Frequency and reasons for unplanned transfer to the primary acute care service of inpatient rehabilitation glioblastoma multiforme patients

To determine the percentage of and factors associated with unplanned transfer to the acute care service of glioblastoma multiforme acute rehabilitation inpatients. Retrospective review of glioblastoma multiforme acute rehabilitation inpatients admitted 4/1/2016–3/31/2020 at a National Cancer Institute Comprehensive Cancer Center. One hundred thirty-nine consecutive admissions of unique glioblastoma multiforme acute rehabilitation inpatients were analyzed. Fifteen patients (10.7%, 95% confidence interval 6.5–17.1%) were transferred to the acute care service for unplanned reasons. The most common reasons for transfer back were neurosurgical complication 6/15(40%), neurologic decline due to mass effect 4/15(26.7%), and pulmonary embolism 2/15(13.3%). Older age (p = 0.010), infection prior to acute inpatient rehabilitation transfer (p = 0.020), and lower activity measure of post-acute care 6-click basic mobility scores (p = 0.048) were significantly associated with transfer to the acute care service. Patients who transferred to the acute care service had significantly lower overall survival than patients who did not transfer off (log-rank test p = 0.001). Acute inpatient physiatrists should closely monitor patients for neurosurgical and neurologic complications. The variables significantly associated with transfer to the acute care service may help identify patients at increased risk for medical complications who may require closer observation.


Introduction
Of the estimated 17,000 primary brain tumors diagnosed in the USA each year, approximately 60% are World Health Organization (WHO) grade IV astrocytomas (otherwise known as glioblastoma multiforme, glioblastoma, or GBM) [1,2]. They are the most common primary brain tumor in adults and one of the deadliest primary cancers with a median survival of 12-15 months with treatment [3,4].
Brain tumor patients often have impairments such as hemiplegia, aphasia, dysphagia, cognitive deficits, cancer related fatigue, and generalized deconditioning that may require neurologic acute inpatient rehabilitation. Previous studies have demonstrated statistically significant functional improvements in acute inpatient rehabilitation (IRF) of brain tumor patients with similar functional independence measure (FIM) efficiencies to stroke and traumatic brain injury (TBI) populations [5][6][7][8]. Neurologic cancer patients account for 53% of all American acute inpatient rehabilitation cancer patients [9]. This may be in part because they are Medicare 60% rule compliant [10] and the established neuro-rehabilitation infrastructure and experience in IRFs. Given the aggressiveness of GBM relative to other brain tumor types, GBM patients are commonly found on acute inpatient rehabilitation units, and focused research examining risk factors for transfer to the acute care service of GBM brain tumor IRF patients would be clinically useful.
Multiple researchers have explored the topic of risk factors for transfer to the acute care service of general cancer patients from IRF's (reported between 17 and 35%) [11][12][13][14][15]. Additional research on the transfer rate from IRF to the acute care service for specific primary cancer types has also been published. This research has focused mostly on hematologic malignancies who are among the most medically fragile [16][17][18][19]. Studies regarding the transfer to acute care service rates of stroke (reported between 7.6 and 11.1%) [20][21][22][23] and TBI (reported between 9 and 22%) [24,25] have also been studied. One recent study has focused on the transfer to acute care service of brain tumor acute rehabilitation inpatients (reported rate of transfer of 20.2%) with a cohort of various primary and metastatic brain tumors. Prior chemotherapy and steroid use were associated with transfer back [24]. Limitations of this study include the diverse primary and metastatic brain tumors of the analyzed cohort and that it was performed at a single site free-standing IRF. Brain tumors can arise from a variety of cell type origins with varying degrees of aggressiveness (from benign to malignant). In other brain tumors studies (again not evaluating GBM's exclusively) where transfer to the acute care service was not the primary outcome, rates of transfer between 7.5 and 24% have been reported [7,26,27]. To our knowledge, there are no previous publications that have focused on the transfer rate to the acute care service of solely GBM patients or of brain tumors patients on an IRF within a general hospital.
It is challenging to extrapolate risk factors for GBM patients based on the existing research on the risk of transfer to the acute care service of other brain injury and diverse primary cancer populations. While glioblastoma patients have a poor survival, their rate of transfer to the acute care service anecdotally appears lower than other cancer populations with poor survival. However, glioblastoma patients do require transfer from acute inpatient rehabilitation for serious and life-threatening issues.
The primary objectives of this study were to determine the rate of and variables associated with unplanned transfer to the acute care service of GBM acute rehabilitation inpatients. A secondary objective was to explore factors at inpatient rehabilitation admit that are associated with decreased survival.

Methods
One hundred fifty-six consecutive GBM admissions to the acute inpatient rehabilitation unit at a National Cancer Institute (NCI) Comprehensive Cancer Center acute inpatient rehabilitation unit from were retrospectively reviewed. Inclusion admission criteria were age 18 and older, having pathology confirmed WHO grade IV glioblastoma multiforme brain tumor, and having been admitted to the inpatient rehabilitation unit between April 1, 2016, and March 31, 2020. Exclusion admission criteria for our study admissions included more than one admission to our acute inpatient rehabilitation unit (13 patients were admitted twice, and 1 patient was admitted 5 times during this period. Only first admissions to the IRF per patient were reviewed). Additionally, because the focus of this study was to evaluate unplanned transfers, planned transfers were excluded. One patient who was transferred to the acute care service for a planned surgical resection was excluded. The remaining 139 unique patient admissions were analyzed.
Approval for retrospective data collection was obtained from our institutional review board (IRB). Due to the retrospective nature of the study, the IRB granted a waiver of informed consent and authorization in compliance with federal and institutional guidelines. Data was collected by two board-certified physiatrists. Relevant medical records data were recorded and organized into five categories: demographic information, glioblastoma characteristics, medications, laboratory values, and hospital admission characteristics. Demographic information included age, race, sex, marital status, payer source, and date of death (if applicable) or date of last contact (if no date of death available). Glioblastoma multiforme characteristics included time since diagnosis, brain lobe involvement, number of previous neurosurgical procedures, dates of radiation treatments, if the patient had evidence of recurrence, and if the patient had previously been diagnosed with a WHO lower grade astrocytoma. Rehabilitation data included neurologic impairments noted, discharge disposition, rehabilitation admit activity measure of post-acute care (AMPAC) 6 clicks basic mobility (version 2) scores (a 6 item physical therapy reported measure of inpatient mobility) [28,29], and rehabilitation admit FIM scores (a therapist scored instrument from 1, total assistance, to 7, independent) [30]. for physical therapy transfers and mobility as well as rehabilitation admit arm and leg manual muscle strength testing (MMT) [31] scores. The lowest MMT scores in the affected arm and leg were used for analysis. Medication data included the presence of antimicrobials, anticoagulants, seizure medications, types and dates of chemotherapy treatments, and dexamethasone dosage per day at inpatient rehabilitation admission. Laboratory values consisted of peripheral white blood cell and platelet counts as well as serum sodium, creatinine, and prealbumin on the day of transfer to inpatient rehabilitation. Hospital admission characteristics included the patient's length of hospitalization prior to transfer to rehabilitation, length of inpatient rehabilitation stay, whether the patient had a documented infection on the acute care service prior to rehab transfer, if the patient had transferred to the acute care service, reason for transfer to the acute care service, co-morbidity information at hospital admission, and if the patient had an indwelling Foley catheter at the time of admission to rehabilitation.
Data was extracted from our institutional Epic electronic medical record system and recorded into a Microsoft Excel spreadsheet. Data collected for this study was summarized using descriptive statistics: frequency and proportion for categorical variables and mean, standard deviation, median, and range for continuous variables. 95% Agresti-Coull confidence intervals were used to estimate the proportion of patients transferring to the acute care service. Chi-squared test or Fisher's exact test was applied to assess associations between discrete variables and transfer to the acute care service, as appropriate. For continuous variables, differences were assessed by two-sided two-sample t-tests with the exception of days from diagnosis, which was assessed by the Mann-Whitney test due to skewness. Overall survival from inpatient rehabilitation admission was summarized by Kaplan-Meier methods with relation to transfer to the acute care service status. Differences with relation to transfer to the acute care status were assessed by the log-rank test. A 95% level of statistical significance was assumed in all statistical testing. Statistical analyses were performed using R statistical software (R Core Team, 2021, version 4.1.2). [32] Additionally, overall survival from inpatient rehabilitation was modeled using parametric survival model (accelerated failure time, with log-logistic distribution selected as optimal by selection among Weibull, exponential, Gaussian, logistic, log-normal, and log-logistic distributions based upon Akaike information criteria and residual fit to the distribution). A Cox proportional hazards model was also considered but was not feasible due to violation of the proportionality of hazards assumption. Exhaustive variable selection based upon the Akaike information criteria was used to determine the optimal combination of discrete and continuous variables for the parametric survival model; inclusion of two-way interactions with transfer to the acute care service was also considered, but all worsened the model. The optimal model included the covariates: infection prior to transfer status (no/yes), previously lower grade (no/yes), hand dominance (ambidextrous, left, right), radiation before inpatient rehabilitation (no/yes), cardiac co-morbidity (no/yes), pulmonary co-morbidity (no/ yes), endocrine co-morbidity (no/yes), neurologic co-morbidity (no/yes), rheum/ortho co-morbidity (no/yes), genitourinary/gynecologic co-morbidity (no/yes), total number of organ systems involved in co-morbidities (continuous), age (continuous), admit transfer FIM (continuous), admit creatinine (continuous), and other admit platelet count (continuous). Tukey-adjusted p values were reported for the nonbinary discrete variables such as hand dominance.

Results
15/139 of GBM patients transferred to the acute care service from IRF, corresponding to a rate of 10.7% with 95% confidence interval spanning 6.5-17.1%. 107/139 (77.0%) were discharged directly home from IRF, 12/139 (8.7%) were transferred to a skilled nursing facility, and 5/139 (3.6%) were transferred to hospice. 5/15 (33%%) were later transferred back to the acute inpatient rehabilitation service after the acute care issues necessitating transfer were resolved. Of the remaining 10 patients who transferred to the acute care service but did not return to acute inpatient rehab, discharge destinations included 4 to home, 3 to hospice, 2 to a skilled nursing facility, and 1 to an outside acute care hospital.
Of the 10.7% who transferred to the acute care service, reasons for transfer included neurosurgical complication requiring transfer to the neurosurgery service 6/15 (40%) (3/6 intracranial bleeding, 1/6 cerebrospinal fluid leak, 1/6 hydrocephalus, and 1 craniotomy wound infection), neurologic decline due to mass effect 4/15 (26.7%) (all 4 later improved neurologically with aggressive corticosteroids), pulmonary embolism 2/15 (13.3%), aspiration pneumonia 1/15 (6.7%), abdominal pain 1/15 (6.7%), and transfer to the acute palliative care unit for end of life symptom management 1/15 (6.7%) ( Table 1). Table 2 displays the results of transfer to the acute care service analysis for demographics. The median cohort age was 58. Most patients were male (82/139), white (107/139), married (113/139), and had private insurance (72/139). Table 3 displays the results of transfer to the acute care service analyses for neurologic, rehabilitation, and tumor variables. 20/139 were previously diagnosed with a lower grade astrocytoma (11 with grade III and 9 with grade II astrocytoma) before their diagnosis of grade IV glioblastoma. Most were left hemiplegic (66/139 vs 52/139 right hemiplegic). The mean arm and leg manual muscle testing weakest scores were 3.0 and 2.5, respectively. The mean transfer and mobility admit FIM scores were 3.6 and 3.0, respectively. The mean AMPAC score improved from 14.0 at admit to 15.7 at discharge. Table 4 displays the results of transfer to the acute care service analyses for medical and hospitalization variables. The mean dose of dexamethasone at inpatient rehab admit was 10.3 mg (standard deviation = 6.5 mg). Approximately half (67/139) of patients were on anticoagulation, and most (119/139) were on antiepileptics at acute inpatient rehabilitation admit. The mean days on the acute care service prior to acute inpatient rehabilitation were 8.1 days. Infection prior to transfer was significantly associated with transfer to the acute care service (Fisher's exact test p = 0.020). Infection prior to inpatient rehabilitation was present in 46.7% of patients who did transfer to the acute care service (versus 18.5% who did not transfer). Age was significantly associated with transfer to the acute care service (t-test p = 0.010). Patients who transferred to the acute care service averaged 7 years older than those who did not (63.9 versus 56.5 years). The AMPAC 6 clicks basic mobility admission scores were significantly associated with transfer to the acute care service (t-test p = 0.048). Patients who transferred to the acute care service averaged 2.1 units lower in AMPAC score than patients who did not (12.1 versus 14.2). Other variables showed no significant evidence of association with transfer to the acute care service.
Patients who transferred to the acute care service had significantly lower overall survival than patients who did not transfer off (log-rank test p = 0.001). Figure 1 summarizes the Kaplan-Meier survival analysis. The median survival of patients who transferred to the acute care service was 0.3 versus 0.7 years for patients who did not. Table 5 presents results of the parametric survival model. 25/139 patients were alive as of 6/1/2021. For those who were deceased, the average time from acute inpatient rehab admission until death was 259.7 days (standard deviation = 241.7 days). The survival model showed significant evidence of association between survival and previously lower grade status (p = 0.047, where survival time of patients with previously lower grade was 1.53 times that of those without), radiation before acute inpatient rehabilitation (p = 0.001, where survival time of patients with radiation before inpatient rehabilitation was 0.58 times that of those who had not received prior radiation), a rheumatologic or orthopedic co-morbidity (p = 0.038, where survival time of patients with rheumatologic/orthopedic co-morbidity was 1.70 times that of those without), age (p = 0.023, where survival time declined by 1% per year as age increased), and lower admit transfer FIM score (p = 0.003, where survival time increased by 20% for each unit increase of transfer FIM).

Discussion
Lower functional status (measured from the AMPAC 6 clicks score or admit FIM transfer score) and older age were associated with increased likelihood of transfer to the acute care service and poorer survival of glioblastoma patients. Previous infection on the acute care service before inpatient rehabilitation was also associated with increased rate of transfer to the acute care service.
The most common reasons for transfer to the acute care service were neurosurgical complication 6/15 (40%) and neurologic decline due to mass effect 4/15 (26.7%). Inpatient rehabilitation physiatrists should monitor for neurologic and incisional changes as well as maintain contact with the neurosurgery and neuro-oncology teams during acute inpatient rehabilitation. Patients who have factors associated with increased risk of transfer to the acute care service may benefit from daily (7 times/week) physician visits instead of the minimum number of IRF physician visits (5 times/week).

Rate of transfer to acute care
The rate of unplanned transfer to primary of our 139 GBM patients was 10.7%. Although on the lower end, this is within the reported range (7.5-24%) in previous studies of transfer to the acute care service of general cohorts of brain tumor patients [6,26,33]. Knowlton et al. recently published regarding the transfer to the acute care service of brain tumors (including metastatic and primary tumors) at a single free-standing academic acute inpatient rehabilitation facility. In that study, 41% (85/208) were primary GBM tumors, with a transfer to primary rate of 20% for GBM [33]. Our acute inpatient rehabilitation program is located within the larger hospital and on the same unit as the neurosurgery and neuro-oncology services. Our lower transfer to primary rate may indicate some value in this arrangement to reduce transfer off. While these services do not routinely follow patients during inpatient rehabilitation service, their close proximity facilitates communication and enables these services to evaluate our patients more easily and potentially address concerning issues (through "curbside" discussion or in-person evaluation) before necessitating transfer off of the rehabilitation service. Because of the shared unit arrangement, our nursing staff is also experienced in caring for neurosurgical and neuro-oncology patients which may make them more vigilant in identifying potential complications earlier. Future case-controlled studies comparing acute inpatient rehabilitation rates of transfer between free-standing rehabilitation hospitals, programs within general hospitals, and on the same unit as neurosurgery and neurology services would be useful.

Factors associated with transfer to the acute care service
Infection on the acute care service before inpatient rehabilitation was found to be associated with eventual transfer to the acute care service (p = 0.020). This relationship may be indicative of a more complicated acute care hospitalization. Infection has been found to be the most common reason for transfer to the acute care service of cancer acute rehabilitation inpatients in multiple studies [11,[15][16][17][18][19]. However, only 2/15 (13.3%) patients in our study transferred to the acute care service due to infection. One was a wound infection (categorized under neurosurgical complications), and the other was an aspiration pneumonia.
Older age and lower performance status have consistently been found to be associated with a poor prognosis in GBM patients [34][35][36]. Our study found similar findings. With respect to function, several other factors were near statistically significant associated with transfer to the acute care service including admit FIM transfer scores (p = 0.054), admit FIM mobility scores (p = 0.057), and the presence of a Foley catheter (p = 0.08). Foley catheters are often present in low-functioning patients due to the difficulty with transfers and mobility to the bathroom or commode. Previous studies of acute inpatient  [20][21][22][23]. In our study, patients who transferred to the acute care service were found to have statistically significant shorter survival than patients who did not. 3/15 (20%) of the patients who transferred to the acute care service were discharged from the hospital to hospice. A similar finding was found in a study of transfer to the acute care service of multiple myeloma patients by Fu et al. [19]

Study limitations
Study limitations include the retrospective design and the limited number of patients. Karnofsky performance scale and Eastern Cooperative Oncology Group are not collected at our acute inpatient rehabilitation unit and thus were unavailable for analysis. Another limitation that can  There was no evidence of association between survival and RTP after controlling for other covariates limit the generalizability of these results is that it was performed at a single site acute inpatient rehabilitation unit on the neuro-oncology/neurosurgery floor within a prospective payment system exempt NCI Comprehensive Cancer Center.

Conclusion
The rate of transfer to the acute care service of GBM acute rehabilitation inpatients was 10.7% and is within the reported ranges of other brain neurorehabilitation populations. Infection prior to IRF transfer, older age, and lower function (measured from the AMPAC 6 clicks basic mobility) was found to have significantly associated with transfer to the acute care service. Acute inpatient rehabilitation physiatrists should maintain communication with referring neurosurgical and neuro-oncology services and monitor closely for neurosurgical and neurologic complications.