Article Text
Abstract
Background Duration of post-traumatic amnesia (PTA) correlates with global outcomes and functional disability. Russell proposed the use of PTA duration intervals as an index for classification of traumatic brain injury (TBI) severity. Alternative duration-based schemata have been recently proposed as better predictors of outcome to the commonly cited Russell intervals.
Objective Validate a TBI severity classification model (Mississippi intervals) of PTA duration anchored to late productivity outcome, and compare sensitivity against the Russell intervals.
Methods Prospective observational data on TBI Model System participants (n=3846) with known or imputed PTA duration during acute hospitalisation. Productivity status at 1-year postinjury was used to compare predicted outcomes using the Mississippi and Russell classification intervals. Logistic regression model-generated curves were used to compare the performance of the classification intervals by assessing the area under the curve (AUC); the highest AUC represented the best-performing model.
Results All severity variables evaluated were individually associated with return to productivity at 1 year (RTP1). Age was significantly associated with RTP1; however, younger patients had a different association than older patients. After adjustment for individually significant variables, the odds of RTP1 decrease by 14% with every additional week of PTA duration (95% CI 12% to 17%; p<0.0001). The AUC for the Russell intervals was significantly smaller than the Mississippi intervals.
Conclusions PTA duration is an important predictor of late productivity outcome after TBI. The Mississippi PTA interval classification model is a valid predictor of productivity at 1 year postinjury and provides a more sensitive categorisation of PTA values than the Russell intervals.
- Head injury
- rehabilitation
- cognitive neuropsychology
- post-traumatic amnesia
Statistics from Altmetric.com
Traumatic brain injury (TBI) is a leading cause of death and disability worldwide.1 Severity of TBI influences both morbidity and mortality.2–4 Accordingly, early determination of TBI severity helps guide initial medical management and is useful in determining prognosis. TBI severity is most commonly classified based on evaluation of initial depth of altered consciousness as measured by the Glasgow Coma Scale (GCS).5 GCS scores are helpful with early management and early prognosis, but may be less helpful in predicting late outcome relative to other injury severity indices.6–10 Other severity indices predictive of late TBI outcome include time to follow commands (TFC) and duration of post-traumatic amnesia (PTA). TFC, also referred to as length of coma (LOC), is the interval from injury until the patient can follow commands.10 PTA is the interval from injury until that patient is oriented and can form and later recall new memories.10 Most studies have found that TFC and PTA are superior to GCS in predicting late functional outcome.11 PTA and TFC are strongly correlated, partially because PTA intervals include TFC.6 11 PTA appears to be more accurate than TFC in predicting long-term global outcome as measured by the Glasgow Outcome Scale (GOS) and in predicting long-term functional disability in rehabilitation samples.6 12–15
Controversy exists regarding the validity of duration intervals used in schema for classifying injury severity based on PTA duration. The most commonly used schema was developed by Russell and Smith.16 This schema, classifies individuals with <1 h of PTA as ‘mild TBI,’ 1–24 h as ‘moderate TBI,’ 1–7 days of PTA as ‘severe TBI’ and >7 days as ‘very severe TBI.’16 This has since been modified to label 1–7 days of PTA as ‘moderate’ and >7 days as ‘severe.’17 It has been reported that this schema classifies many persons with mild or moderate injuries based on GCS scores as having severe (or very severe) injuries based on PTA duration.11 In addition, several studies found that PTA intervals longer than 7 days may be associated with outcomes that are quite favourable.6 11 13 18 19
Recently, the Mississippi PTA Intervals were developed in which injury severity was categorised as moderate (0–14 days), moderate severe (15–28 days), severe (29–70 days) and extremely severe (>70 days) based on relative proportions of persons returning to productive activity.18 Preliminary results indicated that patient severity groups based on these intervals differed in productivity outcomes at 1 year postinjury.20
The present investigation sought to further validate the new Mississippi injury classification intervals and to compare their predictive ability with that of the criterion standard Russell classification intervals for determining 1-year postinjury return to productive activity. This investigation used a much larger, multicentre sample compared with the single centre sample used in the development of the Mississippi schema. We hypothesised that injury-severity groups based on the Mississippi classification intervals would differ in the likelihood of return to productive activities at 1 year postinjury. We also hypothesised that the Mississippi schema would predict productivity outcomes more accurately than the Russell classification, reflecting the former schema's empirically based development. We chose productivity outcome as the endpoint of interest because of the significance of return to work for persons with TBI and because this outcome has been successfully used in numerous previous studies on TBI outcome.
Methods
Participants
Participants were enrolled in the TBIMS National Database—a multicentre, longitudinal study of traumatic brain injury funded by the National Institute on Disability and Rehabilitation Research (NIDRR).21 Currently, there are 16 sites across the United States of America enrolling subjects in the database, which has been in existence since 1988. All TBIMS enrollees are age 16 or older, received medical care in a TBIMS-affiliated acute care hospital within 72 h of injury and were transferred directly from acute care to an affiliated comprehensive rehabilitation programme. All participants provide informed consent directly or by legal proxy.21
All participants in the TBIMS database who enrolled between 1988 and September 2008 were considered for this study. Inclusion criteria were (1) non-penetrating mechanism of injury and (2) engaged in productive activity preinjury. Exclusion criteria included (1) penetrating mechanism of injury; (2) 1-year follow-up not completed; (3) missing 1-year follow-up productivity status; (4) unknown PTA status from acute-care hospitalisation; (5) deceased prior to 1 year follow-up; and (6) missing mechanism of injury.
Measures
Galveston Orientation and Amnesia Test
The Galveston Orientation and Amnesia Test (GOAT) is a 10-item measure that assesses orientation as well as memory for events preceding and following TBI. Two consecutive administrations of the GOAT with a score ≥76 are consistent with emergence from PTA. Inter-rater reliability has been found to be excellent (correlation coefficient=0.99). The GOAT has been used to determine the duration of PTA in numerous studies of TBI.22
Orientation Log
The Orientation Log (O-Log) is a brief measure used to assess orientation that includes questions related to place, time and situational factors.23 24 The O-Log requires that a person obtains 25 points or greater on two consecutive occasions over 72 h in order to clear PTA. Reliability and validity estimates have be found to be quite high (correlation coefficients in the 0.90s) in multiple studies.23 25
Disability Rating Scale
The Disability Rating Scale (DRS) is an eight-item scale that incorporates the GCS as well as items assessing capability for feeding, toileting and grooming. Separate items rate need for assistance/supervision and potential for employment. Higher scores represent a greater level of disability.26
Glasgow Coma Scale
The Glasgow Coma Scale (GCS) is a brief screening measure of altered consciousness. There are three domains assessed: eye opening, spontaneous speech and best motor response. The measure is scored from 3 to 15, with 15 being the best score.5
Functional Independence Measure
The Functional Independence Measure (FIM) is composed of 18 items designed to operationally measure functional independence in self-care, mobility and cognition. Higher scores represent a greater level of independence.27
Procedure
Trained TBIMS research assistants collected information regarding injury severity (GCS, TFC) and medical course from hospital and emergency medical service records. Demographic information such as date of birth, education and employment was collected in interviews with the subjects or family/significant others. Emergence from PTA was assessed prospectively by repeated administration of the GOAT or O-Log 24- 72 h apart until two consecutive scores were achieved at or above the threshold for clearing PTA.28 29 For persons who were admitted to rehabilitation having already emerged from PTA, a chart-review procedure was used that documented two consecutive evaluations indicating orientation within 72 h.28 29 To minimise missing data, the length of PTA was calculated as the length of stay plus 1 day for persons discharged from inpatient rehabilitation still in PTA.20 30 While this procedure underestimates true PTA duration for many participants, it permits inclusion of participants with the most severe injuries and has been used in previous research.20 30 Exclusion of persons whose PTA did not resolve by rehabilitation discharge would bias the sample to those with less severe injuries.
Subjects were contacted by telephone 1 year postinjury to assess their level of functioning including productivity status. When unable to reach the database participants by telephone, data collection was attempted through a printed questionnaire. Productivity status was coded as productive if the person with TBI had returned to full- or part-time competitive employment, full- or part-time school enrolment, or full-time homemaker status, as previously defined in TBI Model System outcomes research.31 32
Data analysis
Examination of the association between PTA and productivity was examined along with the influence of other predictor variables of 1-year productivity outcome. Logistic regression models were used to compare the individual association between PTA and return to productivity with an adjusted association; the individual models accounted only for PTA, while the adjusted model accounted for both PTA and other variables associated with productivity. Individual associations between productivity and other injury severity and demographic variables (DRS at admission, GCS, FIM at admission and age at injury) were evaluated using logistic regression models. Variables found to be significantly associated with productivity were included in the adjusted model evaluating PTA. The odds of being productive, for each additional week of being in PTA, were compared between the individual model and the adjusted model to verify the individual association remained significant after accounting for other variables.
To compare the Russell intervals with the Mississippi intervals, productivity outcomes at 1-year postinjury were first evaluated. For each interval, a logistic regression model was generated to evaluate how well the injury severity scale modelled 1-year productivity. From the logistic regressions, receiver operating characteristic (ROC) curves were used to compare the performance of the two classification intervals by assessing the area under the curve (AUC). The AUCs were compared using non-parametric testing methods to determine statistically significant differences.33 In addition, the intervals were evaluated against an ROC curve modelling true PTA. In general, the categorisation of any continuous variable results in a loss of information and therefore a poorer model fit, so this assessment evaluated how well the severity scales compared with the true PTA values.
In the original Mississippi paper, the percentage of patients who were productive postinjury differed only slightly in the first two categories; therefore, a third classification scheme referred to as the Collapsed Mississippi scheme was added to the analysis with categories of moderate (0–28 days), severe (29–70 days) and extremely severe (>70 days).
Results
Study population
The study population included participants enrolled in the TBIMS database between 1988 and September 2008. See Gordon et al for TBI Model System inclusion and exclusion criteria.21 Of 8369 qualified participants, individuals who were not due for a 1-year follow-up evaluation (n=725), had a penetrating head injury (n=431), were non-productive premorbidly (2326), died during inpatient rehabilitation (n=8) or at one-year follow-up (n=172), or had a missing PTA status from acute hospitalisation records (n=541) were excluded. Of those excluded, 654 participants met two or more exclusion criteria. Of 4864 eligible subjects, an additional 769 were lost to follow-up (15.8%). Of the 4095 participants with follow-up data, the productivity status was missing for 249. Thus, 3846 met inclusion criteria with complete follow-up data were used in analyses. Table 1 lists the demographic and injury severity information for study participants and those lost to follow-up. Persons who were lost-to follow-up were more likely to be Hispanic, single and with higher GCS scores (less injury severity) with shorter lengths of stay than persons retained in the study sample at follow-up.
Associations with return to productivity
The results of the logistic regression models are shown in table 2. The individual associations between return to work and the injury severity and demographic variables are presented in terms of the odds of returning to productivity; for example, the unadjusted odds of returning to productivity are decreased by 16.4% (95% CI 14.6% to 18.8%; p<0.0001) for each additional week of PTA. All severity variables evaluated were individually associated with return to productivity. Age at injury was also significantly associated with return to productivity; however, younger patients have a significantly different association than older patients (see table 2), and therefore the associations are presented for those under 30 and those 30 and older. After adjusting for individually significant variables, the odds of returning to productivity were still significantly associated with PTA; the adjusted odds of being productive are decreased by 14% with every additional week of PTA (95% CI 12% to 17%; p<0.0001). The adjusted odds of being productive are slightly different from the unadjusted odds because the addition of other variables in the model can account for productivity status; however, the association remains significant, and the estimates are not shifted greatly, verifying a true and significant association between PTA and productivity.
Severity-of-injury scales
Of the 3846 eligible participants, 718 (17.1%) were discharged from inpatient rehabilitation still in PTA and thus had PTA imputed using length of stay plus 1 day. Overall, 49.4% (n=2082) of patients were productive at 1 year postinjury. Table 3 shows the percentage of patients who were productive at 1-year postinjury by injury-severity category for the Russell intervals and the Mississippi intervals. The proportion of patients who were productive did not reflect expected values for the Russell intervals, with more patients productive in the Russell ‘severe’ category than in the ‘mild/moderate’ category. As expected, the proportion of productive patients did drop as injury severity category increased for the Mississippi and Collapsed Mississippi schemes.
The AUC values and 95% CIs for the three classification schemes and the true PTA values are displayed in table 4. The area calculated for the Russell intervals was the lowest of the three, while the Mississippi scheme provided the highest area of the three schemes with the highest AUC for continuous PTA values. The AUC for the Russell scheme was significantly smaller than the Mississippi scheme (0.546 vs 0.666; p<0.001), indicating that the Mississippi scheme provided a relatively more accurate categorisation of PTA values. AUC for the collapsed Mississippi scheme was comparable with the Mississippi scheme (0.650 vs 0.666; p<0.097). The AUC for the full Mississippi scheme was not statistically different from the AUC for the true PTA values (0.666 vs 0.675; p<0.406), but the collapsed MS scheme was different from true PTA (0.650 vs 0.675; p<0.018). Findings support the full Mississippi PTA scheme as the interval superior to other schema and comparable with continuous PTA.
Discussion
Results from logistic regression and AUC values revealed that the Mississippi PTA scheme was a better predictor of productivity at 1 year postinjury than the Russell classification scheme. The reason for this may be that the intervals for the Mississippi classification system were developed based on observation of the distribution of PTA duration in a large sample of prospective rehabilitation admissions, while the intervals for the Russell system were somewhat arbitrary. In the current study, the actual duration of PTA was comparable with Mississippi PTA intervals as a predictor of productivity outcomes. This is consistent with prior studies showing that PTA duration is related to productivity outcomes.33
The use of the Mississippi system to categorise injury severity may be more practical than actual PTA durations for two reasons. First, serial assessment of PTA over time is labour-intensive. Use of the Mississippi classification system could allow assessments to be separated by greater intervals, thus saving time and cost. Second, researchers are often faced with estimating PTA retrospectively, often with limited medical records available. For example, a neurosurgical note from the day of injury may indicate that the person with injury was disoriented, and the next available progress note may not be until days later. If the most recent available progress note indicates that the person is oriented, the Mississippi intervals may be easily assessed, while actual duration of PTA would be impossible to obtain. In the absence of medical records, persons with injury or their family members may be able to estimate duration of PTA within the Mississippi intervals but may not know the actual number of days of PTA. Thus, the use of the Mississippi classification system can allow for fewer missing data with regard to PTA duration, while still providing good prediction of outcomes.
Although all injury severity predictors were associated with return to productivity, length of PTA demonstrated unique predictive validity in productivity after accounting for individual and injury severity variables. Additionally, one of the most important clinical findings from this study is that there was a decrease in productivity for every week of increased duration of PTA. This dose–response relationship is important from a clinical perspective, as it not only allows practitioners to plan for vocational outcomes in a very linear and direct manner, but also provides an opportunity to predict a person's expected level of productivity at a very early time point in the recovery process.
In the current study, persons in the ‘mildest’ Russell classification system group (PTA <1 day) were less likely to be productive at 1 year compared with the next category (1–7 days). One explanation for this counterintuitive result may be that in the TBIMS database, patients with very short PTA may have little diffuse axonal injury but serious focal injury, for example, subdural haematoma that may or may not have lasting effects. In fact, since all TBIMS patients are seriously injured enough to warrant inpatient rehabilitation for TBI, the likelihood is great that PTA of <1 day would be accompanied by a focal brain injury that would need to be considered in prognostic estimations. Individuals with a PTA of <1 with poor productivity at 1 year may be explained by other factors such as the presence of a psychological comorbidities and/or pending legal issues. In these cases, the Mississippi intervals would be less likely to result in classification error.
A differential effect in productivity status was noted for persons under the age of 30, as opposed to those over 30. For younger persons with brain injury, the odds of being productive at 1 year were decreased by about 35% for every 5 years of age, whereas for older individuals who sustained a traumatic brain injury, age decreased productivity by only 6% per every 5 years. This age difference indicates that for those who are younger when they are injured, there is a greater challenge for return to work and that rehabilitation efforts need to be customised to the unique needs of this particular cohort.
A significant strength of this study is the large, representative sample from the TBI Model Systems National Database. The 3846 individuals had complicated mild, moderate or severe TBI, and represented various geographic regions of the country that reflect significant educational, social and vocational diversity. These characteristics increase the generalisability of the findings.
There are several limitations that are important to consider. The proportion of persons lost to follow-up were more likely to be Hispanic, single and with less injury severity as measured by GCS. Further, the results are generalisable only to persons who receive comprehensive inpatient rehabilitation following TBI. Persons who do not receive inpatient rehabilitation may differ from those who do in regard to injury severity, socio-economic status, family support and other factors that may impact productivity outcomes. This sample did not include individuals who sustained penetrating injuries, and the results may not generalise to this group, who may differ in regard to mechanism and pathology of injury, risk factors for injury, demographics and levels of preinjury productivity. Another point to bear in mind is that ‘return to productivity’ is a broad term that encompasses a wide range of outcomes. Finally, only participants who were productive prior to injury were included in the current analysis. As PTA duration is commonly used to predict outcomes for all persons with TBI, generalisation of current findings to those who were not productive prior to injury should be established. The current study provides preliminary evidence for the usefulness of the Mississippi Classification System. Future research should further validate the system by assessing its predictive utility for widely used measures of outcome, including the Glasgow Outcome Scale—Extended, the Disability Rating Scale and the Community Integration Questionnaire. Its predictive ability should also be compared with other widely used measures of injury severity, including the GCS and time to follow commands. Future research should also include other variables that might influence both injury severity and recovery. As noted in a recent systematic review, studies on the TBIMS database have confirmed that multiple variables have an impact on TBI outcome.34 These variables include age, education, substance abuse and other psychiatric conditions, and level of family support, among others. Thus, PTA duration is only one variable to include in the process of clinical prognosis.
Conclusion
In conclusion, the current study shows that the use of the Mississippi intervals for classifying PTA duration results in a more accurate prediction of productivity outcomes compared with the traditional Russell intervals. Furthermore, use of the Mississippi system resulted in only slightly less predictive accuracy than use of actual days of PTA. While there are clinical indicators for serial assessment of PTA, use of the Mississippi system appears practical for research purposes, particularly when limited medical records are available.
Acknowledgments
The authors wish to thank M Grote, J Kittelson, C Harrison-Felix, D Mellick and S Barnett. Statistical support was provided by the HSR&D/RR&D Center of Excellence for Maximising Rehabilitation Outcomes (COE-HFP 09-156).
References
Footnotes
See Editorial commentary, p 474
Linked articles 238956.
Funding The contents of this manuscript were developed under grants from the National Institute on Disability and Rehabilitation Research, Department of Education (H133A060038 (TBI Model System National Data and Statistical Center), H133B090023 (Rehabilitation Research and Training Center on Developing Stategies to Foster Community Integration and Participation for Individuals With Traumatic Brain Injury), H133A070043 (Texas TBI Model System of TIRR), H133A070040 (Moss TBI Model System), H133A-080044 (Southeastern Michigan TBI System), H133A070036 (Virginia Commonwealth TBI Model System), H133A070027 (North Texas TBI Model System), H133A070042 (Carolinas Traumatic Brain Injury Rehabilitation and Research System)).
Competing interests None.
Ethics approval Ethics approval was provided by the Methodist Rehabilitation Center, Jackson, Mississippi; James A Haley Veterans Hospital, Tampa, Florida; TIRR Memorial Hermann, Baylor College of Medicine, Houston, Texas; Shepherd Center, Atlanta, Georgia; Moss Rehabilitation Research Institute, Philadelphia, Pennsylvania; Wayne State University School of Medicine, Detroit, Michigan; Virginia Commonwealth University, Richmond, Virginia; Baylor Institute of Rehabilitation, Dallas, Texas; Carolinas Rehabilitation, Charlotte, North Carolina.
Provenance and peer review Not commissioned; externally peer reviewed.