Anda di halaman 1dari 7


Stroke Rehabilitation: Clinical Predictors of Resource Utilization

Richard L. Harvey, MD, Elliot J. Roth, MD, Allen W. Heinemann, PhD, Linda L. Lovell, BS, John R. McGuire, MD, Sylvia Diaz, MSN, RN, CRRN-A
ABSTRACT. Harvey RL, Roth EJ, Heinemann AW, Lovell LL, McGuire JR, Diaz S. Stroke rehabilitation: clinical predictors of resource utilization. Arch Phys Med Rehabil 1998;79: 1349-1355. rehabilitation patients using medical history, physical examination, and functional assessment.

Objective: To identify predictors of rehabilitation hospital resource utilization for patients with stroke, using demographic, medical, and functional information available on admission. Design: Statistical analysis of data prospectively collected from stroke rehabilitation patients. Setting: Large, urban, academic freestanding rehabilitation facility. Participants: A total of 945 stroke patients consecutively admitted for acute inpatient rehabilitation. Main Outcome Measures: Resource utilization was measured by rehabilitation length of stay (LOS) and mean hospital charge per day (CPD). Methods: Independent variables were organized into categories derived from four consecutive phases of clinical assessment: (1) patient referral information, (2) acute hospital record review and patient history, (3) physical examination, and (4) functional assessment. Predictors for LOS and CPD were identified separately using four stepwise multiple linear regression analyses starting with variables from the first category and adding new category data for each subsequent analysis. Results: Severe neurologic impairment, as measured by Rasch-converted NIH stroke scale and lower Rasch-converted motor measure of the Functional Independence Measure (FIM) instrument predicted longer LOS (/72,824 = 231.9, p < .001). Lower Rasch-converted motor FIM instrument measure, tracheostomy, feeding tube, and a history of pneumonia, coronary artery disease, or renal failure predicted higher CPD (F6.s20 = 90.2,p < .001). Conclusion: Stroke rehabilitation LOS and CPD are predicted by different factors. Severe impairment and motor disability are the main predictors of longer LOS; motor disability and medical comorbidities predict higher CPD. These findings will help clinicians anticipate resource needs of stroke

1998 by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation
COST OF STROKE CARE in the United States has T HE been estimated at $17 billion annually. 1 Most of this cost is incurred during the acute phase of care, which includes initial hospital management, diagnostic evaluation, and inpatient rehabilitation. Patients who are referred to acute rehabilitation have varying degrees of medical comorbidity, physical impairment, and disability. How medical and functional severity impact the cost required to achieve functional goals is not clear, but accurate prediction of the resources needed to care for patients has become an important management issue for rehabilitation facilities interested in maintaining quality care within an increasingly competitive healthcare market. Although total costs of care are measurable, how that cost is distributed can vary among patients with similar disability over time. For example, two patients with stroke might have equivalent total costs for inpatient rehabilitation, but their hospital resource utilization could differ if one requires more hospital services on a daily basis but can achieve inpatient goals over a shorter length of stay (LOS). These different utilization patterns require careful planning for availability of human resources, supply inventory, and outside hospital services. Predicting cost of inpatient care has only recently been investigated in stroke rehabilitation. Two studies have specifically examined the use of clinical information to stratify hospital LOS as a measure of resource utilization. Harada and associates 2 found that admission functional status, change in functional status, and time between stroke onset and rehabilitation admission explained 16.9% of the variance in stroke rehabilitation LOS. Stineman and colleagues 3 noted that the summed admission score on 13 motor items of the Functional Independence Measure (FIM) instrument and patient age explained 23% of variance of LOS for patients in stroke rehabilitation. Although these studies indicate that admission functional status is a strong predictor of rehabilitation LOS, neither investigator included impairment measures or medical comorbidities as independent variables. These variables are lacking in most stroke outcome research, despite the frequent clinical observation and preliminary scientific evidence that severity of neurologic deficit and certain associated medical comorbidities (especially heart disease) might influence the course and outcome of stroke rehabilitation. 4,5 We investigated the utility of various types of clinical information for predicting inpatient rehabilitation resource utilization among patients with stroke, using data from a comprehensive database. Our purpose was to examine the relative importance of demographic, medical, and functional information available at rehabilitation admission for predicting rehabilitation LOS and mean hospital charge per day (CPD).
Arch Phys Med Rehabil Vol 79, November 1998

From the Department of Physical Medicine and Rehabilitation, Northwestern University Medical School (Drs. Harvey, Roth, Heinemann), and The Rehabilitation Institute of Chicago (Drs. Harvey, Roth, Heinemann, Ms. Lovell, Ms. Diaz), Chicago, IL; and the Medical College of Wisconsin, Milwaukee, WI (Dr. McGuire). Submitted for publication October 10, 1997. Accepted in revised form May 25, 1998. Supported by The US Department of Education, The National Institute on Disability and Rehabilitation Research, grant H 133B30024, through the Rehabilitation Research and Training Center on Enhancing Quality of Life of Stroke Survivors, and by The Rehabilitation Institute of Chicago. Presented at the 58th Annual Assembly of the American Academy of Physical Medicine and Rehabilitation, October 13, 1996, Chicago, IL. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the authors or upon any organization with which the authors are associated. Reprint requests to Richard L. Harvey, MD, 345 East Superior Street. Chicago, IL 6O611. 1998 by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation 0003-9993/98/7911-467953.00/0


STROKE RESOURCE UTILIZATION, Harvey Table 2: Preexisting Medical Conditions Recorded on Admission to Acute Inpatient Rehabilitation and Frequency of Occurrence Medical Condition Hypertension Smoking history Previous stroke Diabetes mellitus Coronary artery disease Feeding tube Pneumonia Cardiac arrhythmia Cancer Substance abuse Congestive heart failure Seizure disorder Peptic ulcer disease and gastrointestinal bleed Renal failure Hypothyroidism Venous thromboembolism Tracheostomy Peripheral vascular disease Chronic obstructive pulmonary disease n 631 440 323 280 262 179 162 140 97 96 93 91 83 61 58 57 51 51 36

Predicting these outcomes might facilitate more accurate planning of medical rehabilitation resource use, guide appropriate modification in clinical programming, and assist development of equitable reimbursement methods. METHOD

Persons eligible for the study were 945 stroke rehabilitation patients admitted consecutively to a freestanding urban rehabilitation hospital between December 1993 and December 1997. Criteria for admission to acute rehabilitation included ability to tolerate at least 3 hours of therapy daily in at least two of three disciplines (physical, occupational, and speech therapy) and need for 24-hour nursing care. The average payer mix at our institution during the study period included Medicare (56%), Medicaid (19%), HMO/PPO (17%), and fee for service (8%). Included in the study were patients 18 years or older who had a stroke within the previous 3 months, whose primary reason for rehabilitation was disability resulting from stroke, and who remained rehabilitation inpatients for 3 days or longer. Stroke, defined as an acute event of cerebrovascular origin, causing focal or global dysfunction lasting more than 24 hours, was confirmed by both clinical and radiographic evaluation. Of the original 945 patients in the sample, 75 were eliminated due to insufficient data collection. Four additional patients had CPD values greater than or less than three standard deviations from the mean. Examination of these outliers revealed data entry errors and they were eliminated from the analysis. Thus a total of 866 patients were included in all or part of the data analysis.

Using acute hospital records available on admission to inpatient rehabilitation, physicians documented patient demographic characteristics (age, gender, race, and marital status); stroke etiology (ischemic vs hemorrhagic); and location of lesion (cortical vs subcortical; and right, left, or bilateral). A summary of patient demographics and stroke characteristics is found in table 1. Based on acute hospital transfer records and the patient rehabilitation admission interview, physicians identified all preexisting medical conditions and comorbidities for each patient by circling choices on a list of 124 diagnoses. Statistical analysis was limited to the 19 medical conditions that were present in more than 5% of the sample (table 2). A research assistant reviewed each patient's rehabilitation record and documented values for admission temperature, initial systolic and diastolic blood pressure (mmHg), and seven laboratory tests most commonly performed in most of the patients (n = 830) at rehabilitation admission (table 3).
Table 1: Demographic and Stroke Characteristics (n = 866) Age, mean _+ SD Gender, male Race Caucasian Black Married Type of stroke Ischemic Hemorrhagic Bilateral stroke Cortical stroke Days after stroke onset, mean +_ SD Rehabilitation length of stay, mean _+ SD 64.3 _+ 15.5yrs 407 (47%) 489 (57%) 298 (34%) 414 (48%) 624 (72%) 242 (28%) 91 (11%) 458 (53%) 17.3 _+ 14.6 days 28.3 _+ 13.8 days

Attending physicians scored neurologic impairment found on initial physical examination using the National Institutes of Health Stroke Scale (NIHSS). Each physician rater completed training with a standard test videotape defining score criteria for test items. The NIHSS includes 15 neurologic test items graded on a 3- to 5-point scale. 6-8 Higher values on the NIHSS indicate more severe neurologic impairment. This scale has been used extensively to assess efficacy of pharmacologic interventions in acute stroke management trials, 9,t and has also been applied recently in the rehabilitation setting. 11,12 Uniform Data System for Medical Rehabilitation-certified nurses rated disability using the FIM instrument, a validated functional assessment tool used to rate patient performance on each of 18 activities of daily living. 13-19 Lower scores on the FIM instrument indicate more severe disability. Analysis Scores on the 18-item FIM instrument were divided into two subscales that have been previously identified. 19 The motor subscale of the FIM instrument includes ratings on eight self-care activities, three mobility skills, and two sphincter control items. The cognitive subscale of the FIM instrument includes two language and three psychosocial ratings. To meet assumptions for parametric statistics, scores on the NIHSS and the subscales of the FIM instrument were converted from
Table 3: Definition of Laboratory Abnormalities Noted on Admission to Acute Inpatient Rehabilitation* Anemia Male: Hemoglobin <14 mg/dL Female: Hemoglobin <12 mg/dL Renal insufficiency (Creatinine >1.7 mg/dL) Leukocytosis (white blood cell count >10.5) Hypoalbuminemia (albumin <3.5 mg/dL) Abnormal electrolytes Sodium <135 or >148 mg/dL Potassium <3.5 or >5 mg/dL Bicarbonate <24 or >32 mg/dL *Cutpoint for laboratory values are based on 2 standard deviations from the mean.

Arch Phys Med Rehabil Vol 79, November 1998



ordinal scores into interval measures with rating scale (Rasch) analysis using BIGSTEPS software.lJ,18 Rasch-converted FIM instrument measure is weighted by item difficulty with values ranging from 0 to 100 for each subscale. Converted NIHSS measure ranged from - 6 to 2. Laboratory values were coded as normal or abnormal (table 3); systolic and diastolic blood pressure were analyzed using raw values. Elevated temperature was defined as a value of >98.6E The goal of data analysis was to identify predictors of LOS and CPD from medical and functional information available at admission. Of particular interest was the amount and type of information required to enhance prediction of resource utilization. We classified independent variables into categories that represent four phases of initial clinical assessment: (1) patient referral information--demographic and stroke characteristics; (2) medical record review and history--preexisting medical conditions and laboratory abnormalities; (3) physical examination--neurologic impairment (NIHSS); and (4) functional assessment--disability (FIM instrument). We ran four separate stepwise multiple regression analyses with LOS as the dependent variable. The first analysis included only demographic and stroke characteristics as independent variables. The second analysis added variables of preexisting medical conditions to the demographic and stroke characteristics used in the first analysis. Converted NIHSS and FIM instrument scores were added in the third and fourth analyses, respectively. Thus each analysis included new category variables plus all other variables used in the previous analysis. We then repeated this strategy using CPD as the dependent variable. Interaction terms were not included in any regression analysis. A predictor was entered into the model at a significance o f p < .01 and was eliminated at p > .05. Eigenvalues of the predictor variable correlation matrices revealed no problems with colinearity in any analysis. Data were analyzed using SPSS 7.5 for Windows 95. RESULTS LOS Statistically significant predictors for rehabilitation LOS from four linear regression analyses are listed in table 4. Among the demographic and stroke characteristics studied, a longer interval from stroke onset to rehabilitation admission and cortical stroke were significantly associated with longer LOS, but these factors explained only 6.2% of the variance (F2,s63 = 29.5, p < .001). The presence of a feeding tube was the strongest predictor of longer LOS when information about preexisting medical conditions and laboratory abnormalities were added to the analysis. Cortical stroke and a longer interval from stroke onset to rehabilitation admission remained in the

model. These three factors explained 15.5% of the variance in L O S (F3,826 = 5 1.8, p < .001). Adding knowledge about the severity of neurologic impairment significantly improved prediction of LOS. More severe neurologic impairment (higher NIHSS measure), the presence of a feeding tube, and longer interval from stroke onset predicted longer rehabilitation LOS, explaining 32.8% of the variance (F3,823 = 135.3, p < .001). Cortical stroke location did not appear in this model. The addition of functional information demonstrated that a lower motor subscale measure on the FIM instrument, representing more severe physical disability, predicted longer rehabilitation LOS. However, neurologic impairment remained the strongest predictor of LOS and no other factors entered the model. These two variables explained 35.9% of the variance in LOS (F2,824 = 231.9, p < .001). CPD Table 5 displays the statistically significant predictors of mean rehabilitation hospital CPD from four linear regression analyses. Among the demographic and stroke characteristics, a longer interval from stroke onset to rehabilitation admission and bilateral stroke location predicted higher daily hospital charge (F2,863 = 46.7, p < .001). These two predictors explained 9.5% of the variance in CPD. When preexisting medical conditions and laboratory abnormalities were added to the next analysis, the strength of the CPD model improved significantly. The presence of a feeding or tracheostomy tube and a history of CAD, pneumonia, or renal failure predicted higher CPD (F5,824 = 83.8, p < .001). Demographic predictors no longer appeared in the model. These five preexisting medical conditions explained 33.3% of the variance in CPD. The final two analyses improved the explained variance marginally. The five preexisting medical conditions continued to predict higher hospital charges through the remaining analyses. An elevated NIHSS measure, indicating more severe impairment, predicted higher CPD in the third model (F6,820 = 77.7, p < .001). When disability was added to the last analysis, a low motor subscale measure on the FIM instrument at admission predicted higher CPD, but NIHSS was no longer an important predictor. This model explained 39.3% of the variance in CPD (F6.820 = 90.2, p < .001). In summary, admission neurologic impairment as measured by the NIHSS is a stronger predictor of LOS than the admission motor subscale of the FIM instrument. Medical information was better at predicting hospital CPD than LOS. The presence of a tracheostomy or feeding tube and a history of CAD, pneumonia, or renal failure along with severe physical disability predicted high daily hospital charge.

Table 4: Standardized Regression Coefficients (l~) and Adjusted Coefficients of Variation (R2) for Predicting Length of Rehabilitation Hospitalization (LOS) in Four Cumulative Information Categories
Information Categories n Predictors of LOS 13 Change R2 p Value Adjusted R2

Demographic and stroke characteristics Preexisting medical conditions

866 830

+ Neurologic impairment


+ Disability


Days after stroke onset Cortical Stroke Feeding tube Cortical stroke Days after stroke onset NIH stroke score Feeding tube Days after stroke onset NIH stroke score FIM instrument motor measure

.22 .10 .32 .11 .10 .47 .15 .09 .36 -.32

.055 .009 .134 .015 .009 .296 .028 .007 .296 .064

<.001 .004 <.001 .001 .003 <.001 <.001 .004 <.001 <.001

.062 .155



Arch Phys Med Rehabil Vol 79, November 1998



Table 5: Standardized Regression Coefficients (l~) and Adjusted Coefficients of Variation (R2) for Predicting Mean Hospital Charges per Day (CPD) in Four Cumulative Information Categories
Information Categories n Predictors of CPD 13 Change R2 p Value Adjusted R2

Demographic and stroke characteristics + Preexisting medical conditions

866 830

+ Neurologic impairment


+ Disability


Days after stroke onset Bilateral stroke Feeding tube Tracheostomy Coronary artery disease Pneumonia Renal failure Feeding tube Tracheostomy NIH stroke score Pneumonia Coronary artery disease Renal failure FIM instrument motor measure Tracheostomy Feeding tube Pneumonia Coronary artery disease Renal failure

.30 .09 .34 .26 .11 .12 .09 .28 .25 .17 .13 .10 .09 -.30 .22 .20 .12 .10 .08

.090 .008 .236 .067 .013 .013 .008 .238 .068 .025 .013 .011 ,008 .260 .078 .029 .013 .011 .006

<.001 .006 <.001 <.001 <.001 <.001 .002 <.001 <.001 <.001 <.001 <.001 .001 <.001 <.001 <.001 <.001 <.001 .003

.095 .333



DISCUSSION Rehabilitation service providers and researchers have given serious thought to the effectiveness of therapeutic interventions in stroke care. But with declining financial reimbursement levels and increasing competition in the health care market, clinicians are also recognizing that more efficient and costeffective rehabilitation programs are vitally important. Physicians, nurses, and therapists find it particularly challenging to control medical costs when patients with realistic functional goals have severe neurologic impairment, multiple medical comorbidities, and significant nursing needs. How to plan for the care of these patients in rehabilitation with limited funding is an important issue. Research exploring how clinical factors may predict resource utilization during inpatient rehabilitation will begin to address this problem. Researchers have used rehabilitation hospital LOS for many years as a surrogate measure of functional severity and rate of recovery, z More recently, studies have used LOS as a measure of hospital resource utilization. These studies began to be published shortly after the US Congress passed the 1990 Omnibus Budget Reconciliation Act recommending that the Secretary of Health and Human Services propose a Medicare prospective payment system (PPS) for hospital services exempt from diagnostic related groups (DRGs), including medical rehabilitation. 21,22 An appropriate PPS for rehabilitation has been a challenge to develop, especially with the recent emergence of subacute rehabilitation and its different patterns of cost and LOS. 23 However, two PPS methods have been seriously considered. The first method proposes to stratify payment using Function Related Groups (FRGs) within impairment categories. Harada 2 and Stineman and coworkers 3,24 have proposed different FRG classification schemes using clinical data to predict resource utilization as measured by rehabilitation LOS. The second method is resource utilization groups (RUG-III) developed to measure daily resource use in long-term care facilities. 25,26 The most accurate measure of resource utilization in medical rehabilitation would combine these two approaches that stratify daily cost of care in addition to hospital LOS, providing a complete measure of total cost (total hospital cost = LOS daily cost). This study differs from others by identifyArch Phys Med Rehabil Vol 79, November 1998

ing clinical predictors for both LOS and mean daily inpatient rehabilitation hospital charge among patients with stroke. We chose to examine average charges as a measure of daily resource utilization because data on total hospital costs were not readily available. Since charges are not directly comparable between facilities or different geographic areas, we limited data collection to a single facility so that charges between patients could be compared. Using hospital charges from a single facility is useful because each charge is associated with a hospital service or supply delivered for patient care. However, restricting data collection to a single facility limits our ability to generalize findings to other rehabilitation centers.

Inpatient Rehabilitation LOS

Our study showed that more severe neurologic impairment on admission, as measured by the NIHSS, was the most powerful predictor of rehabilitation LOS followed by admission motor FIM measure. Only three other studies have examined both neurologic and functional status as potential LOS predictors and all found that severe impairment predicts longer LOS. Feigenson and associates20 collected demographic, medical, and functional information on 566 patients with acute stroke on admission to rehabilitation. They found that severe weakness, perceptual dysfunction, hemianopsia, sensory deficit, and multiple neurologic deficits were the only predictors of a longer rehabilitation stay. Brosseau and associates 27 measured sociodemographic, physical, functional, and neuropsychologic characteristics of 15 2 patients admitted for acute stroke rehabilitation and reported that perceptual deficits and poor balance predicted longer rehabilitation LOS. These neurologic factors, along with older age and lower score on the FIM instrument 1 week after rehabilitation admission, explained 43.6% of the variance in LOS. In a study by Liu and colleagues, 2~ weakness in ankle dorsiflexion as scored on the Stroke Impairment Assessment Set (SIAS) predicted longer LOS, but was not as strong a predictor as admission score on the FIM instrument. In these two studies, motor and cognitive score on the FIM instrument were stronger predictors of LOS than neurologic impairment. Although the motor measure of the FIM instrument predicted LOS in this study, it was not as powerful as the NIHSS.



This study is the first to demonstrate the association between cortical injury and longer LOS. This may reflect the association between cortical injury and multiple impairments including motor, sensory, visual, cognitive, and language deficits. Alternatively, the cortical strokes tend to be of larger volume than subcortical or brainstem injuries. Other studies have demonstrated an inverse relationship between infarct size and functional o u t c o m e . 29'3 Stroke location was not a strong predictor of LOS in our study; it was no longer statistically significant once NIHSS was added to the model. A longer interval between stroke onset and admission to rehabilitation predicted longer LOS in our sample until the final analysis when disability measures were added. Harada 2 also found that stroke onset to rehabilitation admission interval was useful for predicting LOS in her FRG model: an acute hospital LOS of ->15 days predicted longer rehabilitation LOS in patients with severe disabilities. Stineman 3 did not use onset to admission interval in her FRG model and Feigenson~- found no association. Longer rehabilitation LOS might be required to manage the added disability of generalized physical deconditioning after a prolonged acute hospitalization. Alternatively, concurrent medical comorbidities may lengthen both acute and rehabilitation hospital LOS. However, this seems unlikely because preexisting medical conditions did not predict LOS in this study. Cognitive measure of the FIM instrument/'ailed to predict rehabilitation LOS in this study. Galski and associates 31 reported a negative correlation between cognition and LOS in a sample of 86 patients with ischemic stroke who were given the Neurobehavioral Cognitive Status Examination (NCSE). Lower judgment, orientation, and comprehension, and lower cognitive measure on the FIM instrument predicted longer LOS. Bohannon and colleagues 32 published a letter in response to Galski, presenting similar results, indicating that lower cognition was related to longer LOS. Because cognitive and motor measures on the FIM instrument correlate well, it is not surprising that longer LOS would be associated with low scores on both. Yet stroke patients with severe physical disability and intact cognition can participate actively and benefit more from an intense rehabilitation program. This may explain the absence of cognition as a predictor in our analysis. Although cognition may have a clinical influence on rehabilitation LOS, physical impairment and disability are the most important predictors. Preexisting medical conditions and abnormal laboratory values on admission had very little influence on stroke rehabilitation hospital LOS. The presence of a feeding tube was initially associated with longer LOS, but it did not predict LOS in the analysis that included FIM-measured disability. In contrast, Liu 2s developed a novel comorbidity index that predicted LOS during stroke rehabilitation, but admission score on the FIM instrument and ankle strength were stronger predictors than comorbidity. Finestone and colleagues 33 found that malnutrition on admission predicted longer LOS. Malnutrition may be associated with prolonged acute hospitalization, which predicted longer rehabilitation LOS in our study. Hypoalbuminemia has been associated with increased risk for medical complications and poor functional outcome in rehabilitation, but has not been related to longer LOS in this or other studies. 3334 Concurrent medical comorbidity appears to have a more significant influence on acute stroke hospital LOS. Monane and associates 35 showed a relationship between high Charlson comorbidity index and longer hospital LOS in a retrospective study of 745 patients with acute stroke. Age did not predict LOS in our study. Stineman 3 used age in the FIM-FRG stroke model; older patients were found to have a

shorter LOS. In contrast, Brosseau 27 found that older age predicted a longer LOS in stroke rehabilitation, whereas Feigenson 2 found no association.

Mean Rehabilitation Hospital CPD Our study demonstrates a strong relationship between the presence of preexisting medical conditions and higher daily hospital charges for stroke rehabilitation. In the final analysis, admission motor measure on the FIM instrument was the strongest of six CPD predictors. The remaining predictors were medical conditions. Of these, the presence of a feeding or tracheostomy tube were the strongest predictors of higher mean CPD. Although they are clinically associated with severe disability, a feeding or tracheostomy tube provides unique information about daily healthcare resource needs, which may include charges associated with respiratory care and parenteral feeding. In contrast, severity of neurologic impairment, also associated with greater disability, was no longer a predictor of CPD after FIM instrument measures were added to the analysis. The negative impact on function of aspiration pneumonia, renal failure, and coronary artery disease in stroke rehabilitation has been previously noted. 5,36-3: We have demonstrated that a history of these conditions predicts higher CPD in stroke rehabilitation. Odderson and colleagues 36 significantly reduced the incidence of aspiration pneumonia in patients with stroke during acute care by using a clinical pathway incorporating early swallow assessment. They showed significant cost saving through reduction of LOS and charged hospital services. Cowen and associates 37 noted no difference in total hospital charges between patients with stroke and end-stage renal disease and patients with stroke alone. Data from the Copenhagen Stroke Study 3s failed to demonstrate a relationship between heart disease and resource utilization measured by LOS; neither costs nor charges were examined. Two other studies have explored the relationship between medical severity of illness and total rehabilitation charges for stroke patients. Osberg and coworkers 39 retrospectively scored 73 stroke rehabilitation patients using the Horn severity of illness index and found that patients with greater disease severity had higher total hospital charges. Severity of illness explained 32% of the variance in hospital charge. McGinnis and colleagues4 studied the effectiveness of DRG category, severity of illness, function at admission as measured using Barthel Index and Kenny Scale, and age for predicting inpatient rehabilitation charges. DRG category, functional status, and age were poor predictors. However, greater severity of illness predicted higher total hospital charges, explaining 26% of the variance in all diagnostic categories. For stroke patients in particular, severity of illness explained 23% of the variance in hospital charge. Although Siegler and associates41 did not examine rehabilitation hospital charges, they demonstrated that the likelihood of suffering a medical complication (events that interrupted, delayed, or prolonged rehabilitation) increased with more comorbid conditions. Medical complications would be expected to increase costs due to added diagnostic testing, staff attention, and treatment programs. Thus the need for more intensive medical management required for stroke rehabilitation patients with comorbid conditions is the logical reason for higher daily hospital charges.
CONCLUSION This study has demonstrated that the clinical predictors of LOS and CPD are different and neither measure can represent
Arch Phys Med Rehabil Vol 79, November 1998



resource utilization independently. Similarly, a single measure of total hospitalization cost will not adequately distinguish the important clinical differences among patients undergoing rehabilitation care. Future studies on medical resource utilization should consider using separate analysis for LOS and daily cost to distinguish important outcome predictors. This study identified clinical predictors of resource utilization in acute stroke rehabilitation. Neurologic impairment and physical disability were the best predictors of inpatient rehabilitation LOS. Patients with more severe impairment require longer hospitalization to gain sufficient motor recovery, functional skill, and family training to achieve a safe discharge to home. Severe motor disability, a feeding tube or tracheostomy, and other preexisting medical conditions predicted higher mean rehabilitation hospital CPD. Comorbid conditions and severe disability require more intensive medical management, diagnostic evaluation, and costly treatment. These findings will help clinicians anticipate the resource needs of stroke rehabilitation patients after obtaining a thorough medical history, performing a physical examination, and evaluating motor disability. This will also enhance communication between clinicians and third-party payers as they plan the rehabilitation course and hospital setting for the care of patients with stroke. 42 Finally, prospective payment systems for stroke rehabilitation should factor in the patient's severity of neurologic impairment, functional status, and comorbid medical conditions to provide fair reimbursement for the care of those who can benefit from acute inpatient rehabilitation.












1. Matchar DB, Duncan PW. Cost of stroke. Stroke Clin Update 1994;5:9-12. 2. Harada N, Kominski G, Sofaer S. Development of a resourcebased patient classification scheme for rehabilitation. Inquiry 1993;30:54-63. 3. Stineman MG, Escarce JJ, Goin J, Hamilton BB, Granger CV, Williams SV. A case-mix classification system for medical rehabilitation. Med Care 1994;32:366-79. 4. Lehmann JF, DeLateur B J, Fowler RS, Warren CG, Arnhold R, Schertzer G, et al. Stroke rehabilitation: outcome and prediction. Arch Phys Med Rehabil 1973;56:383-9. 5. Roth EJ, Mueller K, Green D. Stroke rehabilitation outcome: impact of coronary artery disease. Stroke 1988;19:42-7. 6. Brott T, Adams HE Olinger CP, Marler JR, Barsan WG, Biller J, et al. Measurements of acute cerebral infarction: a clinical examination scale. Stroke 1989;20:864-70. 7. Goldstein LB, Bertels C, David JN. Inter-rater reliability of the NIH stroke scale. Arch Neurol 1989;46:660-2. 8. Muir KW, Weir C J, Murray GD, Povey C, Lees KR. Comparison of neurological scales and scoring systems for acute stroke prognosis. Stroke 1996;27:1817-20. 9. Brott TG, Haley EC, Levy DE, Barsan W, Broderick J, Sheppard GL, et al. Urgent therapy for stroke: part I. Pilot study of tissue plasminogen activator administered within 90 minutes. Stroke 1992;23:632-40. 10. Wityk RJ, Pessin MS, Kaplan RF, Caplan LR. Serial assessment of acute stroke using the NIH stroke scale. Stroke 1994;25:362-5. 11. Heinemann AW, Harvey RL, McGuire JR, Ingberman D, Lovell L, Semik P, et al. Measurement properties of the NIH stroke scale during acute rehabilitation. Stroke 1997;28:1174-80. 12. Roth EJ, Heinemann AW, Lovell LL, Harvey RL, McGuire JR, Diaz S. Relationship between impairment and disability during stroke rehabilitation. Arch Phys Med Rehabil 1998;79:329-35. 13. Hamilton BB, Granger CV, Sherwin FS, Zielenzy M, Tashman JS. A uniform national data system for medical rehabilitation. In: Fuhrer MJ, editor. Rehabilitation outcomes: analysis and measureArch Phys Med Rehabil Vol 79, November 1998









32. 33.

ment. Baltimore (MD): Brooks Publishing Company; 1987. p. 137-47. Dodds TA, Martin DP, Stolov WC, Deyo RA. A validation of the functional independence measure and its performance among rehabilitation inpatients. Arch Phys Med Rehabil 1993;74:531-6. Heinemann AW, Linacre JM, Wright BD, Hamilton BB, Granger CV. Relationships between impairment and physical disability as measured by the functional independence measure. Arch Phys Med Rehabil 1993;74:566-73. Wright BD, Linacre JM, Heinemann AW. Measuring functional status in rehabilitation. Phys Med Rehabil Clin North Am 1993;4: 475-91. Hamilton BB, Laughlin JA, Fiedler RC, Granger CV. Interrater reliability of the 7-level functional independence measure (FIM). Scand J Rehabil Med 1994;26:115-9. Heinemann AW, Linacre JM, Wright BD, Hamiliton BB, Granger CV. Measurement characteristics of the functional independence measure. Top Stroke Rehabil 1994; 1:1-15. Linacre JM, Heinemann AW, Wright BD, Granger C, Hamilton BB. The structure and stability of the functional independence measure. Arch Phys Med Rehabil 1994;75:127-32. Feigenson JS, McCarthy ML, Greenberg SD, Feigenson WD. Factors influencing outcome and length of stay in a stroke rehabilitation unit. Part 2. Comparison of 318 screened and 248 unscreened patients. Stroke 1977;8:657-62. Wilderson DL, Batavia AI, DeJong G. Use of functional status measures for payment of medical rehabilitation services. Arch Phys Med Rehabil 1992;73:111-20. Chan L, Keopsell TD, Deyo RA, Esselman PC, Haselkorn JK, Lowery JK, et al. The effect of Medicare's payment system for rehabilitation hospitals on length of stay, charges and total payments. N Engl J Med 1997;337:978-85. Kramer AM, Steiner JF, Schlenker RE, Eilersten TB, Hrincevich CA, Tropea DA, et al. Outcomes and costs after hip fracture and stroke. A comparison of rehabilitation settings. J A M A 1997;277: 396-404. Stineman MG, Fielder RC, Granger CV, Maislin G. Functional task benchmarks for stroke rehabilitation. Arch Phys Med Rehabil 1998;79:497-504. Carpenter GI, Ikegami N, Ljunggren G, Carrillo E, Fries BE. RUG-Ill and resource allocation: comparing the relationship of direct care time with patient characteristics in five countries. Age Ageing 1997;26 Suppl:61-5. Stineman MG. Measuring casemix, severity, and complexity in geriatric patients undergoing rehabilitation. Med Care 1997;35: JS90-JS 105. Brosseau L, Philippe P, Potvin L, Boulanger YL. Post-stroke inpatient rehabilitation. I. Predicting length of stay. Am J Phys Med Rehabil 1996;75:1-10. Liu M, Domen K, Chino N. Comorbidity measures for stroke outcome research: a preliminary study. Arch Phys Med Rehabil 1997;78:166-72. Miller LS, Miyamoto AT. Computed tomography: its potential as a predictor of functional recovery following stroke. Arch Phys Med Rehabil 1979;60:108-9. Hertanu JS, Demopoulos JT, Yang WC, Calhoun WE Fenigstein HA. Stroke rehabilitation: correlation and prognostic value of computerized tomography and sequential functional assessments. Arch Phys Med Rehabil 1984;65:505-8. Galski T, Gruno RL, Zorowitz R, Walker J. Predicting length of stay, functional outcome and aftercare in the rehabilitation of stroke patients. The dominant role of higher-order cognition. Stroke 1993;24:1794-800. Bohannon RW, Cooper J. Predicting length of stay in the rehabilitation of stroke patients [letter]. Stroke 1994;25:1295. Finestone HM, Green-Finestone LS, Wilson ES, Teasell RW. Prolonged length of stay and reduced functional improvement rate in malnourished stroke rehabilitation patients. Arch Phys Med Rehabil 1996;77:340-5.



34. Aptaker RL, Roth EJ, Reichhardt G, Duerden ME, Levy CE. Serum albumin level as a predictor of geriatric stroke rehabilitation outcome. Arch Phys Med Rehabil 1994;75:80-4. 35. Monane M, Kanter DS, Glynn RJ, Avorn J. Variability in length of hospitalization for stroke. The role of managed care in an elderly population. Arch Neurol 1996;53:875-80. 36. Odderson IR, Keaton JC, McKenna BS. Swallow management in patients on an acute stroke pathway: quality is cost effective. Arch Phys Med Rehabil 1995;76:1130-3. 37. Cowen TD, Huang CT, Lebow J, DeVivo MJ, Hawkins LN. Functional outcomes after inpatient rehabilitation of patients with end-state renal disease. Arch Phys Med Rehabil 1995;76:355-9. 38. Jorgensen HS, Nakayama H, Raaschou HO, Olsen TS. Acute stroke care and rehabilitation: an analysis of the direct cost and its



41. 42.

clinical and social determinants. The Copenhagen Stroke Study. Stroke 1997;28:1138-41. Osberg JS, Haley SM, McGinnis GE, DeJong G. Characteristics of cost outliers who did not benefit from stroke rehabilitation. Am J Phys Med Rehabil 1990;69:117-25. McGinnis GE, Osberg JS, DeJong G, Seward ML, Branch LG. Predicting charges for inpatient medical rehabilitation using severity, DRG, age, and function. Am J Public Health 1987;77:826-9. Siegler EL, Stineman MG, Maislin G. Development of complications during rehabilitation. Arch Intern Med 1994; 154:2185-90. Retchin SM, Brown RS, Yeh SJ, Chu D, Moreno L. Outcomes of stroke patients in Medicare fee for service and managed care. JAMA 1997;278:119-24.

Arch Phys Med Rehabil Vol 79, November 1998