The Effect of a Stroke Unit: Reductions in Mortality, Discharge Rate to Nursing Home, Length of Hospital Stay, and Cost
A Community-Based Study
Background and Purpose Treatment of stroke patients in specialized stroke units has become more frequent, yet the effect of this treatment has not been determined.
Methods In a community-based, prospective, and consecutive study of 1241 unselected acute stroke patients, we compared outcome of stroke treatment between two neighboring communities within Greater Copenhagen: the Bispebjerg community, where all acute stroke patients are treated and rehabilitated on a stroke unit, and Frederiksberg community, where all acute stroke patients are treated and rehabilitated on general neurological and medical wards. Except for the different organization of stroke treatment, the two communities and the two patient groups were comparable. Specifically, age, sex, marital status, prestroke residence, and stroke severity were not statistically different between patients treated on the stroke unit and those treated on the general neurological and medical wards. Multivariate regression analyses were used to estimate the independent influence of stroke unit treatment on outcome.
Results Stroke unit treatment significantly reduced in-hospital mortality (odds ratio [OR], 0.50; 95% confidence interval [CI], 0.34 to 0.74; P<.001), case-fatality rate (OR, 0.45; CI, 0.28 to 0.71; P<.001), 6-month mortality (OR, 0.57; CI, 0.39 to 0.82; P=.002), 1-year mortality (OR, 0.59; CI, 0.42 to 0.84; P=.003), and discharge rate to a nursing home (OR, 0.61; CI, 0.38 to 0.98; P=.04). Discharge rate to the patient’s own home was significantly increased (OR, 1.90; CI, 1.30 to 2.70; P<.001). The length of hospital stay (including rehabilitation) was reduced significantly by 30% in patients treated on the stroke unit despite their lower mortality (P<.001). The savings due to stroke unit treatment were estimated at 1313 bed-days and three places at a nursing home per 100 stroke patients.
Conclusions Treatment of unselected acute stroke patients on a stroke care unit saved lives, reduced the length of hospital stay, reduced the frequency of discharge to a nursing home, and potentially reduced cost.
Evidence favoring treatment of stroke patients on specialized stroke care units is accumulating.1 2 3 4 5 6 7 However, small samples and selection bias hamper the applicability of results obtained from previous studies. Nevertheless, in view of these data the favored prospective design of complete randomization of unselected patients to specialized stroke care units or routine care raises a serious ethical dilemma.
The unique design of the healthcare delivery system in Denmark has made it possible to overcome some of the problems in the design of a test of the effectiveness of dedicated stroke unit treatment of unselected stroke patients. The Danish healthcare system requires delivery of free, high-quality hospital care to all patients regardless of income, disease, and home community. All acute stroke patients are treated at their respective home community hospitals. Until recently stroke treatment was organized identically at two neighboring communities within the Copenhagen area; both communities offered routine management on medical and neurological wards to patients with acute stroke. However, restructuring of the medical wards in one of the communities, but not the other, required treatment for all stroke patients on a dedicated stroke care unit. This local development made it possible to test, at a community-based level, whether stroke unit treatment is more effective than routine management on medical and neurological wards.
Subjects and Methods
The study was designed to compare outcome after stroke in two neighboring communities within the city of Greater Copenhagen, the communities of Frederiksberg and Bispebjerg. The admission rate of acute stroke patients in the area was high, at 88%.8 Each community was served by a single hospital, and all stroke patients were treated and rehabilitated at the hospital that served their home community. The incidence rates of stroke patients admitted to the hospital were identical in the two communities, at 3.6/1000 inhabitants.
In the community of Bispebjerg acute treatment as well as all stages of rehabilitation took place exclusively on the stroke unit at Bispebjerg Hospital, regardless of the age of the patient, the severity of the stroke, and the condition of the patient before the stroke. The stroke unit is located in the Department of Neurology. The department has 72 beds, of which 61 beds are occupied by the stroke unit.
In the community of Frederiksberg acute treatment as well as rehabilitation took place exclusively on a general neurological ward (patients aged 70 years or younger) and on two medical wards (patients older than 70 years) at Frederiksberg Hospital. No stroke patient from the community of Frederiksberg received treatment on a stroke unit during the study period.
Both Bispebjerg and Frederiksberg hospitals are public hospitals serving urban communities of 124 000 and 85 600 inhabitants, respectively. The two communities share a border, and the distance between the two study hospitals is 2 miles. Expenditures per bed-day are comparable between the hospitals. They both possess all major medical and surgical specialties as well as radiological and laboratory facilities needed for modern evaluation of stroke patients. There are no differences in the quality of the staff employed in the caretaking of patients at the two hospitals, and salaries for doctors, nurses, and training staff are identical at the two hospitals. Both hospitals have a good reputation, and one is not considered superior to the other. Home care policies after discharge are similar in the two communities. Frederiksberg is a more wealthy community than Bispebjerg, and the average income is higher.
In patients treated at the stroke unit a plan for evaluation, medical treatment, and rehabilitation was made on admission. A standardized diagnostic evaluation program including routine blood tests, electrocardiogram (ECG) on admission, chest x-ray, and CT scan was performed in each patient. Other diagnostic procedures such as Doppler examination of the carotid arteries, single-photon emission computed tomography, and echocardiography were performed when required. Aspirin (150 mg/d) was given to all patients with ischemic strokes. Anticoagulation treatment with warfarin was given to ischemic stroke patients with atrial fibrillation if not contraindicated by the overall clinical condition. Anticoagulation treatment with heparin and warfarin was given to patients with stroke in progression if not contraindicated. As prophylaxis against pulmonary embolism, patients with severe lower extremity paresis were given thromboembolic disease support hose. Low-dose heparin was not used. Rehabilitation was given to all patients by a trained stroke team of medical and nursing staff, physiotherapists, occupational therapists, speech therapists, and neuropsychologists. The rehabilitation program was individualized according to the needs of the patient and started on the patient’s arrival at the stroke unit.
Patients on general neurological and medical wards at Frederiksberg Hospital were given the traditional treatment offered to patients at Danish hospitals without a stroke unit; there was no standardized program for diagnostic evaluation except for CT scan. Physical therapy, occupational therapy, and speech therapy were given when prescribed by physicians.
Inclusion Criteria and Data Collection
The principal investigator (H.S.J.) was responsible for the prospective collection of data in the two communities. Patients from Frederiksberg community were studied from April 1, 1989, to March 31, 1990, during H.S.J.’s engagement at Frederiksberg Hospital. Patients from Bispebjerg community were studied from the establishment of the stroke unit on September 1, 1991, to September 30, 1993, during H.S.J.’s engagement at Bispebjerg Hospital. An analogous approach to data collection was thus ensured by the fact that the same neurologist was responsible for the data collection in the two populations. Funding made it possible to extend the length of the second inclusion period.
The following information was obtained on patients in both communities: (1) age, sex, marital status, residence (home/nursing home); (2) previous myocardial infarction (a clinical event diagnosed as a myocardial infarction confirmed by hospital records and ECG and/or enzymes), ischemic heart disease (a history of ischemic heart disease), hypertension (in treatment with antihypertensive drugs at the time of admission), diabetes (known diabetes before stroke), smoking (daily smoking of any kind of tobacco), former stroke, atrial fibrillation (if present on ECG obtained on admission), and blood pressure on admission; and (3) consciousness (lowered/not lowered), paresis of extremity (present or not), and aphasia (present or not).
This information was entered into a computerized database together with CT data (see below) and information regarding length of hospital stay, discharge status (home/nursing home), and date of death within the first year after stroke.
Patients admitted more than 2 weeks after stroke were excluded from the study. Ninety-five percent of patients in both communities were admitted within the first week after onset.
To avoid observer bias, only strictly objective outcome measures were chosen: mortality and discharge to a nursing home. The length of hospital stay was chosen to evaluate the speed of recovery and cost-effectiveness.
Definition of Acute Stroke
Stroke was defined according to World Health Organization criteria: rapidly developed clinical signs of focal disturbance of cerebral function, lasting more than 24 hours or leading to death, with no apparent cause other than vascular origin.9 Traditionally, patients with subarachnoid hemorrhage are treated in neurosurgery departments in Denmark and were therefore not included.
CT was performed with a Siemens Somatom DR scanner at Bispebjerg Hospital and with a Phillips Somoscan 310 at Frederiksberg Hospital. The time from the onset of stroke to CT examination depended on the accessibility of the scanner. In most cases the scan was done within 2 weeks after the stroke. All scans were interpreted by two radiologists participating in the study (H.O.R., Bispebjerg, and K.L., Frederiksberg), both blinded to the clinical data. Description included type and size of lesion. Size was determined as the largest visible diameter of the lesion on CT.
Statistical analysis was performed with the use of the spss package (Microsoft Windows, version 5). In the univariate analyses Student’s t test was used for the comparison of continuous data between two groups. The χ2 test was used for noncontinuous data. For dichotomous data, multiple logistic regression was used to test for independent variables. All variables of interest were tested with the backward procedure. For continuous data, multiple linear regression was used to test for independent variables. All variables of interest were tested with the backward procedure. The level of significance was chosen to be P<.05.
The study was approved by the Ethics Committee of Copenhagen.
Comparison of Demography, Comorbidity, and Stroke Severity
Patient demographics were essentially similar between the hospitals. Specifically, there were no differences in age, sex, marital status, and nursing home residence (Table 1⇓).
The risk factor distribution in the two groups is shown in Table 2⇓. Hypertension and diabetes were significantly more frequent in patients treated at the stroke unit. The groups were comparable regarding frequency of previous myocardial infarction, ischemic heart disease, smoking, previous stroke, atrial fibrillation, and blood pressure on admission.
A comparison of initial stroke severity, type of stroke, and size of lesion is shown in Table 3⇓. Similar to the demographic comparison, there were no differences between groups. Level of consciousness, frequency of extremity paresis and aphasia, frequency of stroke subtypes, and lesion diameter were similar.
In-hospital mortality was 21% lower in patients treated on the stroke unit versus on the general wards (Table 4⇓). Furthermore, in patients treated on the stroke unit the case-fatality rate was 24% lower at 30 days, 20% lower at 6 months, and 18% lower at 1 year after stroke. The discharge rate to a nursing home was 20% lower, and the discharge rate to the patient’s own home was increased by 16% for those treated on the stroke unit (Table 4⇓). These differences were all statistically significant. Logistic multiple regression analyses demonstrated that these relations were independent of age, sex, presence of atrial fibrillation, former myocardial infarction, former stroke, ischemic heart disease, diabetes, hypertension, blood pressure on admission, level of consciousness, aphasia, and extremity paresis. Treatment on the stroke unit did in fact reduce the relative risk of death by approximately 50%, reduced the relative risk of discharge to nursing home by approximately 40%, and almost doubled the relative chance of discharge to home (Table 5⇓). The regression equations predicted 85.1% of the in-hospital mortality, 89.4% of the 30-day mortality, 80.6% of the 6-month mortality, 77.2% of the 1-year mortality, 87.2% of the discharges to a nursing home, and 78.4% of the discharges to a patient’s own home.
Length of Hospital Stay
The length of hospital stay was reduced by 30% in patients treated on the stroke unit in the univariate analysis (Table 4⇑). In Denmark, patients who are to be discharged to a nursing home usually spend some time in hospital waiting for a place. Because the waiting times for a nursing home can be different, we also calculated the length of hospital stay excluding patients discharged to a nursing home. The length of hospital stay was still lower (23%) in patients treated at the stroke unit. Again, the differences were statistically significant. Linear multiple regression analysis (R2=.31; significance of F: P<.0001) demonstrated that the influence of stroke unit treatment on the length of hospital stay was independent of the contribution of age, sex, presence of atrial fibrillation, former myocardial infarction, former stroke, ischemic heart disease, diabetes, hypertension, blood pressure on admission, level of consciousness, aphasia, extremity paresis, and discharge to a nursing home. Treatment in the stroke unit independently reduced the length of hospital stay by a mean of 13.1 days (SE=3.3, β-coefficient [standardized]=−.11, P=.0001).
An analysis comparing patients treated on the neurological ward (aged ≤70 years; n=85) with patients treated on the stroke unit (aged ≤70 years; n=267) showed results similar to the overall results. Because of the relatively small number of patients analyzed, only the length of hospital stay was reduced significantly. In such patients treated on the stroke unit, the discharge rate to a nursing home was 16% lower (6.9% versus 8.2%) compared with patients treated on the neurological ward. Mortality during hospital stay was 22% lower (14% versus 18%). The length of hospital stay was reduced by 40% (P=.001; Table 4⇑).
This is the first study to report on the effect of stroke unit treatment in unselected stroke patients. Previous stroke unit trials1 2 3 5 6 have been limited by studying small numbers of patients (n<300) in selected hospital settings and by excluding the most severe strokes. This community-based study of 1241 acute stroke patients demonstrates that both mortality and the discharge rate to a nursing home were markedly lower in unselected patients treated on a dedicated stroke care unit compared with patients treated on general wards. Stroke unit treatment was not only more effective in terms of reducing mortality and nursing home discharge; patients treated on the stroke unit also spent on average 13 fewer days in the hospital, thereby reducing the average length of hospital stay by 30%.
The special organization of the Danish healthcare system made this study possible. This allowed us to study the effect of stroke unit treatment at an unselected, community-based level. Patients were allocated to treatment on the stroke unit/general ward not by chance but by home community. This design possesses several advantages over the traditionally randomized stroke unit studies. A randomized study would require informed consent from every patient. The patients with the most severe strokes would thus not qualify for inclusion (eg, because of decreased consciousness, disorientation, or severe aphasia). Randomized studies of stroke units cannot be blinded, and it is therefore impossible to exclude the possibility of observer bias in regard to discharge policy and evaluation of functional outcome. Furthermore, in comparing stroke unit treatment and treatment on general wards at the same hospital, it would be difficult to prevent dissemination of the program and methods used on the stroke unit to the general wards (treatment contamination). Finally, because of the ethical problems of denying some patients a therapy considered to be effective, it has become impossible to conduct a traditional randomized study of the effect of stroke unit treatment. These serious problems have been overcome by the present design because we compared stroke unit treatment with general ward treatment in two different communities.
The allocation-by-community method could have resulted in incomparable groups of stroke patients if the two community populations had differed in demographic or medical factors, but the two groups were comparable in all relevant aspects: age, sex, marital status, living place before stroke, and stroke severity. Patients treated at the stroke unit had a higher rate of comorbidity, which could have biased the results toward lower significance. When all possible differences between groups were accounted for in the multivariate analysis, the conclusion of the study remained unchanged.
A difference in general quality of treatment at the two hospitals could also have biased the results due to the allocation-by-community method used. However, the quality of treatment was directly comparable except for the presence of the stroke unit. Hospital expenditures, staff wages, the quality of physicians, nurses, training staff, and the reputation of the two hospitals were comparable.
Secular changes in mortality could theoretically interfere with results because of the small difference in study periods between the two communities. However, a profound spontaneous change in mortality in the study period is not probable, and no new treatment was introduced in the study period. Thus, stroke unit treatment is a likely explanation of the better outcome and the shorter hospital stay found in patients from the Bispebjerg community.
All studies from stroke units have reported a decrease in mortality,1 2 3 5 6 but the decrease was significant in only one study and referred only to 6-week mortality.2 A sustained and significant reduction of mortality as shown in our study has not been reported before, most likely because of the small samples studied in previous studies. However, in a statistical review4 of other studies a significant reduction of 1-year mortality (odds ratio [OR], 0.79) was estimated. In our study the reduction was even more pronounced, as the OR was almost halved.
The reduction in mortality in patients treated at the stroke unit did not lead to the increased discharge rate to a nursing home that might have been expected. The discharge rate to a nursing home was reduced. A reduction in the need of institutionalization after stroke has been demonstrated in four of five stroke unit trials,1 2 3 5 6 but the reduction was significant in only two studies.2 3 In a recent statistical review of seven stroke unit/stroke rehabilitation trials, it was not possible to demonstrate a significant reduction of the need of institutionalization after stroke.7 The combined reduction of mortality and need of institutionalization after stroke was higher (OR, 0.67) than the reduction of mortality alone (OR, 0.79), indicating that the decrease in mortality does not result in an increase in the need of institutionalization. In our study the odds for being discharged to a nursing home were markedly reduced in patients treated on the stroke unit (OR, 0.61). The unselected inclusion of patients from the entire spectrum of stroke severity, particularly the most severely affected patients with a high risk of being institutionalized after the stroke, is a likely explanation of the large reduction of the need of institutionalization found in patients treated in the stroke unit in the present study.
All stroke unit trials report that patients treated at a stroke unit are more frequently discharged to home.1 2 3 5 6 Nevertheless, the studies are difficult to compare because of such factors as differences in design, inclusion/exclusion criteria, and duration of treatment in the stroke unit. Our study, however, confirms the results of these studies. Odds for home discharge from the stroke unit were nearly doubled in this study.
We found that stroke unit treatment reduced the length of hospital stay by 25% in an unselected stroke population when confounding factors were considered. Although significant in only one previous study,3 the length of hospital stay was reduced by 25% to 50% in three recent stroke unit trials.2 3 6 Our study confirms that the treatment of stroke patients on stroke units not only influences recovery by improving outcome but hastens recovery considerably.
The reduction in mortality, discharge rate to a nursing home, and length of hospital stay were independent of whether patients not treated in the stroke unit were treated in neurological or medical wards. This finding indicates that neurological departments per se are not equivalent to a stroke unit.
Stroke units are not standardized entities. They may differ in several aspects from country to country or even within the same country. Most European stroke units (the source of much of our data) handle the patients from the time of admission to the end of rehabilitation,1 6 8 as was also the case in this study. Other stroke units may care for the patients only during the hyperacute phase of stroke (eg, in the first few days after onset), after the hyperacute phase, or only in the subacute phase of stroke3 5 (eg, from 2 to 4 weeks after onset). Results may therefore not be readily generalizable.
The reason(s) why organized stroke rehabilitation works remains largely undetermined. In this study stroke unit rehabilitation differed from traditional stroke treatment on general wards in several aspects: All patients went through a standardized evaluation program, intensive observation was performed regularly in the first 2 days of hospital stay, treatment was initiated as soon as the patient arrived at the stroke unit and performed within the frame of a multidisciplinary integrated approach, treatment was goal-oriented and guided by weekly multidisciplinary sessions, physiotherapy and occupational therapy were performed within the unit, activities of daily living were consistently used as part of the training by the nursing staff, and weekly assessments of neurological deficits (Scandinavian Stroke Scale) and functional disabilities (Barthel Index) were performed to monitor treatment effect. The mere concentration of stroke patients on stroke units leads to an increased experience of physicians and teams in handling stroke patients, and this may in turn improve outcome. Indredavik et al2 have shown that earlier start of therapy at the stroke unit appears to be the most important factor for home discharge of stroke patients. Stroke units do not spend more time on physiotherapy or occupational therapy.6 10 11 Therapy is, however, more intensive in the initial phase and, as shown by Kalra,11 is directed more specifically toward the physical/cognitive deficits of the patient than is the case for stroke patients treated on general wards. Treatment of medical complications of stroke was similar on the stroke unit and the general wards and cannot account for the difference in mortality. Early mobilization through early training, which may prevent pneumonia, pulmonary embolism, and bedsores, is a likely explanation of the reduction in mortality.
From an economic perspective, stroke units appear promising. A 25% reduction in the length of hospital stay, as found in this study, implies a considerable saving of bed-days: 1313 bed-days per 100 stroke patients. Three places at nursing homes per 100 stroke patients were saved. We conclude that stroke units not only save lives and improve and hasten outcome but may also save money.
This study was supported by grants from the Danish Health Foundation, the Danish Heart Foundation, the Ebba Celinders Foundation, and the Gangsted Foundation. The authors wish to thank Jørgen Hilden for statistical advice.
Reprint requests to Henrik Stig Jørgensen, MD, Department of Neurology, Bispebjerg Hospital, Bispebjerg bakke 23, DK-2400 Copenhagen NV, Denmark.
- Received February 10, 1995.
- Revision received April 17, 1995.
- Accepted April 17, 1995.
- Copyright © 1995 by American Heart Association
Strand T, Asplund K, Eriksson S, Hägg E, Lithner F, Wester PO. A non-intensive stroke unit reduces functional disability and the need for long-term hospitalization. Stroke. 1985;16:29-34.
Indredavik B, Bakke F, Solberg F, Rokseth R, Haaheim LL, Holme I. Benefit of a stroke unit: a randomized controlled trial. Stroke. 1991;22:1026-1031.
Kalra L, Dale P, Crome P. Improving stroke rehabilitation: a controlled study. Stroke. 1993;24:1462-1467.
Stevens RS, Ambler NR, Warren MD. A randomized controlled trial of a stroke rehabilitation ward. Age Ageing. 1984;13:65-75.
Garraway WM, Akthar AJ, Hockey L, Presscott RJ. Management of acute stroke in the elderly: preliminary results of a controlled trial. BMJ. 1980;280:1040-1044.
Langhorne P. Specialist stroke unit care improves survival and functional outcomes: a statistical overview. Cerebrovasc Dis. 1994;4:207. Abstract.
Jørgensen HS, Plesner AM, Larsen K, Hübbe P. Marked increase of stroke incidence in men between 1972 and 1990 in Frederiksberg, Denmark. Stroke. 1992;23:1701-1704.
Report of the WHO Task Force on Stroke and Other Cerebrovascular Disorders. Stroke—1989: recommendations on stroke prevention, diagnosis, and therapy. Stroke. 1989;20:1407-1431.
Indredavik B, Haaheim LL, Bakke F. Treatment in an acute stroke unit: which aspects are important? Cerebrovasc Dis. 1994;4:207. Abstract.
Kalra L. Does age affect benefits of stroke unit rehabilitation? Stroke. 1994;25:346-351.