions were investigated prior to data analysis. Only violations of assumptions will be reported. To identify subgroups with homogeneous HRQoL profiles, a two-phase approach to cluster analysis was undertaken. Initially hierarchical agglomerative cluster analysis, using Ward’s minimum variance method and Squared Euclidean Distance, was undertaken. To determine the number of clusters to retain, inspection of the dendrogram and coefficients of agglomeration schedule was conducted [42]. In the subsequent phase a k-means iterative partitioning technique was used to optimise the solution determined by the initial phase [19]. Repeated measures multivariate analysis of variance (MANOVA) was undertaken to analyse the profile of the clusters across the 4 HRQoL domains. Between-subject analyses were conducted to identify differences in HRQoL between cluster groups, with significant findings investigated with posthoc analyses using Games-Howell correction as equal variances could not be assumed. Flatness and parallelism of each profile was determined with Greenhouse-Geisser correction as Mauchly’s Test of Sphericity was violated. Pairwise comparisons were used to investigate any significant deviations in flatness or parallelism. Independent one-sample t-tests were used to compare cluster groups with normative data.

To investigate the external validity of the HRQoL cluster groups, multinomial logistic regression analyses were undertaken to compare the groups across a variety of demographic and sample characteristics. Initially, bivariate regression analyses were used to identify significant associations between each variable and cluster group membership. Significant bivariate predictors of cluster group membership (p < 0.05) were included into a single multinomial logistic regression model to identify independent predictors of cluster group membership. Multicollinearity was examined and found to be within statistically acceptable limits [43]. The odds ratio (OR), 95% confidence intervals of the OR, and the Wald test were determined to identify the extent of association. No adjustment for multiple comparisons was made as this can increase the rate of Type II errors, the likelihood of missing important findings as well as reducing power [44].

Results

Study recruitment details have been published previously [25,45]. Briefly, approximately 50% of those who were approached for initial recruitment agreed to participate. Of those who were not initially recruited, approximately 90% of these individuals did not fulfil eligibility criteria. The primary reason that some individuals did not eventually participate in the study, despite successful initial recruitment, was an inability to contact them to arrange assessment. Demographic details, injury characteristics and sample characteristics for the key variables of interest are presented in Supplementary Digital Content tables (Table S1 and S2). Only 11% of the sample were involved in litigation at the time of examination. Point biserial correlations between litigation status and RPQ, PSQI, MFI, SF-MPQ-2, the Psych Distress Index and measures of cognition revealed no significant relationships. Given that 90% of litigants were in two of the final cluster groups (see Table 2), point biserial analyses were re-run for each group following cluster analysis. No significant correlations were evident between litigation status and any independent variable for any cluster grouping. Consequently, litigation status was not included in subsequent analyses. No participant failed the assessment of performance validity. As expected with a hospital-admitted cohort, all participants had experienced systemic trauma injuries in addition to their mTBI.

Table 2. Sample characteristics of the separate QoL clusters.

Sample Characteristics	Fully Recovered (n = 38)	Partially Recovered (n = 28)	PartRecovPoor-PhysH (n = 25)
Age $\overline{\mathrm{X}}$ (sd)	38.92 (14.86)	34.89 (13.89)	37.48 (13.39)
Gender (%)
Female	21.05	21.43	32.00
Male	78.95	78.57	68.00
Educ. (yrs) $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	13.55 (2.14)	14.00 (2.34)	12.36 (2.22)
Litigants Total (%)	2.63	14.29	20.00
Days since injury $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	56.50 (10.01)	64.86 (9.38)	63.36 (10.39)
Injury cause (%)
MVA/MBA	18.42	32.15	52.00
Cycling	39.47	28.57	12.00
Fall	28.95	7.14	20.00
Sport	5.26	10.71	8.00
Other	7.90	21.43	8.00
GCS (%)
13	10.53	7.14	8.00
14	23.68	21.43	32.00
15	65.79	71.43	60.00
LOC (%)
None	13.16	10.72	8.00
<5 min	78.95	82.14	68.00
5–10 min	2.63	0.00	16.00
10–20 min	2.63	0.00	0.00
20–30 min	2.63	7.14	8.00
PTA (%)
<5 mins	50.00	64.29	60.00
5–60 mins	21.06	14.29	12.00
1–6 h	7.89	3.57	12.00
6–18 h	15.79	10.71	4.00
18–24 h	5.26	7.14	12.00
QoL Domains $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$
PhysFxn	87.63 (14.13)	65.54 (17.34)	18.80 (10.83)
EmotWellBeing	83.16 (10.89)	69.71 (16.38)	69.28 (17.27)
SocialFxn	77.63 (14.58)	38.84 (18.11)	36.50 (18.01)
Energy/Fatigue	61.71 (13.52)	37.68 (15.30)	43.40 (18.91)
RPQ $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	7.63 (8.45)	14.75 (10.14)	18.20 (11.73)
PSQI $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	6.53 (3.25)	8.83 (4.42)	10.45 (4.16)
MFI $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	42.13 (11.98)	62.15 (12.46)	56.80 (14.19)
SF-MPQ $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	0.73 (1.07)	1.76 (1.21)	2.38 (1.85)
Psych Distress $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	16.81 (11.36)	38.41 (25.18)	39.28 (26.04)
SDMT $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	68.32 (11.52)	65.36 (14.77)	62.26 (12.67)
DSp $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	10.71 (2.49)	11.39 (2.95)	9.44 (2.65)
RAVLT $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	53.97 (8.48)	56.79 (8.83)	49.96 (9.73)
TMT Part B (sec) $\overline{\mathrm{X}}\mathrm{ }(\mathrm{\text{sd}})$	58.47 (21.62)	57.29 (18.99)	67.58 (27.94)

Educ: Education; MVA: Motor Vehicle Accident; MBA: Motorbike Accident; GCS: Glasgow Coma Scale Score; LOC: Loss of Consciousness; PTA: Post traumatic amnesia; QoL: Quality of Life; PhysFxn: Physical Functioning; EmotWellBeing: Emotional Well Being; SocialFxn: Social Functioning; Enerygy/Fatigue: Energy/Vitality/Fatigue; RPQ: Rivermead Post Concussion Symptoms Questionnaire; PSQI: Pittsburgh Sleep Quality Index; MFI: Multidimensional Fatigue Inventory; SF-MPQ: Short Form McGill Pain Questionnaire; Psych Distress: Psychological Distress Index; SDMT: Symbol Digit Modality Test; DSp: Digit Span subtest of the WAIS-IV; RAVLT: Rey Auditory Verbal Learning Test; TMT: Trail Making Test.

After conducting the hierarchical agglomerative cluster analysis, inspection of the dendrogram (Appendix A, supplementary material) and scree plot of the agglomeration schedule coefficients (Appendix B, supplementary material) was undertaken. A three-cluster solution was considered to demonstrate good interpretability and profile separation and was therefore retained. A k-means iterative partitioning method using a three-cluster solution revealed three groups that had distinct quality of life profiles. The three-cluster solution obtained from the k-means method showed excellent agreement with the solution provided by the hierarchical cluster analysis, as illustrated by a Cohen’s κ = 0.91 [46]. Group separation, flatness and parallelism is demonstrated in Figure 1 together with published normative data [47].

Figure 1. Quality of life profiles for the three separate clusters and normative data.

This figure shows that the Fully Recovered group’s performances are equivalent to normative data, whereas the two partially recovered groups have significantly poorer health-related quality of life than normative data. Further with respect to physical function, the PartRecovPoorPhysH group has significantly poorer health-related quality of life than the Partially Recovered group.

On the basis of the profiles, the groups were named in terms of the degree of HRQoL recovery they appeared to have made from their injuries (i.e., systemic plus mTBI): Fully Recovered HRQoL, Partially Recovered HRQoL and Partially Recovered with Poor Physical Health HRQoL (PartRecovPoorPhysH). Comparison of cluster performances with normative data [47] is presented in Table 1.

Table 1. Comparison of mTBI clusters with normative data.

QoL Domains	Normative Data $\overline{\mathrm{X}}$ (SD)	Fully Recovered $\overline{\mathrm{X}}$ (SD) (n = 38)	Cohen’s d	Partially Recovered $\overline{\mathrm{X}}$ (SD) (n = 28)	Cohen’s d	PartRecovPoorPhysH $\overline{\mathrm{X}}$ (SD) (n = 25)	Cohen’s d
PhysHealth	83.20 (22.85)	87.63 (14.13)	0.202	65.54 (17.34)	0.789***	18.80 (10.83)	2.914***
EmotWellbeing	77.84 (16.49)	83.16 (10.89)	0.334	69.71 (16.38)	0.495*	69.28 (17.27)	0.519*
SocFxn	84.08 (23.09)	77.63 (14.58)	0.290	38.84 (18.11)	1.998***	36.50 (18.01)	2.098***
Energy	64.77 (20.43)	61.71 (13.52)	0.155	37.68 (15.30)	1.355***	43.40 (18.91)	1.055***

*< 0.05; ***< 0.001.

The Fully Recovered group demonstrated equivalent levels of HRQoL on all domains relative to normative data, indicating that this group’s HRQoL were at “normal” levels relative to the general population. In contrast, both the Partially Recovered group and PartRecovPoorPhysH group had significantly poorer HRQoL than normative data on all subdomains, indicating that all aspects of HRQoL were reduced for both groups.

Profile analysis revealed significant differences between the 3 clusters on all of the HRQoL domains. Within subjects analyses showed that the groups did not demonstrate flat profiles, F(2.66, 234.01) = 61.34, p < 0.001, η_p² = 0.411. A significant interaction between HRQoL domains and cluster group also indicated that the profiles were not parallel, F(5.32, 234.01) = 31.20, p < 0.001, η_p² = 0.415. The result of pairwise comparisons, which show deviations from flatness for each of the cluster groups are shown in Supplementary Digital Content Table S3. Differences between each cluster group on each domain of HRQoL are shown in Supplementary Digital Content Table S4.

These results indicate that both of the poorly recovered groups had reduced HRQoL on all domains relative to the Fully Recovered group (n = 38; 41.76%) and that the PartRecovPoorPhysH group (n = 25; 27.47%) had even poorer Physical Health QoL than the Partially Recovered group (n = 28; 30.77%). Finally, whereas those in the Fully Recovered group had significantly better Social Fxn QoL than Energy QoL, in both of the Partially Recovered groups, Social Fxn QoL was equally as poor as Energy QoL.

External validity

The sample characteristics of the separate clusters are presented in Table 2.

The results of bivariate multinomial regression analyses, which were used to compare the partially recovered groups with the Fully Recovered group on a range of sample characteristics, are shown in Table 3. The Fully Recovered group was used as the reference group, as it demonstrated levels of HRQoL that were equivalent to normative data.

Table 3. Results of bivariate regression analyses.

	Fully Recovered vs. Partially Recovered					Fully Recovered vs. PartRecovPoorPhysH
Variable	b (Std Err)	OR	95% LCI : UCI	Wald	p-value	b (Std Err)	OR	95% LCI : UCI	Wald	p-value
Age	−0.021 (0.018)	0.979	0.945 : 1.015	1.315	0.251	−0.007 (0.018)	.993	0.958 : 1.029	0.158	0.691
Sex	−0.022 (0.609)	0.978	0.297 : 3.223	0.001	0.971	−0.568 (0.585)	.567	0.180 : 1.783	0.943	0.332
Education	0.084 (0.108)	1.088	0.880 : 1.346	0.604	0.437	−0.273 (0.133)	.761	0.587 : 0.987	4.251	0.039
MVA/MVB	−0.442 (0.868)	0.643	0.117 : 3.526	0.259	0.611	1.025 (1.026)	2.786	0.373 : 20.819	0.997	0.318
Cycling	−1.386 (0.829)	0.250	0.049 : 1.270	2.795	0.095	−1.269 (1.109)	0.281	0.032 : 2.471	1.309	0.253
Fall	−2.303 (1.049)	0.100	0.013 : 0.781	4.820	0.028	−0.288 (1.065)	0.750	0.093 : 6.043	0.073	0.787
Sport	−0.288 (1.155)	0.750	0.078 : 7.210	0.062	0.803	0.405 (1.354)	1.500	0.106 : 21.312	0.090	0.765
RPQ	0.084 (0.030)	1.088	1.026 : 1.153	7.979	0.005	0.112 (0.031)	1.118	1.052 : 1.188	12.973	<0.001
PSQI	0.174 (0.079)	1.190	1.019 : 1.390	4.838	0.028	0.266 (0.084)	1.305	1.107 : 1.539	10.018	0.002
MFI	0.118 (0.026)	1.125	1.068 : 1.184	20.014	<0.001	0.083 (0.023)	1.087	1.039 : 1.137	13.037	<0.001
SF-MPQ	0.878 (0.284)	2.407	1.380 : 4.197	9.585	0.002	1.106 (0.291)	3.023	1.709 : 5.347	14.450	<0.001
Psy Distr	0.067 (0.019)	1.069	1.031 : 1.108	12.820	<0.001	0.068 (0.019)	1.070	1.032 : 1.110	13.214	<0.001
SDMT	−0.018 (0.020)	0.982	0.945 : 1.021	0.839	0.360	−0.038 (0.022)	0.963	0.923 : 1.005	3.032	0.082
DSp	0.093 (0.093)	1.097	0.915 : 1.315	1.001	0.317	−0.192 (0.105)	0.825	0.672 : 1.014	3.346	0.067
RAVLT	0.036 (0.029)	1.037	0.980 : 1.097	1.593	0.207	−0.051 (0.030)	0.950	0.896 : 1.008	2.879	0.090
TMT B	−0.003 (0.013)	0.997	0.973 : 1.022	0.057	0.812	0.016 (0.011)	1.016	0.994 : 1.039	2.016	0.153

Std Err: Standard Error; OR: Odds Ratio; CI: Confidence Interval; 95% LCI: UCI: 95% Confidence Interval of Odds Ratio – Lower bound and Upper bound; Sex: 1 was coded as Male, 2 as Female; MVA: Motor Vehicle Accident; MVB: Motorbike Accident; RPQ: Rivermead Post Concussion Symptoms Questionnaire; PSQI: Pittsburgh Sleep Quality Index; MFI: Multidimensional Fatigue Inventory; SF-MPQ: Short Form McGill Pain Questionnaire; Psy Distr: Psychological Distress Index; SDMT: Symbol Digit Modality Test; DSp: Digit Span subtest of the WAIS-IV; RAVLT: Rey Auditory Verbal Learning Test; TMT: Trail Making Test; Category of “Other” for mechanism of injury was not included as there was insufficient sample size to undertake this analysis in two of the groups (n < 4).

Bivariate significant predictors of group membership were evident for education, cause of injury, postconcussion symptom reporting, psychological distress, pain, fatigue and sleep disturbance. Age, sex, and cognitive variables did not significantly predict group membership. A multivariate model, incorporating significant bivariate predictors, was developed. The results of these analyses are presented in Table 4.

Table 4. Results of multivariate regression analyses.

	Fully Recovered vs. Partially Recovered					Fully Recovered vs. PartRecovPoorPhysH
Variable	b (Std Err)	OR	95% LCI : UCI	Wald	p-value	b (Std Err)	OR	95% LCI : UCI	Wald	p-value
Education	0.325 (0.180)	1.384	0.973 : 1.969	3.276	0.070	−0.192 (0.188)	0.825	0.573 : 1.193	1.043	0.307
Inj Cause	0.045 (0.273)	1.046	0.613 : 1.786	0.028	0.868	−0.220 (0.293)	0.803	0.452 : 1.425	0.564	0.453
RPQ	−00.046 (0.051)	0.955	0.864 : 1.056	0.801	0.371	−0.007 (0.050)	0.993	0.900 : 1.096	0.021	0.885
PSQI	0.037 (0.109)	1.038	0.838 : 1.285	0.114	0.736	0.144 (0.109)	1.154	0.932 : 1.430	1.728	0.189
MFI	0.144 (0.041)	1.155	1.065 : 1.252	12.182	<0.001	0.083 (0.042)	1.087	1.000 : 1.181	3.847	0.050
SF-MPQ	0.360 (0.382)	1.433	0.678 : 3.028	0.887	0.346	0.950 (0.344)	2.587	1.318 : 5.078	7.628	0.006
Psy Distr	0.034 (0.033)	1.035	0.970 : 1.105	1.065	0.302	0.004 (0.036)	1.004	0.935 : 1.078	0.011	0.916

Std Err: Standard Error; OR: Odds Ratio; CI: Confidence Interval; 95% LCI: UCI: 95% Confidence Interval of Odds Ratio – Lower bound and Upper bound; Inj Cause: Injury Cause; RPQ: Rivermead Post Concussion Symptoms Questionnaire; PSQI: Pittsburgh Sleep Quality Index; MFI: Multidimensional Fatigue Inventory; SF-MPQ: Short Form McGill Pain Questionnaire; Psy Distr: Psychological Distress Index.

Pain and fatigue were the only variables to significantly predict group membership independently of other variables. Specifically, pain was the only significant independent predictor of PartRecovPoorPhysH group membership relative to the Fully Recovered group, although fatigue strongly trended towards significance (p = 0.05). In contrast, fatigue was the sole significant independent predictor of the Partially Recovered group relative to the Fully Recovered group.

Discussion

This study revealed that three distinct groups of premorbidly healthy individuals were distinguishable with respect to their self-reported physical, mental, social and energy HRQoL, approximately 10 weeks after a mTBI that was associated with a systemic injury and hospital admission. Further, the factors that significantly predicted group membership differed between groups.

Although the largest group’s HRQoL (Fully Recovered: 42%) was commensurate with the “normal” population on all domains, the majority of individuals (Partially Recovered + PartRecovPoorPhysH) demonstrated reduced quality of life on all domains relative to normative data. Further, while both of the groups with reduced HRQoL had similarly low levels of emotional wellbeing, social function and energy, these groups could be meaningfully differentiated; the PartRecovPoorPhysH group (27%) had significantly poorer physical HRQoL than the Partially Recovered group (31%).

Examination of the groups further revealed that social HRQoL was particularly impacted in both groups that had partially recovered HRQoL. The Fully Recovered group had significantly better social HRQoL compared to energy HRQoL. In contrast, profile analysis revealed that social and energy HRQoL were equivalently poor, relative to each other, in each of the partially recovered groups. This suggests that an individual’s social functioning may be disproportionately impacted in those whose overall HRQoL is only partially recovered and may deserve particular clinical attention.

The present finding, that more than 50% of individuals with mTBI and systemic injury, who have been admitted to hospital, do not experience full HRQoL recovery 10 weeks after injury, is consistent with recent findings. Previous studies have shown 44–53% of individuals who present to hospital with mTBI have reduced life satisfaction, ongoing symptom reporting and functional limitations, 6 to 12 months after injury [17,48,49]. The present findings also support the notion that significant heterogeneity in outcome continues to be evident among individuals who are in the end stages of the “normal” recovery period after mTBI [1,15]. That is, more than 50% of individuals who have presented to/been admitted to hospital continue to experience significant ongoing impact on their HRQoL at a time when research has traditionally indicated that approximately 80% of individuals should be fully recovered [1]. Modifying clinician and patient expectations regarding the expected timeframe of a return to the subjective experience of “normal” health after mTBI that is associated with a hospital presentation or admission will likely benefit management and intervention approaches for these individuals.

Only two variables significantly predicted group membership of the partially recovered individuals, independently of education, post-concussion symptoms, sleep quality and psychological distress. Fatigue significantly predicted membership of the Partially Recovered group, and pain significantly predicted membership of the PartRecovPoorPhysH group. Given that the PartRecovPoorPhysH group had significantly poorer physical function HRQoL than both other groups, it seems logical that pain may be playing an important role in this group’s recovery after mTBI. In light of the dominance of pain in predicting membership of the PartRecovPoorPhysH group, the current findings indicate that careful management of pain should be a particular priority for this group.

In this study, recruitment occurred on the ward for all individuals admitted with a traumatic injury (i.e., systemic and/or head) and all individuals had experienced systemic injury, in addition to their mTBI. It is likely therefore that the pain associated with participants’ systemic injuries was contributing to the PartRecovPoorPhysH group’s experience of significantly poorer physical function in at least some of the sample. Unfortunately, assessment of the nature and severity of systemic injury was beyond the scope of the current study. Given that the present study has identified a meaningful subgroup for whom poorer physical function is a key contributor to their reduced HRQoL after mTBI, however, further investigation of the relative contribution of systemic injury factors to poor physical function in individuals with mTBI is warranted. Of note, however, it is also possible that physically painful mTBI symptoms, such as headache, may have contributed to reduced HRQoL in the ParRecovPoorPhysH group. Consequently, mTBI-related physical factors should also be explored when considering physical factors that contribute to poor HRQoL in this group.

Given that fatigue significantly predicted Partially Recovered group membership, it is interesting that sleep quality did not. This suggests that the subjective experience of feeling tired [50], rather than the subjective experience of having good quality sleep [51], is an important characteristic of this group. Treating clinicians may benefit from focussing their attention on fatigue management/intervention [52], rather than sleep specifically.

Both partially recovered groups had commensurate profiles with respect to emotional well-being, social function and energy HRQoL. It might therefore be expected that some overlap in factors independently predicting group membership might occur. Although not formally significant, fatigue very closely approached significance (p = 0.05) as a predictor of PartRecovPoorPhysH group membership, independently of pain. This suggests that fatigue may be a relevant predictive factor for the PartRecovPoorPhysH group, in addition to the impact of pain. Thus, clinical focus on fatigue management/intervention, may be warranted for anyone whose HRQoL has not normalised by 10 weeks after hospital-admitted mTBI.

The lack of predictive relationship between cognitive variables and group membership is consistent with the literature indicating that cognitive recovery is broadly complete by 3 months post injury [1]. The lack of an independent predictive role for age, sex, psychological distress and post-concussion symptom reporting is more unexpected, given the important role these variables have been shown to have in HRQoL after mTBI in previous studies [5,7–12]. Given that more than 80% of these studies comprised participants who presented to or were admitted to hospital, it is unlikely the present lack of predictive relationship is due to the fact that the present sample were admitted to hospital. Rather, most previous studies have treated individuals with mTBI as a unitary group, and no previous study has characterised group membership in terms of detailed HRQoL profiles. Therefore, it may be that the aforementioned factors (age, sex, psychological distress and post-concussion symptom reporting) are broadly related to HRQoL after mTBI but have no independent role in predicting specific HRQoL outcomes. It is also noteworthy, however, that psychological factors were condensed into a single index in the current study to facilitate statistical power. It is possible that separation of psychological distress into the individual components might reveal a predictive relationship between aspects of psychological distress and HRQoL that is not apparent in the current findings. Future research to investigate this possibility is warranted.

Although premorbid HRQoL was not measured in this study, it is unlikely that the finding of reduced HRQoL in the current study was due to pre-existing health issues. Firstly, participants were carefully recruited to be premorbidly healthy individuals. Secondly, in the general Australian population, less than 15% of individuals self-report “fair” or worse levels of health [53]. Given both these factors, it is very unlikely that the present findings were significantly biased by substantial numbers of individuals with pre-existing health problems.

The primary limitation of the present study is the moderate sample size, as it meant that not all HRQoL domains from the SF-36 could be incorporated into the cluster analysis. Given that the most reliable measure of each aspect of HRQoL (physical, mental and social functioning) were included, however, the key components of the HRQoL construct were investigated and a meaningful HRQoL profile could be obtained in the analyses. An additional consequence of the moderate sample size was the impact on the cluster sample sizes. As three clusters were derived from the analyses, the modest sample size of each cluster limited the number of independent variables that could be included in the analyses. In particular, although it is meaningful to conceptualise both pain and fatigue as multi-dimensional constructs in mTBI populations [25,32], the limited sample size of the individual clusters prevented detailed examination of these constructs. As both pain and fatigue were independent predictors of group membership, future research examining the relationship of these multidimensional constructs to HRQoL outcome is clearly warranted.

The current sample of premorbidly healthy individuals are not representative of the broader mTBI population with respect to psychiatric history, drug and alcohol history and past TBIs, which could raise questions regarding the usefulness of the findings. It has been shown, however, that these factors can impact self-report of outcome after mTBI [54] and therefore impact the ability to draw inferential conclusions regarding the independent role a given variable may have on outcome. By excluding individuals with these historical and comorbid factors, the present study is able to provide useful insights into the independent role of a range of variables on HRQoL 10 weeks after hospital-admitted mTBI. Further research is worthwhile, however, to understand the extent to which those historical and co-morbid factors, which are common after mTBI, might impact on individual differences in HRQoL in this group.

While none of the sample had previously experienced a concussion event that was either medically documented or associated with self-reported consequence, 25% of the sample reported experiencing 1–2 non-specific “concussion” events, which were not associated with any medical attention, diagnosis or self-reported consequence. This could be considered a limitation of the study in terms of examining HRQoL outcome in a sample who were premorbidly healthy. Given the lack of objective evidence of whether these “concussion” events constituted mTBIs, or merely headstrikes without mTBI, it is not possible to confidently argue that the subsample with 1–2 self-reported concussions experienced a previous neurological event. Nevertheless, it is possible that the sample did contain individuals with minor previous concussion history. Although there is no evidence in the literature that individuals with this degree of limited “concussion” history have long-term consequences that are different to those with a single mTBI event, it is not possible to confidently state that there has been no effect of these previous events. Future research into pre-morbidly healthy individuals with mTBI would benefit from removing these “questionable” concussion individuals from their samples to address this issue.

It is certainly the case that the current sample of participants do not reflect the broader mTBI population with respect to hospital-admission. It is well understood that approximately 40–50% of individuals with mTBI do not present to hospital at all [55,56]. While this means that the present pattern of HRQoL clusters cannot be considered representative of all individuals with mTBI, it does not undermine the relevance or importance of this study’s findings. Individuals with mTBI who also sustain systemic injury comprise a significant proportion of the broader mTBI population who present to hospital [13], and likely occur in many individuals with mTBI who do not present to hospital. While the nature and severity of the systemic injury likely differs between these subgroups of individuals with mTBI, the presence of some degree of systemic injury is likely to be relatively common. Consequently, while consideration of the nature and severity of systemic injury is an important consideration for future study, the fact that all participants in the current sample experienced systemic injury at time of mTBI does not undermine the relevance of these findings. Rather, this study highlights the importance of considering the impact of any systemic injury when considering HRQoL after mTBI, whether the individual was admitted to hospital or otherwise. It also highlights the importance of investigating the separate effects of systemic trauma from mTBI with respect to HRQoL outcome; this could be investigated by incorporating a head injury-free trauma control group in future studies.

In conclusion, the present study has demonstrated that three distinct groups of individuals can be identified with respect to their HRQoL 10 weeks after a hospital-admitted mTBI event that included systemic trauma. Further, two of these groups (containing the majority of participants) demonstrated significantly reduced HRQoL relative to normative data. In addition, this study showed that pain or fatigue independently predicted membership of the partially recovered groups, which highlights the importance of considering systemic and head-related physical symptoms in HRQoL outcome after mTBI. The findings suggest that individuals with mTBI who are experiencing pain and fatigue may represent an “at risk” group for good HRQoL recovery by 10 weeks after injury. Consequently, a particular focus on fatigue management and intervention remains relevant for those with ongoing HRQoL difficulties 10 weeks after injury; ongoing focus on pain management is also warranted for those who are experiencing particular difficulties with physical function HRQoL.

Ethics approval

This study was performed in line with the Declaration of Helsinki. Approval was granted by Ethics Committees of The Alfred hospital (251/08) and Royal Melbourne Hospital (MH-311851).

Acknowledgements

The author would like to acknowledge the contribution of post-graduate students and research assistants: Georgia Bolt, Emily Cockle, Nicolette Ingram, Arielle Levy, Courtney Lewis, Joshua Nash, Lucy Oehr, Katie Priestley, Aimee Savage, Nicola Singleton and Patrick Summerell for their assistance in collecting this data. This research received no specific grant from any funding agency, commercial or not-for-profit sectors.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by grants from the Melbourne School of Psychological Science [2017 – 2021].