Care transitions in the first 6 months following traumatic brain injury: Lessons from the CENTER-TBI study

Background: No large international studies have investigated care transitions during or after acute hospitalisations for traumatic brain injury (TBI). Objectives: To characterise various TBI-care pathways and the number of associated transitions during the first 6 months after TBI and to assess the impact of these on functional TBI outcome controlled for demographic and injury-related factors. Methods: This was a cohort study of patients with TBI admitted to various trauma centres enrolled in the Collaborative European NeuroTrauma Effectiveness Research in TBI (CENTER-TBI) study. Number of transitions and specific care pathways were identified. Multiple logistic regression analyses were used to assess the impact of number of transitions and care pathways on functional outcome at 6 months postinjury as assessed by the Glasgow Outcome Scale-Extended (GOSE). Results: In total, 3133 patients survived the acute TBI-care pathway and had at least one documented inhospital transition at 6-month follow-up. The median number of transitions was 3 (interquartile range 2–3). The number of transitions did not predict functional outcome at 6 months (odds ratio 1.08, 95% confidence interval 1.09–1.18; P = 0.063). A total of 378 different care pathways were identified; 8 were identical for at least 100 patients and characterized as ‘‘common pathways’’. Five of these common care pathways predicted better functional outcomes at 6 months, and the remaining 3 pathways were unrelated to outcome. In both models, increased age, violence as the cause of injury, pre-injury presence of systemic disease, both intracranial and overall injury severity, and regions of Southern/Eastern Europe were associated with unfavourable functional outcomes at 6 months. Available online at


Introduction
Traumatic brain injury (TBI) is a major cause of death and longterm disability worldwide [1]. Many patients with TBI are admitted to hospital in the acute phase, representing approximately 1.5 million hospitalisations in the European Union annually [2]. Guidelines for acute neurosurgical and intensive care have been widely adopted [3], but other healthcare and rehabilitation interventions following such hospitalisations are variable [4]. Transitions between inpatient and outpatient care are at risk for both quality and continuum of care in patients with TBI [5], and to exacerbate this, older patients with TBI have a higher risk of inappropriate discharge planning [6]. Previous Scandinavian studies have reported that direct transfers from hospitals to rehabilitation units improved outcomes and reduced length of hospital stay for patients with severe TBI [7][8][9], but the effect of a reduced number of transitions was not addressed. For TBI, no large international studies have investigated care transitions during and after acute hospitalisation.
The Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) project has reported large variations in care structure among countries [10] in neurosurgical services [11], in-hospital acute rehabilitation, and referrals to post-acute rehabilitation services [12]. Even larger variations may be expected because the healthcare context can profoundly affect care pathways [13]. Hospital structure, organisation, and the training of staff can all affect care transitions between intensive care units (ICUs) and regular wards in addition to patient-related characteristics [14]. TBI severity and the presence of other injuries also affect outcomes [15] and may also affect length of stay and care pathways. Comorbidities are high, especially in older patients, and may have profound effects on both care pathways and discharges [16] and need to be considered when evaluating differences between countries.
Informed planning for care transitions is important to avoid adverse effects in patients with complex health care needs [17]. Transitions from hospitals to homes for patients with complex health care needs after TBI are especially vulnerable and require careful planning and support [18]. Acute-care hospitals are often under pressure to transfer patients from ICUs to regular wards or to discharge patients [19]. Consequently, rapid decisions may lead to inadequate healthcare assessments and inappropriate care transitions [17]. Planning discharges and future care for patients with cognitive impairment is particularly challenging and has not been studied in large international cohorts.
The present work addressed the burden of care transitions during the first 6 months after TBI, with a focus on various care pathways, number of care transitions, and assessments of their appropriate timing. The study also aimed to address the hypotheses that both the number of transitions and care pathways affect functional outcomes at 6 months.

Study design and participants
This paper adheres to the STROBE-guidelines for reporting cohort studies [20].
The study was conducted within the context of the Core study of the CENTER-TBI project. This was multi-centre, observational, longitudinal, cohort study of patients with TBI (registered at ClinicalTrials.gov: NCT02210221) who presented (between December 2014 and December 2017) to 59 medical and research centres from 19 European countries and Israel [21]. Appendix A provides a full list of the CENTER-TBI participants and investigators. The CENTER-TBI inclusion criteria were: clinical diagnosis of TBI; indication for CT imaging; presentation within 24 hr of injury; informed consent obtained.
Patients with severe pre-existing neurological disorders that could have confounded outcome assessments were excluded.
Enrolled patients were stratified into 3 groups according to initial clinical care pathway: emergency room (ER) stratum: evaluated in the ER, then discharged; admission (ADM) stratum: admitted to a hospital ward; ICU stratum: admitted directly to an ICU, from the emergency department or another hospital.
Initially, 4559 patients were enrolled, but 43 withdrew consent and 7 centres were excluded because of enrolment of < 5 patients. Thus, records for 4509 patients were available for analysis. See Steyerberg et al. [22] for the flowchart and specific details.

Ethical approval
The CENTER-TBI study was conducted in accordance with all relevant local and national ethical guidelines, regulatory requirements for recruiting human subjects, relevant data protection and privacy regulations. Informed consent was obtained from all patients or their legally acceptable representative. The study obtained ethical clearance from the institutions involved in the project (see https://www.center-tbi.eu/project/ethical-approval for details).

Data collection, handling and storage
Patient data were entered into a clinical database from an electronic Case Report Form with a Global Unique Patient Identifier used to ensure adequate de-identification. Data were stored at the International Neuroinformatics Coordinating Facility (INCF) in Stockholm, Sweden. The Neurobot data management tool was developed by the INCF for data extractions. Data curation was performed by a multidisciplinary data curation team.
This study used care-transition data from hospital admission to discharge home and post-acute care during the first 6 months. Care transitions were defined as points during a care pathway at which the patient was transferred from one treatment facility to another or discharged from organised TBI care. Seven categories were used to describe transitions from hospital ERs to an intensive or high care unit (CU), neurosurgical or neurological ward (WN), other Conclusions: A high number of different and complex care pathways was found for patients with TBI, particularly those with severe injuries. This high number and variety of care pathway possibilities indicates a need for standardisation and development of ''common data elements for TBI care pathways'' for future studies. ward (WO), rehabilitation unit (REHAB), nursing home (NH), home (HOME) and other hospital. Each patient was assigned a specific care pathway. Their last registered transition was designated as their post-acute discharge destination. The number of transitions between destinations was registered. Transitions to and from CT imaging, MRI, or surgery were excluded. Treatment centres classified the timing of each transition as appropriate, premature, or delayed as follows: appropriate transition: a physician judged a patient's condition to be appropriate for transfer; premature transition: for example, a patient was discharged from an ICU due to limited bed capacity but would have remained longer if possible; delayed transition: for example, a patient remained on a ward because of lack of beds at the receiving rehabilitation unit. Geographical region was classifed by the Eurovoc classification scheme [23] as North/West (Austria, Belgium, Denmark, Finland, France, Germany, Latvia, Lithuania, Netherlands, Norway, Sweden, and the United Kingdom) or South/ East and Israel (Hungary, Israel, Italy, Romania, Serbia, and Spain). Living arrangements were assessed by data collected on the number of co-habitants, using a yes or no designation for the patient living alone. Pre-injury somatic health problems were classified according to the American Society of Anesthesiologists Physical Status assessment system (ASAPS) [24] and were divided into 3 categories in the present study: healthy, mild systemic disease, and severe systemic disease. This classification was used to depict the functional impact of the medical comorbidity the patient had before the head injury (e.g., cardiovascular or endocrine disorders). Causes of injury were classified as a fall, road traffic accident (RTA), violence, suicide, or other. The Glasgow Coma Scale (GCS) score was used to evaluate injury severity (3-8, severe; 9-12, moderate; 13-15, mild). The Injury Severity Scale (ISS) was used to evaluate overall trauma severity. Whether or not cranial surgery was conducted was registered as yes or no. The Glasgow Outcome Scale-Extended (GOSE) score [25] was used to assess 6-month outcomes as favourable (5)(6)(7)(8) or non-favourable (1-4), in accordance with Steyerberg et al. [22]. To evaluate transitions and outcome at 6 months, we excluded patients who had died. Hence, patients with a GOSE score of 1, signifying death, were excluded.

Statistical analysis
Data were retrieved from the CENTER-TBI Core 2.0 final sample (May 2019). Analyses were performed with R v3.6.2. Data are described with median (interquartile range [IQR]) or number (%). During descriptive data analyses, patients were classified by age groups: 0-15, 16-20, 21-40, 41-60, 61-70 and > 70 years. To investigate the predictive value of the number of transitions for a non-favourable GOSE classification and to adjust for covariates, we used multivariable logistic regression. The sample (n = 3133) included patients who: survived the acute TBI-care pathway; had been discharged by the time of their 6-month follow-up; had at least one documented in-hospital transition.
Analyses controlled for age, geographical region, living alone, pre-injury health status, cause of injury, GCS and ISS scores, and cranial surgery. Correlation analyses were used to determine possible multicollinearity between the covariates. Another multivariable logistic regression analysis was used to evaluate the impact of care pathway on non-favourable GOSE category and to adjust for covariates by using the same procedure as described above. Care pathways shared by < 100 patients were aggregated and termed ''other'' and used as a reference in the analyses (n = 1197). Odds ratios (ORs) > 1 increased the probability of a favourable functional outcome, and ORs < 1 decreased the probability of a favourable outcome.
Missing covariate values were imputed under the assumption of missing at random by using multiple imputations with IBM SPSS Statistics v25.0. For multiple imputations, all available data on variables used in the models and sex were used to generate 30 imputed data sets. The results from each complete dataset were combined to present single estimates. Sensitivity analyses were performed for the number of imputations for missing values. These multiple imputed models are presented in the results, with complete-case analyses in Appendix A.

Results
In total, 4029 patients were alive at 6 months and deemed eligible for study inclusion. Demographic and injury characteristics by patient strata are in Table 1. Median age overall was 48 years (IQR 29-64); most patients were male (67%), had mild TBI (71%), and showed favourable outcomes at 6 months (70%). Patients with mild TBI typically belonged to the ER and ADM strata, whereas the ICU stratum included mostly patients with more severe injuries and a non-favourable GOSE category. The median number of  Fig. 1 displays a visual representation of the care pathways that occurred for at least 20 patients. Table 2 displays the distribution of the most common care pathways ( 100 patients) and post-acute discharge destinations by strata and GOSE category, including the length of stay and number of transitions. As expected, the most frequent care pathways varied by strata, as did the registered post-acute discharge destinations. Hospital length of stays were longest for the ICU-ward (neuro)-rehab pathway, which suggests that patients ending up in rehabilitation had the longest hospital stays. The ''other pathways'' had both the highest mean number of transitions and the highest frequency of non-favourable GOSE score at 6 months, with 45% of the total ICU strata belonging to this group.
For the 3133 patients with at least one documented in-hospital transition, the median age was 49 years (IQR 29-64). TBI severity was similar to the overall cohort (65% mild, 10% moderate, 21% severe, and 4% unknown), as was sex (69% male), geographical region (76% North/West), and GOSE category (9% vegetative/lower severe disability, 5% upper severe disability, 23% moderate disability, 46% good recovery, and 17% unknown). The median number of transitions among these patients was 3 (IQR 2-3). The median number of transitions did not vary across demographic subgroups except for patients > 70 years old (2 [IQR 2-3]). Of note, the median number of transitions was similar across the different    tes, increased age, pre-injury presence of systemic disease, both intracranial and overall injury severity, injury caused by violence and regions of Southern/Eastern Europe were associated with unfavourable functional outcomes at 6 months. Table 4 displays results of the imputed multivariable logistic regression analysis of the effect of 8 typical pathways on GOSE category. Complete-case results were similar (Appendix A). Five of these 8 most-common pathways showed decreased odds of nonfavourable GOSE category as compared with all other pathways. The exception was the ''CU-Other hospital'' pathway, which showed a 45% increased likelihood of a non-favourable GOSE category. We found no association between the ''CU-Ward (neuro)-Rehab'' pathway and non-favourable GOSE category. This model also showed increased likelihood of non-favourable GOSE category with increased age, South/East region, premorbid systemic disease, injury caused by violence, and more severe intracranial and overall injuries.
Only 293 patients (10%) were reported to have at least one premature or delayed transition. Of these, 244 (8%) had at least one documented delay, and 57 (2%) had documented premature transitions. Demographic and injury-severity characteristics of this group were similar to those for the full sample, except for higher percentages of severe TBI and major trauma. The median age was 51 years (IQR 32-64), 68% of patients were male, 87% were from the North/West region, 37% had severe TBI, 81% belonged to the ICU stratum and 85% had an ISS score 15. Significant differences between the premature/delayed transition group and the remaining patient group were confirmed for GCS and ISS scores, cranial surgery, and region of residence (data not shown). In the premature-transition subgroup, the main transitions were to home or other hospitals. In the delayed transition subgroup, we found a mixture of transitions from/to high CUs, other hospitals, neurosurgical wards and rehabilitation facilities.

Discussion
This is one of the first TBI studies to describe the care-transition burden at the patient level during the first 6 months after TBI. Transition number varied across patient strata and was highest in the ICU stratum. This finding may represent injury severity in patients admitted to the ICU (median GCS score 10), and their transitions related to prolonged care in hospital settings. However, 24% of patients with mild TBI were also admitted to ICUs [22] in line with a US study that reported 24% of mild-TBI patients requiring ICU admission at some stage after injury [26]. Other factors, such as extracranial injuries, might also play a role.
The median transition number showed little variation by demographic characteristics, including age. All age groups had medians of 3 transitions, except patients > 70 years, who had a median of 2. Pressure to free acute-care beds can lead to faster discharge of older patients [27].
Furthermore, patients with more severe TBIs and major trauma were transferred more often between different wards/facilities as compared with patients with less severe injuries. Patients with severe TBI and high disability levels often need prolonged stays in hospital [28].
Most transitions were rated appropriate, with only 9% considered delayed or premature, and the ICU stratum had the highest number of these. Premature transitions were to homes or other hospitals and may also reflect pressure to free acute-care beds [19]. In contrast, the delayed subgroup of patients was characterised by a mixture of care pathways. However, previous studies have reported that delayed transitions could be related to waiting times for destination beds or to other non-clinical care decisions [7].
The present study documented 378 care pathways. This finding may be a reflection of not only TBI complexity but also different care organisations and the decision-making processes involved in management in trauma hospitals involved in the CENTER-TBI project [10]. In the ADM stratum, approximately two-thirds of patients received treatment in wards before being discharged home, whereas one-quarter were in the heterogeneous otherpathways group. The ICU stratum was different: one-quarter of patients were transferred from CUs to hospital wards before discharge home, and almost half of these patients belonged to the other-pathways group. This finding may reflect that ICU stratum patients were more severely injured and thus needed more complex medical treatment and more frequent transitions; their median transition number was 4, and median length of stay was 15 days.
The hypothesis that number of transitions would influence functional outcome was not confirmed by the multivariate regression analysis; the number of transitions was not a significant predictor of non-favourable functional outcomes (GOSE) at 6 months in the imputed model. However, increased age, South/ East regions, presence of premorbid disease, violence-related injuries, most-severe TBI and overall trauma were associated with non-favourable functional outcomes at 6 months. Yet, the complete-case analyses revealed a very small but statistically significant effect suggesting that increased number of transitions predicted unfavorable outcome at 6 months. The ORs were rather similar in both models. Previously, 3 factors were reported to influence transitions for individuals with brain injuries: personal/ individual characteristics, family/caregiving factors, and professional/service factors [29].
The hypothesis that care pathways influence functional outcome was partly confirmed. The multivariable logistic regression analysis confirmed that 5 pathways significantly predicted favourable outcomes at 6 months. The 3 remaining pathways were unrelated to outcomes. Similarly, this model showed that the same covariates predicted outcomes at 6 months.
The better functional outcomes of the most frequent care pathway in both the ADM and ICU strata (transfer to wards and thereafter discharged home) are not surprising. This pathway reflects the patients with less severe TBI who recover faster. The present results suggest poorer functional outcomes for patients with severe TBI, major trauma, and increased age. They support previous findings that the burden of TBI-care pathways is determined by not only TBI severity but also overall injury severity and socio-biological factors such as age [30]. The management of severe TBI is lifelong, and a better understanding is needed of the impairments, available treatments, and optimal care, as highlighted in French guidelines for care pathways with adults with severe TBI [31].
In both multivariate regression models, we found more nonfavourable outcomes for patients in South/East regions. Whether this finding is due to differences in TBI care is unclear, so our results should be interpreted with caution. However, variability in the number of hospital beds among European countries exists. For example, Germany has 2.5 times the number of curative-care beds and 50 times the number of rehabilitative-care beds (per 100,000 population) than Spain [32]. Both the contexts and systems of care assessed here were heterogeneous, and thus, the number of possible care pathways was high. Previous reports from this project highlighted substantial variations in the processes of TBI care [10] and suggested that these variations were opportunities to study specific aspects of TBI-care effectiveness. However, comparing 378 pathways across 59 institutions is a challenge. Thus, there is an urgent need for defining and standardising transitions and TBI-care pathways by integrating these into ''common data elements'' for future studies. For example, future studies should Annals of Physical and Rehabilitation Medicine xxx (xxxx) [1][2][3][4][5][6][7][8] ensure that centres use a standardized definition of what constitutes delayed or premature transitions. Common TBI pathways could be identified for each country and data collected on both full care pathways and TBI-specific transitions and whether the patient was considered to receive standardized TBI care or not. This move would allow for identifying differences in care between patients but further provide a clearer picture of how transitions were affected by individual factors such as comorbid disorders, caregiving factors and service-related factors such as need to free beds or unavailability of rehabilitation services.
The strengths of the study are the large sample size and the number of participating countries, rendering a robust overview of care-pathway variations in Europe and Israel. However, local logistics and academic interests of participating centres as well as low numbers/non-consecutive enrolments in some centres may have resulted in selection bias for patient recruitment. Furthermore, differences in data registration among study sites and countries and organisational differences in discharge timing need to be taken into consideration when interpreting the present results. In addition, we did not analyse care transitions in patients who died within the 6 months (n = 473) nor those still in-hospital 6 months after TBI onset (n = 7) because the focus of the study was to evaluate the completed care trajectories of the patient group. Although the number of patients lost to follow-up was considered low, there was enough missing data to warrant imputed analyses, which is often an issue in longitudinal studies. In-depth studies across countries that follow care pathways in trauma hospitals are highly warranted.

Conclusions
The most important finding in this study was the highly diverse and complex TBI care pathways. The number of transitions, including delayed or premature, was highest in the ICU stratum and showed little variation by demographics. Patients with more severe TBI and major extracranial trauma were transferred more often between different wards and facilities than those with less severe injuries. The high number and variety of care pathway possibilities indicates a need for standardisation and development of ''common data elements for TBI care pathways'' for future studies.

Disclosure of interest
The authors declare that they have no competing interest.

Funding
Data used in preparation of this manuscript were obtained in the context of CENTER-TBI, a large collaborative project with the support of the European Union 7th Framework program (EC grant no. 247 602150). Additional funding was obtained from the Hannelore Kohl Stiftung (Germany), OneMind (USA) and Integra LifeSciences Corp. (USA).