Abstract
Background Little is known about the use of rehabilitation in the acute care setting and its impact on hospital readmissions.
Objective The objective of this study was to examine the association between the intensity of rehabilitation services received during the acute care stay for stroke and the risk of 30-day and 90-day hospital readmission.
Design A retrospective cohort analysis of all acute care hospitals in Arkansas and Florida was conducted.
Methods Patients (N=64,065) who were admitted for an incident stroke in 2009 or 2010 were included. Rehabilitation intensity was categorized as none, low, medium-low, medium-high, or high based on the sum and distribution of physical therapy, occupational therapy, and speech therapy charges within each hospital. Cox proportional hazards regression was used to estimate hazard ratios, controlling for demographic characteristics, illness severity, comorbidities, hospital variables, and state.
Results Relative to participants who received the lowest intensity therapy, those who received higher-intensity therapy had a decreased risk of 30-day readmission. The risk was lowest for the highest-intensity group (hazard ratio=0.86; 95% confidence interval=0.79, 0.93). Individuals who received no therapy were at an increased risk of hospital readmission relative to those who received low-intensity therapy (hazard ratio=1.30; 95% confidence interval=1.22, 1.40). The findings were similar, but with smaller effects, for 90-day readmission. Furthermore, patients who received higher-intensity therapy had more comorbidities and greater illness severity relative to those who received lower-intensity therapy.
Limitations The results of the study are limited in scope and generalizability. Also, the study may not have adequately accounted for all potentially important covariates.
Conclusions Receipt of and intensity of rehabilitation therapy in the acute care of stroke is associated with a decreased risk of hospital readmission.
Each year, approximately 795,000 Americans experience a new or recurrent stroke, with a greater incidence and prevalence in older adults, especially among those living in southern states.1–3 In 2009, the costs of providing acute, inpatient care to patients with stroke was estimated at $10.7 billion,2 with a portion of these costs due to hospital readmissions.4 Hospital readmissions, particularly within the first 30 days following discharge, can be a marker of poor health care quality and inefficient care.5,6 In an attempt to reduce preventable hospital readmissions, the Centers for Medicare & Medicaid Services (CMS) has implemented a policy to reduce payments to hospitals with 30-day readmissions for certain diagnoses.7
Although some readmissions are not preventable, many are preventable. Causes for preventable hospital readmissions include inadequate discharge planning, lack of communication among health care providers both within and outside the hospital, lack of a timely follow-up visit, and lack of clarity regarding who is caring for the patient after discharge.8 Rehabilitation specialists (ie, physical therapists, occupational therapists, and speech-language pathologists) serve an important role in the acute care of patients with stroke. After an evaluation to determine the extent of functional limitations of the patient, the rehabilitation specialist begins treatment to address these limitations and, perhaps more importantly, assists with discharge planning to determine the need and most appropriate setting for postacute care. Furthermore, the rehabilitation specialist provides education to the patient, family members, and other caregivers, including safety considerations during mobility, activities of daily living, and swallowing.
Several clinical practice guidelines for acute stroke recommend that rehabilitation provided by occupational therapists, physical therapists, and speech-language pathologists begin soon after admission to the hospital.9–14 These guidelines, which have been developed in several different countries, generally contend that the earlier the rehabilitation professionals can begin providing care to a patient, the better the quality of care the patient will be receiving. Using data from the Danish National Indicator Project, Ingeman and associates15 examined the medical records of hospitalized patients with stroke. They found that the 30- and 90-day mortality rates were significantly lower for patients who received all 7 of their identified indicators of quality of care. Two of these quality indicators were an early assessment by a physical therapist and an early assessment by an occupational therapist.
Other investigators have sought to explore the relationship between functional status and care during the index hospitalization and readmission in patients with acute stroke.16–21 Some of these studies showed that poorer physical function was predictive of hospital readmission.16,17,19–21 Little research, however, has specifically focused on the use of rehabilitation in the acute care setting and its impact on hospital readmissions. The purpose of our study was to examine the association between the intensity of rehabilitation services received during the acute care stay for stroke and the risk of 30- and 90-day hospital readmission. We hypothesized that more intensive rehabilitation would be associated with a decreased risk of hospital readmission.
Method
Study Design
We utilized a retrospective cohort design with a 3-month baseline period (where we examined prior hospitalizations), an exposure period (defined by the hospitalization length of stay), and a 3-month follow-up period (beginning the day after hospital discharge) (Fig. 1).
Study design.
Data Source
We examined the State Inpatient Databases (SID)22 from Arkansas and Florida. These databases contain data on all hospital discharges from short-term acute care and specialty hospitals in the states. Arkansas and Florida were selected because the SIDs for these states contained the necessary linkage information needed to examine readmissions and the detailed revenue codes to identify therapy intensity. These states are also in the southern United States, where the incidence of stroke is higher than in other census regions.23,24 Data from 2009 and 2010 were examined. Using the American Hospital Association (AHA) ID number available in the SID, we linked the SID data to the 2009 and 2010 AHA Annual Survey Database to obtain information on hospital characteristics.25
Cohort Identification
We identified patients admitted to short-term, acute care hospitals with a primary discharge diagnosis of stroke (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 430.xx–434.xx and 436.xx–438.xx) from April 1, 2009, to September 30, 2010. We eliminated patients with strokes in the first quarter of 2009 and last quarter of 2010 to allow for the baseline and follow-up periods. Patients were excluded if they were younger than 45 years of age, did not live in Arkansas or Florida, died during the hospitalization, had a stroke within the previous 3 months, or were transferred from the admitting hospital to another short-term, acute care hospital. Patients younger than 45 years of age were excluded because stroke characteristics, outcomes, and mortality differ in younger adults.26 Excluding these patients did not appreciably diminish the sample size (lost ∼5% of sample). Patients with a stroke in the prior 3 months also were excluded because the incident stroke may have been a readmission related to the previous stroke. The 3-month time frame was selected because most hospital readmissions for stroke occur within the first 100 days16 and to maximize the use of available data. Use of a longer baseline period would have decreased the period of time for identifying an index admission for stroke. Records with missing data (∼4%) also were excluded (Fig. 2). In preliminary analyses, records with and without missing data were similar in regard to demographic characteristics.
Identification of cohorts for study states: (A) Arkansas cohort, (B) Florida cohort.
Study Variables
Our primary outcome variable was readmission to a short-term, acute care hospital during a 30- or 90-day follow-up period. Readmissions were not further characterized as to whether they were unexpected, planned, or preventable. For these analyses, patients who died prior to an acute care readmission were excluded. We also examined our results with composite outcomes of 30-day readmission or death and 90-day readmission or death to account for the competing risk of death.
Our primary exposure variable was therapy intensity, defined as the amount of therapy received during the acute care stay. Consistent with methods used in a previous study,27 we created this variable by identifying the UB 92 revenue codes for physical therapy, occupational therapy, and speech therapy and summing the charges associated with these codes. Because the amount billed for the same services varies across hospitals and states,28 we categorized therapy intensity within each hospital as: none, low, medium-low, medium-high, or high based on the quartile distribution of the charges and controlled for length of stay to account for the fact that patients with longer lengths of stay potentially could receive more therapy.
We created several covariates to control for potential confounding. These covariates were based on content expertise of the team, literature, and available data and included demographic characteristics (age, sex, race, insurance, residence, income), comorbidities (based on the Elixhauser index29 as well as relevant condition-specific comorbidities), number of chronic conditions,30 and measures of illness severity. Our measures of illness severity included specific diagnoses in the hospital discharge record (eg, aphasia, hemiparesis) as well as proxy measures (eg, length of stay and discharge disposition). We also included several hospital quality measures (eg, volume of patients admitted with stroke, registered nurse full-time equivalents, for-profit status, teaching status, Commission on Accreditation of Rehabilitation Facilities [CARF] accreditation).31–33 See eTable 1 for definitions of the study variables.
Data Analysis
We used multivariate, Cox proportional hazards regression with a random effect for hospital to estimate hazard ratios. The random effect for each hospital accounted for the nonindependence of measures within the hospital (ie, patients within the same hospital will be more similar to each other than patients across hospitals). Failure to account for this “clustering” within hospitals in statistical models can lead to inappropriate inferences.34 We first estimated an unadjusted ratio and then examined how the estimate changed with the addition of covariate blocks: (1) facility random effect, (2) demographic characteristics, (3) illness severity characteristics, (4) comorbidities, and (5) hospital quality indicators and state variable. Although many of our covariates were likely correlated (eg, illness severity and comorbidities), we did not conduct formal testing for collinearity, as we were primarily interested in controlling for these variables and not in making conclusions regarding the individual parameter estimates for these variables. We also conducted a descriptive analysis to determine the primary diagnosis associated with the readmission. Analyses were performed using SAS, version 9.3 (SAS Institute Inc, Cary, North Carolina).
Role of the Funding Source
This study was funded by Elon University Faculty Research and Development Fund.
Results
A total of 64,065 patients met the study entry requirements, with the majority of the sample (77%) from Florida. Figure 2 presents our sample creation for each state. Table 1 presents selected characteristics of the cohort, stratified by therapy intensity (see eTab. 2, for a listing of all characteristics of the cohort by therapy intensity). Two-thirds of the participants (66.7%) received rehabilitation therapy during their inpatient stay. Participants who did not receive rehabilitation were younger and more likely to be male, white, and have private insurance. They also had fewer comorbidities and chronic conditions and were more likely to be discharged home. Participants who received high-intensity rehabilitation therapy were more likely to be black, have Medicaid coverage, and have more comorbidities and chronic conditions relative to those who received low-intensity therapy. Participants who received high-intensity rehabilitation therapy also were more likely to be discharged to a skilled nursing or inpatient rehabilitation facility.
Characteristics of Sample by Therapy Intensity
Table 2 presents the crude rates for hospital readmission by state. Rates of readmission were similar overall for Florida and Arkansas at approximately 15% for 30-day readmission and 26% for 90-day readmission. State differences, however, were apparent when these data were stratified by therapy intensity. In Florida, rates of readmission increased as therapy intensity increased. This trend was less apparent in the Arkansas data, where readmission rates for individuals who received no therapy were higher than readmission rates for low, medium-low, and medium-high therapy intensity. For both states, the prevalence and rates of readmission were highest for individuals who received high-intensity therapy.
Crude Readmission Rates by State
In Table 3, we present the unadjusted and adjusted effects of therapy intensity on 30- and 90-day hospital readmission. After adjusting for demographic characteristics, comorbidities, number of chronic conditions, measures of illness severity, and hospital quality measures, individuals who received higher-intensity therapy had a decreased risk of hospital readmission relative to those who received low-intensity therapy. These effects were greater for the 30-day readmission outcome and generally increased as therapy intensity increased. Individuals who received no therapy were at an increased risk of readmission relative to the low-therapy group. See eTables 3 and 4 for the full model results for 30- and 90-day readmission outcomes, respectively.
Hazard Ratios for Therapy Intensitya
Because less than 1% of the sample died prior to a 30- or 90-day hospital readmission, the frequencies of our composite outcomes (ie, readmission or death within 30 or 90 days) were only slightly greater than our readmission outcomes. Our results using the composite outcomes were almost identical to our findings presented in Table 3.
Based on the stepwise analyses, the greatest changes in the parameter estimates for therapy intensity occurred with the addition of the illness severity measures. The estimates were fairly stable with the addition of other covariate blocks. Several covariates were associated with readmission in the expected direction. For example, patients with more comorbidities and chronic conditions were more likely to be readmitted, as were patients discharged to a skilled nursing or inpatient rehabilitation facility relative to patients discharged to home. Patients with longer lengths of stay during the initial admission also were more likely to be readmitted relative to those with shorter lengths of stay.
The primary diagnoses associated with hospital readmissions during the 90-day follow-up period are presented in Table 4. The top 15 diagnoses accounted for 41.4% of the readmissions. The top 2 diagnoses—occlusion/stenosis of precerebral arteries without mention of cerebral infarction (ICD-9-CM code 433.10) and cerebral artery occlusion, unspecified with cerebral infarction (ICD-9-CM 434.91)—were related to the cerebrovascular system, and they accounted for 16% of the readmissions.
Top Primary Diagnoses for Readmission in the 90-Day Follow-up Period (N=16,412)
Discussion
We examined the association among therapy intensity and risk of 30- and 90-day hospital readmission following an incident stroke. After adjusting for demographic characteristics, comorbidities, various measures of illness severity, and hospital factors, we found that intensity of rehabilitation therapy received during the acute care admission was associated with a decreased risk of hospital readmission. These effects were greatest for the 30-day readmission outcome, which is more proximal to the care received during the initial hospitalization.
Participants who received no therapy during their acute care admission were at the greatest risk of 30- and 90-day readmission relative to those who received low-intensity therapy. These results were somewhat surprising, as individuals who did not receive therapy were generally healthier than those who did (Tab. 1). Their overall unadjusted risk of readmission also suggested that these individuals were not at risk of readmission. However, when we adjusted for illness severity, comorbidities, and other factors, the parameter estimate on no therapy changed considerably (Tab. 3), indicating confounding and suggesting that some individuals in this group had more comorbidities and greater illness severity. Although these results suggest that use of low-intensity therapy versus no therapy decreases the risk of readmission, after controlling for illness severity, we cannot rule out unmeasured confounding. Inclusion of measures of impairments in body structure and function as well as measures of activities limitations may have altered this finding.
Our more compelling findings relate to the dose-response relationship we observed with therapy intensity. Individuals who received more intensive therapy, controlling for length of stay as well as several other covariates, were at a decreased risk of readmission. These findings support current clinical practice guidelines for stroke,9–11,13,14 which recommend early referral and evaluation for patients with stroke in the acute care setting. One role of therapists in this setting is to maximize patient mobility, which has been identified as a potential physical biomarker for hospital readmission.35 Because more than 75% of the patients in our sample were in the hospital for ≤3 days, our findings may relate more to adequate time for discharge planning36 and communication among the health care team. In addition to providing examinations and interventions, rehabilitation providers also educate patients, caretakers, and family members, and they make recommendations for safe engagement in activities and participation after discharge, all of which may affect readmission. Furthermore, consultative services and effective interprofessional communication may have a significant impact on minimizing readmissions. Effective decision making and communication in the hospital setting should be a primary objective of clinicians, and policies and procedures should foster implementation of these skills.
Our 30-day readmission rate of 15% is generally similar to what has been reported in the literature from US studies.37 Lichtman et al38 reported a 30-day readmission rate of 14% in their analysis of Medicare data. Ottenbacher et al20 reported a 30-day readmission rate of 18%, although their sample was limited to individuals discharged to inpatient rehabilitation facilities in 6 states.
Our study had several limitations. First, our results are limited in scope and generalizability. We targeted southern states, given the higher incidence of stroke in this region. However, we were only able to utilize data from the Florida and Arkansas State Inpatient Databases, as none of the other southern state databases contained the outcome and exposure variables needed to answer our question. Furthermore, although Florida is geographically a southern state, the demographics and cultural norms of many parts of Florida may not be similar to those of other southern states. We also lost approximately 4% of the records on stroke hospitalization due to missing covariates or linkage IDs. Although the demographic characteristics of patients with and without missing data were similar, it is possible that those with missing data differed in illness severity. The small percentage of missing records, however, would likely have a minimal impact on our findings.
A second limitation is that we may not have adequately accounted for all potentially important covariates. For example, although we were able to account for the presence of comorbidities, we were unable to account for the duration and severity of these comorbidities. Also, as noted earlier, our database did not contain measures of specific measures of body structure and function impairments or activities limitations; thus, we depended on indirect measures of stroke severity, such as length of stay and discharge disposition. Furthermore, our exposure variable was a crude measure based on charges and did not reflect the content of the rehabilitation care received, nor did it distinguish between the amounts of therapy delivered by each discipline.
Lastly, most hospital readmissions were unplanned, but some were planned. Potentially, patients who received less rehabilitation services could have been readmitted at a higher rate for follow-up medical or surgical procedures such as carotid stenting or cardiac ablation.
Despite these limitations, our findings provide the basis for additional research on understanding the role of rehabilitation in the acute care of stroke. Future research should replicate our findings using data from different states and national data. Describing the exposure variable in more detail (eg, by discipline and content of care) also would be useful in gaining a better understanding of what is effective. Further research on factors that affect receipt of rehabilitation in the acute care setting is needed. Lastly, comparing subgroups of readmitted patients based on their original discharge destination (eg, home, inpatient rehabilitation, skilled nursing facility) also would prove insightful.
In conclusion, in this retrospective, observational cohort study, we found that patients with stroke who received higher-intensity therapy were at a decreased risk of hospital readmission relative to those who received lower-intensity therapy. We also found that individuals who received no therapy were at an increased risk of hospital readmission relative to those who received low-intensity therapy. These results support the contention of clinical practice guidelines that recommend patients hospitalized with an acute stroke receive rehabilitation services as soon and as much as practicable.
Footnotes
Dr Andrews and Dr Freburger provided concept/idea/research design, writing, and project management. Ms Li provided data collection. Dr Freburger and Ms Li provided data analysis. Dr Andrews provided fund procurement, participants, facilities/equipment, and administrative support. The authors thank Lily Wang at the Cecil G. Sheps Center for Health Services Research, The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, for her help with the statistical analyses.
This study was approved by the Institutional Review Board of Elon University.
Results from this study were presented as an abstract at the Combined Sections Meeting of the American Physical Therapy Association; February 3–6, 2014; Las Vegas, Nevada.
This study was funded by Elon University Faculty Research and Development Fund.
- Received January 20, 2015.
- Accepted June 11, 2015.
- © 2015 American Physical Therapy Association
References
Issue highlights
