Readmission to Acute Care Hospital during Inpatient Rehabilitation for Traumatic Brain Injury

Abstract

Objective

To investigate frequency, reasons, and factors associated with readmission to acute care (RTAC) during inpatient rehabilitation for traumatic brain injury (TBI).

Design

Prospective observational cohort.

Setting

Inpatient rehabilitation.

Participants

2,130 consecutive admissions for TBI rehabilitation.

Interventions

Not applicable.

Main Outcome Measure(s)

RTAC incidence, RTAC causes, rehabilitation length of stay (RLOS), and rehabilitation discharge location.

Results

183 participants (9%) experienced RTAC for a total 210 episodes. 161 patients experienced 1 RTAC episode, 17 had 2, and 5 had 3. Mean days from rehabilitation admission to first RTAC was 22 days (SD 22). Mean duration in acute care during RTAC was 7 days (SD 8). 84 participants (46%) had >1 RTAC episode for medical reasons, 102 (56%) had >1 RTAC for surgical reasons, and RTAC reason was unknown for 6 (3%) participants. Most common surgical RTAC reasons were: neurosurgical (65%), pulmonary (9%), infection (5%), and orthopedic (5%); most common medical reasons were infection (26%), neurologic (23%), and cardiac (12%). Older age, history of coronary artery disease, history of congestive heart failure, acute care diagnosis of depression, craniotomy or craniectomy during acute care, and presence of dysphagia at rehabilitation admission predicted patients with RTAC. RTAC was less likely for patients with higher admission Functional Independence Measure Motor scores and education less than high school diploma. RTAC occurrence during rehabilitation was significantly associated with longer RLOS and smaller likelihood of discharge home.

Conclusion(s)

Approximately 9% of patients with TBI experience RTAC during inpatient rehabilitation for various medical and surgical reasons. This information may help inform interventions aimed at reducing interruptions in rehabilitation due to RTAC. RTACs were associated with longer RLOS and discharge to an institutional setting.

Individuals with traumatic brain injury (TBI) are at risk for a range of medical complications,^1-8 which may result in the need for readmission to acute care (RTAC) after the original acute hospital discharge to acute inpatient rehabilitation. Some rehopitalizations may be planned for elective procedures (such as for cranioplasty, orthopedic, or reconstructive surgery), while others may be unplanned and disrupt the intended course of acute inpatient rehabilitation. Some RTACs that occur soon after acute care discharge are considered “potentially preventable” and an indicator of substandard quality-of-care by some payer sources and quality monitoring organizations. ⁹ Identifying “potentially preventable” disorders that cause RTAC during acute rehabilitation and reducing these RTAC occurrences is of particular relevance under the Affordable Care Act (ACA). ¹⁰ In accordance with the ACA, the Centers for Medicare and Medicaid Services (CMS), a primary payer for rehabilitation services in the United States, has established a goal to reduce inappropriate re-admission rates while improving quality of care and safety. ¹¹

In a general inpatient rehabilitation sample (mixed diagnoses) from 1980-1986, Siegler et al. ¹² found 33% of rehabilitation patients developed acute complications and 44% of those (or 14.7% of the entire rehabilitation sample) were severe enough to warrant RTAC, with the most common conditions noted as surgical (23%), infection (17%), and thromboembolic events (16%). In a more recent study, Faulk et al. ¹³ (2009-2011) observed 10.9% RTAC with top causes including respiratory (27%), infections (22%), and cardiac (11%). Comorbid medical conditions, such as those that may lead to RTAC may impact rehabilitation outcome, ^14,15 length of stay, ^14,16 and treatment costs. ¹⁷

Published studies of RTAC rates and causes during TBI inpatient rehabilitation are sparse. Deshpande et al. ¹⁸ reported that 22 out of 100 TBI subjects returned to acute care at least once during inpatient rehabilitation. Of note, this single-site study reviewed cases in 1992-1994 when the average rehabilitation length of stay (RLOS) for individuals had only recently decreased by several months attributable to greater attention by insurance payers on RLOS. Each year from 2005-2008, the Uniform Data System for Medical Rehabilitation of inpatient brain injury rehabilitation observed a range of 10.1 – 10.9% discharges to acute care and 1.0 – 1.3% program interruptions. ¹⁹ Deshpande found statistically significant associations with a recent history of pneumonia or recent surgery during the acute care stay. Following TBI, the common medical complications noted in the acute care setting include sepsis, respiratory infections, hypertension, severe respiratory failure, acute kidney injury, diabetes, cardiac arrhythmias, fluid and electrolyte disorders, and extremity fractures, ^1-4 and during inpatient rehabilitation common disorders include hydrocephalus, seizures, paroxysmal autonomic dysfunction, ventricular dilatation, abnormal liver function, hypertension, thrombophlebitis, respiratory infections, heterotopic ossification, fractures, and pituitary-hypothalamic dysfunction, psychiatric and behavioral disturbances, and problems with eyes, ears, nose, and throat.^5-8,18,20

Among individuals with disordered consciousness receiving acute inpatient rehabilitation for severe TBI, Whyte et al. ²¹ found rehabilitation patients (who were deemed medically stable for a research trial) on average experienced a rate of 0.4 complications per week per patient with more than 80% experiencing at least 1 new medical complication during the 6 week observation period during inpatient rehabilitation. Newly documented medical complications included hypertonia, agitation/aggression, urinary tract infection, and sleep disturbance, and new complications that were considered severe included hydrocephalus, pneumonia, gastrointestinal problems, and paroxysmal sympathetic hyperactivity. In addition to potentially causing RTAC, these medical conditions can pose significant barriers to successful rehabilitation, as well as lead to increased RLOS, cost of care, and mortality. ^22,23

The potential costs, disrupted rehabilitation progress, and impact on outcome that may stem from RTAC, make it critical to identify and understand the medical factors, resources, and processes that contribute to RTAC following admission to inpatient rehabilitation. The relationship between medical complications and other factors leading to RTAC is complex with potential contribution from underlying TBI severity, acute medical complications, chronic medical conditions, functional status, and patient care processes. ^{12,13,17,18,24} The primary aim of this paper is to describe the medical complications that precipitate a RTAC during inpatient rehabilitation for patients with TBI by assessing the incidence of and reasons for RTACs and the factors associated with RTACs. A secondary aim is to evaluate the relationship between RTACs and outcome at rehabilitation discharge as measured by rehabilitation disposition and RLOS.

METHODS

This study is based on the prospective observational TBI-PBE cohort. The TBI-PBE Project was a five-year, multi-center investigation of the TBI inpatient rehabilitation process for 2,130 patients. ²⁵ The introductory paper to this series of articles about the TBI-PBE project describes the study design, including the PBE research methodology, inclusion criteria, data sources, and analysis plan. ²⁵ The study sites included 9 inpatient rehabilitation facilities in the United States and 1 in Canada. Institutional Review Board approval was obtained at each center. Participants were 14 years of age or older, gave (or their parent/guardian gave) informed consent, and were admitted to the facility's adult Brain Injury unit for initial rehabilitation following TBI. Trained personnel recorded collected data from patient and family interview and abstraction of inpatient rehabilitation medical records using standardized procedures and >95% accuracy.

Variables

Premorbid variables studied for association with RTAC included: gender, race, age, education level, employment status, drug and alcohol use, and cause of injury. It was noted whether the original (at the time of injury) acute medical care was provided at an in-system (associated with rehabilitation center) or an out-of-system hospital (not associated with the rehabilitation center). The Glasgow Coma Scale (GCS) after resuscitation in the Emergency Department, CT scan findings during the first seven days after injury, duration of posttraumatic amnesia (PTA), and time from injury to rehabilitation admission were used to describe the injury severity. Height and weight at the time of rehabilitation admission were collected to calculate Body Mass Index (BMI), and the World Health Organization's BMI category classifications were used. The Comprehensive Severity Index (CSI^®), the study's principal medical severity measure, was used to score the extent of deviation from normal physiological status for each diagnosis including each medical complication and comorbidity present during the first 3 days after rehabilitation admission.²⁶ Higher CSI scores denote increased medical severity. We also noted primary payer and several medical conditions or procedures that were present at the time of rehabilitation admission or had occurred during acute care prior to the initial rehabilitation admission, including craniotomy, craniectomy, hypertension, renal failure, diabetes, hypercholesterolemia, anxiety, depression, dysphagia, aphasia, ataxia, and paralysis. The Functional Independence Measure (FIM) was used to describe a patient's independence in motor and cognitive abilities at rehabilitation admission. All FIM data were Rasch-adjusted to 0 to 100 ratio level scores as described in prior publications.^25,27 The RLOS excluded days during which the patient was returned to acute care.

Definition of RTAC

All interruptions of inpatient rehabilitation requiring readmission to an acute care hospital were considered to represent an RTAC. For each such event, the dates of interruption and cause(s) were abstracted from the medical chart. In many instances, multiple reasons were listed along with the presenting signs, symptoms, and diagnoses recorded by the clinical team. In such cases, the authors reviewed the data and selected the one primary cause that best represented the reason for the RTAC episode. Hence, no RTAC episode is counted more than once and only one reason is represented even when multiple causes may have prompted the RTAC. The RTAC and supporting descriptions abstracted from the chart were reviewed using clinical judgment and experience. When several reasons were listed, the cause that would most likely require management in acute care was selected. In some cases we used the main sign or symptom of instability as the main reason (e.g., acute mental status change or seizure) In cases when specific diagnoses that precipitated the RTAC (e.g., infection or intracranial hemorrhage) were not identifiable in the available records, the main sign or symptom of instability was used as the main reason (e.g., acute mental status change or seizure). Since the unit of analysis is the patient, a patient with one or multiple returns to acute care was counted only once in the RTAC group. The RTAC causes were grouped into one of three broad categories of surgery, medical, or unknown, and then further subdivided.

Data analysis

Descriptive statistics were used to provide frequencies and percentages for categorical variables describing patients, treatments, and outcomes, and means, medians, quartiles, and SDs to summarize continuous measures. For discrete variables, we used the chi-square test to determine significance of associations. For continuous variables we used t-tests or analysis of variance (ANOVA). A two-sided p value <0.05 was considered statistically significant.

RTAC during inpatient rehabilitation was analyzed as to probable cause, associated factors, and relationship with rehabilitation discharge disposition and RLOS. Independent variables in the prediction of RTAC included demographic and premorbid characteristics, injury severity, medical conditions, and functional status. We explored models allowing sites to enter the models in addition to the other predictors. When data were missing, adjustments were made depending on the variable and its intended use. Sometimes we categorized values simply as “unknown” (and included the category in analysis as a dummy variable representing missingness); sometimes we excluded patients with missing data from analysis; and sometimes we collapsed continuous variables with missing data into categorical variables and placed the cases with missing information into a category using corroborating data available.

Logistic regression analyses were used for binary predictions of whether patients experienced an RTAC or if they were discharged home. Separate models predicting whether patients experienced a medical or surgical RTAC were also performed. Independent variables for prediction of discharged home and RLOS allowed the inclusion of the same variables above, as well as presence or absence of RTAC during rehabilitation. Ordinary least squares regression models were used to determine the influence of RTACs during acute inpatient rehabilitation on rehabilitation length of stay. We used stepwise selection with an entry and exit p value level of 0.05 to produce the most parsimonious models. Variables entering into regression models were checked for multicollinearity; all correlations were <0.67 in absolute value. We used c-statistics and a rescaled R²,²⁸ in logistic models, to indicate discrimination or variation explained in outcomes. In the prediction model for RLOS, data from the Canadian site were not included due to differences in the system of care, resulting in markedly different RLOS and that site not allowing some important predictor variables to be collected. Analyses were performed using SAS version 9.2 (SAS Institute, Inc., Cary, NC).

RESULTS

The demographic and injury characteristics are summarized for the sample as a whole and for each injury group defined by admission FIM Cognitive score in the first article in this series of manuscripts.²⁵ The sample was 73% male, 74% white, 37% married, 83% not obese (BMI <30), and 51% employed at the time of injury. Average age was 45 (SD 21) years. The most common cause of injury was vehicular accidents (56%), followed by falls or flying objects (32%), violent etiologies (7%), and sports (2%). Mean RLOS was 26.5 days (SD 19.9). The mean Rasch-adjusted FIM Motor score at admission was 33 (SD 19) and mean Rasch-adjusted FIM Cognitive score was 37 (SD 20). Mean time from injury to rehabilitation admission was 29.3 days (SD 34.3).

Table 1 provides the characteristics of the participants who did and did not have a RTAC during rehabilitation. After excluding 10 missing admission FIM scores, the remaining 2,120 study participants included 183 (9%) individuals who had 1 or more RTAC episodes for a total of 210 episodes. Most participants (161 or 161/183=88%) experienced only 1 RTAC, while 17 (9%) had 2 RTACs, and 5 (3%) had 3 RTACs. Mean days from rehabilitation admission to first RTAC was 22.0 days (SD 21.5). Mean duration in acute care during the RTAC was 7.0 days (SD 7.6). Eighty-four (46%) participants had at least 1 RTAC for medical reasons, 102 (56%) participants for surgical reasons, and in 6 participants (3%) the reason for RTAC was unknown. Nine participants had 1 RTAC episode classified as medical and a different RTAC episode classified as surgical, and thus, these participants were included in both the surgical and medical groups for logistic regression analyses.

Table 1.

Comparison of Patient and Injury Characteristics for patients with and without an RTAC event

Characteristics	RTAC N=183	No RTAC N=1947	P
Predictors present prior to injury
Age at rehabilitation admission (mean, SD)	51.5 (22.4)	43.8 (21.1)	<0.001†
Employment prior to injury (%)			<0.001*
Employed and student	2	4
Employed only	39	48
Student only	11	11
Unemployed	10	14
Retired	37	22
Student only	11	11
Unknown	1	1
Highest education achieved (%)			0.004*
Some high school, no diploma	15	24
High school diploma	25	26
Work towards or Associate's degree	16	16
Work towards or Bachelor's degree	19	19
Work towards or Master's/Doctoral degree	16	9
Unknown	8	5
History of alcohol abuse before injury (%)	40	45	0.180*
Male (%)	67	73	0.100*
Number of previous brain injuries (mean, SD)	0.1 (0.5)	0.1 (0.4)	0.311†
Payer (%)			0.008*
Medicare	29	19
Medicaid	14	16
Private insurance	25	24
Centralized (single payer system)	7	7
Worker's compensation	6	7
Self pay/None	1	5
MCO/HMO	10	15
No-fault auto insurance	7	4
Other/unknown	2	4
Race/Ethnicity (%)			0.138*
White	78	74
Black	15	15
White Hispanic	2	7
Other and unknown	4	4
Predictors present during acute care prior to first rehabilitation admission
Anxiety (%)	26	19	0.032*
Congestive heart failure (%)	10	3	<0.001*
Coronary artery disease (%)	18	8	<0.001*
Craniectomy (%)	19	6	<0.001*
Craniotomy (%)	36	19	<0.001*
Days from injury to rehabilitation admission (mean, SD)	35.6 (41.7)	28.7 (33.5)	0.030†
Depression	44	28	0.032*
Diabetes	27	15	<0.001*
Epidural hematoma (%)	7	8	0.481*
Facial fracture (%)	9	14	0.054*
GCS score immediately after injury or upon arrival in acute care (%)			0.006*
Mild (13-15)	13	15
Moderate (9-12)	4	8
Severe (3-8)	29	33
Intubated/sedated	9	13
Unknown	45	32
Hypercholesterolemia	27	15	<0.001*
Injury cause (%)			<0.001*
Fall	45	31
Sports	1	2
Motor vehicle crash	44	57
Violence	4	7
Miscellaneous	6	3
Intraventricular hemorrhage (%)	19	19	1.0*
Midline shift (%)			<0.001*
No midline shift	19	32
Any Midline shift	48	34
Unknown	33	34
Non-free standing rehabilitation facility (%)	30	31	0.363*
Paralysis (%)	53	37	<0.001*
Skull fracture (%)	30	27	0.384*
Subdural hematoma (%)	62	45	<0.001*
Predictors present at time of first rehabilitation admission
Body mass index (%)			0.517*
<16	1	2
16-<=18.5	10	8
>18.5-<=25	48	50
>25-<=30	25	24
>30-<=35	9	8
>35-<=40	2	2
>40	2	1
Unknown	3	6
Brain injury component of CSI score (mean, SD)	51.4 (24.9)	44.1 (23.5)	<0.001†
Non-brain injury component of CSI score (mean, SD)	20.8 (19.3)	16.6 (14.4)	<0.001†
Moderate to severe dysphagia (%)	70	52	<0.001*
FIM motor score - Rasch transformed (mean, SD)	24.4 (18.3)	34.0 (19.2)	<0.001†
FIM cognitive score - Rasch transformed (mean, SD)	30.5 (20.9)	37.9 (19.2)	<0.001†
FIM cognitive score category (%)			<0.001*
1 Score <=6	26	15
2 Score 7-10	20	17
3 Score 11-15	24	23
4 Score 16-20	17	19
5 Score >=21	14	25
Rehabilitation length of stay - excludes interruptions (mean, SD)	42.6 (32.3)	25.0 (17.6)	<.001†
Discharge disposition home (%)	60%	86%	<.001*

NOTE: Abbreviations: MCO/HMO, Managed care organization/Health maintainance organization; GCS, Glasgow Coma Scale score; CSI, Comprehensive Severity Index; FIM, Functional Independence Measure

^*Chi-Square analysis.

^†Two sample t test.

The primary causes for RTAC episodes are summarized in table 2 (please see the supplemental digital content data [provide link to SDC table here] for a complete breakdown of RTAC causes). Considering medical and surgical causes together, the most common RTAC reasons were neurosurgical (34%), infection (14%), neurologic (10%), pulmonary (9%), and cardiac (5%).

Table 2.

Primary causes for 210 RTAC episodes

RTAC Reason	Frequency (%)
Surgical	110 (52%)
Neurosurgical	65%
Pulmonary	9%
Orthopedic	5%
Dermatologic (wound)	5%
Infectious	5%
Otolaryngological	5%
Gastrointestinal	3%
Ophthalmologic	2%
Lymphatic	1%
Vascular	1%
Medical	94 (45%)
Infectious	26%
Neurological	23%
Cardiac	12%
Pulmonary	10%
Vascular	9%
Gastrointestinal	6%
Psychiatric	3%
Genitourinary	2%
Orthopedic	2%
Dermatologic (wound)	1%
Otolaryngological	1%
Electrolytes	1%
Tumor	1%
Unknown	3%
Unknown	6 (3)
Total	210*

^*Some patients had more than one RTAC episode.

For RTACs occurring due to medical issues (N=94), mean days from rehabilitation admission to first RTAC was 17.9 days (SD 20.6). Mean days from rehabilitation admission to first RTAC for surgical cause was 25.5 days (SD 21.8). A large variation in RTAC incidence was observed across facilities with a range of 4 to 18%. Two sites had an RTAC incidence of less than 5%, whereas 4 sites had an incidence greater than 10%.

Prediction of RTAC

Table 3 presents logistic regression models for the prediction of any type of RTAC event and RTAC for medical and surgical reasons separately. Any RTAC and medical RTAC had c statistics >0.74, which indicates acceptable discrimination, and surgical RTAC had c statistic >0.80, which indicates excellent discrimination. Any RTAC was predicted as more likely for patients with older age, history of coronary artery disease, history of congestive heart failure, acute care diagnosis of depression, craniotomy and craniectomy during acute care, and moderate/severe dysphagia present at the time of rehabilitation admission. RTAC for any reason was less likely with higher Rasch-adjusted admission FIM Motor score and education less than high school diploma. Medical RTAC was predicted as more likely by older age at the time of injury, mild GCS score, subdural hematoma on neuroimaging during the first seven days post-injury, and a diagnosis of diabetes or congestive heart failure. RTAC for medical reason was less likely with higher Rasch-adjusted admission FIM Motor score. Predictors of greater likelihood of surgical RTAC included: missing GCS score, craniectomy performed during acute care, skull fracture, average BMI >40, and a diagnosis during acute care or at rehabilitation admission of coronary artery disease, hypercholesterolemia, moderate-to-severe dysphagia, anxiety, and depression. RTAC for surgical reason was again less likely with higher Rasch-adjusted admission FIM Motor score and education less than high school diploma, as well as no midline shift on neuroimaging during the first seven days post-injury. Note that the significant predictors of medical and surgical RTACs do not overlap except for Rasch-adjusted admission FIM Motor score.

Table 3.

Prediction of outcome of RTAC based on patient and medical characteristics present on or before rehabilitation admission

Outcome:	Any RTAC*				Medical RTAC				Surgical RTAC
Predictor:	Parameter Estimate	Standardized Parameter Estimate	Odds Ratio (95% Confidence Interval)	P	Parameter Estimate	Standardized Parameter Estimate	Odds Ratio (95% Confidence Interval)	P	Parameter Estimate	Standardized Parameter Estimate	Odds Ratio (95% Confidence Interval)	P
Intercept	−3.27	0.00		<.001	−3.76	0.00		<.001	−4.03	0.00		<.001
Predictors present prior to injury
Age	0.02	0.14	1.01 (1.00,1.02)	0.006	0.01	0.16	1.01 (1.00,1.03)	0.017
Highest education achieved:
Some high school, no diploma	−0.54	−0.13	0.59 (0.38,0.91)	0.016					−0.67	−0.16	0.51 (0.28,0.92)	0.026
Predictors present during acute care prior to first rehabilitation admission
Anxiety									0.61	0.13	1.84 (1.09,3.10)	0.022
Coronary artery disease	0.62	0.10	1.85 (1.13,3.02)	0.014					0.75	0.12	2.12 (1.10,4.09)	0.025
Congestive heart failure	0.82	0.09	2.28 (1.23,4.23)	0.009	1.10	0.12	3.01 (1.48,6.15)	0.002
Craniectomy	1.36	0.19	3.89 (2.47,6.11)	<.001					1.72	0.25	5.61 (3.32,9.48)	<.001
Craniotomy	0.52	0.11	1.67 (1.18,2.37)	0.004
Depression	0.69	0.17	1.99 (1.44,2.75)	<.001					0.71	0.18	2.04 (1.26,3.30)	0.004
Diabetes					0.53	0.11	1.70 (1.04,2.80)	0.036
GCS score in field or arrival in acute care: Mild (13-15)					0.69	0.13	1.99 (1.13,3.51)	0.018
GCS score in field or arrival in acute care: Unknown									0.54	0.14	1.71 (1.09,2.68)	0.019
Hypercholesterolemia									0.58	0.12	1.79 (1.02,3.12)	0.042
Neuroimaging during first 7 days: No midline shift									−0.78	−0.20	0.46 (0.25,0.86)	0.015
Neuroimaging during first 7 days:									0.62	0.15	1.86 (1.17,2.96)	0.009
Skull fracture
Neuroimaging during first 7 days: Subdural hematoma					0.69	0.19	2.00 (1.24,3.22)	0.005
Predictors present at time of first rehabilitation admission
Body mass index >40									1.95	0.12	7.01 (2.15,22.86)	0.001
FIM motor score – Rasch adjusted	−0.02	−0.19	0.98 (0.97,0.99)	<.001	−0.03	−0.30	0.97 (0.96,0.98)	<.001	−0.02	−0.19	0.98 (0.98,1.00)	0.005
Moderate/severe dysphagia	0.525	0.14	1.69 (1.15,2.48)	0.008					0.75	0.21	2.12 (1.24,3.63)	0.006
Number of observations used		2120 (yes=183, no=1947)				2120 (yes=84, no=2036)				2120 (yes=102, no=2018)
c Statistic			.757				.743				.806
Maximum rescaled R2†			.148				.108				.200

Note: Abbreviations: GCS, Glasgow Coma Scale score

^*The unit of analysis is the patient. In all three columns, patients characterized as YES may have had 1 or more RTACs.

^†A scaled coefficient of determination (R²).

RTAC as Predictor of Outcomes

Statistically significant associations were found for patients with RTAC occurrence for both the outcomes of rehabilitation disposition and RLOS. See tables 4 and 5.

Table 4.

Predictors of Rehabilitation Length of stay

Outcome:	Rehabilitation length of stay
Predictors:	Parameter Estimate	Standardized Parameter Estimate	P Value
Intercept	35.95	0	<.001
Predictors present prior to injury
Employment prior to injury: Unknown	7.49	0.04	0.012
Highest education achieved: Work towards or completed Bachelor's degree	2.12	0.04	0.009
Highest education achieved: Work towards or completed Master's/Doctoral degree	2.51	0.04	0.024
History of alcohol use before injury	1.52	0.04	0.023
Payer: Medicare	−2.94	−0.06	<.001
Payer: Worker's Compensation	4.51	0.06	<.001
Race: Asian, other and unknown	5.69	0.05	0.003
Race: Black	2.19	0.04	0.014
Predictors present during acute care prior to first rehabilitation admission
Days from injury to rehabilitation admission	0.073	0.12	<.001
Depression	2.95	0.07	<.001
Neuroimaging during first 7 days: Intraventricular hemorrhage	1.89	0.04	0.025
Neuroimaging during first 7 days: Midline shift unknown	2.04	0.05	0.010
Neuroimaging during first 7 days: No midline shift	−2.13	−0.05	0.009
Paralysis	2.09	0.05	0.004
Predictors present at time of first rehabilitation admission
Body mass index 16-18.5	3.91	0.06	<.001
FIM cognitive score – Rasch adjusted	−0.18	−0.18	<.001
FIM motor score – Rasch adjusted	−0.38	−0.35	<.001
Predictors emerging during rehabilitation stay
Return to acute care during rehabilitation	11.09	0.17	<.001
# Observations Used		1971*
R-Square		0.444
Adjusted R-Square		0.438

Note: Abbreviations: FIM, Functional Independence Measure

^*149 patients from one site were removed due to not allowing some important predictor variables to be collected related to this model.

Table 5.

Predictors of Discharge Home

Outcome:	Discharged Home
Predictor:	Parameter Estimate	Standardized Parameter Estimate	Odds Ratio	P Value
Intercept	3.37	0.00	NA	<.001
Predictors present prior to injury
Age	−0.03	−0.36	0.97	<.001
Highest education achieved: Work towards or completed Master's/Doctoral degree	0.64	0.10	1.90	0.006
Number of previous brain injuries	−0.34	−0.08	0.71	0.009
Predictors present during acute care prior to first rehabilitation admission
Cause of injury: Motor vehicle crash	0.68	0.19	1.97	<.001
Days from injury to rehabilitation admission	−0.01	−0.08	1.00	0.023
Non-free standing rehabilitation facility	−0.89	−0.22	0.41	<.001
Neuroimaging during first 7 days: Epidural hemotoma	0.66	0.10	1.93	0.031
Neuroimaging during first 7 days: Facial Fracture	−0.40	−0.07	0.67	0.046
Neuroimaging during first 7 days: Midline shift: Unknown	−0.40	−0.10	0.67	0.010
Predictors present at time of first rehabilitation admission
Body mass index Unknown	−0.62	−0.08	0.53	0.030
CSI score Brain Injury component	−0.01	−0.18	0.99	<.001
FIM motor score - Rasch adjusted	0.03	0.32	1.03	<.001
Predictors emerging during rehabilitation stay
Return to acute care during rehabilitation	−1.14	−0.18	0.32	<.001
No. of observations used		2120 (yes=1777, no=343)
c statistic		.799
Maximum rescaled R2		.279

Note: Abbreviations: CSI, Comprehensive Severity Index; FIM, Functional Independence Measure;

DISCUSSION

Minimizing acute care re-admissions is both a federal mandate and a major clinical challenge for TBI inpatient rehabilitation providers. This prospective, ten-center, longitudinal study with over 2,000 patients fills a significant void in the TBI inpatient rehabilitation literature by providing empirical evidence on the incidence of, reasons for, and factors associated with acute care –readmissions occurring during inpatient rehabilitation. Acute care transfer was required for either medical complication or surgery (elective or unexpected) for 8.6% of the study participants during initial acute rehabilitation. RTACs were prompted for surgical reasons more often than medical reasons with the majority of surgical RTACs occurring for neurosurgical (primarily craniotomy, cranioplasty, and hydrocephalus), pulmonary, and orthopedic reasons. RTACs for medical reasons were often for infections (most commonly pneumonia and urosepsis) and neurologic reason (mostly mental status change and seizure).

The observation of 8.6% RTACs appear low compared to the sparse comparison data available in the literature. Our low rate of RTAC during acute brain injury rehabilitation may be due to several reasons. There is insufficient prior data for reliable comparison. The largely reduced RTAC rate compared to Deshpande, et al.¹⁸ may be, at least in part, an artifact of a substantially reduced RLOS since this 1992-1994 study, thereby reducing the number of days during which RTAC may occur during rehabilitation. It is also possible that more medical complications are being successfully identified and managed during inpatient rehabilitation or prior to admission to the rehabilitation unit.

Acute care facilities have already been reporting readmissions within 30 days of acute care discharge to CMS. Although there is limited evidence, some acute care RTAC reduction strategies for specifically targeted groups (e.g., patients with congestive heart failure) have shown a decrease in readmissions.^29,30 CMS has released “draft” criteria for a quality measure of all-cause RTAC 30 days post rehabilitation discharges prompting enhanced attention by many rehabilitation providers to the reduction of RTACs.^11,31 With a wide variety of RTAC causes, further RTAC reduction appears challenging and perhaps unlikely. This study identified more than 65 different reasons for RTAC with most occurring in low numbers (5 or less times out of 2120 patients). Instead of needing to address one or two single causes to substantially reduce the RTAC rate, the targets are diverse. Unfortunately, many of the predictors found in this study—age, obesity, history of coronary heart disease, craniectomies and craniotomies, midline shift, skull fracture and subdural hematoma— are not modifiable. In summary, these findings may argue against the prevailing perception that all acute rehabilitation hospitals have to do is practice better medicine to easily reduce these “preventable” problems. However, the RTAC causes can inform for RTAC reduction efforts and medical management in this patient population. The data point toward the need to for prevention and early recognition of the signs and symptoms of infection, and a high index of suspicion for conditions, such as tracheal stenosis and hydrocephalus. In the case of urinary tract infections, established strategies may reduce occurrence through timely removal of indwelling catheters.^31,32 It should be kept in mind that for many conditions (such as infection and seizure) only a portion of the total incidence of occurring during TBI rehabilitation would be expected to prompt RTAC and be represented by this study. The high cranioplasty rate, as well as the need for other neurosurgical and orthopedic procedures, highlights the potential relevance of early and on-going communication between surgeon and rehabilitation team regarding any pending surgical needs and timing or emerging complications. The large number of craniotomies during rehabilitation seems surprising. It is not known if these craniotomies represent emergent treatment of a new or worsening intracranial hemorrhage. With such a high number of RTACs for craniotomy combined with 2 cases of medical RTAC for intracranial hemorrhage, further study may be warranted to determine if there is a relationship with falls or anticoagulation causing new intracranial hemorrhage.

The wide range of medical and surgical complications reported in the literature and the present study may underscore the importance of these patients with brain injuries transitioning from acute care to acute inpatient brain injury rehabilitation prior to discharge to less provider-intensive levels of care (such as skilled nursing facility or home). However, this study is not able to directly determine this as the study only examined RTAC occurrence in the acute inpatient rehabilitation setting.

Acute rehabilitation occurs in a variety of settings (such as, units within hospitals and free-standing hospitals), care delivery patterns (such as, varied use of screening tests, number of physician rounding days), resources, staff expertise, availability of consultants, and proximity to and relationship with acute care facilities. These factors may impact the ability to evaluate and manage the medical complexity and instability commonly present after TBI, and influence whether or not a sick patient is treated in the rehabilitation setting versus transferred out to acute care. The RTAC incidence among facilities in the present study ranged of 4.5 to 17.8%. However, when site variables were allowed to enter the models to predict RTAC, either no site variables entered (for outcomes of medical RTAC and discharge home) or 1 or 2 sites entered (2 sites entered for any RTAC model and 1 site for surgical RTAC model). However, the c statistic only changed by <0.01, site effects were very small, and the other predictor variables were either the same or essentially the same. Since knowing what site a patient was treated in does not tell you anything about patient risk factors for RTACs or discharge home, we do not report the models when a few sites entered. This suggests that RTAC was driven more by patient characteristics, than resources, skill, and other site-specific issues. A comparison of the free standing vs non-free standing inpatient rehabilitation facilities and system vs non-system rehabilitation facilities in our study likewise did not prove to be statistically significant predictors of RTAC.

For the model predicting LOS, 5 sites entered significantly when allowed, 2 with negative and 3 with positive coefficients. However, the R² increased by only 3%, from 44% to 47%. Again, most of the patient risk factor predictors were the same with and without sites being allowed to enter. Sites were expected to have more of an impact on RLOS than for the other outcomes, since RLOS in rehabilitation for a patient with TBI can be influenced by many factors beyond the patient risk factors, such as the resources available in rehabilitation sites.

Rehabilitation facilities are able to benchmark their data against that of other rehabilitation facilities within the region and United States through outcome databases such as the Uniform Data Systems for Medical Rehabilitation, eRehabData, or Exchanged Quality Data for Rehabilitation.^33-35 Based on the findings of this study, risk stratification by age and severity would be important to set realistic, empirically-based expectations for rehabilitation facilities. Future large sample research could then examine within stratification groups for differences in RTAC rates to identify best practices.

This appears to be the first study published on the relationship of RTAC during brain injury rehabilitation to outcomes. RTAC occurrence during rehabilitation was associated with longer RLOS and failure to discharge to home, even controlling for other variables such as injury, medical, and patient factors. This relationship may represent the impact of medical conditions and stability on the achievement of functional milestones for discharge readiness.

Future Research Directions

Research is needed to understand if RTACs are preventable, which ones are preventable, and how they may be prevented. The findings of this study may inform future research on potential surveillance, prevention, and management strategies aimed at increasing medical stability and reducing RTAC occurrence, and possibly improving outcomes. Such strategies would then require rigorous study with formal quality improvement or research to discern which efforts are most effective and efficient in reducing RTACs in clinical practice.

Study Limitations

The TBI-PBE Project took place at specialized brain injury rehabilitation units. Thus, findings of this study may not be generalizable to all inpatient rehabilitation facilities treating TBI. The acute care hospital medical records, either for the stay prior to the first rehabilitation admission or for the RTACs, were not available (other than summaries included in the rehabilitation record), which would have contributed to a more complete assessment of the reasons for RTACs and their potential avoidability. Non availability of records also contributed to an “unknown” categorical dummy variable entering regression models in a few instances, like Glasgow Coma Scale score in field or upon arrival in acute care or employment status prior to injury. Unknown categorical dummy variables can mean several different things, such as a patient could not communicate or family or friends were not available to provide information. Also, some of our study sites were not permitted to receive detailed information about the patient from the patient's acute care setting. Determining the exact reason for medical instability is often difficult as there may be multiple, co-occurring issues at play. In order to avoid counting an episode more than once, clinical judgment of the physician researchers was applied to the selection of the primary reason. For some RTACs only the main sign or symptom of instability was identifiable in the records, and thus, used as the RTAC cause (e.g., acute mental status change or seizure) when other diagnoses may have precipitated the RTAC (e.g., infection or intracranial hemorrhage). Thus, the reasons for return to acute care may under-represent some of the medical problems that contributed to the need for RTAC. In this study we were not able to determine if the surgical RTACs were scheduled versus unscheduled. Also, site specific information (such as in-house acute care resource availability, transfer policies, and acute care relationships) that might explain a portion of the differences in RTAC incidence among the 9 sites was not obtained. However, the regression models found that site differences had no or only a very small effect on medical and surgical RTAC occurrence.

CONCLUSION

Approximately 9% of TBI patients experienced RTAC during the course of rehabilitation for a variety of medical and surgical reasons. RTACs were associated with longer RLOS and non-home discharge. The details provided in this study may help inform interventions aimed at reducing interruptions in rehabilitation for RTAC. The large number of low frequency complications will make further RTAC reduction challenging. Prevention protocols and timely response to complications in patients with predictable risk for RTAC should be developed, and will require rigorous assessment of effectiveness. Patient characteristics, complexity, and critical aspects of care identified in this study may provide insight into developing such protocols. The findings also point to the high medical complexity that accompanies TBI and underscores the value of inpatient rehabilitation with specialized, interdisciplinary brain injury care.

Abbreviations

ACA

Affordable Care Act

BMI

Body mass index

CMS

Centers for Medicare and Medicaid Services

CSI

Comprehensive Severity Index

FIM

Functional Independence Measure

GI

Gastrointestinal

GCS

Glasgow Coma Scale

PTA

Post-traumatic amnesia

RLOS

Rehabilitation length of stay

RTAC

Readmission to acute care

TBI

Traumatic Brain Injury