Recovery of ambulation during inpatient rehabilitation: physical therapist prognosis for children and adolescents with traumatic brain injury.

One of the first questions parents frequently ask physical therapists when their child is transferred from the acute care setting to the rehabilitation setting is, "What is the likelihood that my child will regain the ability to walk?" After an acquired injury, particularly for children who previously walked, families often expect their children to regain their preinjury ambulation status. Recovery of premorbid ambulation ability is a primary goal for children admitted to inpatient rehabilitation programs with traumatic brain injury (TBI) and often is a primary focus of the individualized rehabilitation plan of care. (1,2)

Physical therapists often focus primarily on recovery of mobility skills. Physical therapist examination during inpatient rehabilitation of children and adolescents following TBI includes a evaluation of premorbid and current (injury-related) history. Because the clinical sequelae of TBI are multifaceted, the systems review for a child with TBI includes the cardiovascular/pulmonary, integumentary, musculoskeletal, and neuromuscular systems as well as the child's general cognitive and communication abilities. (3,4) As part of the evaluation of a child within a comprehensive inpatient rehabilitation program, physical therapists routinely examine function as part of their testing. (5) The results of the physical therapist's examination and evaluation are used to establish a physical therapist diagnosis and to determine a prognosis. (3)

A physical therapist prognosis includes "the determination of the predicted optimal level of improvement in function." (3(p46)) Knowledge of prognostic factors about mobility recovery contributes to formation of a plan of care and theoretically to effective intervention programs and the establishment of realistic goals for children and their families. We know of no studies in which predictors or prognostic factors of recovery of the ability to ambulate have been identified during inpatient rehabilitation for children and adolescents following TBI.

Descriptive accounts of recovery of the ability to ambulate in children and adolescents after TBI are varied. Although many children with TBI are unable to ambulate at the time of discharge from the hospital, (6-10) others have only minor residual deficits, such as performing standardized gross motor test items requiring speed. (11-13) In a study of the ambulation status of 98 children with traumatic and anoxic brain injuries at the time of discharge from inpatient rehabilitation, 71 children (72%) were considered "community walkers." (6) Other authors (14) who examined the outcomes of 25 children with TBI reported that 16 of the children were able to walk outdoors independently and that 4 other children could ambulate within their homes at the time of discharge from an inpatient rehabilitation program.

Previous descriptive studies of children with TBI provide a myriad of factors to consider in developing a predictive model for ambulation recovery during inpatient rehabilitation. For example, children who experience a TBI at a later age have been described as regaining more motor skills, including ambulation capability, than younger children. (9) The presence of extracranial injuries, specifically lower-extremity injuries, (15,16) and type of brain injury (6,10,17) have been related to poor mobility (including ambulation) outcomes in children with TBI. Injury severity reports (8,10,18) indicate that the more severe the injury (operationally defined as the amount of time the child is unconscious following injury), the less likely the child with TBI will achieve full recovery of physical function. Hypertonicity, a commonly reported motor impairment following TBI in children, (14,18,19) also has been reported to influence a child's ability to ambulate after injury. (14,19) Diminished cognitive outcomes (level of responsiveness) have been correlated with greater deficits in wheelchair mobility and ambulation at the time of discharge from inpatient rehabilitation. (6) Lastly, as has been found with adults following a stroke, (20,21) the level of physical functioning, as determined by an initial functional assessment score at hospital admission, also may assist with the development of an ambulation prognosis in the pediatric patient with TBI.

Positive and negative prognostic factors for ambulation have been reported in pediatric conditions other than TBI. In children with cerebral palsy (CP), (22-25) the type of CP, (24) the age at which children achieved the ability to sit, (23-25) the the persistence of selected primitive reflexes, (22-24) and a history of epilepsy (23,24) have been described as predictors of ambulatory ability. The presence of hypertonicity in muscles around the knee and hip joints, the number of shunt revisions, and the occurrence of balance disturbances have been negatively correlated with ambulation in children with myelomeningocele, regardless of lesion level. (26)

Information about recovery of ambulation in children with TBI during inpatient rehabilitation is derived from descriptive studies. It is not clear, however, which clinical information should be used routinely by physical therapists to develop a prognosis regarding the ability to ambulate. Based on our synthesis of the literature and on the clinical experience of the rehabilitation team at Franciscan Children's Hospital and Rehabilitation Center (FCH), Boston, Mass, we hypothesized that potential predictors for recovery of ambulation are: (1) demographic factors (eg, age at the time of injury (9)), (2) injury-related factors (eg, extracranial injuries, (15,16) type of brain injury, (10,17) injury severity (8,10,18)), (3) impairments (eg, hypertonicity in the lower extremities (14,18,19) and cognitive status (6)), and (4) functional limitations (eg, physical functioning) at admission to the rehabilitation program. (20,21)

The achievement of ambulation signifies a major milestone in the recovery of physical functioning following TBI in children. It is also a primary goal for discharge from the inpatient rehabilitation setting because it allows children to return home without restrictions in mobility. The aim of our study was to delineate a predictive model and determine the value of the following variables: age at the time of injury, presence of lower-extremity injury, type of brain injury, injury severity, presence of hypertonicity in the lower extremities, cognitive status, and physical functioning at admission to the rehabilitation facility in establishing a physical therapist prognosis for ambulation following inpatient rehabilitation.

Method

Subjects

Children and adolescents discharged from the Inpatient Pediatric Physical Rehabilitation Program at FCH between November 1994 and January 2002 with a primary diagnosis of TBI were eligible for inclusion in this retrospective study. We chose this period to include as many children as possible in the study because we believed it would maximize the validity of our predictive model. Data were excluded if the children were less than 2 years of age when they were admitted to the inpatient rehabilitation program, if they were not independent ambulators prior to the TBI due to a pre-existing condition, or if they experienced a complete spinal cord injury (a nonreversible condition) in addition to TBI. Data also were excluded if the children were capable of ambulation indoors for greater than 15.24 m (50 ft) when they were admitted to the rehabilitation facility or if they had restrictions due to the presence of multiple trauma of impairments in which ambulation was not expected to occur during the time the children were in the inpatient rehabilitation setting. For example, data from children who remained non-weight bearing on their lower extremities throughout their inpatient stay were not included. For children who were transferred to an acute care hospital and returned to FCH to complete their rehabilitation, all inpatient rehabilitation admission-discharge episodes were combined. We had full data on all eligible subjects. Demographic characteristics of the 95 (86%) children and adolescents who were discharged during the study time frame and met the inclusion criteria are shown in Table 1.

Procedure

Following approval by the Institutional Review Board at FCH, retrospective data were collected from hospital medical records and a pre-existing program database of patient demographics and program outcome data. A data collection form (27) was developed by the study's primary investigator (HMD) to collect all information extracted from the medical records (Appendix). Instructions indicating where the information was located in the medical record and definitions of all independent variables were developed prior to data collection and were attached to the data collection form. Medical record reviews were completed by 2 of the authors (HMD, TMC) and entered into a Statistical Package for the Social Sciences (28), * database designed for this study. Random checks were conducted by the primary investigator on a total of 12% of the records reviewed to ensure absolute agreement of data extraction from the medical records and to ensure accuracy of data entry. Five records were reviewed initially, 3 records were reviewed halfway through the data collection period, and an additional 3 records were reviewed near the end of the data collection period. Absolute agreement of values from the medical record was greater than 95% for all records at all 3 checkpoints.

Information regarding ambulation ability, the dependent variable for this study, was collected from each individual child's Physical Therapy Discharge Examination note in the medical record. If discharge ambulatory status was unclear, the attending physiatrist's discharge summary of team conference reports in the child's medical record were reviewed. Ambulation was defined as being able to walk indoors (with or without deviations in gait pattern) with no assistive gait device (eg, cane, walker), support (person or surroundings), of guarding or supervision for balance on level surfaces. Supervision for behavioral problems such as impulsivity or for other potential medical conditions such as seizures was not considered as an inability to ambulate. Orthoses (eg, ankle-foot orthosis) could be used. The child had to be capable of ambulating greater than 15.24 m on a flat, level surface in an indoor setting. Distance determination was based on assessment of ambulation in one of several previously measured hallways throughout the hospital. Capability to ascend or descend stairs was not considered.

The independent variables were dichotomized and coded systematically for the study. For example, we classified the independent variable of lower-extremity hypertonicity as "present" or "not present." Due to the prestudy establishment of variable definitions and development of a method of data extraction, we believe we avoided the need for data interpretation by the chart reviewers. Presence or absence of lower-extremity injury (injury-related variable) included the presence of absence of any unilateral or bilateral lower-extremity fracture of joint subluxation or dislocation. Classifying children as having a diffuse or nondiffuse type of brain injury (injury-related variable) was intended by us to distinguish between noncentralized axonal injuries and subdural hematomas, confusions, of hemorrhages due to trauma, injury severity (injury-related variable), defined as the length of time of unconsciousness following injury, was classified as either more or less than 24 hours. This was done due to inconsistency in the number of hours of unconsciousness after injury documented in the medical records. Presence or absence of lower-extremity hypertonicity at the time of admission to the rehabilitation setting (motor impairment) was based only on the presence of absence of hypertonia in any lower-extremity muscles, as documented by the physical therapist during the admission examination. Classification of cognitive status at the time of admission to the inpatient facility was based on the ability of the child to respond or not to respond to verbal or physical commands (impairment variable) as documented by the physical therapist during the admission examination. In addition, age and physical functioning, based on Pediatric Evaluation of Disability Inventory (29) (PEDI) Functional Skills and Caregiver Assistance Mobility summary scale scores, at the time of admission to the rehabilitation facility were used as continuous variables.

The PEDI is used to measure functional mobility at the time of admission to and discharge from the inpatient rehabilitation facility. The PEDI is a functional assessment that is designed to measure both capability and performance of daily living skills based on general observation of the child. Physical therapists complete the PEDI's Mobility domain at the time each child is admitted to and discharged from the rehabilitation program. Physical therapists are trained in the use of the PEDI by senior physical therapy staff in the inpatient physical rehabilitation program by use of case studies in the PEDI manual (29) and by joint administration of the assessment before independent use. An annual review of the case studies provided for training in the PEDI manual and a review of the scoring criteria with all program and physical therapy staff is designed to contribute to staff competence and reliability of data obtained with the PEDI as an outcome measure.

Data Analysis

Data analysis and interpretation of results were performed using SPSS. (28) Descriptive statistics were generated for demographic characteristics of the study sample. A multiple logistic regression analysis (30) was used to determine the contributions of each of the variables identified as potential predictors of recovery of the ability to ambulate for children and adolescents with TBI.

The analysis was conducted in 2 phases. In the first phase, a bivariate correlational analysis was completed with all 8 potential predictor variables (Phi coefficient for dichotomous variables and Pearson correlation coefficient for interval-level variables) and the dependent variable (ambulation). This was done to account for the large number of potential predictor variables that could be entered into the logistic regression model relative to the sample size and to inspect for multicolinearity between the independent variables. With 95 subjects, we had 80% power to detect a correlation of .25 at an alpha level of .05. (31) In the second phase, one potential independent variable (PEDI Caregiver Assistance score) was removed due to multicolinearity, and the independent variables that reached statistical significance in the first phase ([alpha]=.05) were then entered into a final logistic model.

To facilitate further interpretation of the predictive model of ambulation recovery, the positive and negative predictive values and the predictive utility of the model were calculated. We report the Nagelkerke [R.sup.2], (32) which has an approximate variance interpretation to [R.sup.2] in multiple linear regression. The Nagelkerke [R.sup.2,32] is based on log-likelihoods and is adjusted so that a value of 1.0 can be achieved. (33)

The overall percentage of accuracy in classification, the specificity (classification of true negatives), and the sensitivity (classification of true positives) also were determined. To help interpret the results to the specific elements of physical therapist patient/client management, the odds ratios (ORs) with 95% confidence intervals (CIs) for each of the significant independent variables were calculated. Finally, we organized the results into a prognosis stratification table to help display the clinical value of various patterns of predictors.

Results

Fifty-three children (56%) met the criteria to be considered as ambulating at the time of discharge from the inpatient rehabilitation program. Results of the bivariate correlational analysis are presented in Table 2. Based on these results, we eliminated 2 of the 8 potential predictors (type of brain injury and age at time of admission) from entry into the regression solution. These variables were eliminated because they had near zero correlations with the dependent variable and were therefore not expected to contribute to the prediction model. The results of the logistic regression analysis for the predictor model are depicted in Table 3. Nagelkerke [R.sup.2] for the overall prediction model was 44%, indicating that the presence of hypertonicity in the lower extremities, lower-extremity injury, and more severe injury contributed to 44% of the variance in achieving ambulation. The absence of lower-extremity hypertonicity (P<.01), low injury severity (P=.02), and the absence of lower-extremity injury (P=.01) were all predictors of ambulation capability at the time of discharge from the inpatient rehabilitation program.

The absence of lower-extremity hypertonicity was the strongest predictor of recovery of the ability to ambulate. The OR for the absence of lower-extremity hypertonicity indicates that those children who achieved ambulation at the time of discharge were 8.9 times (95% CI=2.17-36.30) more likely to have had a history of no lower-extremity hypertonicity when admitted to the inpatient rehabilitation facility. In addition, absence of lower-extremity injury (OR=5.27, 95% CI=1.65-16.82) and low injury severity (ie, loss of consciousness for less than 24 hours) (OR=4.77, 95% CI=1.27-17.88) contributed to the overall prediction solution.

Sensitivity (proportion of children who ambulated and who were predicted to do so) was 87%, but the specificity (proportion of children who did not ambulate and who were not predicted to do so) was 67%. Fourteen children were predicted to walk and did not (false positives), while 7 children were not predicted to walk and did (false negatives). Positive (77%) and negative (80%) predictive values were found to be nearly identical (Tab. 4). Table 5 provides a prognostic stratification created by probability calculations for each combination of dependent variables left in the regression model to predict ambulation.

Discussion

Prediction Model

We developed a model for predicting the recovery of ambulation for children and adolescents with TBI at the time of discharge from an inpatient rehabilitation facility that includes one impairment variable (lower-extremity hypertonicity) and 2 injury-related variables (injury severity and lower-extremity injury). The absence of lower-extremity hypertonicity was by far the strongest predictor of recovery of the ability to ambulate for children admitted to the inpatient rehabilitation facility following TBI. Hypertonicity is evident in approximately 65% of children with severe brain injuries, (18) and the presence of hypertonicity, has previously been reported to be related to the inability to ambulate following inpatient rehabilitation for children with TBI. (14) In our sample, children who achieved ambulation at the time of discharge were at least 8 times more likely to have had a history of no lower-extremity hypertonicity on admission to the inpatient rehabilitation facility. Due to the sample size, however, the 95% CIs around each OR were large. Further validation of this finding in other samples or in a larger series would provide additional confidence in the strength of this predictor.

Additional variables beyond the absence of lower-extremity hypertonicity were important to predict ambulation at the time of discharge. Injury severity, measured as whether the child was unconscious for more than 24 hours immediately after injury, was related to the prognosis of whether children could ambulate as defined. Injury severity has been shown to be a predictor for survival and morbidity. The more severe the injury, the greater the cognitive and behavioral deficits may be, thus potentially affecting ambulation ability. (17)

The absence of lower-extremity injury was shown in our study to be an important predictor for recovery of the ability to ambulate. In our sample, assistive devices were needed for ambulation due to lower-extremity weight-bearing restrictions, balance, and energy limitation for long-distance (>15.24 m) ambulation. Fourteen children used an assistive device such as a cane, a walker, or crutches for home ambulation at the time of discharge from the inpatient rehabilitation setting. Although they may have been able to ambulate with an assistive device and without assistance of others, children using an assistive device were not considered ambulatory for the purposes of our study. The experience at FCH has been that children with TBI are almost always independent ambulators without an assistive device before the TBI, and the intent of this prediction model was to examine the return to this level of ambulation ability.

Following the first phase of our analysis, we removed functional mobility independence (PEDI Caregiver Assistance Mobility scale score) as a potential predictor variable because of the relationship with the PEDI Functional Skills Mobility scale. Several other variables with correlations to the dependent variable of ambulation, however, were not significant in the logistic regression model. We were surprised by the lack of relationship between ambulation at the time of discharge and initial functional mobility status as measured by the PEDI. The initial correlation between ambulation and the PEDI Functional Skills Mobility scale score was .37 (P<.001). Presence of hypertonicity and PEDI scores were related (Functional Skills Mobility scale=.59, P<.001), and therefore the PEDI does not show up in the predictor model. We were also surprised that age was not a factor for children in our study. Age has been described as a predictor of recovery of cognitive, behavioral, and social functioning following TBI in children. (9,17) We may have attained these findings, in part, due to the large age range of our sample and, in part, due to the inclusion of only those children older than 2 years at the time of injury, thus excluding any child who did not ambulate prior to the injury.

Accuracy of Prediction Model

Based on the values for sensitivity and specificity we found, we believe the prediction equation we generated will be more useful in identifying children who are expected to ambulate at the time of discharge than those who are not expected to ambulate at the time of discharge. Sensitivity, the proportion of children who ambulated at the time of discharge and who were predicted to ambulate, was 87%. This sensitivity value indicates that the model is very good at identifying which children are likely to ambulate because there were only 7 false negatives. Specificity, the proportion of children who did not ambulate at discharge and were not predicted to ambulate, was only 67%, with 14 false positives. We therefore predicted a greater likelihood of ambulation in 14 children, and we predicted a lesser likelihood of ambulation in 7 children. Approximately half of these children for whom we had incorrect predictions were children who ambulated using an assistive device. Although it might seem that being surprised at the recovery of ambulation would be better than providing false hope for recovery of unassisted, short-distance ambulation at the time of hospital discharge, both errors potentially misguide the use of physical therapy resources, may cause anguish or disappointment for a child and family, and may bring into question the physical therapist's plan of care with the child, family, and rehabilitation team.

The overall accuracy of the predictions made was 78%, indicating that, for approximately 1 out of every 4 children, outcomes were not correctly predicted by use of the overall model. The values for positive and negative predictive value are intended to be useful when determining probability of ambulation or nonambulation for individual clients. The nearly similar positive (77%) and negative (80%) prediction values we achieved suggest to us that the model is just as good at predicting ambulation for an individual child as it is at predicting nonambulation for an individual child. Positive and negative predictive values, however, are dependent on the prevalence of ambulation at the time of discharge in a given setting. The prevalence of ambulation at the time of discharge in our setting was 56%. In other samples, higher or lower prevalence rates may affect the positive predictive value of the prediction model. (34) A formula to calculate new values for positive predictive value, based on the prevalence of unassisted ambulation for 15.24 m at the time of inpatient discharge in other settings, is available. (35)

The probability of ambulation for children who had no lower-extremity hypertonicity and no lower-extremity injury and who were unconscious for less than 24 hours was quite high (92%). Although the probability of recovery of ambulation for the children who had no lower-extremity hypertonicity or injury but were unconscious for greater than 24 hours was even higher at 100%, there were only 4 children with this profile and we are less confident with this profile. Probability models are intended to help physical therapists establish realistic goals with the children and families and design appropriate management plans. These models also may be a useful tool in future research for classification of children's ability to ambulate in more homogeneous groups.

Because the presence of hypertonicity was such a dominant predictor of ambulation, we believe that improvements in the prediction model may be achieved through more sensitive measures of hypertonicity. In this study, we extracted just the presence or absence of lower-extremity hypertonicity from each child's medical record. Changing the variable to the presence of bilateral or unilateral extremity hypertonicity, scores from a scale to measure abnormalities in muscle tone such as the Modified Ashworth scale, (36) of documentation of hypertonicity for individual lower-extremity muscle groups could be used. In addition, we used loss of consciousness for more than 24 hours at the time of injury as a measure of injury severity, another predictive variable. More commonly, Glasgow Coma Scale (GCS) scores are used to measure injury severity (37-41) however, consistent GCS scores were not available in the medical records used for our retrospective study. Lower-extremity injuries also could be further defined or classified as bilateral injuries or leg injuries versus hip and knee injuries to strengthen the predictive utility of the model.

Many of the children for whom there were false negatives (children predicted to ambulate but did not) used an assistive device for ambulation at the time of discharge. In addition, some of these children stayed in the rehabilitation facility less than the mean length of time for the remainder of the sample. These children may have been discharged when they achieved safe indoor ambulation with an assistive device, but they did not remain in the hospital long enough to attain ambulation without an assistive device. Inpatient rehabilitation length of stay was not entered into the multiple regression analysis as a predictor variable. We believe, however, that it warrants further consideration as a covariate in the models of recovery of ambulation ability following TBI in children and adolescents, as does ambulation with an assistive device.

Limited evidence is available to guide prognosis in pediatric practice. In a report on ambulation predictors for children with cerebral palsy, Montgomery (42) suggested that operationally defined levels of "meaningful, functional ambulation" are needed to further understand predictive variables. In our study, we identified a predictive model for ambulation, which we believe is meaningful to the children and their families and physical therapists practicing in inpatient pediatric rehabilitation settings. We believe that the ability to predict a return to this level of ambulatory ability is a critical factor in rehabilitation because it influences decision making about the plan of care and the establishment of appropriate short-term and long-term discharge goals.

We found that 1 of the body system variables (the absence of lower-extremity hypertonicity) and 2 injury-related variables (injury severity [ie, loss of consciousness for less than 24 hours after injury] and absence of lower-extremity injury) together were predictors of recovery of the ability to ambulate. Theoretically, these predictors not only may assist with early determination of whether ambulation will be achieved by children and adolescents with TBI at the time of discharge from an inpatient rehabilitation facility but may influence the type and frequency of physical therapy intervention.

As is common for inpatient pediatric rehabilitation programs, (43,44) FCH accepts children with moderate to severe TBIs and with residual functional deficits that prevent them from returning home following hospitalization in an acute care setting. We have demonstrated that more than half of the children are capable of ambulation at the time of discharge from the inpatient rehabilitation program. These results are consistent with those of previous reports about children with TBI who were discharged from inpatient rehabilitation programs. (6,14) According to some authors (6,7,14,35,45,46) and in our personal experience with the post-discharge follow-up clinic at FCH, children continue to demonstrate improvements in motor abilities following discharge from inpatient rehabilitation facilities.

Recommendations for Prognostic Studies

We used available retrospective data from one program in our study. The data were dichotomized and required secondary coding. This line of research could be enhanced with prospective multicenter studies in which multiple measurement tools using ordinal- or interval-level scores that are more reliable and show greater validity are used. In addition, including intervention data in predictive models could further our understanding of the role of intervention type and intensity on recovery.

We studied only those children who were older than 2 years at the time of admission to the inpatient rehabilitation facility, who did not have a complete spinal cord injury (ie, a nonreversible condition), and whose restrictions due to the presence of multiple trauma or identified impairments were such that ambulation could still be expected to occur during inpatient rehabilitation. The results of our study, though worthy of note to physical therapist practice in any setting, may have limited generalizability beyond inpatient rehabilitation.

We believe that meaningful outcome measures that could be used to examine both capability and performance, in the home and in the community rather than in a hospital setting, warrant further study. The results of our study, however, can add to discussions of how physical therapists can use predictive models and individual predictors when choosing tests and measures to determine physical therapist diagnoses and prognoses.

Conclusion

The absence of lower-extremity hypertonicity was the most important predictor of the recovery of ambulation ability following TBI in children and adolescents. Injury severity (loss of consciousness for less than 24 hours) and absence of lower-extremity injury also were factors in helping to predict if a child will ambulate at the time of discharge from an inpatient rehabilitation facility. Awareness of clinical predictors of recovery of ambulation ability that are gathered as part of a physical therapist's examination, we contend, can allow for a determination of ambulation prognosis and thus could enable physical therapists to establish appropriate plans of care.

Appendix. Predicting Ambulation Data Collection Sheet (a)

Medical Record #: --

Child Name (Last, First): --

Admission Date: --
Discharge Date: --
Date of Birth: --
Date of Injury: --

Mechanism of Injury:         Passenger/Driver in an MVA
                             Passenger/Bike Rider in an MVA
                             Fall
                             Other

Sex:                         Male   Female

Discharge Disposition:       Home   Other

Injury Severity:             Coma (>24 hours unconscious)   No Coma
Ambulation (at discharge)?   Yes        No
  Use of Assistive Device?   Yes        No
    If yes, what type?       Cane, Walker, Other --
  Use of LE Orthoses?        Yes        No
    If yes, what type?       SMO, AFO, KAFO, HKAFO, Other --
  Supervision Required?      Yes        No
  Distance (Indoors)         <15.24 m   >15.24 m
  Lower-Extremity Injury?    Yes        No
    If yes, type: --
  Type of Brain Injury?      Diffuse    Diffuse plus other
  Cognitive Status at Time of Admission (responds to verbal or
    physical command)?   Yes   No
  Motor Impairment (LE hypertonicity present at time of
    admission?)   Yes   No
  Functional Mobility Capability? (PEDI Functional Skills
    Mobility scale score at time of admission)--

(a) MVA=motor vehicle accident, LE=lower extremity,
SMO=supramalleolar orthosis, AFO=ankle-foot orthosis,
KAFO=knee-ankle-foot orthosis, HKAFO=hip-knee-ankle-
foot orthosis, PEDI=Pediatric Evaluation of Disability
Inventory.

Table 1.
Demographic Characteristics of Study Sample (N=95)

                                                       No. of
Variable                                               Subjects (a)

Sex
  Male                                                 63 (66%)
  Female                                               32 (34%)
Ethnicity
  White                                                61 (64%)
  Asian/Pacific Islander                                3 (3%)
  Black                                                 9 (10%)
  Hispanic                                             19 (20%)
  Other                                                 3 (3%)
Cause of traumatic brain injury
  Pedestrian/bicycle rider struck by a motor vehicle   45 (47%)
  Passenger/driver in a motor vehicle accident         37 (39%)
  Fall                                                  7 (7%)
  Other                                                 6 (6%)
Discharge to the home                                  80 (84%)
Days from injury to admission to rehabilitation
    facility
  [bar.X]                                              30.2
  SD                                                   56.7
  Range                                                 4-407
Age at time of admission to inpatient rehabilitation
    facility (y)
  [bar.X]                                              10.5
  SD                                                    4.9
  Range                                                 2.1-18.9
Length of inpatient rehabilitation stay (d)
  [bar.X]                                              73.1
  SD                                                   91.3
  Range                                                 3-664

(a) Values shown are number of subjects, with
percentage in parentheses, unless otherwise stated.

Table 2.
Correlation (Pearson Product Moment) Matrix
Among the Independent and Dependent Variables (a)

                                         Absence of      Injury
                                         Lower-          Severity
                                         Extremity       (Coma
Variable                    Ambulation   Hypertonicity   <24 hr)

Ambulation                               .49 **          .40 **
Lower-extremity
  hypertonicity                                          .44 **
Injury severity
Lower-extremity injury
Type of brain injury
Cognitive status at time
  of admission
Age at time of admission
PEDI Functional Skills
  Mobility scale score at
  time of admission
PEDI Caregiver Assistance
  Mobility scale score
  at time of admission

                                                   Follows
                                                   Commands
                            Absence of   Diffuse   at Time of
                            Lower-       Type of   Admission
                            Extremity    Brain     (Cognitive
Variable                    Injury       Injury    Status)

Ambulation                   .25 *       -.08       .33 **
Lower-extremity
  hypertonicity             -.01         -.13       .56 **
Injury severity             -.04         -.02       .58 **
Lower-extremity injury                   -.07       .02
Type of brain injury                               -.14
Cognitive status at time
  of admission
Age at time of admission
PEDI Functional Skills
  Mobility scale score at
  time of admission
PEDI Caregiver Assistance
  Mobility scale score
  at time of admission

                                                          PEDI Care-
                                        PEDI (b)          giver
                                        Functional        Assistance
                                        Skills Mobility   Mobility
                            Age at      Scale Score at    Scale Score
                            Time of     Time of           at Time
Variable                    Admission   Admission         of Admission

Ambulation                  -.18         .37 **            .35 **
Lower-extremity             -.11         .59 **            .61 **
  hypertonicity
Injury severity             -.32 *       .38 **            .39 **
Lower-extremity injury      -.03         .23 *             .20
Type of brain injury         .11        -.34 **           -.35 **
Cognitive status at time
  of admission              -.19         .47 **            .50 **
Age at time of admission                -.10              -.05
PEDI Functional Skills
  Mobility scale score at
  time of admission                                        .91 **
PEDI Caregiver Assistance
  Mobility scale score
  at time of admission

(a) Asterisk indicates P [less than or equal to] .03, df=93;
double asterisk indicates P [less than or equal to] .001,
df=93 (significance assessed using Pearson product moment
correlation coefficient).

(b) PEDI=Pediatric Evaluation of Disability Inventory.

Table 3.
Results of Logistic Regression Analysis Examining
the Effect of Predictor Variables on Ambulation (a)

                                         Regression
Variable                                 Coefficient (b)   SE (b)

Lower-extremity hypertonicity            2.15              0.71
Lower-extremity injury                   1.66              0.59
Injury severity                          1.57              0.68
Cognitive status at time of admission   -0.30              0.74
PEDI Functional Skills Mobility scale    0.00              0.02
  score

                                        Odds Ratio   95% CI for
Variable                                (Exp B)      Odds Ratio   P

Lower-extremity hypertonicity           8.61         2.14-34.61   <.01
Lower-extremity injury                  5.26         1.65-16.79    .02
Injury severity                         4.79         1.28-17.98   <.01
Cognitive status at time of admission   0.74         0.17-3.15     .68
PEDI Functional Skills Mobility scale   1.00         0.97-1.03     .87
  score

(a) Only variables included in the logistic analysis
are listed. SE=standard error, CI=confidence interval,
PEDI=Pedriatric Evaluation of Disability Inventory.

Table 4.
Classification Outcome Table of Observed
Versus Predicted Results for Ambulation

                             Observed

                             Ambulation   Not Ambulation

Predicted   Ambulation       46           14
            Not Ambulation    7           28

Percentage of ambulators: 55/95=56%,
Overall accuracy of prediction=78% (with default value of 0.5)
Sensitivity=46/53=87%
Specificity-28/42=67%
Positive predictive value=46/60=77%
Negative predictive value-28/35-80%

Table 5.
Probability of Regaining Ambulation

No. of Children With
These Characteristics   Lower-Extremity   Injury Severity
at Admission            Hypertonicity     (Coma >24 hr)

26                      No                No
 4                      No                Yes
20                      No                No
 3                      No                Yes
10                      Yes               No
 9                      Yes               No
14                      Yes               Yes
 9                      Yes               Yes

No. of Children With
These Characteristics   Lower-Extremity
at Admission            Injury            P

26                      No                 .92
 4                      No                1.00
20                      Yes                .60
 3                      Yes                .33
10                      No                 .60
 9                      Yes                .33
14                      No                 .14
 9                      Yes                .11

* SPSS Inc, 233 S. Wacker Dr, Chicago, IL 60606.

References

(1) Michaud LJ, Duhaime AC, Batshaw ML. Traumatic brain injury in children. Pediatr Clin North Am. 1993;40:553-565.

(2) Molnar GE, Easton JKM, Badell A, et al. Pediatric rehabilitation, 2: brain damage causing disability. Arch Phys Med Rehabil. 1989; 70(suppl 5):S166-S169.

(3) Guide in Physical Therapist Practice. 2nd ed. Phys Ther. 2001;81:9-746.

(4) Kerkering GA, Phillips WE. Brain injuries: traumatic brain injuries, near-drowning, and brain tumors. In: Campbell SK, Vander Linden DW, Palisano RJ, eds. Physical Therapy for Children. 2nd ed. Philadelphia, Pa: WB Saunders Co; 2000:597-620.

(5) 2000 Medical Rehabilitation Standards Manual. Tucson. Ariz: CARF--The Rehabilitation Accreditation Commission; 2000.

(6) Vander Schaaf P, Kriel R, Krach L, et al. Late improvements in mobility after acquired brain injuries in children. Pediatr Neurol. 1997;16:306-310.

(7) Boyer MG, Edwards P. Outcome 1 to 3 years after severe traumatic brain injury in children and adolescents. Injury. 1991;22:315-320.

(8) Kriel RL, Krach LE, Sheehan M. Pediatric closed head injury: outcome following prolonged unconsciousness. Arch Phys Med Rehabil. 1988;69:678-681.

(9) Kriel RL, Krach LE, Panser L. Closed bead injury: comparison of children younger and older than 6 years of age. Pediatr Neurol. 1989;5:296-300.

(10) Kviel RL, Krach LE, Jones-Saete C. Outcome of children with prolonged unconseiousncss and vegetative states. Pediatr Neurol. 1993; 9:362-368.

(11) Jaffe KM, Polissar NL, Fay GC, et al. Recovery trends over three years following pediatric traumatic brain injury. Arch Phys Med Rehabil. 1995;76:17-26.

(12) Jaffe KM, Fay GC, Polissar NL, et al. Severity of pediatric traumatic brain injury and early neurobehavioral outcome: a cohort study. Arch Phys Med Rehabil. 1992;73:540-547.

(13) Chaplin D, Deitz J, Jaffe K. Motor performances in children after traumatic brain injury. Arch Phys Med Rehabil. 1993;74:161-164.

(14) Emanuelson I, von Wendt L, Lundaly E, et al. Rehabilitation and follow-up of children with severe traumatic brain injury. Childs Nerv Syst. 1996; 12:460-465.

(15) Greenspan AI, MacKeozie EJ. Functional outcome after pediatric head injury. Pediatrics. 1994;94:425-432.

(16) Wesson DE, Williams JI, Spence LJ, et al. Functional outcome in pediatric trauma. J Trauma. 1989;29:589-592.

(17) Michaud LJ, Rivara FP, Grady MS, et al. Predictors of survival and severity of disability after severe brain injury in children. Neurosurgery. 1992;31:254-264.

(18) Brink JD, Imbus C, Woo-Sam J. Physical recovery after severe closed head trauma in children and adolescents. J Pediatr. 1980;97:721-727.

(19) Costeff H, Grosswasser Z, Goldstein R. Long-term follow-up review of 31 children with severe closed head trauma. J Neurosurg. 1990;73:684-687.

(20) Korner-Bitensky N, Mayo N, Cabot R, et al. Motor and functional recovery after stroke: accuracy of physical therapist predictions. Arch Phys Med Rehabil. 1989;70:95-99.

(21) Sanchez-Blanco I, Ochoa-Sangrador C, Lopez-Munain L, et al. Predictive model of functional independence in stroke patients admitted to a rehabilitation programme. Clin Rehabil. 1999;13:464-475.

(22) Bleck EE. Locomotor prognosis in cerebral palsy. Dev Med Child Neurol. 1975;17:18-25.

(23) Watt JM, Robertson CMT, Grace MGA. Early progmosis for ambulation of neonatal intensive care survivors with cerebral palsy. Dev Med Child Neurol. 1989;31:766-773.

(24) Trahan J, Marcoux S. Factors associated with the inability of children with cerebral palsy to walk at six years: a retrospective study. Dev Med Child Neurol 1994;36:787-795.

(25) Campos de Paz A Jr, Burnett SM, Braga LW. Walking prognosis in cerebral palsy: a 22-year retrospective analysis. Dev Med Child Neurol. 1994;36:130-134.

(26) Bartonek A, Saraste H. Factors influencing ambulation in myelomeningocele: a cross-sectional study. Dev Med Child Neurol. 2001;43: 253-260.

(27) Labelle J, Swaine BR. Reliability associated with the abstraction of data from medical records for inclusion in an information system for persons with a traumatic brain injury. Brain Inj. 2002;16:713-727

(28) SPSS 9.0. Chicago, Ill: SPSS Inc; 1999.

(29) Haley SM, Coster WJ, Ludlow LH, et al. Pediatric Evaluation of Disability Inventory: Development, Standardization and Administration Manual. Boston, Mass: Trustees of Boston University; 1992.

(30) Hosmer DW, Lemeshow S, Applied Logistic Regression. New York, NY: John Wiley & Sons Inc; 2000.

(31) Kraemer HC, Thiemann S. How Many Subjects? Statistical Power Analysis in Research. Newbury Park, Calif: Sage Publications Inc; 1987.

(32) Tabachnick BG, Fidell LS. Using Multivariate Statistics. 4th ed. Boston, Mass: Allyn & Bacon; 2001.

(33) Nagelkerke NJD. A note on a general definition of the coefficient of determination. Biometrika. 1991;78:691-692.

(34) DiScala C, Osberg JS, Savage RC. Children hospitalized for traumatic brain injury: transition to postacute cree. J Head Trauma Rehabil. 1997;12:1-10.

(35) Fletcher RH, Fletcher SW, Wagner EH. Clinical Epidemiology: The Essentials. 3rd ed. Baltimore, Md: Williams & Wilkins; 1996.

(36) Bohannon RW, Smith MB. Interrater reliability of a modified Ashworth scale of muscle spasticity. Phys Ther. 1987;67:206-207.

(37) Massagli TL, Michaud LJ, Rivara FP. Association between injury indices and outcome after severe traumatic brain injury in children. Arch Phys Med Rehabil. 1996;77:125-132.

(38) Pillai S, Praharaj SS, Mohanty A, et al. Prognostic factors in children with severe diffuse brain injuries: study of 74 patients. Pediatr Neurosurg. 2001;34:98-103.

(39) White JR, White MD, Farukhi Z, et al. Predictors of outcome in severely head-injured children. Crit Care Med. 2001;29:534-540.

(40) Ong L, Selladurai BM, Dhillon MK, et al. The prognostic value of the Glasgow Coma Scale, hypoxia and computerized tomography in outcome prcdiction of pediatric head injury. Pediatr Neurosurg. 1996; 24:285-291.

(41) Chiaretti A, Piastra M, Pulitano S, et al. Prognostic factors and outcome of children with severe head injury: an 8-year experience. Childs Nerv Syst. 2002;18:129-136.

(42) Montgomery PC. Predicting potential for ambulation in children with cerebral palsy. Pediatric Physical Therapy. 1998;10:148-155.

(43) Osberg JS, Unsworth CA. Trauma-rehabilitation connections: discharge and admission decisions for children. Pediatr Rehabil. 1997;1: 131-146.

(44) Osberg JS, DiScala C, Gans BM. Utilization of inpatient rehabilitation services among traumatically injured children discharged from pediatric trauma centers. Am J Phys Med Rehabil. 1990;69:67-72.

(45) Eiben CF, Anderson TP, Lockman L, et al. Functional outcome of closed head injury in children and young adults. Arch Phys Med Rehabil. 1984;65:168-170.

(46) Greenspan AI. Functional recovery following head injury among children. Curr Probl Pediatr. 1996;26;170-177.

HM Dumas, PT, MS, is Manager, Research Center for Children with Special Health Care Needs, Franciscan Children's Hospital and Rehabilitation Center, 30 Warren St, Boston, MA 02135 (USA) ([e-mail]). Address all correspondence to Ms Dumas.

SM Haley, PT, PhD, is Director, Center for Rehabilitation Effectiveness, and Associate Professor of Physical Therapy, Sargent Conege of Health and Rehabilitation Sciences, Boston University, Boston, Mass.

LH Ludlow, PhD, is Professor and Chair, Department of Educational Research, Measurement, and Evaluation, Lynch School of Education, Boston College, Boston, Mass.

TM Carey, PT, MSPT, was a Research Assistant, Research Center for Children With Special Health Care Needs, Franciscan Children's Hospital and Rehabilitation Center, at the time of the study.

Ms Dumas, Dr Haley, and Dr Ludlow provided concept/idea/research design, writing, and data analysis. Ms Dumas and Ms Carey provided data collection. Ms Dumas provided project management, fund procurement, subjects and facilities/equipment. All authors provided consultation (including review of manuscript before submission).

This study was approved by the Institutional Review Board at Franciscan Children's Hospital and Rehabilitation Center.

Funding for this project was provided by a Clinical Research Grant from the Section on Pediatrics, American Physical Therapy Association.

This article was received May 15, 2003, and was accepted September 30, 2003.