Article Text

Original research
Discharge disposition to skilled nursing facility after endovascular reperfusion therapy predicts a poor prognosis
  1. Samir R Belagaje1,2,
  2. Chung-Huan J Sun1,
  3. Raul G Nogueira1,2,
  4. Brenda A Glenn3,
  5. Lisa Ann Wuermser4,
  6. Vishal Patel1,2,
  7. Michael R Frankel1,2,
  8. Aaron M Anderson1,2,
  9. Tommy T Thomas1,2,
  10. Christopher M Horn1,2,
  11. Rishi Gupta3
  1. 1Department of Neurology, Emory University School of Medicine, Atlanta, Georgia, USA
  2. 2Marcus Stroke and Neuroscience Center, Grady Memorial Hospital, Atlanta, Georgia, USA
  3. 3Wellstar Neurosurgery, Wellstar Health Systems, Marietta, Georgia, USA
  4. 4Department of Rehabilitation Medicine, Emory University School of Medicine, Atlanta, Georgia, USA
  1. Correspondence to Dr R Gupta, Wellstar Neurosurgery, Wellstar Health Systems, 61 Whitcher Street, Suite 3110, Marietta, GA 30060, USA; rishi.gupta{at}


Objective We explore the impact of discharge disposition (independent rehabilitation facility (IRF) vs skilled nursing facility (SNF)) on 90 day outcomes in persons with stroke who received acute endovascular treatment.

Methods Using a database from a single primary care stroke center, discharge disposition, National Institutes of Health Stroke Scale (NIHSS), Totaled Health Risks in Vascular Events (THRIVE), Houston Intra-Arterial Therapy 2 (HIAT-2), and Acute Physiology and Chronic Health Evaluation (APACHE II) scores, and successful reperfusion were obtained. Univariate analysis was performed to assess predictors of good clinical outcome, as defined by 90 day modified Rankin Scale (mRS) scores ≤2. A binary logistic regression model was used to determine the impact of placement to an IRF versus an SNF on clinical outcomes.

Results 147 subjects were included in the analysis with a mean age of63±14 years and median NIHSS of 18 (IQR 14–21). Final infarct volumes, and modified APACHE II, THRIVE, and HIAT-2 scores were similar between those discharged to an IRF and those discharged to an SNF.However, their 90 day outcomes were significantly different, with far fewer patients at SNFs achieving good clinical outcomes (25% vs 46%; p=0.023). Disposition to SNF was significantly associated with a lower probability of achieving an mRS score of 0–2 at 90 days (OR = 0.337 (95% CI 0.12 to 0.94); p<0.04).

Conclusions Subjects discharged to SNFs and IRFs after thrombectomy have similar medical and neurological severity at admission and similar final infarct volumes at discharge. Despite these similarities, patients discharged to an SNF had a significantly lower probability of achieving a good neurological outcome. These results have implications for future acute stroke trial design.

  • Stroke
  • Thrombectomy

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


There exist a number of different variables along the entire spectrum of stroke which impact optimization of stroke outcomes. For instance, one of the best examples of such a variable is administration of intravenous tissue plasminogen activator (IV tPA) in a person with stroke who meets appropriate inclusion/exclusion criteria in the acute setting.1 ,2 The impact of intra-arterial therapy (IAT) on clinical outcomes is unclear,3–7 but successful reperfusion has been linked to a reduction in final infarct volume.8 Treatment effect of endovascular reperfusion may be diluted by subsequent care and discharge disposition but has not been examined systematically in previous trials.

This observation points to another variable in the stroke spectrum—namely, type of discharge disposition. Previous studies have demonstrated that patients’ post-stroke outcomes can vary depending on the type and nature of post-stroke rehabilitation they receive. In particular, these studies focused on the outcomes following post-hospital discharge to an independent rehabilitation facility (IRF) compared with skilled nursing facilities (SNF) and consistently, patients who went to an IRF had improved outcomes compared with an SNF.9–11 However, many of these studies were undertaken at a time when acute stroke treatment was not widely available or implemented.Furthermore, in the current US healthcare system, patients without insurance or family support may not be eligible for inpatient rehabilitation services despite meeting criteria from a physical or medical aspect.

In this analysis,we explore the impact of discharge disposition on the eventual 90 day post-stroke outcomes in persons with stroke who receive IAT. In particular, we hypothesize that patients discharged to an IRF will have better outcomes than those discharged to an SNF, even if they have similar characteristics and stroke severity.


We reviewed a prospectively maintained database of 221 consecutive patients treated with endovascular reperfusion therapy for acute ischemic stroke from October 1, 2010 to May 30, 2012. Data pertaining to patient demographics, baseline laboratory values, neurological assessments, radiographic imaging, time to reperfusion, and medical comorbidities were collected. Additionally, patient disposition at the time of discharge was collated into four possible scenarios: acute rehabilitation (AR), SNF, home, or hospice/death. Information on length of stay, requirement of a gastrostomy tube, and requirement of a tracheostomy was also captured to control for hospital complications.

The probability of a good clinical outcome was calculated using the Totaled Health Risks in Vascular Events (THRIVE)12 and Houston IAT (HIAT-2) scales.13 ,14 These scales were also used to assess the sensitivity of each scale for patients discharged to an SNF or AR. The modified Acute Physiology and Chronic Health Evaluation (APACHE II) score was tabulated for each patient to account for acute medical comorbidities and to control for such comorbidities in a multivariate analysis.15 Age was analyzed as a separate predictor of patient outcome, and subsequently excluded from the calculations of the modified APACHE II scores. Final infarct volumes were collected as previously described.8 ,16

Reperfusion success was denoted as achieving a Thrombolysis in Cerebral Infarction score of 2B or 3, which corresponded to reperfusion of more than two-thirds of the vascular territory.17 All patients with parenchymal hematomas within the first 36 h post-procedure were classified as having symptomatic hemorrhages whereas hemorrhagic infarctions were deemed asymptomatic according to the guidelines of the European Cooperative Acute Stroke Study (ECASS).18 The Alberta Stroke Program Early CT Scores (ASPECTS) were evaluated from CT images obtained at our institution for all patients prior to endovascular therapy. All ASPECTS >7 were defined as ‘favorable’ in the analysis.19 Patients that went directly to interventional radiology from outside hospitals did not have ASPECTS assessed. Lastly, data regarding the side of ischemic stroke burden as well as patient insurance status were analyzed as possible determinants of patient outcome. Ninety day modified Rankin Scale (mRS) scores were adjudicated at 3 month follow-ups by an independent certified examiner blinded to the initial National Institutes of Health Stroke Scale (NIHSS) scores and recanalization results. A ‘good outcome’ was defined as mRS score of 0–2 and a ‘poor outcome’ as mRS score of 3–6.

Statistical analysis

A univariate analysis was conducted comparing baseline characteristics and 90 day outcomes between patients discharged to home, AR, or SNF. Only anterior circulation strokes were used in the analysis. All neurological assessments and medical scoring systems (ie, NIHSS, THRIVE, HIAT-2, APACHE II, and ASPECTS) were compared using the Mann–Whitney U or Kruskal–Wallis test, as appropriate. Student t tests, ANOVAs, and χ2 tests were calculated for all continuous and categorical variables. A secondary univariate analysis was performed to identify predictors of good outcome among patients discharged to AR or SNF. All variables with p<0.15 on univariate analysis were included in a binary logistic regression model, with odds ratios and 95% CIs reported. Previously established predictors of outcome such as age and NIHSS were entered into the model.20 ,21 Modified APACHE II scores were also included to control for possible differences in medical comorbidities at the time of discharge. A Pearson correlation matrix was used to identify variables for exclusion from the multivariate model, based on high degrees of correlation (r>0.4, p<0.01) with other independent predictors of outcome.


A total of 221 consecutive patients receiving endovascular therapy for acute large vessel occlusions were reviewed for this study. Seventy-four patients (33.4%) were excluded from the analysis on the basis of: 51 expired or hospice bound patients at the time of discharge (23%); 18 posterior infarct patients (8%); four incomplete disposition/mRS records (2%); and one federal penitentiary patient. Of the 147 patients analyzed, 39 (26.5%) were discharged home, 48 (32.7%) were admitted to AR, and 60 (40.8%) were placed in SNF. Mean age and median NIHSS of the study cohort were 63±14 years and 18 (IQR 14–21), respectively.

Table 1 summarizes the univariate analysis of patients stratified by discharge disposition with regards to their demographic, radiographic, and angiographic characteristics, as well as final outcomes. When comparing patients discharged to SNFs compared with AR, there were no significant differences with respect to age (66±13 vs 63±17 years; p=0.218), modified APACHE II scores (8 vs 7; p=0.17), or predicted THRIVE and HIAT-2 scores (5 vs 4, p=0.182; and 4 vs 4, p=0.468, respectively). Admission NIHSS was significantly higher among patients discharged to SNF (20 vs 17; p=0.001), yet the postprocedural infarct volumes remained the same between the two cohorts (33 cm3 vs 43 cm3; p=0.675).

Table 1

Baseline characteristics of patients discharged to acute rehabilitation or skilled nursing facilities after endovascular therapy for ischemic stroke

Despite the relative similarities in medical and neurological severity at discharge between patients admitted to SNFs versus AR, their final 90 day outcomes were significantly different, with far fewer patients at SNFs achieving good clinical outcomes (25% vs 46%; p=0.023), as shown in figure 1A. More specifically, as seen in figure 1B, there was a shift of patients towards lower mRS in the AR group compared with the SNF group (p<0.001, χ2). Furthermore, discharge disposition to an SNF was associated with longer hospital stays (15 days vs 9 days; p<0.001), lower rates of IV tPA administration (43% vs 65%; p<0.028), greater rates of percutaneous endoscopic gastrostomy tube placements (37% vs 10%; p=0.002), and higher rates of uninsured patients (38% vs 21%; p=0.05), compared with those in AR.

Figure 1

(A) Total number of patients and good outcomes (modified Rankin Scale (mRS) score of 0–2 at 90 days) following endovascular reperfusion therapy stratified by disposition status. (B) Per cent of patients with mRS scores of 1–6 at 90 days post-procedure, following discharge to acute rehabilitation or skilled nursing facilities.

Of the 108 patients discharged to an SNF or AR, 37 (34%) had good outcomes and 71 (66%) had poor outcomes at the 90 day follow-up. The following variables were identified as possible predictors of good clinical outcome on univariate analysis: male gender (p=0.006), favorable ASPECTS (p=0.042), postprocedure infarct volume (p=0.010), insurance coverage (p=0.146), IV tPA delivery (p=0.152), and disposition to AR (p=0.023). After adjusting for these variables in a binary logistic regression model, disposition remained an independent predictor of patient outcome (p<0.038) (table 2). Discharge to an SNF was associated with lower probabilities of achieving good clinical outcomes (OR 0.337; 95% CI 0.12 to 0.94), whereas acceptance to AR was associated with improved probabilities for good outcomes (OR 2.97; 95% CI 1.06 to 8.31). ASPECTS were withheld from the model due to significant collinearity with postprocedure infarct volume, while age, NIHSS, and modified APACHE II scores were forced into the model based on previous literature.20 ,21

Table 2

Binary logistic regression model adjusting for predictors of good outcome after endovascular therapy in patients discharged to acute rehabilitation or skilled nursing facilities

In a separate analysis including patients discharged home following IAT, disposition to home was associated with lower rates of hypertension (home 49% vs AR 73% vs SNF 78%; p<0.011), higher percentages of favorable ASPECTS (87% vs 58% vs 55%; p=0.003), smaller postprocedural infarct volumes (18 cm3 vs 43 cm3 vs 33 cm3; p<0.001), and higher rates of good clinical outcomes (97% vs 46% vs 25%; p<0.001), compared with the AR and SNF groups, respectively. The THRIVE and HIAT-2 outcome scales were also significantly better among the home cohort, with both lower scores on admission NIHSS and modified APACHE II (see online supplementary table S1).


Our results demonstrate that patients receiving IAT have better outcomes if they are discharged to AR compared with SNF. From a macroscopic viewpoint, it is not surprising that the post hospital discharge is related to outcomes given the differences in the nature of the rehabilitation as well as how disposition is often determined. The IRF, where AR occurs, is a facility designed to provide comprehensive multidisciplinary rehabilitation for up to 3 h of therapy per day under supervision of a physician trained in rehabilitation and a team of therapists in a variety of disciplines (physical therapy, occupational therapy, speech therapy). On the other hand, SNFs are facilities that provide less intensive therapy services. In a 2006 study of Medicare services, 78% of SNF patients received therapy ranging from 45 min per week to 12 h per week.22 Disposition is often determined by a patient's ability to participate in therapy and demonstrate consistent functional improvement towards returning home. Patients with more comorbidities and severe neurological illness are often discharged to an SNF, thus having a lower probability of favorable 3 month outcomes. The current analysis shows that patients discharged to an SNF did worse despite having matched medical comorbidities and final infarct volumes to the AR group.

In our model, post-hospital discharge between AR and SNF was demonstrated as a predictor of neurological outcome, as seen in table 2. There may be some criticisms of the model as typical predictive factors, such as age and NIHSS, were insignificant. However, it is important to remember that this model was created between patients who were discharged to either an AR or SNF. By excluding patients who had dispositions of either home or deceased, the impact of age and NIHSS was diluted. Infarct volumes remained significant in the model and are a logical predictor.

Our findings are consistent with previous studies. In a study of 92 IRFs and SNFs from multiple centers in different states, Kramer et al demonstrated the IRFs are more likely to produce improvements in activities of daily living as well as discharge to the community.9 In a second study, stroke patients discharged to IRFs had a significantly higher functional recovery than those discharged to an SNF.10 Furthermore, in another study focusing on outcomes by motor function and cognitive abilities, IRFs were superior.11 The main issue with these studies is that they involved patients in an era where acute stroke care treatment was not widely available, standardized, or optimized. Therefore, it is likely that sicker patients with unfavorable prognoses were being discharged appropriately to an SNF and were not having a good prognosis. The current analysis differs as it uses a group of patients receiving interventional stroke treatments along with further advances in neurocritical care not employed at the time of the previous studies. Yet, despite these advances, similar findings are being demonstrated, thus highlighting how best stroke outcomes can be achieved by optimizing care in all phases of stroke care, not just the acute phase.

We found that patients with similar acute treatments and stroke volumes were being discharged to different types of institutions. One reason for this discrepancy involves the socioeconomic factors of our patients. In previous stroke epidemiological studies, socioeconomic status and the presence of family support have been implicated as factors in outcomes.23 ,24 While the exact reasons in those studies are not clear, it is possible that under or uninsured patients lack access to IRFs and are therefore discharged to an SNF. Furthermore, current guidelines require that patients discharged to IRFs go home following their AR. This is not possible for patients who do not have family support to provide adequate supervision and continued rehabilitation support; therefore, in such instances, patients are sent to SNFs even though they are fully able to tolerate the intensity of an IRF. The characteristics of our subjects seem to support this idea as there were a significantly larger number of uninsured patients in SNF compared with IRF. In this cohort, one-fifth of the uninsured or under insured population was discharged to AR through charity care. Further analysis is needed to determine whether the outcome of these subcohorts of patients were consistent with insured patients who went to AR. In addition, there were a significantly higher number of patients with percutaneous endoscopic gastrostomy tubes in the SNF placed population. One can hypothesize that maintenance, feeding, and flushing of tubes requires skills, and patients without adequate family or social support would be unable to return to a community setting precluding them from going to AR. Other reasons, such as cognitive status and ability to participate in therapy, are possible but such data are not available in the current database. Further analysis of the above mentioned factors is required to better understand the reasons by which post-hospital discharge was determined.

Our study suggests that re-evaluation of protocols in selecting patients for AR may need to be investigated in order to optimize the patient's chances of a good outcome, particularly after aggressive upfront care. For instance, at our institution, a physiatrist is now consulted acutely in the management of stroke patients with the intent to ensure patients are discharged appropriately and initiating aggressive rehabilitation early in the process. Future analyses will examine whether these measures alter the discharge disposition of our patients and if they improve outcomes.

Another implication is in the design of future acute stroke trials. Assessing the impact of an acute intervention in ischemic stroke has the challenge of ensuring the treatment group and placebo group have similar care in the days after their stroke. The potential impact of discharge disposition has not been previously assessed in patients undergoing endovascular reperfusion therapy and may explain some of the reasons why recent acute stroke intervention trials failed to show benefit for endovascular reperfusion. We recommend that future acute stroke trials account for post-hospital discharge disposition by including discharge disposition as a patient characteristic to ensure there is not an imbalance among the treatment arms. In addition, we recommend that acute trials maximize opportunities for AR placement for qualifying patients.

This exploratory analysis does have some limitations. It is a single center study with the inherent biases of a retrospective study. The decision tree for discharge disposition cannot be ascertained in a retrospective manner effectively. The small numbers may not be powered to detect differences in medical comorbidities and final infarct volumes between the two groups. Lastly, the use of final infarct volumes as a surrogate for neurological impairment at discharge is limited by regional and hemispheric differences. For example, a relatively small infarct involving the brainstem or internal capsule may have enormous impact on patient functionality during disposition. Such limitations may be alleviated by the addition of a clinical measure, such as 24 h NIHSS or discharge NIHSS, which could confirm the similarities in neurological morbidity between the two cohorts. Nevertheless, our study of acutely treated patients provides an exploratory framework of the role of discharge disposition in persons with stroke who receive IAT. It also supports data from previous studies and stresses the importance of stroke care along the entire spectrum of stroke.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors Writing of manuscript: SRB, C-HJS, and RG. Conception of research: SRB, C-HJS, RG, RGN, VP, and LAW. Statistical analysis: SRB, C-HJS, and RG. Revision of manuscript: RGN, AMA, TTT, MRF, and CMH. Data collection: C-HJS and BAG. RG and SRB had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

  • Competing interests SRB serves as a consultant for Neural Stem. RGN serves on the scientific advisory board for Stryker Neurovascular, Covidien, and CoAxia. He also serves on the Data Safety Monitoring Board for Rapid Medical and Imaging Core Lab for Covidien and Reverse Medical. He is an editor of Interventional Neurology. RG serves on the scientific advisory board for Stryker Neurovascular, Covidien, and CoAxia. He is also a member of the Data Safety Monitoring Board for Reverse Medical and Rapid Medical. He is an associate editor for the Journal of Neuroimaging, and associate editor for Interventional Neurology. MRF provides legal consult as an expert witness.

  • Ethics approval The study was approved by the Emory University institutional review board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Any inquiries regarding access to data should be addressed directly to the corresponding author.