Factors That Predict Acute Hospitalization Discharge Disposition for Adults With Moderate to Severe Traumatic Brain Injury

721

ORIGINAL ARTICLE

Factors That Predict Acute Hospitalization Discharge Disposition for Adults With Moderate to Severe Traumatic Brain Injury Jeffrey P. Cuthbert, MPH, MS, John D. Corrigan, PhD, Cynthia Harrison-Felix, PhD, Victor Coronado, MD, MPH, Marcel P. Dijkers, PhD, Allen W. Heinemann, PhD, ABPP, Gale G. Whiteneck, PhD ABSTRACT. Cuthbert JP, Corrigan JD, Harrison-Felix C, Coronado V, Dijkers MP, Heinemann AW, Whiteneck GG. Factors that predict acute hospitalization discharge disposition for adults with moderate to severe traumatic brain injury. Arch Phys Med Rehabil 2011;92:721-30. Objective: To identify factors predicting acute hospital discharge disposition after moderate to severe traumatic brain injury (TBI). Design: Secondary analysis of existing datasets. Setting: Acute care hospitals. Participants: Adults hospitalized with moderate to severe TBI included in 3 large sets of archival data: (1) Centers for Disease Control and Prevention Central Nervous System Injury Surveillance database (n⫽15,646); (2) the National Trauma Data Bank (n⫽52,012); and (3) the National Study on the Costs and Outcomes of Trauma (n⫽1286). Interventions: None. Main Outcome Measure: Discharge disposition from acute hospitalization to 1 of 3 postacute settings: (1) home, (2) inpatient rehabilitation, or (3) subacute settings, including nursing homes and similar facilities. Results: The Glasgow Coma Scale (GCS) score and length of acute hospital length of stay (LOS) accounted for 35% to 44% of the variance in discharges to home versus not home,

while age and sex added from 5% to 8%, and race/ethnicity and hospitalization payment source added another 2% to 5%. When predicting discharge to rehabilitation versus subacute care for those not going home, GCS and LOS accounted for 2% to 4% of the variance, while age and sex added 7% to 31%, and race/ethnicity and payment source added 4% to 5%. Across the datasets, longer LOS, older age, and white race increased the likelihood of not being discharged home; the most consistent predictor of discharge to rehabilitation was younger age. Conclusions: The decision to discharge to home a person with moderate to severe TBI appears to be based primarily on severity-related factors. In contrast, the decision to discharge to rehabilitation rather than to subacute care appears to reflect sociobiologic and socioeconomic factors; however, generalizability of these results is limited by the restricted range of potentially important variables available for analysis. Key Words: Brain injuries; Healthcare disparities; Hospitalization; Nursing homes, Patient discharge; Rehabilitation; Rehabilitation centers. © 2011 by the American Congress of Rehabilitation Medicine RAUMATIC BRAIN INJURY is one of the leading causes T of disability in the United States, and it is a contributing factor in approximately one-third of all injury-related deaths.

1

From the Research Department, Craig Hospital, Englewood, CO (Cuthbert, Harrison-Felix, Whiteneck); Department of Physical Medicine and Rehabilitation, Ohio State University, Columbus, OH (Corrigan); National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, GA (Coronado); Department of Rehabilitation Medicine, Mt. Sinai School of Medicine, New York, NY (Dijkers); Department of Physical Medicine and Rehabilitation, Feinberg School of Medicine, Northwestern University, Chicago, IL (Heinemann); and the Rehabilitation Institute of Chicago, Chicago, IL (Heinemann). Supported by a supplemental grant to the Traumatic Brain Injury Model Systems National Data and Statistical Center from the National Institute on Disability and Rehabilitation Research, Office of Special Education and Rehabilitative Services, U.S. Department of Education (grant no. H133A060038); and Traumatic Brain Injury Model System Centers grants to Craig Hospital (grant no. H133A070022), Ohio State University (grant no. H133A070029), Mount Sinai Medical Center (grant no. H133A070033), and the Rehabilitation Institute of Chicago (grant no. H133A080045). However, the contents do not necessarily represent the policy of the Department of Education, and the reader should not assume endorsement by the Federal Government. This article does not reflect the official policy or opinions of the Centers for Disease Control and Prevention (CDC) or the U.S. Department of Health and Human Services (HHS) and does not constitute an endorsement of the authors or their programs— by CDC, HHS, or other components of the federal government—and none should be inferred. All interpretations of the data are the responsibility of the current authors only. The content reproduced from the National Trauma Data Bank remains the full and exclusive copyrighted property of the American College of Surgeons. The American College of Surgeons is not responsible for any claims arising from works based on the original data, text, tables or figures. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit on the authors or on any organization with which the authors are associated. Reprint requests to Jeffrey P. Cuthbert, MPH, MS, 3425 S Clarkson St, Englewood CO 80113, e-mail: [email protected]. 0003-9993/11/9205-00715$36.00/0 doi:10.1016/j.apmr.2010.12.023

Each year at least 1.7 million Americans incur a TBI, and of these injuries, 275,000 are severe enough to require hospitalization.1 Recently, hospitalizations related to TBI have risen sharply, with an increase of 19.5% from 2002 to 2006.1 Health care providers, patients, and families must decide, on completion of acute medical care, which postacute care setting will maximize outcomes and minimize morbidity and mortality.2 Potential settings include the patient’s home (with or without outpatient or home health services), inpatient rehabilitation, and skilled or extended nursing care facilities, with the level and type

List of Abbreviations AIS CDC CNSIS GCS ICD-9-CM IRF LOS NSCOT NTDB NTDS SCI TBI

Abbreviated Injury Scale Centers for Disease Control and Prevention Central Nervous System Injury Surveillance Glasgow Coma Scale International Classification of Diseases, 9th Revision, Clinical Modification inpatient rehabilitation facility length of stay National Study on the Costs and Outcomes of Trauma National Trauma Data Bank National Trauma Data Standard spinal cord injury traumatic brain injury

Arch Phys Med Rehabil Vol 92, May 2011

722

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert

of care varying across settings. In making the decision, acute care professionals must assess the severity of injury, the degree of recovery, ability to function independently in daily tasks with or without family support, and ability to actively participate in rehabilitation. Other factors also must be weighed, including familial and social supports, availability of funding, and the postacute discharge options available within the community. Research examining the factors influencing postacute care discharge options for persons with TBI have suggested that decisions may be based on factors other than clinical criteria.2-5 While some investigations have suggested that discharge disposition is mostly related to indicators of premorbid functioning, overall injury severity and recovery, other studies have indicated possible disparities, with discharge destination related to demographic, biologic, and socioeconomic factors.3 Numerous studies have noted age as a significant predictor of mortality, morbidity, and recovery after TBI, with older persons at risk for less desirable outcomes.6-18 Older adults differ in their sex distribution (higher proportion of women to men) and insurance coverage (higher rates of public funded insurance) from younger adults.19 The older group has greater numbers of comorbidities at injury, often requiring longer and more complex medical treatment.8,20 These differences suggest that age may be a significant factor in determining discharge location. In a retrospective study of 1059 persons with TBI discharged from acute care between 1996 and 1997, Mellick et al5 found that persons older than 65 were more likely to be discharged to long-term care facilities; however, the role of age may have been confounded by insurance type. A populationbased study of persons with TBI ages 65 years or older in 15 states revealed that 37% were discharged home with no or unskilled assistance.8 This percentage, however, decreased substantially with age. In contrast, the proportions of persons discharged to home with home health or outpatient rehabilitation services tended to increase with age. The proportion of those discharged to IRFs decreased with age. This study, however, did not take into account injury severity or insurance payer.8 Leblanc et al6 reported that older persons with moderate to severe TBI had higher rates of discharge to long-term care and lower rates of discharge to inpatient and outpatient rehabilitation, compared to young- and middle-aged groups matched on severity. In contrast, Mosenthal et al12 found that people over age 60 with mild TBI were more likely to be discharged to inpatient rehabilitation than younger persons. Recently published reviews suggest disparities by race and ethnicity after TBI, including differences in medical, functional, social, psychosocial, and reintegration outcomes3-4; however, similar effects are not as clear in regards to the effect of race and ethnicity on discharge disposition after acute medical care. A study of 344 patients (114 persons from minority backgrounds, 230 non-Hispanic whites) with severe TBI discharged from an urban trauma center found no ethnic differences in discharge setting; rather, differences in discharge disposition were related to insurance type, which is associated with age.21 A similar result was found in a study examining the effect of insurance type on discharge location for 5550 adults ages 18 to 65, with Medicaid-insured persons less likely to be discharged to inpatient rehabilitation, after accounting for other patient characteristics including age, sex, and race.22 Malec et al2 conducted a prospective study of 230 adults with moderate to severe TBI and found no differences between ethnic groups in regards to discharge setting; however, age was a significant predictor of discharge disposition. The Mellick study5 found that TBI severity was the strongest predictor of discharge disposition; however, the study also found that minorities, as compared to nonminorities, were more likely to be discharged Arch Phys Med Rehabil Vol 92, May 2011

home than to any other setting and once home, less likely to receive outpatient services. The hypothesis motivating the current study was that injury severity was the predominant factor contributing to discharge from acute care for persons with moderate or severe TBI, and that persons discharged home were likely to have less severe injuries. If this hypothesis held true, the opportunity arises to determine what, if any, factors contribute to a specific nonhome discharge. Thus, the purpose of the current study was: (1) to determine if injury severity-related variables are the primary predictors of acute care discharge to home for persons with moderate and severe TBI and (2) to determine if, after controlling for injury severity, differences related to sociobiologic or socioeconomic factors exist for both home and nonhome discharge dispositions. METHODS Three sets of archival data were selected to address the study questions: (1) CDC CNSIS23; (2) the NTDB24; and (3) the NSCOT.25 These datasets contain adequate data to: (1) identify hospitalized patients with moderate to severe TBI, and (2) categorize hospital discharge disposition. Because no national dataset of acute care for persons with TBI exists, these databases were selected because they are the only datasets that met the study criteria and were appropriate to address the study purpose. Each database was analyzed separately and results were compared across databases. If results determined across databases were found to converge, confidence in each increased, as each database was collected for different purposes and across different years. For the protection of participants included in these datasets, this study was reviewed and approved by the HealthOne Inc Institutional Review Board. Data Sources The CDC CNSIS was designed for TBI and SCI surveillance and is a multi-year, population-based dataset of standardized medical and injury-related variables from patients with TBI, SCI, and combination TBI and SCI injuries occurring in specific states. The CDC CNSIS includes persons of all ages. Core components are obtained from computerized ICD-9-CM26 codes recorded from standard hospital discharge databases using ICD-9-CM codes for case identification. Extended component data are obtained via medical record review and abstraction from a sample of cases identified through the core component. Since its inception in 1995, the number of states reporting cases to the CDC with extended data has fluctuated between 4 and 12, with each state submitting information from a stratified sample of cases based on hospital size (small and large hospitals [⬍100 and ⱖ100 licensed beds for acute care, respectively]). Data from 1997 to 2003 were available for analysis; however, population-based weights were not available for all years. As a result, each case was treated as a single incidence, and we made no attempt to generalize results to a national level. Furthermore, only states with less than 40% missing data on any of the variables of interest were selected for analysis; these criteria limited the dataset to cases from Alaska, Colorado, Minnesota, South Carolina, and Utah. Beginning in 2007, data submitted to the NTDB followed the guidelines of the NTDS, a data dictionary comprised of standardized variables and response categories. In 2007, the NTDB consisted of over 500,000 records with valid trauma codes. Data were submitted by 435 hospitals with American College of Surgeons designation as a Trauma Center. Most contributing hospitals (85%) were accredited as level I or level II, and represent two-thirds of the accredited trauma centers. Only data from the year 2007 were available.

723

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert

The NSCOT is comprised of extensive data on over 5000 adult trauma patients treated at 1 of 69 hospitals in 12 states. As this dataset was created for prospective research purposes, case inclusion occurred only during 2001 and 2002. The NSCOT was conducted in 15 regions selected from among the 25 largest population centers in 19 states. All level I trauma centers and large nontrauma centers were identified in each region. A sample was selected representing approximately equal numbers of small, medium, and large centers on the basis of the annual volume of patients with major trauma. Eighteen (67%) of the trauma centers and 51 (41%) of the nontrauma centers selected agreed to participate and received approval from their institutional review board. Patients included in the study were 18 to 84 years of age, arrived alive at a participating hospital, and were treated for a moderate to severe injury, defined by at least 1 injury with a score of 3 or higher on the AIS27 between July 2001 and November 2002. Patients who presented with no vital signs and were pronounced dead within 30 minutes after arrival were excluded, as were patients who delayed seeking treatment for more than 24 hours, were 65 years of age or older with a first listed diagnosis of hip fracture, presented with major burns, spoke neither English nor Spanish, were non-U.S. residents, or were incarcerated or homeless at the time of injury. A total of 18,198 patients met these initial eligibility criteria. From this sample, 1438 subjects who expired in-hospital were included, and medical records for these cases were abstracted. Among those persons discharged alive, a quota sampling strategy was used to enroll approximately 3000 patients who were 18 to 64 years of age and 1300 patients 65 to 84 years of age, evenly distributed across trauma and nontrauma centers, and across categories of injury severity and principal body region injured.25 Variables of Interest To ensure that analyses were comparable across datasets, we selected similar variables within each dataset. Using codebooks from each, we recoded and reorganized variables into equivalent classifications, so that the analyses would be consistent across datasets. We standardized ICD-9-CM codes, as these fields are entered as text and displayed a great deal of variation. Study Population Cases from each of the 3 datasets were selected to include adults aged 16 and older, with a diagnosis of moderate to severe TBI, and who were discharged alive from an acute care hospital with a known discharge disposition (the oldest person was found to be 104). Each dataset was limited to cases that had at least 1 ICD-9-CM code that met the CDC case definition for TBI: 800.0 – 801.9, 803.0 – 804.9, 850.0 – 854.1, and 959.01.23 A final criterion reduced cases to those determined to have a moderate to severe TBI, as defined by previously published definitions of severity, on at least 1 of 3 measures: (1) an AIS score of the head between 3 and 6,28 (2) a GCS score between 3 and 12,29 and/or (3) a fifth digit of any associated TBI-related ICD-9-CM code indicating a loss of consciousness equal to or longer than 31 minutes.30 The results of each of the case reduction steps, as well as the resultant comparable groups of discharged patients by disposition are listed in table 1. These reductions in the number of cases are not the result of missing data; rather, the selected samples represent the specific range of TBI diagnoses, which is only a portion of the comprehensive retrospective databases. Outcome Measure The outcome measure was discharge disposition from acute hospitalization. For purposes of comparison between the 3

Table 1: Remaining Cases After Stepwise Case Reduction Data Selection Steps

CDC CNSIS

NTDS

NSCOT

Full dataset Cases remaining when eligibility criteria were applied sequentially ICD-9-CM code for TBI Age ⱖ16 (y) Discharged alive to known disposition At least 1 TBI severity measure known Moderate/severe TBI Discharge to home, subacute or rehabilitation Home Subacute* IRF

107,052

507,762

5043

92,098 78,020 64,938

175,816 150,265 125,876

2447 2447 1632

42,879

122,988

1632

16,854 15,646

54,468 52,012

1333 1286

10,029 2087 3530

35,028 9546 7438

766 135 385

*Subacute setting includes skilled nursing facilities, residential facilities, and/or nursing homes.

datasets, we defined 3 categories of discharge disposition: home, subacute, and rehabilitation. The home group included all person who returned home after acute hospitalization, regardless of the need for assistance or the receipt of services once at home. Patients discharged from acute care who were sent to a nursing home or similar facility were designated as the subacute group (even though they may have received some rehabilitation services in the subacute facility). The rehabilitation group only included discharges to IRFs. The specific discharge disposition codes assigned to each of the 3 discharge dispositions from each of the 3 datasets are presented in table 2. Cases with all remaining discharge dispositions (eg, transferred to another acute care facility) were removed from the datasets. Predictors of Discharge Disposition We selected variables that could be associated with discharge disposition based on published research.2-6,17,21-22 Predictors were grouped into 3 categories: injury severity, sociobiological, and socioeconomic factors. Three variables were included in the injury severity category: GCS score (GCS scores from the NTDS and NSCOT were measured at admission to acute care; GCS scores for the CDC were the lowest score obtained during emergency care or admission to acute care), length of acute hospital LOS, Injury Severity Score calculated from AIS.26 Sociobiologic factors were age and sex, while socioeconomic factors were race/ethnicity and payment source. We acknowledge that race/ethnicity is not a perfect fit within the socioeconomic category; however, the relationships between race/ethnicity and access to health care have been well established.31-34 These results suggest that race/ethnicity can be a good proxy for poverty and therefore a proxy for health care access. Statistical Analysis We completed descriptive analysis of the demographic and injury-related characteristics for each discharge disposition group. Logistic regression was used to calculate odds of association between discharge disposition and predictors. Models were constructed that used the following dichotomies: home versus not home (not home was comprised of the discharge dispositions of subacute and rehabilitation), and subacute verArch Phys Med Rehabil Vol 92, May 2011

724

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert Table 2: Categorization of Discharge Disposition Codes, by Dataset

Discharge Category

CDC CNSIS

NTDS

NSCOT

Home Home, self-care Home, nonskilled assistance required Home, skilled assistance required Left AMA

Home, no services Home with home health Left AMA

Home Home with services Left AMA Left without treatment

Residential facilities with no skilled nursing Residential facility with skilled nursing

Intermediate care facility Skilled nursing

Residential facility Skilled nursing Nursing home

Inpatient rehabilitation

Rehabilitation or long-term care*

Inpatient rehabilitation

Acute care Correctional facility Other

Short-term general hospital Hospice

Other acute care Other

Subacute

Rehabilitation Removed

Abbreviation: AMA, against medical advice. *We make the assumption that this code refers to inpatient rehabilitation centers and other care facilities that include rehabilitation, as long-term care facilities that do not include rehabilitation (eg, skilled nursing) are designated with specific codes.

sus rehabilitation (for those not going home). This analysis plan was based on an assumed 2-step decision making process where the discharge disposition was first decided by whether the person was independent enough to go home; and if not, followed by the decision whether to discharge the person to a subacute or to an IRF. Models were evaluated separately within each dataset, using the discharges of not home and rehabilitation as the outcomes of interest in each model, respectively. We applied a consistent framework in the construction of the overall models for both of the discharge dichotomies. Each predictor variable was modeled independently. Continuous predictors were assessed by creating linear, quadratic, and cubed versions because it was assumed that their relationship with the outcomes might not be linear. The appropriate relationship between each of the continuous predictor variables and the outcome was determined by adding the squared and cubed variables to the linear variable in sequence using the following criteria: a significant relationship with the dependent variable (Pⱕ.05), a minimum 2% increase in Nagelkerke pseudo R2 value, and an overall reduction in Akaike’s Information Criterion score was required (as compared to the previously tested relationships) to retain the higher order relationship (only LOS had a nonlinear relationship with each outcome, an example of which is presented graphically in fig 1). Categorical predictors were entered into each model using the most common response as the referent group. Two-way interactions between GCS, LOS, age, and payment source were tested; however, these terms did not remain, as none of the interactions added value to the models after the individual variables had been included. The overall models were constructed using the previously described process for modeling predictor variables. Variables were stepped into the final models in order of highest to lowest associative value. Due to the size of the databases, standard stepwise methods were not used, as every variable was significantly associated with the dependent variable. To make the variable inclusion process stricter, predictors were sequentially entered into each model based on their predictive strength as long as each added variable was significantly related to the outcome, the Nagelkerke value increased .02 points over the previous combination of predictors, and the Akaike’s Information Criterion value was reduced. Within rehabilitation research, the AIS of the head has not been commonly used as a tool for designation of moderate to Arch Phys Med Rehabil Vol 92, May 2011

severe TBI. To better understand the utility of this measure for classification purposes, a separate series of analyses was conducted to determine whether the AIS performed differently than the other 2 measures of severity and to evaluate how sensitive our results were to the inclusion of the AIS. These analyses are available through supplemental content displayed on the Archives journal website: www.archives-pmr.org (sup-

Fig 1. LOS functions for home versus not home discharge fit using NTDS data (progression: linear, linear and quadratic, and cubed). (A) Percent chance of discharge homeⴝLOS. (B) Percent chance of discharge home ⴝ LOSⴙLOS2ⴙLOS3.

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert

plemental Appendix 1, supplemental fig 1, supplemental tables 1–3). We concluded that the AIS performs no differently than the GCS and LOC in defining moderate to severe TBI. All analyses were completed using SPSS version 18.0.a RESULTS Patients’ demographic and injury characteristics, by dataset and discharge destination, are provided in table 3. Across all 3 datasets, patients with moderate to severe TBI were likely to be young, white males. Comparisons of the 3 discharge disposition groups showed that the subacute group was older and more likely to be female than both the home and the rehabilitation groups. Logistic regression models for the discharge disposition home versus not home had moderately strong Nagelkerke values. Included in each of the 3 models (for the 3 datasets) were LOS, age, and race/ethnicity, with the risk of nonhome discharge increasing as LOS and age increased. In regards to race/ethnicity, persons of black, Hispanic, Native American, and other backgrounds were more likely to be discharged home. The models for the NTDS and NSCOT had additional variables in common, including GCS and payor source. People with less severe injuries were more likely to be discharged home, as were persons whose medical expenses were self-paid. Detailed results for each model are displayed in table 4. Results of the logistic regressions predicting the discharge dispositions subacute versus rehabilitation are included in table 5; these models had mildly robust Nagelkerke values. Age was the only covariate common to all models; older age was associated with a decreased likelihood of discharge to rehabilitation. The inclusion of the remaining predictors varied across the datasets. On completion of each of the overall models, the effect of the factors in each predictor category (injury severity, sociobiologic factors, and socioeconomic factors) was assessed to determine the factors that predicted discharge disposition. Table 6 presents the Nagelkerke values for each sample within each discharge disposition comparison, as each predictor category is added to the previous one. The percent of the total association, including all 3 categories of predictor variables, that is added at each step, is also reported. The injury severity variables accounted for the greatest proportion of Nagelkerke values for the models comparing home versus not home. Sociobiologic predictors accounted for the next highest percentage of explanatory power. Models predicting the discharge dispositions of subacute versus rehabilitation showed a major role for the sociobiologic variables, followed by socioeconomic variables. DISCUSSION Injury severity, sociobiologic, and socioeconomic variables provided a moderately strong prediction of discharge disposition to home versus not home for adults with moderate to severe TBI admitted to acute care hospitals. Across the 3 datasets, injury severity-related variables accounted for the greatest amount of association. These results are logical and consistent with previously reported research,2,5 and they support the utility and appropriateness of the severity measures themselves. Age was associated with discharge disposition, with older persons being discharged to some type of postacute medical facility (subacute or rehabilitation) more often than younger ones. In previously published research, this association has been explained through the relationship between age and government-funded insurance5,22; however, here, both vari-

725

ables (age and payment source) added considerably, and separately, to the overall percent of the dependent variable explained in the models for both the NTDS and NSCOT databases, demonstrating that age plays a role in discharge disposition that goes beyond what can be explained by payment source. The socioeconomic variables, race/ethnicity and payment source, were significant factors in predicting home discharge for adults with a moderate to severe TBI. Across each of the 3 datasets, 4 of the 5 nonwhite ethnicity/race groups were more likely to be discharged home, even after accounting for severity-related predictors. These results demonstrate quite clearly that whites are more likely than their counterparts to be discharged to a facility in which continued medical treatment and follow-up occur (subacute or rehabilitation). Payment source has often been shown to be a predictor of discharge disposition, though this variable is often inextricably linked with age or other demographics.22,35 In the NTDS and NSCOT models, payment source was significantly associated with discharge disposition after controlling for both age and ethnicity. Commonalities between these models suggest that persons with Medicare are more likely to be discharged to a setting in which additional medical care can be provided, while persons who must pay out of pocket are less likely to receive inpatient postacute care. Given that the relationship between age and discharge disposition was assessed in linear, cubed, and quadratic formats, it seems unlikely that this association is a result of age-related entitlement to Medicare; access is a separate influence. Prediction of discharge to rehabilitation or subacute disposition was less successful. Although age and LOS were common predictors across datasets, the structure and predictive direction of the LOS predictor varied across models. These results suggest that predicting discharge from acute care to either subacute or rehabilitation is extremely complex. In order to create a more robust model to predict discharge to rehabilitation for persons with moderate to severe TBI who are not discharged home, more specific variables, including those that provide comorbid conditions and the social context of the injury, may be required,2,35-36 The decision whether to discharge home or to additional inpatient services is driven primarily (to the degree that the factors here explain discharge destination) by brain and overall injury severity. The demographic variables and the additional factors that presumably affect access (racial/ ethnic group, type of insurance) play a relatively minor role. However, the decision to discharge to inpatient rehabilitation versus to subacute care, which is less easily explained by these same factors, appears driven primarily by the sociobiologic and socioeconomic factors distinguished here, and to a much smaller degree by injury severity. While this should not be interpreted directly as discharge planning staff (physician and case manager or social worker, together with the patient and family) making 2 sequential decisions driven by completely different factors, nonetheless it would seem that if the TBI exceeds a certain level of severity (which probably has somewhat different cutoffs for various levels of age, comorbidities, etc), then a decision is made to not send the patient home. The second decision, where to send the patient, to subacute care or to an IRF, appears to be driven by other factors, with age an apparent important one. LOS plays a role either as a proxy for injury severity (reflecting severity effects that are not well tapped by GCS), and/or as an indicator of the seriousness of other comorbidities and/or concurrent injuries, whether injuries associated Arch Phys Med Rehabil Vol 92, May 2011

726

Entire Dataset

Characteristic Mean age ⫾ SD Age group (%) 16–19 20–29 30–39 40–49 50–59 60–69 70–79 80–89 ⬎90 Sex (%) Men Women Unknown Race/ethnicity (%) White Black Native American/ Aleut Asian Hispanic Other race Missing Payor Source (%) Private insurance Medicaid Medicare Workers’ compensation Self-pay/no pay Other Missing GCS Score (%) Mild (13–15)* Moderate (9–12) Severe (3–8) Unknown Max AIS, head (%) Mild (1–2)* Moderate (3) Severe (4–6) Unknown ISS score Mean ⫾ SD Unknown (n) Acute LOS Mean ⫾ SD Unknown (n)

CDC CNSIS

NTDS

n⫽15,646

n⫽52,012

44.5⫾22.2

Home NSCOT

CDC CNSIS

NTDS

n⫽1286

n⫽10,029

n⫽35,028

46.2⫾22.0

48.3⫾21.3

40.4⫾19.6

12.5 20.5 15.9 15.2 9.9 6.8 9.1 8.1 2.0

10.0 21.2 12.9 15.2 12.8 8.5 8.5 8.8 2.1

6.5 20.3 13.1 14.2 10.8 10.9 16.9 7.3 0.0

68.3 31.7 0.0

69.9 30.0 0.1

50.0 5.9 3.2

Rehabilitation NSCOT

CDC CNSIS

NTDS

n⫽766

n⫽3530

n⫽7438

42.4⫾20.1

45.4⫾20.2

42.8⫾21.4

14.3 21.0 16.0 15.4 10.0 7.3 7.8 7.2 1.0

10.9 19.3 11.7 15.2 14.3 9.7 9.2 8.2 1.5

6.2 22.9 11.2 14.0 9.9 10.6 17.4 7.8 0.0

68.0 32.0 0.0

71.5 28.5 0.0

73.1 26.8 0.1

64.6 11.4 0.7

65.0 13.3 1.5

50.2 7.2 4.1

1.1 6.2 1.0 32.6

1.3 9.8 2.4 9.8

2.2 12.8 2.9 2.3

30.4 9.2 14.7 3.3

25.4 7.6 16.4 2.1

14.1 23.7 4.6 45.9 12.5 19.0 22.6 4.8 33.4 46.8 15.0 19.2⫾9.8 11,987 9.3⫾14.5 3420

Subacute NSCOT

CDC CNSIS

NTDS

NSCOT

n⫽385

n⫽2087

n⫽9546

n⫽135

46.4⫾21.6

48.3⫾21.7

67.4⫾21.9

59.6⫾23.7

64.70⫾18.5

14.1 23.2 17.9 16.6 10.3 6.2 7.0 4.0 0.7

11.0 24.1 14.7 16.5 12.9 7.9 6.6 5.3 1.0

7.6 21.6 15.3 15.3 11.7 10.7 13.4 4.4 0.0

2.6 6.5 6.0 8.4 7.2 8.7 21.1 29.7 9.8

5.6 11.3 7.5 10.4 11.0 9.8 15.3 22.5 6.6

1.5 5.9 5.9 8.9 8.1 12.6 34.9 22.2 0.0

69.7 30.3 0.0

68.9 31.1 0.0

68.7 30.9 0.4

66.5 33.5 0.0

52.3 47.7 0.0

59.0 40.9 0.1

62.2 37.8 0.0

61.7 12.3 0.8

56.1 14.2 1.0

48.0 2.8 1.8

67.9 8.5 0.8

60.6 14.8 1.8

51.5 4.6 1.3

72.6 10.3 0.4

73.4 6.7 0.7

1.2 6.7 1.3 29.3

1.3 11.9 2.6 9.4

2.0 19.2 5.7 1.8

1.1 6.2 0.5 39.6

1.3 7.0 2.6 11.9

2.0 16.3 2.9 1.6

1.0 4.0 0.7 36.9

1.2 4.5 1.5 9.5

1.5 8.1 1.5 8.1

33.0 9.2 27.3 3.0

31.7 8.8 10.5 3.8

25.3 7.4 11.6 2.3

35.5 8.2 19.5 3.5

31.0 9.9 11.9 3.2

28.1 8.1 14.0 2.3

29.8 11.2 33.2 2.9

22.7 9.8 39.7 1.1

23.0 7.9 35.8 1.4

28.1 8.9 54.8 0.0

17.0 6.6 24.9

17.7 6.6 3.2

18.0 23.3 3.9

21.5 7.0 24.9

22.3 6.8 4.2

8.4 28.4 7.2

8.8 4.7 34.0

13.8 7.5 1.6

5.0 18.0 3.7

7.1 6.7 18.1

3.0 3.0 2.2

57.2 9.5 26.2 7.1

67.5 8.1 23.2 1.2

53.7 13.2 13.6 19.5

64.7 9.0 19.7 6.6

79.8 6.8 13.7 0.7

30.2 11.9 34.4 23.5

34.4 11.7 47.6 6.3

43.7 10.4 43.6 2.3

34.8 10.4 19.0 35.8

48.0 9.4 33.1 9.5

66.0 8.9 24.4 0.7

49.0 34.6 6.5 9.9 19.7⫾10.0 687 10.5⫾14.6 19

3.4 34.4 62.0 0.2 21.0⫾9.7 0 10.9⫾14.0 3

53.9 21.6 3.6 20.9 25.8⫾11.4 123 19.3⫾16.7 1

6.0 23.1 70.4 0.5 26.0⫾10.7 0 18.3⫾18.3 1

65.8 24.0 3.1 7.1 23.8⫾10.6 687 18.7⫾20.5 4

2.2 21.5 75.6 0.7 22.16⫾10.7 0 16.76⫾17.2 0

6.2 41.5 36.5 15.8 16.3⫾7.6 7711 5.8⫾9.8 2285

43.5 40.2 8.0 8.3 17.3⫾8.4 508 6.5⫾9.7 15

2.3 42.3 55.4 0.0 18.3⫾7.7 0 6.0⫾8.0 2

2.6 18.3 68.8 10.3 25.4⫾11.3 2578 14.6⫾11.3 716

Abbreviation: ISS, Injury Severity Scale. *All cases had at least 1 TBI severity score of moderate to severe; however, alternate severity measures may have been coded as mild.

2.0 20.1 58.8 19.1 21.0⫾9.7 1689 16.8⫾27.0 419

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert

Arch Phys Med Rehabil Vol 92, May 2011

Table 3: Demographic and Injury Characteristics by Dataset and Discharge Disposition

727

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert Table 4: Logistic Regression Models Predicting Discharge Dispositions of Home Versus Not Home* CDC CNSIS

NTDS

NSCOT

n⫽12,226

n⫽48,313

n⫽1268

Model Information Nagelkerke

0.42 OR

GCS Score Acute LOS LOS-squared LOS-cubed Injury Severity Scale score Age Sex Women Men Race/ethnic group Black Native American/Aleut Asian Hispanic Other White Payment source Medicare Medicaid Self-pay/no pay Workers’ compensation Other Private insurance

0.48 95% CI

§

1.297† 0.996† 1.000†

0.57 95% CI

†

(1.281–1.313) (0.996–0.996) (1.000–1.000)

0.923 1.283† 0.996† 1.000†

(1.029–1.033)

1.031†

储

1.031†

OR

OR

95% CI

(0.917–0.928) (1.274–1.292) (0.996–0.996) (1.000–1.000)

†

0.901 1.498† 0.991† 1.000†

(0.865–0.938) (1.406–1.595) (0.989–0.994) (1.000–1.000)

(1.030–1.033)

1.023†

(1.011–1.035)

‡

‡

‡

‡

‡

Reference

Reference

Reference

0.262† 0.410† 1.093 0.973 0.382† Reference

(0.203–0.339) (0.295–0.571) (0.714–1.674) (0.792–1.197) (0.222–0.655)

‡ ‡ ‡ ‡ ‡

Reference

0.720† 0.771 0.874 0.419† 0.793† Reference

(0.662–0.782) (0.561–1.059) (0.698–1.096) (0.378–0.464) (0.671–0.938)

0.757 0.068† 1.023 0.275† 0.917 Reference

(0.461–1.243) (0.015–0.312) (0.369–2.835) (0.151–0.501) (0.364–2.315)

1.601† 0.814† 0.380† 0.895 0.758† Reference

(1.476–1.737) (0.735–0.902) (0.347–0.416) (0.749–1.070) (0.678–0.847)

2.222† 1.113 0.449† 0.722 1.360 Reference

(1.347–3.663) (0.612–2.022) (0.264–0.764) (0.264–2.053) (0.698–2.648)

NOTE. A value ⬍1.00 signifies that discharge to home is more likely; a value ⬎1.00 indicates that a discharge to not home is more likely. Reference indicates the category with which other categories within a variable is compared. Abbreviations: CI, confidence interval; OR, odds ratio. *Includes subacute and rehabilitation discharges. † Significant at the .01 level. ‡ Not included in the final model. § Missing for ⬎40% of cases. 储 Injury Severity Scale was dropped as a CDC CNSIS variable after 1999.

with the same etiology as the TBI (eg, fractures) or preexisting disorders. A unique aspect of the analyses reported here was the restriction of the acute care samples to persons with moderate to severe TBI as indicated by at least 1 of 3 severity measures (ie, GCS, AIS, or length of unconsciousness). The majority of patients with moderate to severe TBI went home directly from acute care. The percentage going to inpatient rehabilitation ranged from 13% to 29%, while 57% to 65% went directly home. The ratio of those going home versus rehabilitation ranged from almost 2:1 (in the NSCOT dataset) to more than 4.5:1 (in the NTDS dataset). In 2007, the Uniform Data System for Medical Rehabilitation and eRehabData datasets combined recorded more than 20,000 admissions to rehabilitation of persons 16 years and older who had a primary diagnosis of TBI (S. Flemming, written communication, December 2009; S. Markello, written communication April 2010). These 2 datasets combined include essentially 100% of all civilian admissions to IRFs in the U.S. Using a conservative estimate that for every 1 patient 16 years and older with moderate to severe TBI who goes to rehabilitation, 3 go directly home, we surmise that in the U.S. each year, as many as 60,000 late adolescents and adults with moderate to severe TBI may go home directly from an acute care hospital. The residual effect of these moderate to severe TBIs on brain functioning cannot be precisely estimated; however, Whiteneck et al37 reported

from the Colorado TBI Registry and Follow-up System that 36.6% of Coloradoans hospitalized with any severity of TBI had disability at 1-year postinjury as defined by needing assistance on 1 or more cognitive or physical functional independence (FIM) items. In addition, they reported that among those hospitalized with moderate to severe TBI (as defined by GCS), the percentage was 48.7% (severe 54.2% and moderate 38.7%; mild 33.6%).37 Given Selassie et al’s38 estimate that 43.3% of persons hospitalized in South Carolina with any TBI will experience some disability 1 year later, it is reasonable to expect a higher rate among those with moderate to severe TBI. Those with moderate to severe injuries who go directly home may be a previously unrecognized component of the public health burden created by TBI. Although the findings for the 3 datasets were fairly similar, there were not insignificant differences, in spite of the facts that 2, CNSIS and NTDS, were large databases (with over 15,000 and 52,000 cases, respectively). The discrepancies may be explained by differences in catchment areas of the hospitals involved, the differences between the datasets in inclusion criteria, and a multitude of other factors. However, the discrepancies in the details of findings should be a warning that a large dataset is not sufficient for this type of research. Rehabilitation researchers often draw conclusions based on administrative data, and often assume that size Arch Phys Med Rehabil Vol 92, May 2011

728

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert Table 5: Logistic Regression Models Predicting the Discharge Dispositions Subacute Versus Rehabilitation, for Patients not Returning Home CDC CNSIS

NTDS

NSCOT

n⫽5617

n⫽15,615

n⫽519

Model Information Nagelkerke

0.33 OR

GCS score Acute LOS LOS-squared LOS-cubed Injury Severity Scale score Age Sex Women Men Race/ethnic group Black Native American/Aleut Asian Hispanic Other White Payment source Medicare Medicaid Self-pay/no pay Workers’ compensation Other Private insurance

0.15 95% CI

OR

§

0.970†

†

0.988 0.991†

(0.965–0.976)

‡

‡

‡

‡

储

0.948†

0.25 95% CI

OR

(0.980–0.996) (0.989–0.993)

‡

‡

1.069* 0.998* 1.000

(1.001–1.142) (0.996–1.000) (1.000–1.000)

‡

0.979†

(0.945–0.951)

95% CI

(0.977–0.981)

0.952†

(0.936–0.969)

‡

‡

‡

Reference

Reference

Reference

‡

0.718† 2.459† 1.168 1.327† 1.461† Reference

(0.639–0.807) (1.542–3.923) (0.861–1.583) (1.145–1.537) (1.155–1.850)

1.180 1.103 1.324 0.609 2.156 Reference

(0.524–2.656) (0.105–11.630) (0.259–6.774) (0.248–1.497) (0.444–10.461)

0.622† 0.819† 0.974 1.310* 0.524† Reference

(0.557–0.696) (0.718–0.934) (0.854–1.111) (1.026–1.673) (0.450–0.611)

2.416* 1.039 4.103* Infinite 2.841 Reference

(1.226–4.759) (0.444–2.432) (1.170–14.387)

‡ ‡ ‡ ‡

Reference ‡ ‡ ‡ ‡ ‡

Reference

(0.844–9.569)

NOTE. A value ⬍1.00 signifies that discharge to subacute is more likely; a value ⬎1.00 indicates that a discharge to rehabilitation is more likely. Reference indicates the category with which other categories within a variable is compared. Infinite refers to an unobtainable odds ratio that is the result of a small n for a particular variable category. Abbreviations: CI, confidence interval; OR, odds ratio. *Significant at the .05 level. † Significant at the .01 level. ‡ Not included in the final model. § Missing for ⬎40% of cases after additional variables included. 储 ISS was dropped as a CDC CNSIS variable after 1999.

equates representativeness. It is clear that such may be an erroneous assumption. While the expansion of existing retrospective and administrative databases may improve the accuracy of predicting acute care discharge for persons with moderate to severe

TBI, a more appropriate solution would be the design and implementation of a prospective clinical database including standardized injury severity, demographic, socioeconomic, comorbid condition, and social context variables that would provide a more precise and accurate data source. Analysis of

Table 6: Nagelkerke Value for Sequentially Added Categories of Variables, and the Percent of Total Association Explained Attributable to the Categories of Variables Model Comparisons Sample

Home vs Not Home

Not Home: Subacute vs Rehabilitation

CDC CNSIS

NTDS

NSCOT

0.35 83

0.36 74

0.44 77

0.40 12

0.46 21

0.42 5

0.48 4

CDC CNSIS

NTDS

NSCOT

0.02 7

0.03 22

0.04 16

0.52 14

0.33 93

0.10 44

0.21 68

0.57 9

* *

0.15 33

0.25 16

Predictor category Severity Nagelkerke % of total association Severity plus sociobiologic factors Nagelkerke Added % of total association Severity plus sociobiologic factors plus socioeconomic factors Nagelkerke Added % of total association

*No variables within category included in final model.

Arch Phys Med Rehabil Vol 92, May 2011

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert

such data could be extremely useful in providing accurate discharge disposition predictions for clinicians and family members, and would also provide a trustworthy resource for risk and outcome analysis. These data could also be used to substantiate the estimate from the current study that a large percentage of persons with moderate to severe TBIs are discharged home from acute care, many presumably with no rehabilitation. Such a database would overcome the obstacles experienced in conducting this analysis (missing data, nonstandardized variables) and alleviate the shortcomings associated with analyzing retrospective data for secondary analysis purposes.39 Study Limitations This study has several limitations. Our research is limited to late teens and adults. Future research could be expanded to include all ages to increase the generalizability. The databases analyzed were not population-based and thus, may not allow identification of all cases of TBI, which limits the ability to extrapolate the study findings to a national level. It is also noted that the time frames for hospital admissions included in the 3 datasets were disparate (CDC, 1997–2003; NTDS, 2007; NSCOT, 2001–2002). Numerous variables that may have been pertinent to determining discharge disposition were unavailable (eg, mechanical ventilation, type of health care facility, health status before hospitalization, and transport time to a trauma unit immediately after the injury).40,41 Analyses were limited by missing data, with some variables being unavailable for analysis because of low rates of completion. The lack of availability of these data and the unique features of each dataset may necessitate additional studies. CONCLUSIONS Results of this study determined that the decision to discharge a person home was due largely to severity-related factors, while the decision to discharge to rehabilitation or subacute care was driven by sociobiologic and socioeconomic factors. Within both models (home vs not home; subacute vs rehabilitation), possible disparities, including those by age, race/ethnicity, and payment source, were apparent and should be evaluated further. Future research should investigate additional independent variables. Because data available is limited, current sources of information should be enhanced or new ones should be designed and implemented. Acknowledgements: We thank Ellen MacKenzie, PhD, for the use of the NSCOT database; the Centers for Disease Control and Prevention for the use of the CDC CNSIS database; the National Trauma Data Bank for the use of the NTDS database. References 1. Faul M, Xu L, Wald MM, Coronado VG. Traumatic brain injury in the United States: emergency department visits, hospitalizations and deaths 2002–2006. Atlanta: Centers for Disease Control and Prevention, National Center for Injury Prevention and Control; 2010. 2. Malec JF, Mandrekar JN, Brown AW, Moessner AM. Injury severity and disability in the selection of next level of care following acute treatment for traumatic brain injury. Brain Inj 2009;23:22-9. 3. Arango-Lasprilla JC, Kreutzer JS. Racial and ethnic disparities in functional, psychosocial, and neurobehavioral outcomes after brain injury. J Head Trauma Rehabil 2010;24:128-36. 4. Gary KW, Arango-Lasprilla JC, Stevens LF. Do racial/ethnic differences exist in post-injury outcomes after TBI? A comprehensive review of the literature. Brain Inj 2010; 23:775-89.

729

5. Mellick D, Gerhart KA, Whiteneck GG. Understanding outcomes based on the post-acute hospitalization pathways followed by persons with traumatic brain injury. Brain Inj 2003;17:55-71. 6. LeBlanc J, de Guise E, Gosselin N, Feyz M. Comparison of functional outcome following acute care in young, middle-aged and elderly patients with traumatic brain injury. Brain Inj 2006; 20:779-90. 7. Testa JA, Malec JF, Moessner AM, Brown AW. Outcome after traumatic brain injury: effects of aging on recovery. Arch Phys Med Rehabil 2005;86:1815-23. 8. Coronado VG, Thomas KE, Sattin RW, Johnson RL. The CDC Traumatic Brain Injury Surveillance System. Characteristics of persons aged 65 years and older hospitalized with a TBI. J Head Trauma Rehabil 2005;20:215-28. 9. Harrison-Felix C, Whiteneck G, DeVivo M, Hammond FM, Jha A. Mortality following rehabilitation in the Traumatic Brain Injury Model Systems of Care. NeuroRehabil 2004;19:45-54. 10. Labi ML, Brentjens M, Coad ML, Flynn WJ, Zielezny M. Development of a longitudinal study of complications and functional outcomes after traumatic brain injury. Brain Inj 2003;17:265-78. 11. Cross J, Trent R, Adekoya N. Public health and aging: Nonfatal fall-related traumatic brain injury among older adults—California, 1996 –1999. MMWR Morb Mortal Wkly Rep 2003; 52:276-8. 12. Mosenthal AC, Lavery RF, Addis M, et al. Isolated traumatic brain injury: age is an independent predictor of mortality and early outcome. J Trauma 2002;52:907-11. 13. Peek-Asa C, McArthur D, Hovda D Kraus J. Early predictors of mortality in penetrating compared with closed brain injury. Brain Inj 2001;15:801-10. 14. Goleburn CR, Golden CJ. Traumatic brain injury outcome in older adults: a critical review of the literature. J Clin Geropsychology 2001;7:161-87. 15. Lannoo E, Van Rietvelde F, Colardyn R, et al. Early predictors of mortality and morbidity after severe closed head injury. J Neurotrauma 2000;17:403-14. 16. Luerssen TG, Klauber MR, Marshall LF. Outcome from head injury related to patient’s age: a longitudinal prospective study of adults and pediatric head injury. J Neurosurgery 1988;68:409-16. 17. Frankel JE, Marwitz JH, Cifu DX, Kreutzer JS, Englander J, Rosenthal M. A follow-up study of older adults with traumatic brain injury: taking into account decreasing length of stay. Arch Phys Med Rehabil 2006;87:57-62. 18. Mosenthal AC, Livingston DH, Lavery RF, et al. The effect of age on functional outcome in mild traumatic brain injury: 6-month report of a prospective multicenter trial. J Trauma 2004;56:1042-8. 19. Okor CA, Young SL, Strine TW, Balluz LS, Mokdad AH. Uninsured adults aged 65 years and older: is there health at risk? J Health Care Poor Underserved 2005;16:453-63. 20. Bodofsky E, Schindelheim A, Micarek B, Lachant M, Ross S. Increase in TBI discharges and associated diagnoses in the U.S., 2001-2007. Paper presented at: the International Brain Injury Association 8th World Congress, March 11, 2010; Washington, DC. 21. Shafi S, Marquez de la Plata CM, Diaz-Arrastia R, et al. Racial disparities in long-term functional outcome after traumatic brain injury. J Trauma 2007;63:1263-8. 22. Chan L, Doctor J, Temkin N, et al. Discharge disposition from acute care after traumatic brain injury: the effect of insurance. Arch Phys Med Rehabil 2001;82:1151-4. 23. Marr A, Coronado V. Central Nervous System Injury Surveillance Data Submission Standards –2002. Atlanta: Centers for Disease Control and Prevention, National Center for Injury Prevention and Control; 2004. 24. National Trauma Data Bank. National Trauma Data Standard Data Dictionary 2009. Available at: www.ntdsdictionary.org/index. html. Accessed February 1, 2009. Arch Phys Med Rehabil Vol 92, May 2011

730

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert

25. Mackenzie EJ, Rivara FP, Jurkovich GJ, et al. The National Study on Costs and Outcomes of Trauma. J Trauma 2007;63(Suppl 6):S54-67; discussion S81-6. 26. U.S. Department of Health and Human Services. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Hyattsville: National Center for Health Statistics; 2004. 27. Baker SP, O’Neill B, Haddon W Jr, Long WB. the injury severity score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma 1974;14:187-96. 28. Pickelsimer EE, Selassie AW, Gu JK, Langlois JA. A populationbased outcomes study of persons hospitalized with traumatic brain injury: operations of the South Carolina Traumatic Brain Injury Follow Up Registry. J Head Trauma Rehabil 2006;21:491-504. 29. Teasdale G, Jennett B. Assessment of coma and impaired consciousness: a practical scale. Lancet 1974;13:81-4. 30. National Center for Injury Prevention and Control. Report to Congress on Mild Traumatic Brain Injury in the United States: steps to prevent a serious public health problem. Atlanta: Centers for Disease Control and Prevention; 2003. 31. Richardson LD, Norris M. Access to health and health care: how race and ethnicity matter. Mt Sinai J Med 2010;77:166-77. 32. Sentell T, Shumway M, Snowden L. Access to mental health treatment by English language proficiency and race/ethnicity. J Gen Intern Med 2007;22:S2:289-93. 33. Hadley J, Cunningham P, Hargraves JL. Would safety-net expansions offset reduced access resulting from lost insurance coverage? Race/ethnicity differences. Health Aff (Milwood) 2006;25: 1679-87. 34. Selassie AW, McCarthy ML, Pickelsimer EE. The influence of insurance, race, and gender on emergency department disposition. Acad Emerg Med 2003;10:1260-70.

Arch Phys Med Rehabil Vol 92, May 2011

35. Chang PF, Ostir, GV, Kuo YF, Granger CV, Ottenbacher KJ. Ethnic differences in discharge destination among older patients with traumatic brain injury. Arch Phys Med Rehabil 2008;89: 231-6. 36. Corrigan JD, Selassie AW, Lineberry LA, et al. Comparions of the Traumatic Brain Injury (TBI) Model Systems National Dataset to a population-based cohort of TBI hospitalizations. Arch Phys Med Rehabil 2007;88:418-26. 37. Whiteneck G, Brooks CA, Mellick D, Harrison-Felix C, Terrill MS, Noble K. Population-based estimates of outcomes after hospitalization for traumatic brain injury in Colorado. Arch Phys Med Rehabil 2004;85(4 Suppl 2):S73-81. 38. Selassie AW, Zaloshnja E, Langlois JA, Miller T, Jones P, Steiner C. Incidence of long-term disability following traumatic brain injury hospitalization, United States, 2003. J Head Trauma Rehabil 2008;23:123-31. 39. Shahian, D.M., Silverstein, T., Lovett, A.F., Wolf, R.E., Normand, S.T. Comparison of Clinical and Administrative Data Sources for Hospital Coronary Artery Bypass Graft Surgery Report Cards. Circ 2007;115:1518-27. 40. Meldon SW, Reilly M, Drew BL, Mancuso C, Fallon W Jr. Trauma in the very elderly: a community-based study of outcomes at trauma and non-trauma centers. J Trauma 2002;52: 79-84. 41. Taylor MD, Tracy JK, Meyer W, Pasquale M, Napolitano LM. Trauma in the elderly: intensive care unit resource use and outcome. J Trauma 2002;53:407-14. Supplier a. SPSS Inc, 233 S Wacker Dr, 11th Fl, Chicago, IL 60606.

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert

730.e1

Supplemental Fig 1. Percent of patients discharged to rehabilitation, subacute care, or home, by dataset and criterion used for identification of moderate-severe cases.

SUPPLEMENTAL APPENDIX 1: AIS VERSUS OTHER MEASURE OF TBI SEVERITY: EFFECT ON STUDY RESULTS Within rehabilitation research, the AIS1 of the head has not been commonly used as a tool for designation of moderate to severe TBI. To better understand the utility of this measure for classification purposes, a series of analyses were conducted to determine whether the AIS performed differently than other measures of severity and how sensitive our results were to the inclusion of the AIS. To assess these questions, we compared cases for which the AIS indicated moderate or severe TBI to comparable cases based on GCS scores2 and loss of consciousness. Comparisons were computed both for cases in which only 1 indicator led to a case being considered moderate or severe, as well as for all cases for which an indicator was in the moderate or severe range. We also reran our final regression analyses with and without at least AIS cases to determine whether results would change. The discharge disposition distributions for people in each of the datasets who met only 1 severity inclusion criterion (a moderate to severe score on the AIS, GCS, or ICD-9-CM), as well as cases that met more than 1 criterion were compared. Results of these comparisons, shown in supplemental figure 1, revealed that the distributions of discharges were quite similar across datasets and within severity measures used for designating patients as having moderate or severe TBI (AIS only, GCS only, and ICD-9-CM only), with most cases being discharged home; however, discharge dispositions for the group that met severity criteria for more than 1 measure were distributed differently. It would appear that those included based on 2 or 3 criteria are more severely injured, with the majority of cases being discharged to rehabilitation. To further examine the properties of each of the 3 measures, discharge disposition distributions were compared for any cases in each of the datasets that met inclusion criterion for the

3 measures, including cases that met criteria for 2 or 3 severity measures. We thought that comparisons of these groups (at least AIS, at least GCS, and at least ICD-9-CM) would allow for the variability of severity to be maintained across each of the 3 measures. Within severity measures, discharge distributions appeared similar across the datasets; however, again the discharge distributions varied across severity measures. Complete data for these comparisons can be found in table 1. To determine if the differences across severity measures produced dissimilar outcomes, the final models (home vs not home and rehab vs subacute) for each of the datasets were analyzed using only the inclusion groups of at least AIS, at least GCS, and at least ICD-9-CM. For the logistic regressions of home versus not home, the overall models for each of the severity measures across each of the datasets were found to have similar Nagelkerke values (maximum differences of 0.07), and had similar structure in terms of the direction and significance of predictors. Data for these comparisons are listed in supplemental table 2. Supplemental table 3 provides the comparisons of subacute versus rehabilitation logistic regressions. Again, the overall models, across each of the datasets, were found to have similar Nagelkerke values (maximum difference 0.09) and common predictors, both in regards to significance and direction. Based on these results, the AIS appears no better or worse at determining moderate to severe TBI as compared to more commonly used severity measures, including the GCS or ICD-9-CM. 1. Baker SP, O’Neill B, Haddon W Jr, Long WB. The Injury Severity Score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma 1974;14:187-96. 2. Teasdale G, Jennett B. Assessment of coma and impaired consciousness: a practical scale. Lancet 1974;13:81-4.

Arch Phys Med Rehabil Vol 92, May 2011

721

ORIGINAL ARTICLE

Factors That Predict Acute Hospitalization Discharge Disposition for Adults With Moderate to Severe Traumatic Brain Injury Jeffrey P. Cuthbert, MPH, MS, John D. Corrigan, PhD, Cynthia Harrison-Felix, PhD, Victor Coronado, MD, MPH, Marcel P. Dijkers, PhD, Allen W. Heinemann, PhD, ABPP, Gale G. Whiteneck, PhD ABSTRACT. Cuthbert JP, Corrigan JD, Harrison-Felix C, Coronado V, Dijkers MP, Heinemann AW, Whiteneck GG. Factors that predict acute hospitalization discharge disposition for adults with moderate to severe traumatic brain injury. Arch Phys Med Rehabil 2011;92:721-30. Objective: To identify factors predicting acute hospital discharge disposition after moderate to severe traumatic brain injury (TBI). Design: Secondary analysis of existing datasets. Setting: Acute care hospitals. Participants: Adults hospitalized with moderate to severe TBI included in 3 large sets of archival data: (1) Centers for Disease Control and Prevention Central Nervous System Injury Surveillance database (n⫽15,646); (2) the National Trauma Data Bank (n⫽52,012); and (3) the National Study on the Costs and Outcomes of Trauma (n⫽1286). Interventions: None. Main Outcome Measure: Discharge disposition from acute hospitalization to 1 of 3 postacute settings: (1) home, (2) inpatient rehabilitation, or (3) subacute settings, including nursing homes and similar facilities. Results: The Glasgow Coma Scale (GCS) score and length of acute hospital length of stay (LOS) accounted for 35% to 44% of the variance in discharges to home versus not home,

while age and sex added from 5% to 8%, and race/ethnicity and hospitalization payment source added another 2% to 5%. When predicting discharge to rehabilitation versus subacute care for those not going home, GCS and LOS accounted for 2% to 4% of the variance, while age and sex added 7% to 31%, and race/ethnicity and payment source added 4% to 5%. Across the datasets, longer LOS, older age, and white race increased the likelihood of not being discharged home; the most consistent predictor of discharge to rehabilitation was younger age. Conclusions: The decision to discharge to home a person with moderate to severe TBI appears to be based primarily on severity-related factors. In contrast, the decision to discharge to rehabilitation rather than to subacute care appears to reflect sociobiologic and socioeconomic factors; however, generalizability of these results is limited by the restricted range of potentially important variables available for analysis. Key Words: Brain injuries; Healthcare disparities; Hospitalization; Nursing homes, Patient discharge; Rehabilitation; Rehabilitation centers. © 2011 by the American Congress of Rehabilitation Medicine RAUMATIC BRAIN INJURY is one of the leading causes T of disability in the United States, and it is a contributing factor in approximately one-third of all injury-related deaths.

1

From the Research Department, Craig Hospital, Englewood, CO (Cuthbert, Harrison-Felix, Whiteneck); Department of Physical Medicine and Rehabilitation, Ohio State University, Columbus, OH (Corrigan); National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, GA (Coronado); Department of Rehabilitation Medicine, Mt. Sinai School of Medicine, New York, NY (Dijkers); Department of Physical Medicine and Rehabilitation, Feinberg School of Medicine, Northwestern University, Chicago, IL (Heinemann); and the Rehabilitation Institute of Chicago, Chicago, IL (Heinemann). Supported by a supplemental grant to the Traumatic Brain Injury Model Systems National Data and Statistical Center from the National Institute on Disability and Rehabilitation Research, Office of Special Education and Rehabilitative Services, U.S. Department of Education (grant no. H133A060038); and Traumatic Brain Injury Model System Centers grants to Craig Hospital (grant no. H133A070022), Ohio State University (grant no. H133A070029), Mount Sinai Medical Center (grant no. H133A070033), and the Rehabilitation Institute of Chicago (grant no. H133A080045). However, the contents do not necessarily represent the policy of the Department of Education, and the reader should not assume endorsement by the Federal Government. This article does not reflect the official policy or opinions of the Centers for Disease Control and Prevention (CDC) or the U.S. Department of Health and Human Services (HHS) and does not constitute an endorsement of the authors or their programs— by CDC, HHS, or other components of the federal government—and none should be inferred. All interpretations of the data are the responsibility of the current authors only. The content reproduced from the National Trauma Data Bank remains the full and exclusive copyrighted property of the American College of Surgeons. The American College of Surgeons is not responsible for any claims arising from works based on the original data, text, tables or figures. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit on the authors or on any organization with which the authors are associated. Reprint requests to Jeffrey P. Cuthbert, MPH, MS, 3425 S Clarkson St, Englewood CO 80113, e-mail: [email protected]. 0003-9993/11/9205-00715$36.00/0 doi:10.1016/j.apmr.2010.12.023

Each year at least 1.7 million Americans incur a TBI, and of these injuries, 275,000 are severe enough to require hospitalization.1 Recently, hospitalizations related to TBI have risen sharply, with an increase of 19.5% from 2002 to 2006.1 Health care providers, patients, and families must decide, on completion of acute medical care, which postacute care setting will maximize outcomes and minimize morbidity and mortality.2 Potential settings include the patient’s home (with or without outpatient or home health services), inpatient rehabilitation, and skilled or extended nursing care facilities, with the level and type

List of Abbreviations AIS CDC CNSIS GCS ICD-9-CM IRF LOS NSCOT NTDB NTDS SCI TBI

Abbreviated Injury Scale Centers for Disease Control and Prevention Central Nervous System Injury Surveillance Glasgow Coma Scale International Classification of Diseases, 9th Revision, Clinical Modification inpatient rehabilitation facility length of stay National Study on the Costs and Outcomes of Trauma National Trauma Data Bank National Trauma Data Standard spinal cord injury traumatic brain injury

Arch Phys Med Rehabil Vol 92, May 2011

730.e3

PREDICTING DISCHARGE AFTER TRAUMATIC BRAIN INJURY, Cuthbert

Supplemental Table 3: Full Model With Models for Groups With a Moderate to Severe Score on at Least the AIS, ICD-9-CM, and GCS for Subacute Versus Rehabilitation, for Patients Not Returning Home Dataset CDC Referent Group: Subacute

Nagelkerke GCS LOS LOS-squared LOS-cubed Age Race/Ethnicity Black American Indian Asian Other Hispanic White Payment source Other Medicaid Medicare Workers’ compensation Self-pay/no pay Private

Full Model

At Least AIS

0.33

0.37

NTDS

At Least ICD-9-CM

0.33

NSCOT

At Least GCS

Full Model

At Least AIS

At Least ICD-9-CM

0.24

0.14 0.996 1.009*

0.13 0.929* 1.013*

At Least GCS

Full Model

At Least AIS

0.11 0.988 1.01*

0.25

0.25

†

†

†

†

1.031*

1.026*

1.031*

1.029*

0.15 1.012* 1.009*

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

1.055*

1.061*

1.057*

1.046*

1.021*

1.021*

1.017*

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

1.392* 0.407* 0.856 0.684* 0.754*

1.333* 0.502* 0.861 0.827 0.755*

‡

‡

‡

‡

‡

‡

At Least ICD-9-CM §

At Least GCS

0.26

†

†

§

†

0.928* 1.002* 1.000* 1.050*

§

1.014*

0.936* 1.002* 1.000 1.050*

1.048 1.000 1.000 1.049*

1.774* 0.361 0.843 0.729 1.038

1.496* 0.399* 0.928 0.668* 0.785*

0.848 0.907 0.755 0.464 1.643

0.921 0.879 0.785 0.470 1.934

‡

‡

§ § § §

‡

‡

§ § § § § §

0.943 1.443 2.292 1.015 1.043 ‡

§ †

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

†

‡

‡

‡

‡

1.907* 1.221* 1.607* 0.763* 1.026

2.029* 1.216* 1.535* 0.709* 0.954

0.647* 0.906 1.497* 0.937 1.656*

1.782* 1.091 1.933* 0.846 1.121

0.352 0.962 0.414* 0.000 0.244*

0.364 0.796 0.392* 0.000 0.222*

§

‡

‡

‡

‡

‡

‡

§

§ § § §

0.316 0.998 0.737 0.000 0.181* ‡

*Significant at the Pⱕ.01 level. † Not included in the final model. ‡ Reference group. § Unavailable due to n⫽0 comparative group.

Arch Phys Med Rehabil Vol 92, May 2011