Influence of Early Variables in Traumatic Brain Injury on Functional Independence Measure Scores and Rehabilitation Length of Stay and Charges

Transcription

1 797 Influence of Early Variables in Traumatic Brain Injury on Functional Independence Measure Scores and Rehabilitation Length of Stay and Charges Todd D. Cowen, MD, Jay M. Meythaler, MD, Michael J. DeVivo, DrPH, Clarence S. lvie III, MPH, Joan Lebow, MA, Thomas A. Novack, PhD ABSTRACT. Cowen TD, Meythaler JM, DeVivo M J, Ivie CS III, Lebow J, Novack TA. Influence of early variables in traumatic brain injury on functional independence measure scores and rehabilitation length of stay and charges. Arch Phys Med Rehabil 1995;76: Objective: To determine the relationship between early variables (initial Glasgow Coma Scale [GCS] scores, computed tomography [CT] findings, presence of skeletal trauma, age, length of acute hospitalization) and outcome variables (Functional Independence Measure [FIM] scores, rehabilitation length of stay [LOS], rehabilitation charges) in traumatic brain injury (TBI). Design: Inception cohort. Setting: University tertiary care rehabilitation center. Patients: 91 patients with TBI. Interventions: Inpatient rehabilitation. Main Outcome Measures: FIM, rehabilitatian LOS, and rehabilitation charges. Results: Patients in the severely impaired (GCS = 3 to 7) group showed significantly lower (p =.01) mean admission and discharge motor scores (21.26, 39.83) than patients in the mildly impaired (GCS = 13 to 15) group (38.86, 55.29). Cognitive scores were also significantly lower (17 <.01) in the severely impaired group on admission (26.73 vs 54.14) and discharge (42.28 vs 66.48). These findings continued to be statistically significant (p <.01) after regression analysis accounted for the other early variables previously llisted. Regression analysis also illustrated that longer acute hospitalization LOS was independently associated with significantly lower admission motor (/7 <.01) and cognitive (p =.05) scores, and significantly higher (/1 =.01) rehabilitation charges. Patients with CT findings of intracranial bleed with skull fracture had longer total LOS (70.88 vs days; p <.05), rehabilitation LOS (30.01 vs days; p <.10), and higher rehabifitation charges ($43,346 vs $25,780; p <.05). Paradoxically, those patients in a motor vehicle crash with an extremity bone fracture had significantly higher (p =.002; p =.04 after regression analysis) FIM cognitive scores on admission (48.30 vs 27.28) and discharge (64.74 vs 45.78) than those without a fracture. Finally, data available on rehabilitation admission were used to predict discharge outcomes. The percentage of explained variance for each outcome variable is as follows: discharge FIM motor score, 69.5%; discharge FIM cognitive score, 71.2%; rehabilitation LOS, 54.1%; rehabilitation charges, 61.1%. The most powerful predictor of LOS and charge,~ was the admission FIM motor score (p <.001), followed by CT findings (p =.02) and age (p =.04). Conclusion: Information readily available on rehabilitation admission, particularly the FIM motor score, may be useful in predicting discharge FIM scores as well as utilization of medical rehabilitation resources. Earlier transfer to rehabilitation may result in higher functional status and lower rehabilitation charges, as well as lower acute hospitalization charges. The presence of extremity fractures encountered during a motor vehicle crash is associated with a more favorable outcome in TBI as evidenced by higher discharge FIM cognitive scores by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation Predicting outcome after traumatic brain injury (TBI) has been the focus of numerous studies over the past several decades. Teasdale and Jennett developed the Glasgow Coma Scale (GCS) as a method of describing injury severity and predicting outcome after brain injury] '2 Subsequently, the Glasgow Outcome Scale (GOS) 3 was developed to assess outcome and was found to correlate well with early GCS scores. Many other variables have been used to predict GOS scores after TBI. A complete discussion and bibliography of each of these v~mables is beyond the scope of this article. From the Department of Physical Medicine and Rehabilitation, University of Alabama Birmingham Medical Center, Birmingham, AL. Submitted for publication January 27, Accepted in revised form May 2, No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the authors or upon any organization with which the authors are associated. Reprint requests to Todd D. Cowen, MD, Spain Rehabilitation Center, 1717 Sixth Avenue South, Birmingham, AL, by the American Congress Of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation /95/ /0 Pupillary responses, 2 presence of skull fractures or hematoma, 4'5 initial systolic blood pressure, 5 age, 4'5 mechanism of injury, 5 brainstem reflexes, 6 and duration of post-traumatic amnesia 7 have all been shown to correlate with GOS scores after TBI. In addition, the combination of hypoxia and hypotension 8 has been shown to have a significant negative impact on GOS scores. More recently, intracranial pressure 9 and cerebral perfusion pressure 1 have also been utilized to predict outcomes (GOS) among this patient population. The GOS grossly separates outcome into five categories. Most studies using the GOS simply designate patients into a good outcome (GOS of 4 or 5) or poor outcome (GOS of 1 to 3) category. In response to the limited scoring range of the GOS, Rappaport and colleagues ~1 developed the Disability Rating Scale (DRS), which has been shown to be a more sensitive measure of outcome than the GOS. 1J,12 DRS scores at one year are highly predicted by electroencephalography (EEG) and GCS scores at seven days after injury, brainstem auditory evoked responses, initial computed tomography (CT) findings, and age. ~3 In the same study, the EEG findings and GCS scores combined to account for 75% of the variance

2 798 BRAIN INJURY REHABILITATION, Cowen of the DRS scores at one year. Another study found the Patient Evaluation Conference System (PECS) to be a more powerful predictor of return to work than the DRS and the Level of Cognitive Functioning Scale. 14 The Functional Independent Measure (FIM) 15 is an attempt to further refine outcome measurement and is currently being utilized in inpatient rehabilitation programs with patients having varied medical problems, including TBI. No studies were found that have evaluated predictive variables as they relate to the FIM after TBI. This study evaluates several early variables that are readily available to the clinician (initial GCS scores, etiology of injury, CT findings, associated skeletal trauma, age, and length of acute hospitalization) to determine their relationship to inpatient rehabilitation admission and discharge motor and cognitive FIM scores. 15'16 In addition, rehabilitation length of stay (LOS) and charges will be explored in reference to early variables and the FIM. Study Population METHODS All inpatient admissions at a university-based rehabilitation facility between March, 1992, and December, 1993, were screened for the diagnosis of new (no prior history) TBI. For uniformity, only patients admitted directly from acute care to rehabilitation were included. Patients who had undergone previous inpatient rehabilitation at the study facility or another facility were excluded. In addition, patients with a history of stroke, spinal cord injury, or psychiatric disorders were not included. A total of 110 patients were identified. Of these, 19 did not have initial GCS scores (within 24 hours) documented in the acute hospitalization records. The mean age of the 91 remaining study patients was years (SD = 14.76) and 71% were men. Rehabilitation Program Criteria for admission to the inpatient rehabilitation program included the ability to participate in therapy, stable medical course, and the clinical judgment of the admitting physiatrist regarding the potential benefits of the program to the patient. Time of discharge was decided by the rehabilitation team and was considered when patients reached their functional goals. Because the study was performed retrospectively, it had no impact on the admission or discharge process. Therapies focused on individualized functional goals that were identified on admission to rehabilitation. Patients received three or four hours of physical and occupational therapy each day as well as psychology services. If indicated, speech therapy was utilized for communication and swallowing problems. Average rehabilitation LOS was days (SD = 13.60). Data Collection Because of its proven validity and reliability, and its ability to measure functional capacity as it relates to burden of care, 16 the FIM ~5 was utilized to assess functional outcome. All FIM data was collected using the protocol established by the Uniform Data System (UDS) for Medical Rehabilitation in Buffalo, New York. Members of the rehabilitation team collected this information prospectively at both admission and discharge. Because the study did not exist at the time the measurements were taken, data collection can be considered double-blinded (for purposes of evaluating the study hypothesis) with respect to both the rehabilitation team members and patients. In addition, the facility had successfully completed the UDS reliability certification process before collection of any data used in the study. Initial GCS scores were obtained from the initial emergency room reports and/or the neurosurgical intensive care unit records within the first 24 hours after injury. Patients were divided into severe (GCS = 3 to 7), moderate (GCS = 8 to 12), and mild (GCS = 13 to 15) groups based on these scores. Etiology of the TBI was ascertained by direct examination of the hospital records, including admission notes, emergency room records, and consultation reports. Etiologies were classified as either motor vehicle crash (MVC), blunt trauma, fall, or gunshot wound (GSW). CT information was obtained from radiology reports. Head CT findings were designated as "none" (no abnormalities), intracranial bleed (includes any intracranial hemorrhage), or intracranial bleed with skull fractures (all patients with skull fracture suffered intracranial bleeding). Because associated skeletal trauma could possibly affect FIM scores, the presence of extremity fractures was included as a variable. In particular, the presence of a long bone extremity fracture or a hip or shoulder girdle fracture was considered. This information was obtained from both the acute hospital and rehabilitation charts as well as radiology reports. For simplification, patients were designated as "no fracture" or "with fracture." Information such as age, sex, length of rehabilitation, and length of acute hospital stay were readily available from the UDS data base. Statistical Analysis FIM scores were divided into motor and cognitive subscores and rescaled from 1 to 100 using results of the Rasch analysis previously performed by Heinemann and associates. ~6 This method has been shown to provide two valid measures (motor and cognitive) reflecting burden of care across impairment groups. The raw ordinal FIM scores are converted into interval scales that provide two linear substrates upon which more powerful parametric analyses can be performed. ~6 All statistical analyses were based on these converted scores. Separate one-way analyses of variance were performed with dependent variables being admission FIM motor score, discharge FIM motor score, admission FIM cognitive score, discharge FIM cognitive score, change in FIM motor score, change in FIM cognitive score, acute care LOS, rehabilitation LOS, and rehabilitation charges. Inpatient physician fees were not included in the rehabilitation charges. All charges were adjusted to 1993 dollars using the Hospital and Related Services Component of the Consumer Price Index. These charges represent what was billed by the hospital rather than

3 BRAIN INJURY REHABILITATION, Cowen 799 Table 1: FIM Motor Scores Among GCS Severity Groups on Admission and Discharge to Rehabilitation Mean Admit p Mean Difference p Mean Discharge p Mean Difference p Motor Value (Regression) Value Motor Value (Regression) Value Severe (N = 40) (2.96) (3.75) Moderate (N = 30) (3.39) < (4.51) (2.21) < (4.39) <.01 Mild (N = 21) (1.76) < (5.14) (1.57) < (5.06).02 GCS Severity Groups: Severe, GCS = 3-7; Moderate, GCS = 8-12; GCS = Parentheses indicate 1 standard error; p values indicate comparisons to the severe group. what was actually reimbursed. Grouping (independent) variables included GCS scores, CT findings, etiology of injury, and presence of extremity fracture. Groups were established as defined in the data collection system. When overall differences across all groups were statistically significant (p <.05), post hoc t tests were used to assess the statistical significance of differences between specific groups. Next, several multiple linear regression analyses were performed to control for possible confounding effects among the independent variables and confirm the univariate results. Dependent measures were the same as those used in the univariate analyses. For each independent multi-category grouping variable:, one dichotomous variable coded 0 for no and 1 for yes was created for each category, with one referent category excluded for comparison purposes. For example, because there were three GCS groups, two GCS variables were created (one for mildly impaired and one for moderately impaired). In this way, the regression model coefficient for each category can be interpreted as the adjusted mean difference in the dependent measure between that category and the referent category. This coefficient can then be compared with the univariate mean difference between groups for that dependent measure to assess how much of that difference is due to confounding by other factors included in that regression model. In addition to the grouping variables used in the univariate analyses, age was included as a covariate in all regression analyses. Acute care LOS was included as an independent variable in all regression models except when it served as the dependent measure. Admission motor FIM score and admission cognitive F1M score were also included in the regression models of discharge motor FIM score, change in motor FIM score, discharge cognitive FIM score, change in cognitive FIM score, rehabilitation LOS, and rehabilitation charges. RESULTS When patients were grouped by initial GCS scores, mean FIM motor scores were significantly lower (p <.01) for the severely impaired group compared with the mildly impaired and moderately impaired groups on both admission and discharge (table 1). These differences were only partly the result of confounding by the other variables discussed previously because the observed differences in FIM scores after regression analysis decreased only slightly while remaining statistically significant. There was no statistically significant difference between the mean change in motor scores, although a trend for higher gains was noted in the moderately impaired group (not shown). A similar finding was noted with the FIM cognitive scores (table 2). The mildly impaired group displayed significantly higher scores on admission than the moderately impaired (p <.05) and severe (p <.01) groups. After regression analysis, the difference remained significant between the mildly and severely impaired groups. On discharge, both the mildly and moderately impaired groups demonstrated significantly higher (p <.01) mean scores than the severely impaired group. Once again, the difference remained significant after regression analysis. The gain in FIM cognitive scores was highest among those in the moderately impaired group, with the difference after regression analysis showing a significantly higher mean gain in these patients ( ; p =.01) compared with the severely impaired group (not shown). Length of acute hospitalization was significantly (p <.01) longer in the severely impaired group (table 3), with the difference remaining significant after regression analysis. Rehabilitation LOS showed a stepwise increase as injury severity increased, although no statistically significant findings were noted. Rehabilitation mean charges demonstrated a similar stepwise pattern, with the severely impaired group having significantly higher (p <.01) charges compared with the mildly impaired group (before regression only). Regression analysis was also used to examine the relationship between LOS and outcomes (table 4). As the length of acute hospitalization increased, the admission F1M motor and cognitive scores significantly decreased. Also, a longer acute hospitalization resulted in a longer mean inpatient rehabilitation LOS and significantly (p =.01) higher inpatient rehabilitation charges. A longer inpatient rehabilitation LOS was associated with significantly higher gains in both FIM motor and cognitive scores (p <.01). Table 2: FIM Cognitive Scores Among GCS Severity Groups on Admission and Discharge to Rehabilitation Mean Admit p Mean Difference p Mean Discharge p Mean Difference p Cognitive Value (Regression) Value Cognitive Value (Regression) Value Severe (N = 40) (4.19) (4.43) Moderate (N = 30) (4.73) < (6.59) > (3.33) < (5.92).03 Mild (N = 21) (3.70) < (7.51) (1.57) < (6.83).01 GCS Severity Groups: Severe, GCS = 3-7; Moderate, GCS = 8-12; Mild, GCS = Parentheses indicate 1 standard error; p values indicate comparisons to the severe group.

4 800 BRAIN INJURY REHABILITATION, Cowen Table 3: Comparisons of LOS and Cost Data Among the GCS Severity Groups Mean Acute Mean Mean Mean Hospitalization p Difference p Rehah p Difference p Mean Rehab LOS Value (Regression) Value LOS Value (Regression) Value Charges $ Mean p Difference p Value (Regression) Value S e v e r e (N = 40) (5.43) (I.96) 38,404 (3,226) Moderate (N = 30) (2.50) < (6.55) < (2.83) > (3.55) >.10 27,218 (3,644) Mild (N = 21) (3.30) <.05-15,59 (7.79).05 18,58 (2.55) > (4,05) >.10 23,595 (4255) GCS Severity Groups: Severe, GCS = 3-7; Moderate, GCS = 8-12; Mild, GCS Parentheses indicate ±1 standard error; p values indicate comparisons to the severe group. m >.05-6,149 (5,231) >.05 <.01-9,090 (5,959) >.05 The outcomes as they relate to CT findings were analyzed (patients with a GSW were considered as a group separate from the patients with nonpenetrating injuries and were not included in this part of the analysis). A trend (nonsignificant) for lower admission and discharge FIM scores was associated with intracranial bleed and an intracranial bleed with skull fracture as noted on CT (not shown). No significant differences were observed with regards to changes in FIM motor and cognitive scores. A particularly significant finding was that those patients suffering a bleed with skull fracture had a longer (p <.05) total LOS and higher (p <.05) rehabilitation charges than those with a normal CT (table 5). Also, after regression analysis, there was a persistent trend for longer rehabilitation LOS and higher rehabilitation charges for this group. Next, outcome as it relates to etiology of head injury was examined. A trend for lower scores, longer lengths of stay, and increased charges was observed among the patients suffering from blunt trauma (data not shown). However, because of unequal distribution of patients (MVC = 63, GSW = 11, Fall = 10, Blunt = 7), no statistical significance was observed. Outcome as it relates to presence of extremity fractures was also investigated. Because all but one of the patients with an extremity fracture had an etiology of MVC, only this group was used for this analysis. Any possible variance introduced by the etiology has thus been removed when comparing those with and without an extremity fracture. The admission and discharge FIM motor scores showed a trend for higher scores among the group with fractures, although no statistical significance was noted before or after regression analysis (not shown). However, the admission and discharge FIM cognitive scores were significantly higher for Table 4: Relationships Between Acute Hospitalization LOS and Admission FIM Scores, Rehabilitation LOS, and Rehabilitation Costs Using Regression Analysis Dependent Variable Mean Difference p Value Admission Motor -.25 (.07) <.01 Admit Cognitive -.21 (. 10).05 Rehab LOS +.10 (.06).07 Rehab Charges $ +212 (83).01 Relationships between rehab LOS and gain in FIM scores using regression analysis Gain Motor +.31 (,11).01 Gain Cognitive +.49 (. 12) <.01 The mean difference reflects the change in the dependent variable as the acute hospitalization LOS and the rehabilitation LOS increases by one day, Parentheses indicate + 1 standard error of the difference. the fracture group both before (p =.002) and after (p =.04) regression analysis (table 6). As shown, after accounting (via regression analysis) for the differences among the other variables, those patients with an extremity fracture had a mean FIM cognitive score points higher on admission and points higher on discharge than those patients without an extremity fracture. On examining the raw FIM cognitive data before Rasch conversion (table 7), it was found that the patients with an extremity fracture had significantly higher mean scores on all of the individual cognitive measures on admission and discharge. These raw data were analyzed to ensure that one or a few of the raw cognitive measures were not accounting for the differences noted on the Rasch-converted scores. Indeed, the opposite was true in that all of the raw scores were significantly higher among the patients with an extremity fracture compared with those without an extremity fracture. In addition, before and after regression analysis, the presence of an extremity fracture was associated (table 8) with a shorter acute hospitalization (12.19 days less, p =.01) and total length of stay (18.76 days less, p <.01). The rehabilitation length of stay was also shorter (19.82 vs days, p =.03) and the mean rehabilitation charges were less ($26,105 vs $38,235; p =.03) among those with an extremity fracture. This difference, however, did not remain significant after regression analysis accounted for the other variables previously mentioned. Finally, the data that would be available at the time of admission to rehabilitation was used to prognosticate discharge FIM scores, LOS, and charges. These data would include initial GCS scores, etiology of traumatic brain injury, CT findings, presence of extremity fracture, age, acute hospitalization LOS, and admission FIM motor and cognitive scores (not included in the previous-regressions). Using this information, 69.5% of the variance of discharge FIM motor scores was explained. Most of this was contributed by the admission FIM motor score (p <.001) and the length of acute hospitalization (p =.06). Similarly, 71.2% of the variance of the discharge FIM cognitive score was accounted for by the model. Most of the variance was explained by the admission cognitive score (p <.001). Using the same analysis, 54.l% of the variance of the rehabilitation LOS and 61.1% of the variance of the rehabilitation charges was explained by the model. In both, the most significant predictor was the admission FIM motor score (p <.001). Other significant variables included the presence of a bleed with skull fracture on CT scan (19 =.02) and age (p =.04).

5 BRAIN INJURY REHABILITATION, Cowen 801 Table 5: LOS and Cost Data Among Patients as Grouped by CT Findings Mean Mean Mean Mean Total p Difference p Mean p Difference p Mean Rehab p Difference p LOS Value (Regression) Value Rehab LOS Value (Regression) Value Charges $ Value (Regression) Value CT-Normal (N = 25) (3.83) (2.45) 25,780 (3,443) CT-Blced (N = 38) (4.44) > (8.44) > (2.25) > (3.54) >.10 32,948 (3,266) >.10 +3,100 (5,212) >.10 CT~Bleed + skull fracture (N = 17) (9.56) < (10.75) > (3.61) < (4.52).08 42,346 (6,471) <.05 +I2,346 (6,658).07 Parentheses indicate _+ 1 staladard error; p values indicate comparisons to the "normal CT" group. DISCUSSION Before interpreting the data, the potential pitfalls of the current study must be considered. First, although the vast majority of patients initially were treated at the same university-based acute hospital, some patients were transferred to rehabilitation from outlying medical centers and may have received different acute medical care. Also, because of the retrospective design, not all GCS scores were collected uniformly, although all were within 24 hours. In addition, as previously mentjioned, some of the intragroup comparisons may have been weakened by unequal distribution of patients. As noted previously, the initial GCS scores have been used to help predict GOS scores 1-3 as well as DRS scores 13 after TBI. Intuitively, the lower the initial GCS scores, the poorer the outcome. However, no study was found looking at the relationship between GCS scores and FIM scores before and after inpatient rehabilitation of this patient population. In this study, patients were divided into groups by injury severity based on their initial GCS scores. As expected, those in the severely impaired group had significantly lower admission and discharge FIM motor scores. This difference continued to be significant even after accounting for the other study variables via a multiple regression analysis. The FIM cognitive scores showed similar findings. After regression analysis, the admission and discharge cognitive scores were significantly lower (22.31 and points less, respectively; p <.01) in the :~everely impaired group compared with the mildly impaired group. Regarding the change in cognitive scores during rehabilitation, the moderately impaired group showed the highest mean gain. This may be explained by poor recovery among the severely impaired group and a ceiling effect among the patients in the mildly impaired group. One must also consider that the measurement tool (FIM) may not have the inherent sensitivity to detect subtle changes in cognitive status. The longer total LOS in the severely impaired group mostly resulted fi:om a longer acute hospitalization. This may be a reflection of more medical and treatment complications among the more severely brain-injured patients. Rehabilitation LOS showed a similar trend, but perhaps because of the relatively small numbers of patients, no statistically significant findings were noted. Rehabilitation charges were highest among the severely impaired group, which may simply be a reflection of longer rehabilitation LOS. One of the most important findings was that longer acute hospitalization LOS resulted in significantly lower FIM motor and cognitive scores on admission to rehabilitation. This may be presumed to simply be a reflection of increased severity of injury, but variables such as GCS scores, CT findings, extremity fractures, etiology, and age were controlled by the regression analysis. Using the same analysis, longer acute hospitalization was strongly associated (p =.07) with longer rehabilitation LOS and significantly (p =.01) associated with higher rehabilitation charges. This lends strong support for earliest possible transfer to rehabilitation in an attempt to obtain better functional results as well as lower rehabilitation charges. Cost savings would not only be realized from rehabilitation charges but also from fewer patient days incurred during the acute hospitalization phase. It was encouraging to note that the patients made greater gains of FIM motor and cognitive scores the longer they spent in rehabilitation. This may not only be a reflection of the usefulness of inpatient programs for this population, but also an indication of appropriate timing of admission and discharge as well as optimal utilization of inpatient and outpatient services. Future studies comparing the outcomes of patients treated in an inpatient setting with those discharged earlier to outpatient programs would be useful. Other studies have recognized the significance of CT findings and outcome. 4'5'13 As the number of abnormalities found by CT increased, outcome as measured by GOS scores was poorer. 5 Indeed, 81% of patients with a normal CT scan made a good recovery, but only 45% of patients with only one abnormality improved similarly. In the same study, 67% of patients with a skull fracture had a poor outcome (GOS of 1 to 3). The current study concurred with previous data in that a trend for lower scores was seen in patients suffering Table 6: FIM Cognitive Scores Among Patients With and Without Extremity Fracture Mean Admit p Mean Difference p Mean Discharge p Mean Difference p Cognitive Value (Regression) Value Cognitive Value (Regression) Value No fracture (N = 36) Fracture present (N = 27) (4.23) (4.69) (6.88) (4.09) (3.79) (5.91).04 Parentheses indicate _ 1 standard error.

6 802 BRAIN INJURY REHABILITATION, Cowen Table 7: Comparisons of Mean FIM Cognitive Items Among Patients With and Without Extremity Fractures No Fracture Fracture p FIM Item (N = 36) (N = 27) Value Admit COMP 2.67 (.29) 4.11 (.37) <.01 D/C COMP 4.33 (.30) 5.78 (.26) <.01 Admit EXP 2.61 (.29) 4.33 (.39) <.01 D/C EXP 4.17 (.31) 5.74 (.30) <.01 Admit SOC 2.50 (.30) 3.93 (.37) <.01 D/C SOC 4.08 (.34) 5.48 (.31) <.01 Admit PROB 2.06 (.25) 3.52 (.36) <.01 D/C PROB 3.25 (.30) 4.74 (.30) <.01 Admit MEM 2.08 (.26) 3.41 (.34) <.01 D/C MEM 3.14 (.30) 4.74 (.31) <.01 Parentheses indicate _ 1 standard error. Abbreviations: D/C, Discharge; COMP, comprehension; EXP, expression; SOC, socialization; PROB, problem solving; MEM, memory. an intracranial bleed with a skull fracture. Because of the relatively small number of patients in our study, further subdivisions of CT findings were not statistically practical. Also, because the study was designed to provide a simple and easy method of classifying patients for prognostic purposes, a more complex system was not used. A significant finding was the increased total LOS (70.88 days) in this group (bleed with skull fracture) compared with the group with normal CT scans (43.08 days). Rehabilitation LOS was significantly longer (30.01 vs days) and mean rehabilitation charges higher in this population of patients as well. This will provide important information when predicting utilization of resources. Previous reviews J7 indicated that the presence of multiple injuries may have an effect on initial mortality, but mortality rates of patients with the same GCS scores were not affected by the presence of concomitant multiple system injuries. In another study, the ultimate rehabilitation outcome was unaffected by the presence of skeletal trauma.~8 In the current study, MVC victims with TB! appeared to experience a protective effect from the presence of extremity fractures. Lower FIM motor scores might be expected among patients with a fracture, but our patients actually had slightly higher (though nonsignificant) motor scores. The cognitive scores, however, were significantly higher among those with an extremity fracture. As described, even when examining the raw cognitive FIM variables, those with a fracture had significantly higher mean subscores universally. In addition, those with extremity fractures had shorter acute hospital and rehabilitation lengths of stay, as well as lower rehabilitation charges. There may be several reasons for the above phenomenon. One must consider the possible presence of an unknown variable that simply caused an artificial finding. Perhaps there are metabolic complications associated with complicated skeletal trauma that contributed to a lower initial GCS score. In addition, patients with more severe injuries and multiple skeletal trauma may have been inadvertently selected out and not transferred to rehabilitation. However, the possibility that a protective effect was introduced by the presence of an extremity fracture must also be considered. The energy absorbed by the fracture may have spared the brain of some additional forces. Perhaps the patients with fractures used their extremities to brace and protect themselves during the MVC in a way that lessened cranial trauma. Also, patients with obvious skeletal trauma may have received quicker medical attention (ie, intravenous fluids, oxygen, transport, etc) and were spared from more severe hypotension and hypoxia, which has been associated with poorer outcome. 8 Whatever the reason, FIM cognitive scores were statistically significantly higher among these patients and were not related to GCS scores, CT findings, age, or length of acute hospitalization (all accounted for by the regression model). In addition, etiology was not a causative factor because all of the patients examined in this aspect of the study were involved in an MVC. As described in the results section, discharge FIM motor and cognitive scores can be predicted with a high degree of the variance explained by the data known at time of rehabilitation admission (initial GCS scores, CT findings, etiology, presence of fractures, age, length of acute hospitalization, and admission FIM motor and cognitive scores). As expected, the greatest predictors were the admission FIM scores and the length of acute hospitalization. Length of acute hospitalization has been shown to be predictive of outcome (GOS) in a previous study of severely brain-injured patients. 19 Longer acute care intuitively implies more severe injury, and may also delay early rehabilitation and prolong immobilization. Also, rehabilitation LOS and charges were well predicted (54.1% and 61.1% of variance explained, respectively) with the same admission data listed above. The admission FIM motor score was the most powerful predictor of each. Also, the presence of an intracranial bleed with skull fracture and age was significantly associated with longer rehabilitation LOS and higher charges. With TBI, it might be expected that the admission FIM cognitive score would have a stronger influence on these variables. The lack of this finding may be a reflection of a "motor-goal" approach to rehabilitation. These findings may have an important impact on the allocation and prediction of resource utilization in this patient population. CONCLUSIONS This study illustrates that in TBI information readily available to the clinician on admission to inpatient rehabilitation can be useful in predicting resource utilization as well as Table 8: Length of Stay and Cost Data Among Patients With and Without Extremity Fractures No Fracture Fracture p Variable Measures (N = 36) (N = 27) Value Total LOS (mean) (4.93) (4.24).02 Total LOS (regression) (6.34) <.01 Acute LOS (mean) (3.57) (2.59).08 Acute LOS (regression) (4.18).01 Rehab LOS (mean) (2.49) (2.14).03 Rehab LOS (regression) (3.58).40 Rehab charges $ (mean) 38,235 (3,936) 26,105 (3,547).03 Rehab charges $ (regression) (5,349).40 Parentheses indicate _+ 1 standard error.

7 BRAIN INJURY REHABILITATION, Cowen 803 functional outcomes as measured by FIM scores. In particular, the admission FIM motor scores were highly predictive of rehabilitation LOS and charges. A high percentage of the variance for these two outcome measures was explained by the admission v~iables, mainly by the admission FIM motor scores. As expected, lower initial GCS scores are independently associated with significantly lower admission and discharge FIM motor and cognitive scores. In addition, total LOS was significantly longer among those with lower GCS scores. CT findings of an intracranial bleed with skull fracture were also associated with significantly longer total and rehabilitation lengths of stay and significantly higher rehabilitation charges. Paradoxically, TBI patients from motor vehicle crashes who have concoraitant extremity fractures had significantly higher admission and discharge FIM cognitive scores than those without exllremity fractures. These findings have not previously been reported in the literature and were noted even after controlling for other variables such as GCS scores, CT findings, and acute hospitalization LOS. Finally, independent of the other variables noted, patients who incurred longer acute hospitalizations had lower admission FIM motor and cognitive scores and had longer rehabilitation lengths of stay with higher rehabilitation charges. This may support the idea that quicker transfer to rehabilitation could result not only in better functional outcomes but overall lower rehabilitation and acute hospitalization costs. References 1. Teasdale G, Jennett B. Assessment of coma and impaired consciousness. Lancet 1974;July 13: Teasdale G, Mmxay G, Parker L, Jennett B. Adding up the glasgow coma score. Acta Neurochir 1979; Suppl 28: Jennett B, Snoek J, Bond MR, Brooks N. Disability after severe head injury: observations on the use of the glasgow outcome scale. J Neurol Neurosurg Psychiatry 1981;44: Jennett B, Teasdale G, Braakman R, Minderhoud J, Heiden J, Kurze T. Prognosis of patients with severe head injury. Neurosurgery 1979;4: Waxman K, Sundine M, Young R. Is early prediction of outcome in severe head injury possible? Arch Surg 1991;126: Born JD, Albert AA, Hans P, Bonnal J. Relative prognostic value of best motor response and brain stem reflexes in patients with severe head injury. Neurosurgery 1985; 16: Bishara SN, Partridge FM, Godfrey PD, Knight RG. Post-traumatic amnesia and glasgow coma scale related to outcome in survivors in a consecutive series of patients with severe closed head injury. Brain Inj 1992;6: Kohi YM, Mendelow AD, Teasdale GM, Allardice GM. Extracranial insults and outcome in patients with acute head injury-relationship to the glasgow coma scale. Brain Inj 1984; 16: Alberico AM, Ward JD, Choi SC, Marmarou A, Young HF. Relationship to mass lesions, diffuse injury, and ICP course in pediatric and adult patients. J Neurosurg 1987;67: Changaris DG, McGraw CP, Richardson JD, Garretson HD, Arpin EJ, Shields CB. Correlation of cerebral peffusion pressure and glasgow coma scale to outcome. J Trauma 1987;27: Rappaport M, Hall KM, Hopkins K, Belleza T. Disability rating scale for severe head trauma: coma to community. Arch Phys Med Rehabil 1982; 63: Hall K, Cope DN, Rappaport M. Glasgow outcome scale and disability rating scale: comparative usefulness in following recovery in traumatic head injury. Arch Phys Med Rehabil 1985;66: Thatcher RW, Cantor DS, McAlaster R, Geisler F, Krause P. Comprehensive predictions of outcome in closed head-injury patients. Ann N Y Acad Sci 1991;620: , Rao N, Kilgore K. Predicting return to work in traumatic brain injury using assessment scales. Arch Phys Med Rehabil 1992;73: Granger CV, Hamilton BB, Sherwin FS. Guide for use of the uniform data set for medical rehabilitation. Buffalo, NY: Uniform Data System for Medical Rehabilitation, Heinemann AW, Linacre JM, Wright BD, Hamilton BB, Granger C. Relationship between impairment and physical disability as measured by the functional independence measure. Arch Phys Med Rehabil 1993; 74: Davis RA, Cunningham PS. Prognostic factors in severe head injury. Surg Gynecol Obstet 1984; 159: Groswasser Z, Cohen M, Blankstein E. Polytrauma associated with traumatic brain injury: incidence, nature and impact on rehabilitation outcome. Brain Inj 1990;4: Spettell CM, Ellis DW, Ross SE, Sandel ME, O'Malley KF, Stein SC, et al. Time of rehabilitation admission and severity of trauma: effect on brain injury outcome. Arch Phys Med Rehabil 1991;72: Arch Phys Med Rehabi! Vol 76, September 1995