Background and Purpose We sought to clarify the way in which organized inpatient (stroke unit) care can produce reductions in case fatality and in the need for institutional care after stroke.
Methods We performed a secondary analysis of a collaborative systematic review of all randomized trials that compared organized inpatient (stroke unit) care with contemporary conventional care. Nineteen trials were included, of which 18 (3246 patients) could provide outcome data on death, place of residence, and final functional outcome. Data were less complete (but always available for at least 12 trials; 1611 patients) for subgroup analyses examining timing and cause of death and outcomes in patients with different levels of severity of initial stroke.
Results The reduction in case fatality of patients managed in a stroke unit setting developed between 1 and 4 weeks after the index stroke. The reduction in the odds of death was evident across all causes of death and most marked for those deaths considered to be secondary to immobility. However, data were insufficient to permit a firm conclusion. The relative increase in the number of patients discharged home from stroke units as opposed to conventional care was largely attributable to an increase in the number of patients returning home physically independent. Across the range of stroke severity, stroke unit care was associated with nonsignificant increases in the number of patients regaining independence.
Conclusions Within the limitations of the available data, we conclude that organized inpatient stroke unit care probably benefits a wide range of stroke patients in a variety of different ways, ie, reducing death from secondary complications of stroke and reducing the need for institutional care through a reduction in disability.
A recent systematic review of the randomized trials that have compared organized inpatient (stroke unit) care with contemporary conventional care has indicated that stroke patients who are managed in an organized (stroke unit) setting are less likely to die, remain physically dependent, or require long-term institutional care.1 These observations were surprising because the stroke units studied did not routinely employ any medical or surgical interventions that might be expected to influence the pathological process or immediate neurological complications of stroke disease.1 Furthermore, no routine medical or surgical therapies have yet been shown to be effective in acute stroke.2 3
It has previously been suggested that stroke unit care could be effective through a number of mechanisms.4 5 6 The provision of standardized assessment and early management protocols may allow a more accurate diagnosis to be reached, more appropriate investigations, and more appropriate individualized patient care. Secondary complications of stroke (eg, chest infection, venous thromboembolism) could possibly be prevented through improved assessment procedures and early active rehabilitation. Earlier, more intense, and better coordinated rehabilitation procedures may assist patients in achieving their maximal functional outcome. However, to date there has been a lack of direct evidence to support any of these proposals.
An improved understanding of the way in which stroke units exert their beneficial effect on stroke outcomes is of considerable clinical importance because it could help to elucidate the mechanisms of improved recovery and hence important practical aspects of stroke care.
In this report we use data available from a collaborative systematic review of the available randomized trials of stroke unit care1 to identify the aspects of recovery for which stroke unit care appeared to make the greatest impact. In particular, we wished to identify the following: (1) What causes of death were most likely to be prevented? (2) Did stroke unit care result in more patients surviving in a physically dependent state? (3) Did all groups of patients obtain similar benefit from stroke unit care?
The methods used for the collaborative systematic review of the randomized trials of stroke unit care have been described previously.1 In summary, we identified randomized trials of organized inpatient (stroke unit) care for the period up to December 1995 using a variety of search strategies.1 We aimed to include all trials that compared management in an organized (stroke unit) setting with that of contemporary conventional care (usually provided within general medical wards). The coordinators of all relevant randomized trials were then contacted and invited to join a collaborative review group (Stroke Unit Trialists’ Collaboration). The trial coordinators provided data in a standardized format concerning the trial characteristics, patient selection criteria and characteristics, and the numbers of patients in each outcome group.1
Subgroup data (either in the form of tabular or individual patient data) were also sought, in particular the initial stroke severity, which was defined in terms of the functional status of the patient at the time of randomization.1 In this categorization, a patient with a mild stroke had a functional status approximately equivalent to a Barthel Index score of greater than 50/100 within the first week after the stroke and greater than 65/100 within 2 weeks. Moderate strokes were characterized by a score intermediate between the mild and severe subgroups. A severe stroke was equivalent to a Barthel Index score of less than 15/100 within the first week and less than 20/100 within 2 weeks after stroke.
The main outcomes of interest in this analysis were (1) death, (2) the duration elapsed between the index stroke and death, (3) the certified cause of death (ie, clinician’s diagnosis), (4) the final functional status (Rankin score or equivalent measure of dependency, and (5) the requirement for long-term institutional care (ie, within a residential home, nursing home, or hospital setting).
The certified primary cause of death reported was allocated into the following categories7 : (1) neurological: death attributable to the index stroke or recurrent stroke, ie, stroke, cerebral infarction, brain herniation, cerebral edema, recurrent stroke; (2) cardiovascular: myocardial infarction, congestive cardiac failure, cardiac arrhythmia, cardiac arrest; (3) complications of immobility: any death that might reasonably be considered a complication of immobility, ie, sepsis (particularly chest or urinary tract), venous thromboembolism, decubitus ulceration; and (4) other causes: other illnesses (eg, malignancy).
Patterns of case fatality over time were analyzed as the proportion of patients, for whom information was available, who were known to be dead at specific census times after the index stroke. This simple approach, which provides a series of “snapshots” of outcomes at various census times, was used because insufficient individual patient data were available for more sophisticated survival curve analysis.
Relative differences in dichotomous outcomes were analyzed by calculating the odds ratio (OR) (plus the 95% confidence interval [CI]) of an adverse outcome occurring in the stroke unit group relative to the control (conventional care) group.8 9 We used the risk difference to calculate the absolute outcome rates (ie, the proportion of patients with a particular outcome) because this can provide additional clinical information to relative outcomes. However, when the results of the stroke unit and control groups were compared, all calculations of statistical significance were based on the z statistic of the OR of that comparison.8 9
When data from several trials were used to calculate a summary result, we calculated the heterogeneity between the individual trial data contributing to that summary result using standard techniques.8 9 Nonsignificant heterogeneity tests (P>.05) indicate that the results from the individual trials were all compatible with the summary result. Fixed effects statistical models were used8 unless heterogeneity tests were significant when a “random effects” model was used.9
Description of Trials
A total of 19 trials were identified up to December 1995; 17 were formally randomized with the use of random numbers or sealed envelopes (References 1010 -25 and A. Svensson, P. Harmsen, L. Wilhelmsen, unpublished data, 1988), and an additional 2 trials (containing a total of 405 patients) used informal procedures based on bed availability26 or a strict admission rota.27 Exclusion of the 2 quasi-randomized trials would have no substantial effect on any of the subsequent conclusions. One of the trials13 has not yet been completed; the remaining 18 trials have randomized a total of 3249 patients.
Ten of the 18 trials (which randomized 2063 patients) evaluated units with an immediate admission policy; the remaining 8 trials (1186 patients) were of units in which admission was usually delayed 1 to 2 weeks after stroke.2
Typically all patients allocated to organized inpatient (stroke unit) care received inpatient rehabilitation characterized by a period (of up to several weeks) of coordinated multidisciplinary rehabilitation by a team with a specialist interest in stroke disease and/or rehabilitation that had programs of education and training in stroke.1 Most of the control patients (1346 patients) received conventional care in a general medical ward that did not incorporate the above characteristics. A small number (277) of control patients were exposed to some multidisciplinary rehabilitation in a mixed rehabilitation setting.11 20 21 24 For the purposes of this analysis, they were analyzed with the rest of the control group who were managed in general medical wards.
Data were available on the time of death for 14 trials (2463 patients randomized), of which 10 trials could provide information on exact date of death and 4 gave information at census times. Fig 1⇓ illustrates the proportion of patients who were known to be dead at intervals after the index stroke. In 13 of the 14 trials the rise in case fatality among patients exposed to organized (stroke unit) care was less marked than (or the same as) those exposed to conventional care. The apparent number of lives saved (ie, the proportion dead in the conventional care setting minus the proportion dead in the stroke unit setting) is also shown. This indicates that the observed differences largely developed during the period of 1 to 4 weeks after the index stroke. These results do not differ substantially if we analyze separately data from stroke units that had an acute admission policy and those that routinely employed a delayed admission policy because the majority of death events (507 [76%] of all recorded deaths) occurred in trials in which an acute admission policy was examined.
Cause of Death
Information on cause of death was available for 12 trials (1611 patients randomized). The proportions of stroke unit and control group patients dying within particular categories of certified cause of death were as follows: (1) neurological: 9.2% stroke unit patients, 10.3% control subjects (OR, 0.92; 95% CI, 0.66 to 1.28); (2) cardiovascular: 5.2% stroke unit patients, 7.0% control subjects (OR, 0.72; 95% CI, 0.47 to 1.09); (3) complications of immobility: 3.8% stroke unit patients, 6.3% control subjects (OR, 0.62; 95% CI, 0.39 to 0.97); and (4) other causes: 3.6% stroke unit patients, 4.2% control subjects (OR, 0.90; 95% CI, 0.53 to 1.51). There was no significant heterogeneity between trials within each of the outcome groups examined (P>.2 in each case).
It is clear that the analysis lacks sufficient statistical power to draw unequivocal conclusions. We were also limited to using the certified cause of death because insufficient numbers of autopsies were performed to allow a meaningful sensitivity analysis based on the autopsy-proven cause of death.
Other Patient Outcomes
To estimate the impact of stroke unit care on a variety of patient outcomes, we calculated the proportion of patients in four outcome categories at the end of scheduled follow-up (median, 1 year after stroke). Table 1⇓ contains these data, which were available for 14 trials (2770 patients randomized). Stroke unit care was associated with an increase in the number of patients residing at home in an “independent” state (Rankin score 0 to 2). There was only a marginal increase in the odds of a patient being at home in a “dependent” state (Rankin score 3 to 5), and there were reductions in the odds of death or requiring institutional care.
There was no significant heterogeneity (P>.2) between the individual trials contributing to the analyses with the exception of the home (dependent) subgroup, which showed a χ2 of 27.1 (13 df; P<.05). In absolute terms the increase in independent survival appears to be the most striking consequence of organized inpatient (stroke unit) care.
The main limitation in this analysis is the lack of “blinding” of functional assessments in some trials. However, results were similar if restricted to those trials14 16 17 20 26 that employed an unequivocally blinded outcome assessment.
Fig 2⇓ shows the proportion of subjects living at home and the cumulative difference between stroke unit patients and the control group.
The apparent impact of stroke unit care on patients with different degrees of initial stroke severity is shown in Table 2⇓, which presents the numbers of patients surviving in either a physically dependent (Rankin score 3 to 5) or independent (Rankin score 0 to 2) state. Data were available for 13 trials (2091 patients) at the end of scheduled follow-up (median, 1 year).
Patients with mild stroke managed in a stroke unit showed no net increase in survival but tended to be more likely to regain independence. Patients suffering a stroke of moderate severity showed a trend toward both increased survival and increased independent survival. Stroke unit care resulted in an apparent increase in both independent and dependent survival of severe stroke patients. The severe stroke patient group was the only one in which stroke unit care also resulted in an increase in physically “dependent” survivors. Overall, the trend is toward improvement in all outcome groups, although conclusions are limited by the relatively small patient numbers. There was no significant heterogeneity (P>.1) between trials contributing to the summary results. Restriction of the analysis to those trials14 16 17 20 26 that had used an unequivocally blinded outcome assessment produced a similar pattern of results.
This article presents one part of a larger analysis of the randomized stroke unit trials. The first (hypothesis testing) component of this project1 used an a priori hypothesis that organized inpatient (stroke unit) care is more effective than conventional care and obtained primary outcome data (death, institutionalization, dependency) from all relevant trials. All three primary outcomes were significantly less frequent among patients managed in an organized stroke unit setting compared with conventional care (usually in general medical wards).1 The current (secondary) analysis attempted to identify the way in which these apparent benefits were achieved. This analysis is therefore more exploratory and is limited by two main problems. First, the analyses are frequently not based on predefined a priori hypotheses and therefore could be subject to bias. Second, the analyses were frequently restricted to incomplete data sets (although data were always present for randomly matched stroke unit and control groups and were available for more than half the relevant trials). However, a knowledge of the methodological limitations should serve to qualify the conclusions that can be drawn.
There have been many suggestions about how organized stroke unit care could improve outcomes after a stroke (see 5 6 28 ). It has been considered unlikely that the initial stroke pathology and the immediate neurological consequences of the stroke could be substantially influenced by a nonspecific intervention such as stroke unit care.29 However, a number of authors6 29 have suggested that the frequency of several common complications after stroke (in particular, cardiovascular complications, venous thromboembolism, and infections) may be influenced. Although the current analysis appears to support this view, there is insufficient statistical power to provide an unequivocal conclusion, even within this pooled analysis. It is also important to recognize that we were limited to using the certified cause of death, which is a potentially inaccurate source of information (although in most circumstances it will represent the best information available). Further circumstantial support for the view that stroke unit care may reduce secondary complications of stroke is provided by the observations on the timing of deaths within the stroke unit and control groups (Fig 1⇑). Most of the deaths prevented were those occurring between 1 and 4 weeks after the index stroke, which is the period of time in which many of the complications of immobility are believed to occur.7 29
It is possible to speculate on several possible mechanisms by which organized stroke unit care could reduce deaths due to secondary complications after stroke. A careful and systematic assessment of dysphagia may reduce the risk of aspiration and subsequent chest infections. A reduction in the use of urinary catheters could reduce the risk of urinary tract infection. It is also possible that stroke unit staff offered more aggressive management of infections once these complications had become established. Programs of early activation and mobilization may reduce the risk of venous thromboembolism or cardiovascular events. Unpublished data concerning the management strategies within the stroke unit trials suggest that an insufficient number of patients were exposed to specific drug or surgical therapies for these to have any significant impact on the pooled results from the stroke unit trials.
We had previously observed that organized stroke unit care resulted in a reduction in the need for long-term hospital or institutional care.1 This reduction could in theory have resulted from either a more aggressive discharge policy or from a reduction in the number of patients who remained disabled (and therefore required institutional care). Clearly the former is of doubtful clinical value, while the latter is of great therapeutic interest. The current analysis (Table 1⇑) indicates that the reduction in the need for institutional care is largely attributable to a reduction in patient dependency. Across the range of levels of stroke severity observed, stroke unit care appeared to result in an increase in the numbers of survivors who were judged to be physically independent.
We can speculate on how stroke unit care could reduce disability (dependency) after stroke. A more coordinated and focused program of rehabilitation involving patients and caregivers1 may well allow caregivers to better assist with the rehabilitation process to continue therapeutic strategies beyond formal therapy sessions and thereby allow more patients to achieve independence. Some but not all of the stroke units used a more intensive physiotherapy and occupational therapy input1 than conventional care. In addition, less tangible factors, such as the level of patient motivation and morale, may have been improved in the stroke unit setting. Observational studies30 comparing patient activity within stroke unit and the general ward settings have indicated that stroke unit patients spend more of their time in more appropriate and purposeful activity.
The final question we wished to address was whether the benefits of organized stroke unit care were equally apparent across a range of stroke levels of severity. Our working hypothesis was that patients in the “middle band” of stroke severity would gain most benefit.4 In fact, all groups appeared to benefit, but in different ways (Table 2⇑). Patients with mild strokes, who are at a relatively lower absolute risk of death, showed no net increase in survival, but more survivors regained physical independence. More of the moderate severity stroke patients survived and became independent. Those at highest risk of death (severe stroke patients) showed the largest absolute increase in survival with increases in the numbers of both dependent and independent survivors. Overall, the increase in independent survivors is substantially greater than the increase in dependent survivors, and the net effect of stroke unit care appears to be to shift the distribution of all outcomes in a favorable direction.
In summary, this secondary analysis of a systematic review of the randomized stroke unit trials has indicated that the observed benefits of stroke unit care probably resulted from a reduction in deaths caused by secondary complications of stroke (predominantly complications of immobility) and a reduced requirement for institutional care through a reduction in patient dependency. The net effect of stroke unit care appears to be to shift the distribution of all observed outcomes in a favorable direction.
This study was supported by Chest, Heart, and Stroke, Scotland. Following is a list of collaborators (in alphabetical order): K. Asplund (Professor, Umea University Hospital, Umea, Sweden); P. Berman (Physician, City Hospital, Nottingham, England); C. Blomstrand (Neurologist, Sahlgrenska University Hospital, Goteborg, Sweden); M. Dennis (Secretariat; Senior Lecturer, Western General Hospital, Edinburgh,UK); T. Erila (Neurologist, Tampere University Hospital, Tampere, Finland); M. Garraway (Professor, Public Health Sciences, University of Edinburgh, UK); E. Hamrin (Professor, Linkoping University, Linkoping, Sweden); G. Hankey (Neurologist, Royal Perth Hospital, Perth, Australia); M. Ilmavirta (Neurologist, Central Hospital, Jyvaskyla, Finland); B. Indredavik (Physician, University Hospital, Trondheim, Norway); L. Kalra (Professor, Orpington Hospital, Kent, England); M. Kaste (Professor, University of Helsinki, Helsinki, Finland); P. Langhorne (Coordinator; Senior Lecturer, Royal Infirmary, Glasgow, UK); H. Rodgers (Physician, University of Newcastle, England); J. Sivenius (Professor, University of Kuopio, Kuopio, Finland); J. Slattery (Secretariat; Statistician, University of Edinburgh, UK); R. Stevens (Retired Physician, formerly Dover, England); A. Svensson (Professor, Ostra Hospital, Goteborg, Sweden); C. Warlow (Secretariat; Professor, Western General Hospital, Edinburgh, UK) B. Williams (Secretariat; Physician, Gartnavel General Hospital, Glasgow, UK); S. Wood-Dauphinee (Professor, McGill University, Montreal, Canada). In addition to the listed collaborators, important contributions were also made by D. Deleo (Perth), A. Drummond (Nottingham), R. Fogelholm (Jyvaskyla), N. Lincoln (Nottingham), H. Palomaki (Helsinki), T. Strand (Umea), and L. Wilhelmsen (Goteborg). Carl Counsell and Hazel Fraser (Cochrane Collaboration Stroke Group) provided invaluable assistance with literature searching.
A list of participants in this multicenter study appears at the end of the article.
- Received February 14, 1997.
- Revision received May 6, 1997.
- Accepted June 2, 1997.
- Copyright © 1997 by American Heart Association
Stroke Unit Trialists’ Collaboration. Collaborative systematic review of the randomised trials of organised in-patient (stroke unit) care after stroke. BMJ. 1997;314:1151-1159.
Sandercock P. Managing stroke: the way forward. BMJ. 1993;307:1297-1298.
Garraway WM. Stroke rehabilitation units: concepts, evaluation and unresolved issues. Stroke. 1986;16:178-181.
Ebrahim S. Clinical Epidemiology of Stroke. Oxford, England: Oxford University Press; 1990.
Langhorne P, Dennis MS, Williams BO. Stroke units: their role in acute stroke management. Vasc Med Rev.. 1995;4:33-44.
Peacock PB, Riley CHP, Lampton TD, Raffel SS, Walker JS. Trends in Epidemiology. In: Stewart GT, eds. The Birmingham Stroke, Epidemiology and Rehabilitation Study. Springfield, Ill; Charles C Thomas Publishing; 1972:231-345.
Stevens RS, Ambler NR, Warren MD. A randomised controlled trial of a stroke rehabilitation ward. Age Ageing. 1984;13:65-75.
Garraway WM, Akhtar AJ, Hockey L, Prescott RJ. Management of acute stroke in elderly: follow-up of a controlled trial. Br Med J. 1980;281:827-829.
Kaste M, Palomaki H, Sarna S. Where and how should elderly stroke patients be treated? A randomised trial. Stroke. 1995;26:249-253.
Sivenius J, Pyorala K, Heinonen OP, Salonen JT, Riekkinen P. The significance of intensity of rehabilitation after stroke: a controlled trial. Stroke. 1985;16:928-931.
Wood-Dauphinee S, Shapiro S, Bass E, Fletcher C, Georges P, Hensby V, Mendelsohn B. A randomized trial of team care following stroke. Stroke. 1984;5:864-872.
Aitken PD, Rodgers H, French JM, Bates D, James OFW. General medical or geriatric unit care for acute stroke? A controlled trial. Age Ageing. 1993;22(suppl 2):4-5.
Kalra L, Dale P, Crome P. Improving stroke rehabilitation: a controlled study. Stroke. 1993;24:1462-1467.
Kalra L, Eade J. Role of stroke rehabilitation units in managing severe disability after stroke. Stroke. 1995;26:2031-2034.
Hankey G, Deleo D, Stewart-Wynne EG. Acute hospital care for stroke patients: a randomised trial. Cerebrovasc Dis. 1995;5:228.
Ilmavirta M, Frey H, Erila T, Fogelholm R. Does treatment in a non-intensive care stroke unit improve the outcome of ischaemic stroke? Presented at Nordiska Moter om Cerebrovasculara Sjukdomar; August 23-25, 1993; Jyvaskyla, Finland.
Indredavik B, Bakke R, Solberg R, Rokseth R, Haahein LL, Home I. Benefit of stroke unit: a randomised controlled trial. Stroke. 1991;22:1026-1031.
Strand T, Asplund K, Eriksson S, Hagg E, Lithner F, Wester PO. A nonintensive stroke unit reduced functional disability and the need for long-term hospitalization. Stroke. 1985;16:29-34.
Dennis MS, Langhorne P. So stroke units save lives? Where do we go from here? BMJ. 1994;309:1273-1277.
Millikan CH. Stroke intensive care units: objectives and results. Stroke. 1979;10:235-237.
Lincoln NB, Willis D, Philips SA, Juby LC, Berman P. Comparison of rehabilitation practice on hospital wards for stroke patients. Stroke. 1996;27:18-33.