The stroke mortality rate in Finland has declined steadily

Decreased Severity of Brain Infarct Can in Part Explain the Decreasing Case Fatality Rate of Stroke Heikki Numminen, MD; Markku Kaste, MD; Kari Aho, MD; Olli Waltimo, MD; Mervi Kotila, MD Background and Purpose Case fatality rates for stroke has declined in most Western industrialized countries during recent decades. One possible explanation for this is a decrease in the severity of stroke symptoms. We therefore sought evidence for a change in stroke severity and its relationship with case fatality rates. Methods We compared the severity of symptoms among first-ever stroke patients in 2 population-based prospective stroke registers maintained during 1972 to 1973 and 1989 to 1991 in Finland. Patients who were evaluated by study assistants or the investigator during the first week after the onset of symptoms were included in the study, and their severity of symptoms was assessed with the use of comparable scales modified from the Scandinavian Stroke Scale. Results A total of 244 and 594 patients were registered, and a portion of them (155 [63.5%] and 360 [60.6%]) were included in the analyses in the registers for Espoo-Kauniainen from 1972 to 1973 and for 4 separate districts in Finland from 1989 to 1991, respectively. The death rates during the first week among those who were not included did not differ between the registers. The severity of symptoms decreased significantly between the registers in both patients with brain infarct or intracerebral hemorrhage but not in those with subarachnoid hemorrhage. The severity of symptoms was an independent factor of case fatality at 1 month. Conclusions The severity of symptoms of brain infarcts has decreased and can in part explain the decreased case fatality rate of stroke in Finland. However, the change in patients with intracerebral hemorrhage may be overestimated due to undiagnosed intracerebral hemorrhages in the first register resulting from the lack of brain CT. (Stroke. 2000;31:651-655.) Keys words: cerebrovascular disorders treatment outcome mortality prospective studies The stroke mortality rate in Finland has declined steadily since 1965 in men and since 1955 in women, 1 which is typical for most Western industrialized countries. 2 5 This has occurred despite a stable incidence rate during the 1980s. The decreasing case fatality rates reported in the same studies 3 5 are the most likely cause of this. At least 3 explanations for the observed decrease in case fatality rates have been suggested: (1) enrollment bias during registration includes patients with milder strokes during subsequent years, (2) better supportive and rehabilitative care of acute stroke patients may have decreased the case fatality rate, and (3) the symptoms of acute stroke could have become milder, leading to increased survival rates. To evaluate the impact of stroke symptoms, a valid method for the assessment of stroke severity is necessary. A retrospective method used in previous studies, 6 11 which suggests that strokes are less disabling at the present than they were previously, may be too inaccurate or insensitive for this purpose. The present method, which is based on a prospectively used standard scale to evaluate the severity of symptoms modified from the Scandinavian Stroke Scale, avoids the difficulties involved in the previous studies. The aim of the present study was to compare the severity of stroke in 2 population-based registers during 1972 to 1973 and 1989 to 1991 in Finland and to determine the relationship between stroke severity and case fatality. Subjects and Methods The study area, population, case finding, and case ascertainment for the 2 registers have been published previously. 5,12 The first register covered a semiurban area, Espoo-Kauniainen, from 1972 to 1973 (EK 72 73). 12 The second register was maintained from 1989 to 1991 in 4 separate rural districts in Finland (FHA 89 91). 5 The mean study population consisted of 113 100 and 134 804 inhabitants in the EK 72 73 and FHA 89 91 register regions, respectively. All patients with first-ever stroke and aged 15 years were registered and evaluated by the study assistants for the FHA 89 91 register 5 or the investigator for the EK 72 73 register. 12 The registers were maintained according to the guidelines of the World Health Organization (WHO) Stroke Register. 13 The Scandinavian Stroke Scale 14 was used for assessment of stroke severity in the FHA 89 91 register. It has a 4-grade scale for the evaluation of consciousness and a 5-grade scale for the evaluation of motor function in the arm and leg (Table 1). The WHO Stroke Register form, which assesses consciousness with a 4-grade scale, Received September 15, 1999; final revision received December 2, 1999; accepted December 2, 1999. From the Department of Clinical Neuroscience (H.N., M.K., O.W., M.K.), Helsinki University Central Hospital, Helsinki, Finland; and Department of Neurology (K.A.), Jorvi Hospital, Espoo, Finland. Correspondence to Dr Heikki Numminen, Department of Clinical Neuroscience, Helsinki University Central Hospital, PL 300, 00029 HYKS, Helsinki, Finland. E-mail h.numminen@kimpinen.fimnet.fi 2000 American Heart Association, Inc. Stroke is available at http://www.strokeaha.org 651

652 Stroke March 2000 TABLE 1. Severity Score With Original Descriptions of Definitions Used for Conversion of Patient Status in EK 72-73 Register and Scandinavian Stroke Scale Used in FHA 89-91 Register EK 72-73 FHA 89-91 Consciousness 6 Fully conscious Fully conscious 4 Somnolent Somnolent, can be awakened to full consciousness 2 Semicoma Reacts to verbal command but is not fully conscious 0 Coma Unconscious Eye movements 4 No gaze palsy No gaze palsy 2 Gaze palsy Gaze palsy present 0 Conjugate eye deviation Arm power 6 Normal strength Raises arm with normal strength 5 Raises arm with reduced strength 4.5 Raises arm or reduced strength 4 Raises arm with flexion in elbow 2 Can move, but Can move but not against gravity cannot raise arm 0 No movement Paralysis Leg power 6 Normal strength Raises leg with normal strength 5 Raises leg with reduced strength 4.5 Raises leg or reduced strength 4 Raises leg with flexion of knee 2 Can move, but Can move but not against gravity cannot raise leg 0 No movement Paralysis was used in the EK 72 73 register. In addition, a 5-grade scale for the separate evaluation of motor function in the arm and leg was used in this register. It was possible to convert the raw data recorded in the EK 72 73 register to comparable values in the Scandinavian Stroke Scale (Table 1). In estimation of the severity of stroke, we used only level of consciousness, eye movements, and arm and leg power, which are known to correlate with case fatality. 15 After these values were totaled, they constituted the score used to assess the severity of stroke. This severity score has a minimum of 0 points and a maximum of 22 points (from the most severe to the mildest stroke), which corresponds to the prognostic score used in the Scandinavian Stroke Scale. Only patients who were registered and evaluated during the first week after the onset of stroke symptoms were included in the analyses. The severity scores are expressed as age-adjusted mean values and 95% CIs. We used 2 tests with continuity correction (Statview; Abacus Concepts) and the exact test (Egret; Statistics and Epidemiology Research Corporation) to calculate differences in proportions. An ANOVA (SuperAnova; Abacus Concepts), with age as a covariate, was used to evaluate the differences in severity scores between the registers. The strength of association between the severity score and case fatality was analyzed with linear regression (Statview). Multiple logistic regression (Egret) was used to estimate the effect of various factors on case fatality rates. Factors used in these analyses included patient age and sex, severity score, subtype of stroke TABLE 2. Proportion of Number of Deaths of Stroke Patients Included in Analyses and Case Fatality at 1 Month After Onset of Stroke According to Stroke Subtype and Gender EK 72-73 FHA 89-91 Stroke Subtype Patients, n Case Fatality, % Patients, n Case Fatality, % BI 22/105 21.0 42/304 13.8 Men 11/56 19.6 14/138 10.1 Women 11/49 22.4 28/166 16.9 ICH 8/15 53.3 5/40 12.5 Men 1/5 20.0 5/26 19.2 Women 7/10 70.0 0/14 0.0 SAH 8/31 25.8 4/13 30.8 Men 0/5 0.0 1/3 33.3 Women 8/26 30.8 3/10 30.0 NS stroke 0/4 0.0 1/3 33.3 Men 0/3 0.0 0/1 0.0 Women 0/1 0.0 1/2 50.0 Total 38/155 24.5 52/360 14.4 Men 12/69 17.4 20/168 11.9 Women 26/86 30.2 32/192 16.7 NS indicates nonspecific. (infarct versus other types), and register (time period). Case fatality rates were also compared between the registers in subgroups of patients with brain infarcts (BIs): those with more severe symptoms (severity score 0 to 10), with moderate symptoms (severity score 10.5 to 19.5), and with milder symptoms (severity score 20 to 22). In the FHA 89 91 register, conventional risk factors were recorded. Regression analysis was used to estimate the association of risk factors with the severity score. The available risk factors included treated hypertension, coronary heart disease, heart failure, atrial fibrillation, diabetes, and smoking. A combined factor, cardiovascular disease, was constituted if either hypertension, coronary heart disease, or heart failure was present. Other factors that were evaluated included age, sex, subtype of stroke, and the Rankin scale score to measure independence in activities of daily living before the onset of stroke. Results A total of 244 and 594 patients with first-ever strokes were included in the EK 72 73 and FHA 89 91 registers, respectively, and a portion of them (155 [63.5%] and 360 [60.6%] patients, respectively) were investigated within 1 week after the onset of stroke and included in the analyses. Table 2 presents their number and case fatality according to stroke subtype and sex. The mean ages were 64.0 and 69.8 years for men and 64.5 and 75.3 years for women in the EK 72 73 and FHA 89 91 registers, respectively. In the EK 72 73 register, 15.2% (37 of 244) of the patients were excluded due to death during the first week and 21.3% (52 of 244) were excluded due to a delay in evaluation; respective values for the FHA 89 91 register were 11.3% (67 of 594) and 28.1% (167 of 594) (difference in death rates 3.9%, 95% CI 1.3% to 9.1%, 2 2.06, P 0.152). The severity score for all strokes has increased slightly (ie, strokes have become milder since 1973) (Table 3). The change was seen in BIs and intracerebral hemorrhages (ICHs)

Numminen et al Severity of Brain Infarcts Has Decreased 653 TABLE 3. Age-Adjusted Mean of Severity Score (95% CI) According to Stroke Subtype and Gender and P Values for Significance of Difference Between Registers EK 72-73 FHA 89-91 P BI 13.3 (12.0 14.6) 15.5 (14.8 16.3) 0.004 Men 13.6 (11.9 15.4) 16.7 (15.6 17.8) 0.004 Women 13.1 (11.2 15.1) 14.6 (13.5 15.6) 0.198 ICH 7.5 (3.7 11.3) 14.9 (12.7 17.1) 0.001 Men 8.7 (1.1 16.4) 14.2 (11.8 16.7) 0.078 Women 6.5 (0.8 12.3) 16.4 (11.8 21.0) 0.008 SAH 17.1 (14.2 19.9) 14.6 (9.8 19.4) 0.358 Men 17.8 (12.5 23.2) 10.3 (0.0 21.2) 0.072 Women 16.8 (13.7 20.9) 16.1 (10.3 21.9) 0.818 NS stroke 16.6 (8.8 24.4) 3.5 (0.0 15.7) 0.025 Men 19.0 (5.4 32.6) 6.0 0.289 Women 10.0 2.0 Total 13.4 (12.2 14.5) 15.4 (14.7 16.2) 0.002 Men 13.9 (12.3 15.5) 16.1 (15.1 17.1) 0.025 Women 12.9 (11.3 14.4) 14.9 (13.9 15.9) 0.038 NS indicates nonspecific. but not in subarachnoid hemorrhages (SAHs). The increase in severity score is slightly less in females than in males with BI. The severity score is strongly associated with case fatality (slope 0.028, 95% CI 0.024 to 0.032, P 0.001, r 2 0.27) throughout the patient population (n 515). Table 4 presents the factors associated with case fatality rates for all strokes at 1 month after onset. In addition to the increase in severity score (ie, less severe stroke), low age, and having a stroke later (ie, included in the FHA 89 91 register) were independently associated with decreased case fatality rates. When only BIs were included in the analysis, having a stroke between 1989 and 1991 (ie, included in the FHA 89 91 register) remained an independent factor even after adjustments to the severity score and patient age (Table 4). To analyze whether inclusion in the FHA 89 91 register had any effect on case fatality rates in different subgroups of patients with BIs, these patients were stratified according to their severity score. Case fatality rates for BI patients in the EK 72 73 register were 50% (16 of 32), 9% (5 of 57), and 6% (1 of 16) in the groups of severe, moderate, and mild symptoms, respectively. In the FHA 89 91 register, the respective values TABLE 4. Effect of Various Factors on Case Fatality at 1 Month in Multiple Logistic Regression Analysis in All Patients and in Patients With BI All Patients Patients With BI OR 95% CI OR 95% CI FHA 89-91 0.39 0.21 0.71 0.38 0.21 0.69 Age/1 y 1.03* 1.01 1.06 1.02* 1.00 1.05 Male sex 0.70 0.39 1.24 0.70 0.39 1.23 Stroke other than infarct 1.91 0.95 3.83 Severity score/1 point 0.83* 0.79 0.86 0.82* 0.79 0.86 *Change in the risk for death per unit change in the variable. were 44% (34 of 77), 6% (6 of 96), and 2% (2 of 131). Despite the differences, the exact test did not show any significance in case fatality rate at 1 month between the registers (odds ratio 1.41, 95% CI 0.68 to 2.91, P 0.41). Because changes in stroke risk factors during recent decades have been associated with decreased stroke mortality rates, we wanted to analyze whether the presence of risk factors was associated with the severity of symptoms among patients in the FHA 89 91 register. However, only patient age (slope 0.12, 95% CI 0.17 to 0.07, P 0.001, r 2 0.06) and the Rankin scale score (slope 1.86, 95% CI 2.45 to 1.28, P 0.001, r 2 0.10) before the onset of stroke were significantly associated with the severity of stroke symptoms in a simple linear regression. Neither the stroke subtype nor the presence of cardiovascular disease or atrial fibrillation recorded at the onset of stroke was shown to be a significant factor. Discussion The severity of stroke symptoms among patients with their first-ever stroke has declined between 1973 and 1991 in Finland (ie, the essential signs of stroke, including motor hemiparesis, gaze paresis, and disturbed consciousness, are now milder than during previous decades); however, this change concerns only patients with BI and ICH. The change is overestimated in patients with ICH for methodological reasons discussed later. The results also show that the severity of stroke symptoms is a significant contributor to case fatality rate. In addition, the results indirectly support the hypothesis that better management of patients with BI could be another independent factor in case fatality rate. Our results rely on the appropriate conversion of the severity scores into a modified Scandinavian Stroke Scale. We believe that the converted scores represent the original symptoms of patients as exactly as possible and thus are reliable, although the scales used were not originally the same. The assessment of stroke severity in both registers has occurred prospectively and simultaneously with the enrollment of patients by study assistants 5 or the investigator. 12 In addition, the conversion of scores was performed inflexibly with a microcomputer with the algorithm described in Table 1 without a review of medical records. The appropriate conversion of scales is supported by the observation that neither the severity score in patients with SAH in the present study nor case fatality rates throughout the study population of these patients 5 changed. This is in agreement with the results of another Finnish study, in which the 3-week case fatality rates remained stable from the 1970s to the 1980s despite an intensified neurosurgical treatment policy 16 that favored the hypothesis of a disastrous unchanged course in SAH, which was also observed at the Mayo Clinic. 17 One probable bias arose from the use of brain computed tomography, which was not available at the time of the EK 72 73 register, as a diagnostic tool. In the FHA 89 91 register, a CT scan was performed for 60% of patients. This naturally led to increased sensitivity in finding patients with a small ICH. Thus, the advent of cranial imaging might explain at least in part the changes in case fatality rates related to ICH. We argue, however, against the possibility that this

654 Stroke March 2000 could also be the reason for the observed decrease in severity of symptoms among patients with BI, because patients with low severity scores (ie, severe stroke resulting from the ICH) were also most likely diagnosed in the first register, whereas the patients with slight symptoms and an undiagnosed ICH in the first register could have diluted the trend observed among patients with BI. The argument supporting this judgment is the fact that in the FHA 89 91 register, the mean severity score among patients with ICH was almost as high as the mean severity score among patients with BI. The BI results thus appear reliable, but those for ICH include bias due to the diagnostic differences between the registers. Bias in patient registration is an unlikely reason for our results, because most of the patients enrolled in both registers were hospitalized during the acute phase of their stroke (89% and 86% in the EK 72 73 and FHA 89 91 registers, respectively), and in both time periods, all patients with severe strokes were hospitalized. The possibility, however, that more patients with mild strokes, especially BIs, were included in the FHA 89 91 register than in the EK 72 73 register cannot be ruled out. This is unlikely, however, because the incidence of BI in the FHA 89 91 register has not increased compared with the previous register; on the contrary, it has decreased slightly. 5 In addition, the case fatality rates during the first week among the patients not included in the present analysis did not differ significantly between the registers, supporting the fact that mild strokes in both registers are registered with equal frequency within the first week after the onset of stroke. In the few studies that evaluated trends in the severity of stroke, 6 11 the assessment of symptoms was made retrospectively from medical records, which may have lead to either inaccuracy or insensitivity in the results compared with the present method. All of the investigators reached the same conclusion, however: the severity of stroke has declined during recent decades. The earliest observation is from the Framingham cohort, in which a significant increase in frequency of mild deficits was found from the 1950s to the 1970s. 6 In the Minnesota Heart and Stroke Survey, the decreasing frequency in various stroke symptoms among hospitalized patients continued throughout the 1970s and 1980s. 7,9 The WHO Monica project in northern Sweden also draws the conclusion that the proportion of patients with extensive motor deficits has declined during the late 1980s. 8 Thus, the present prospective study confirms these earlier retrospective observations. What are the reasons for the decreasing severity of symptoms? One possible explanation could be better general health of the population today than 2 decades ago. Reduction in the rate of stroke risk factors could be another reason. In all probability, changes in risk profile may in part be involved in the decreased mortality rates for stroke. In a recent publication from Finland, changes in conventional risk factors during the 1970s and 1980s could predict 71% of the observed fall in mortality rates for stroke in men and 54% of the fall in women. 18 Although we were not able to compare changes between risk factors and case fatalities in the previous study, no association between the case fatality rate at 1 month and the cardiovascular risk factors could be found throughout the FHA 89 91 population, 5 nor could any association be found in the present study between the cardiovascular risk factors and the severity of symptoms in all strokes. A similar conclusion could be drawn in a more detailed risk factor registration. 19 Thus, the risk factors appear to play a minor role both in the case fatality rate and in the severity of symptoms when age, the subtype of stroke, and patient premorbid condition are accounted for. Did medical treatment contribute to the decrease in stroke severity? To date, no evidence is available that any drug treatment, except thrombolytic therapy, which was not used among the registered patients, could influence stroke symptoms during the acute phase. There is some evidence, however, that aspirin treatment can contribute to a reduction in death and dependency caused by ischemic stroke in the long term, 20 and the Chinese Acute Stroke Trial suggested that patients treated with aspirin could have a better early functional outcome. 21 The use of aspirin before the appearance of SAHs could reduce the risk of ischemic complications. 22 On the other hand, neither a population-based study 10 nor a prospective survey 23 that evaluated hospitalized stroke patients within 24 hours after admission could find any supportive evidence for the favorable effect of aspirin on the severity of stroke. The use of anticoagulant or antithrombotic treatment was not registered in the present study, and accordingly this question cannot be evaluated, but this possibility must be considered in future surveys. The results of the logistic regression analysis show that in addition to the severity of symptoms among patients with BI, the time period of the register also exerted an independent positive effect on case fatality rates. Whether this effect is dependent on the above-mentioned possibilities of reducing the rate of stroke risk factors and medical treatments or on better prevention of complications and early rehabilitative care of acute stroke patients cannot be answered here. The lack of evidence for the association of stroke risk factors and case fatality, as well as medical treatments and case fatality, favors the assumption that better overall care of patients with BI could also influence case fatality rates. Our conclusion is that the severity of ischemic BI has decreased and that this change could in part explain the decrease in case fatality rates observed in Western countries during the recent decades. Both the lack of any other evidence for the decrease in case fatality rates and the finding of the register time period as an independent factor contributing to the case fatality rate indirectly support the role of better patient care as another explanation for the decrease in case fatality rates, which tallies with the observation of the Stroke Unit Trialists Collaboration. 24 Acknowledgments This work was supported by the Finnish Heart Association. We would like to thank Associate Professor Seppo Sarna for his statistical assistance and Jaana Immonen, Seija Keskiväli, Eeva Kröger, Pirjo Lyytikäinen, Marja-Liisa Niemi, Oona Tuomi-Nikula, and Markku Karjalainen for registration of the patients. References 1. Aho K. Epidemiology of cerebrovascular diseases in the Nordic countries. Scan J Soc Med Suppl. 1980;14:107 121.

Numminen et al Severity of Brain Infarcts Has Decreased 655 2. Bonita R, Beaglehole R. Stroke mortality. In: Whisnant JP, ed. Stroke: Population, Cohorts, and Clinical Trials. Oxford, UK: Butterworth- Heinemann; 1993:59 79. 3. Bonita R, Broad JB, Beaglehole R. Changes in stroke incidence and case-fatality in Auckland, New-Zealand, 1981 1991. Lancet. 1993;342: 1470 1473. 4. Harmsen P, Tsipogianni A, Wilhelmsen L. Stroke incidence rates were unchanged, while fatality rates declined, during 1971 1987 in Göteborg, Sweden. Stroke. 1992;23:1410 1415. 5. Numminen H, Kotila M, Waltimo O, Aho K, Kaste M. Declining incidence and mortality rates of stroke in Finland from 1972 to 1991: the results of three population based stroke registers. Stroke. 1996;27: 1487 1491. 6. Wolf PA, D Agostino RB, O Neal MA, Sytkowski P, Kase CS, Belanger AJ, Kannel WB. Secular trends in stroke incidence and mortality: the Framingham study. Stroke. 1992;23:1551 1555. 7. McGovern PG, Pankow JS, Burke GL, Shahar E, Sprafka JM, Folsom AR, Blackburn H. Trends in survival of hospitalized stroke patients between 1970 and 1985: the Minnesota Heart Study. Stroke. 1993;24: 1640 1648. 8. Stegmayer B, Asplund K, Wester PO. Trends in incidence, case-fatality rate, and severity of stroke in northern Sweden, 1985 1991. Stroke. 1994;25:1738 1745. 9. Shahar E, McGovern PG, Sprafka JM, Pankow JS, Doliszny KM, Luepker RV, Blackburn H. Improved survival of stroke patients during the 1980s: the Minnesota Stroke Survey. Stroke. 1995;26:1 6. 10. Stegmayer B, Asplund K. Exploring the declining case fatality in acute stroke: population-based observations in the northern Sweden MONICA project. J Intern Med. 1996;240:143 149. 11. Barker BH, Mullooly JP. Stroke in a defined population, 1967 1985: a less lethal and disabling but no less common disease. Stroke. 1997;28: 284 290. 12. Aho K. Incidence, Profile and Early Prognosis of Stroke. Helsinki, Finland: University of Helsinki: 1975. Dissertation. 13. Aho K, Harmsen P, Hatano S, Marquardsen J, Smirnov VE, Strasser T. Cerebrovascular disease in the community: results of a WHO collaborative study. Bull WHO. 1980;58:113 130. 14. Scandinavian Stroke Study Group. Multicenter trial of hemodilution in ischemic stroke: background and study protocol. Stroke. 1985;16: 885 890. 15. Oxbury JM, Greenhall RC, Grainger KM. Predicting the outcome of stroke: acute stage after cerebral infarction. BMJ. 1975;3:125 127. 16. Fogelholm R, Hernesniemi J, Vapalahti M. Impact of early surgery on outcome after aneurysmal subarachnoid hemorrhage: a population-based study. Stroke. 1993;24:1649 1654. 17. Phillips LH, Whisnant JP, O Fallon WM, Sundt TM Jr. The unchanging pattern of subarachnoid hemorrhage in a community. Neurology. 1980; 30:1034 1040. 18. Vartiainen E, Sarti C, Tuomilehto J, Kuulasmaa K. Do changes in cardiovascular risk factors explain changes in mortality from stroke in Finland? BMJ. 1995;310:901 904. 19. Hénon H, Godefroy O, Leys D, Mounier-Vehier F, Lucas C, Rondepierre P, Duhamel A, Pruvo JP. Early predictors of death and disability after acute cerebral ischemic event. Stroke. 1995;26:392 398. 20. International Stroke Trial Collaborative Group. The International Stroke Trial (IST): a randomized trial of aspirin, subcutaneous heparin, both, or neither among 19435 patients with acute ischaemic stroke. Lancet. 1997; 349:1569 1581. 21. Chinese Acute Stroke Trial. CAST: randomised placebo-controlled trial of early aspirin use in 20 000 patients with acute ischaemic stroke. Lancet. 1997;349:1641 1649. 22. Juvela S. Aspirin and delayed cerebral ischemia after aneurysmal subarachnoid hemorrhage. J Neurosurg. 1995;82:945 952. 23. Karepov V, Bornstein NM, Hass Y, Korczyn AD. Does daily aspirin diminish severity of first-ever stroke? Arch Neurol. 1997;54:1369 1371. 24. Stroke Unit Trialists Collaboration. Collaborative systematic review of the randomised trials of organised inpatient (stroke unit) care after stroke. BMJ. 1997;314:1151 1159.