Outcome measurement in brain injury rehabilitation--towards a common language

Brain injury rehabilitation after road trauma in new South Wales, Australia - insights from a data linkage study

Background

Population-based patterns of care studies are important for trauma care but conducting them is expensive and resource-intensive. Linkage of routinely collected administrative health data may provide an efficient alternative. The aims of this study are to describe the rehabilitation pathway for trauma survivors and to analyse the brain injury rehabilitation outcomes in the two care settings (specialist brain injury and non-specialist general rehabilitation units).

Methods

This is an observational study using routinely collected registry data (New South Wales Trauma Registry linked with the Australasian Rehabilitation Outcomes Centre Inpatient Dataset). The study cohort includes 268 road trauma patients who were admitted to trauma services between 2009 and 2012 and received inpatient rehabilitation because of a brain injury.

Results

Of those who need inpatient rehabilitation, 62% (n = 166) were admitted to specialist units with the remainder (n = 102) admitted to non-specialist units. Those admitted to a specialist units were younger (p < 0.001), had a lower cognitive FIM score (p = 0.003) on admission than those admitted to non-specialist units. Specialist units achieved better overall FIM score improvements from admission to discharge (43 vs 30 points, p > 0.001) but at a cost of longer length of stay (median 47 vs 24 days, p < 0.001). There were very few discharges to residential aged care facilities from rehabilitation (2% in non-specialist units and none from specialist units). There was a long time lag between trauma and admission to inpatient rehabilitation with only a quarter of the patients admitted to a specialist unit by end of week four. Few older patients (19%) with brain injury were admitted to specialist units.

Conclusions

It is feasible to use routinely collected registry data to monitor inpatient rehabilitation outcomes of trauma care. There were differences in characteristics and outcomes of patients with traumatic brain injury admitted to specialist units compared with non-specialist units.

The Utility of the Modified Barthel Index for Traumatic Brain Injury Rehabilitation and Prognosis

The aim of this correlational study was to examine the utility of the Modified Barthel Index (MBI) in measuring functional outcome and in determining its predictive ability at the commencement of rehabilitation. All 79 patients with a traumatic brain injury admitted in 12 months were the participants in the study. Functional status, as measured by the MBI total scores on admission to and discharge from comprehensive rehabilitation, and all available independent variables on admission were extracted.

The study found that 85% of admissions were male. The participants' mean age was 28 ± 10.67 years, 73% of them were single and 97% had a diagnosis of closed traumatic brain injury confirmed by a CAT scan. Their rehabilitation stay was 61 ± 65 days. The admission MBI scores were 75 ± 30 and the discharge scores were 93 ± 29. In a stepwise regression, initial deficit as measured by the MBI, delay in commencing rehabilitation and age contributed to explaining 68% of variance. However, when the MBI scores were transformed, the fourth root of the admission score alone explained 84% of variance. This predictive ability is far superior to the findings of many other studies.

Reliability and sensitivity to change of measurement instruments used in a traumatic brain injury population

Objective: To compile a minimum data set for the follow-up of traumatic brain injury patients from discharge from hospital to one year post injury to assess functioning and participation in the physical, cognitive and psychosocial domains, and in quality of life.

Design: Repeated questionnaire interviews by two observers to establish interobserver reliability of the measurement instruments at discharge and at one year post injury, as well as their sensitivity to change over time in traumatic brain injury patients.

Setting: Department of neurosurgery of an academic hospital, department of a rehabilitation centre, and at the patients' homes in the Netherlands.

Subjects: The study at discharge included 25 patients aged 18-50 years with a moderate to severe traumatic brain injury (Glasgow Coma Scale score 3-14), whereas the one year post injury study included 14 patients aged 19-51 years.

Main (outcome) measures: Physical domain: Barthel Index (BI), Functional Independence Measurement (FIM), Glasgow Outcome Scale (GOS), GOS Extended (GOSE). Cognitive domain: Disability Rating Scale (DRS), Functional Assessment Measurement (FAM), Levels of Cognitive Functioning Scale (LCFS), Neurobehavioural Rating Scale (NRS). Psychosocial domain: Community Integration Questionnaire (CIQ), Employability Rating Scale (ERS), Frenchay Activity Index (FAI), Multi Health Locus of Control (MHLC), Rehabilitation Activities Profile (RAP), Social Support List (SSL), Supervision Rating Scale (SRS), Wimbledon Self Reporting Rating Scale (WSRS). Quality of life: Coop/Wonca Charts (Coop), Rand SF-36 (Rand-36), Sickness Impact Profile-68 (SIP-68).

Results: At both discharge and at one year post injury, in the physical domainthe FIM showed excellent squared weighted kappa (SWK ranging from 0.75 to 0.80), and intraclass correlation coefficient (ICC ranging from 0.75 to 0.92), and a relatively small standard error of measurement (SEM 3.22) and smallest detectable difference (SDD 8.92). In the cognitive domain the FAM and the NRS showed excellent SWK, and ICC, and a relatively small SEM and SDD. In the psychosocial domainthe FAI showed excellent SWK (0.89), and ICC (0.87), and a relatively small SEM (2.64) and SDD (7.31). For quality of life, at both discharge and at one year post injury the SIP-68 and the Coop showed excellent SWK (0.87), and ICC (0.89), and a relatively small SEM (3.79) and SDD (10.51). At both time points SWK and ICC ranged from 0.80 to 0.89, SEM ranged from 1.47 to 1.98, and the SDD was 4.07.

Conclusions: An example of a reliable minimum data set that is also able to detect changes over time is: the FIM, the FAM and the Coop for the early stages in recovery, extended with the NRS, the FAI, and the SIP-68 later in recovery, thereby covering all relevant domains after traumatic brain injury.

Use of the Barthel Index and the Functional Independence Measure during early inpatient rehabilitation after single incident brain injury

Objective: To compare the appropriateness and responsiveness of the Barthel Index and the Functional Independence Measure (FIM) during early inpatient rehabilitation after single incident brain injury.

Design: Cohort study.

Setting: A regional neurological rehabilitation unit.

Patients: Two hundred and fifty-nine consecutive patients undergoing inpatient comprehensive neurological rehabilitation following a vascular brain injury due to single cerebral infarction (n=75), spontaneous intracerebral haemorrhage (n=34) and subarachnoid haemorrhage (n=43), and 107 patients who had sustained traumatic brain injury.

Measurements: Admission and discharge FIM total, physical and cognitive scores and the Barthel Index were recorded. Appropriateness and responsiveness in the study samples were determined by examining score distributions and floor and ceiling effects, and by an effect size calculation respectively. Non-parametric statistical analysis was used to calculate the significance of the change in scores.

Results: In all patient groups there was a significant improvement (Wilcoxon's rank sum, P < 0.0001) in the Barthel Index (mean change score: vascular 3.9, traumatic 3.95) and FIM (mean change score: vascular 17.3, traumatic 17.4) scores during rehabilitation, and similar effect sizes were found for the Barthel Index (effect size: vascular 0.65, traumatic 0.55) and FIM total (effect size: vascular 0.59, traumatic 0.48) and physical scores in all patient groups. In each patient group the cognitive component of the FIM had the smallest effect size (0.35-0.43).

Conclusions: All measures were appropriate for younger (less than 65 years of age) patients undergoing early inpatient rehabilitation after single incident vascular or traumatic brain injury. The Barthel Index and the total and physical FIM scores showed similar responsiveness, whilst the cognitive FIM score was least responsive. These findings suggest that none of the FIM scores have any advantage over the Barthel Index in evaluating change in these circumstances.

A Comparison of Rehabilitation Outcome Measures for Traumatic Brain Injury

This study evaluates the appropriateness, responsiveness, and predictive ability of the Modified Barthel Index (MBI), the Disability Rating Scale (DRS), the Barry Rehabilitation In-patient Screening of Cognition (BRISC), and the Glasgow Coma Scale (GCS) for 78 patients with traumatic brain injury referred for in-patient rehabilitation. Appropriateness was evaluated by examining means, standard deviations, coefficients of variation, and ceiling and floor effects. Responsiveness was determined by examining paired t-test results for admission and discharge scores, and on the basis of an effect size calculation. Predictive power was evaluated by performing two stepwise regressions with length of rehabilitation and total length of hospital stay. The results suggest that although the DRS and GCS have some advantages, notably in low ceiling and floor effects, overall the MBI is the most effective measure, particularly for prediction, with a moderate coefficient of determination (r2 = 0.42) and no significant differences between predicted and real length of hospital stay.

Deriving a Barthel Index from the Northwick Park Dependency Scale and the Functional Independence Measure: are they equivalent?

OBJECTIVE

to examine the extent of agreement between Barthel Index scores derived from Northwick Park Dependency Scores (NPDS) and the Functional Independence Measure (FIM) ratings, in an inpatient setting.

DESIGN AND SETTING

previously described conversion criteria were applied in a secondary analysis of a large existing dataset, gathered in a tertiary specialist inpatient neurorehabilitation unit.

SUBJECTS

patients with neurological disabilities (N = 1347), mainly following acquired brain injury.

MAIN MEASURES

comparison of Barthel scores derived from the NPDS (rated by nursing staff) and from parallel FIM scores (rated by the therapy team).

RESULTS

very strong intraclass correlations were observed between the total scores (0.93, P<0.001); 95% limits of agreement ranged from -3.53 to 4.90. Item-by-item agreement (linear-weighted Cohen's kappa coefficients) ranged from 0.41 to 0.77, which represents 'moderate' to 'substantial' agreement. A significant bias towards lower NPDS-derived scores (median 10 (interquartile range (IQR) 6-16) compared with median 11 (IQR 7-16) for the FIM-derived score; Wilcoxon z 11.60, P<0.001) was considered most likely to reflect actual differences in patient performance observed by therapy and nursing staff.

CONCLUSIONS

this study demonstrates good overall agreement between the Barthel Index ratings derived from NPDS and FIM scores. However, scores may be affected by differential performance with nursing and therapy staff, and should not automatically be regarded as equivalent.

Can the Northwick Park Dependency Scale be translated to a Barthel Index?

OBJECTIVE

to develop and test an algorithm for conversion of the Northwick Park Dependency Scale (NPDS) to a Barthel Index.

DESIGN AND SETTING

conversion criteria were developed to derive a Barthel Index from NPDS data. The criteria were then applied in two community-based datasets of NPDS and Barthel scores, gathered from carers/patients via postal questionnaires. A retrospective exploratory analysis (dataset A) was followed by prospective confirmatory analysis (dataset B).

SUBJECTS

patients with neurological disabilities, mainly following acquired brain injury: dataset A (n = 225) and dataset B (n = 96).

MAIN MEASURES

comparison of NPDS-derived Barthel scores with parallel directly rated Barthel scores.

RESULTS

the exploratory analysis identified that one Barthel item ('Stairs') could not be derived from the NPDS items alone, and required minor adaptation of one of the five additional questions that are included to translate the NPDS into an assessment of community care needs (Northwick Park Care Needs Assessment, NPCNA). For the prospective confirmatory analysis, the NPCNA 'Stairs' question was adjusted to support full conversion. Very strong intraclass correlations were observed between the total 'direct' and 'derived' Barthel scores (0.97 (dataset A), 0.95 (dataset B), P<0.001); 95% limits of agreement ranged from -2.52 to 2.56 (dataset A) and -3.29-3.91 (dataset B). Item-by-item agreement (linear-weighted Cohen's kappa coefficients) ranged from 0.68 to 0.85 (dataset A) and 0.59-0.83 (dataset B), which represents 'substantial' to 'almost perfect' agreement.

CONCLUSIONS

this study demonstrates that a Barthel Index can be reliably derived from NPDS and NPCNA data through a conversion algorithm which has now been built into the supporting software package.

The structure of functional and community outcomes following traumatic brain injury

OBJECTIVES

To investigate the dimensionality of functional and community outcomes following serious TBI. To identify items that fit, misfit or are redundant, as well as to assess person misfit.

METHODS

Rating-scale (Rasch) analysis was applied to 1-year follow-up data from 231 cases in the US National TBI Model Systems database. Items selected for analysis included all items indicative of global outcomes, disability, activity or participation.

RESULTS

A powerful singular measurement dimension was identified. Item reliability was very high (0.98), as was person reliability (0.97). The dimension fit over 90% of cases; that is approximately 10% of cases displayed anomalous patterns of functioning that indicated that their functioning was not measurable in terms of the general dimension identified. There was tension within the dimension between ratings of dependency (FIM) and cognitive functioning in everyday life (NFI). Most-but not all-neuropsychological test scores misfit the outcome dimension.

CONCLUSIONS

Actual dimensionality was distinct from the named scales employed. A unidimensional measure model fit the data much better than expected. This outcome dimension might be called 'general community functioning'. In the future, it should be possible to develop more valid and parsimonious measures of community outcomes following TBI.