Q J Med 2002; 95: 143-152
© 2002 Association of Physicians
Survival of unselected stroke patients in a stroke unit compared with conventional care
1 From the Department of Medicine, 2 Department of Neurology, and 3 Foundation for Health Services Research (HELTEF), Central Hospital of Akershus, Nordbyhagen, Norway
Received 7 May 2001 and in revised form 3 January 2002
 |
Summary |
|---|
 Top Summary IntroductionMethodsResultsDiscussionAppendixReferences |
|---|
Background: Recent meta-analyses have reported a beneficial effect of stroke units compared with traditional care, both on patient survival and on dependency after one year.
Aim: To determine whether these results can be reproduced outside a clinical trial setting.
Setting: A medium-sized general hospital.
Methods: From 1993 to 1998, all patients aged >60 years with suspected acute stroke were allocated either to a stroke unit or general medical wards according to date of birth (day of the month). Patients were identified retrospectively, using a discharge diagnosis of ICD-9 codes 431, 434 and 436. We assessed 30-day and 1-year survival.
Results: In the stroke unit, 926/1128 patients survived at 30 days, vs. 905/1141 in the general medical wards (p=0.08). Beyond the first 30 days, there was no difference in survival (p=0.27). Under Cox regression analysis, there was a 20% reduction in mortality in the stroke unit after 30 days compared with the general medical wards (RR 0.80, p=0.02) after adjusting for age, gender, stroke type and season of the year.
Discussion: In this, the largest single-centre study to evaluate the survival benefit of a stroke unit, survival at 30 days was increased, although not significantly so. Survival at one year was unchanged.
|
Introduction |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionAppendixReferences |
|---|
The role of stroke units (SU) in the care of patients with acute stroke has been debated for many years.1,2 This controversy arises because the moderate benefits expected can be reliably detected only in a very large organized trial or by combining data from available small trials.3 According to recent meta-analyses, organized in-patient care in a SU improves early survival and independence after one year.2,3 However, available individual and pooled studies have been clinical trials, with the vigorous follow-up of a study setting. Recently, the report from a large multicentre observational study supported the findings of the meta-analyses, though the magnitude of the benefit from the SUs appeared smaller.4
Since 1 January 1993 the Central Hospital of Akershus has had a special allocation procedure for patients above 60 years admitted for acute stroke.5 Patients were allocated by date of birth (day of the month) either to an acute SU in the Department of Neurology or to general medical wards (GMWs) in the Department of Medicine. We wanted to compare the two treatment groups and assess whether the survival benefit of SUs reported from the meta-analyses can be reproduced in a large scale routine setting, outside the context of a closely supervised clinical trial.
|
Methods |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionAppendixReferences |
|---|
Study design and population
The Central Hospital of Akershus county is the only provider of hospital care for its catchment population of 300 000, of whom 50 000 are 60 years or older. In this area, from 1992, general practitioners have been advised to admit all patients with acute stroke to the hospital as early as possible.
From 1 January 1993, patients aged >60 years presenting with symptoms suggesting acute stroke were admitted either to a SU in the Department of Neurology or one of five GMWs in the Department of Medicine.5 This allocation was instituted by the hospital management after opening a new dedicated SU in the Department of Neurology. Prior to 1993, all stroke patients >60 years of age were admitted to GMWs in the Department of Medicine.
Allocation of stroke patients to the two departments from 1993 onwards was by date of birth (day of the month). Between 1 January and 30 April 1993, patients born on the 1st to the 10th day of the month were allocated to the SU in the Department of Neurology, and the rest were treated in GMWs in the Department of Medicine. Following gradual SU expansion, the SU received patients born on the 1st to the 12th day of the month from 1 May 1993, and finally those born on the 1st to the15th of the month from 1 September 1993.
Patients with all types of stroke, except transitory ischaemic attacks (TIAs; focal neurological deficit with complete restoration of function within 24 h), subarachnoid haemorrhages and subdural haematomas were included in the study. We identified patients retrospectively, initially including all consecutive patients aged >60 years admitted between 1 January 1993 and 31 December 1998 with the principal discharge diagnosis codes 431–433, 434 and 436–437 (excluding 432.2) of the International Classification of Diseases, 9th revision (ICD-9). This sample (n=2321) included both patients with first time and recurrent strokes, independent of the time between onset of the acute stroke and the admission. Based on reports of low probability of stroke among patients with several of these ICD-9 codes,6–9 we excluded the remaining patients with the ICD-9 diagnosis codes 432, 433 and 437 (n=52) (Figure 1
). Hence we included 2269 patients for further analysis. By the allocation rule, 1128 should have been allocated to the SU and 1141 to the GMWs (Figure 1). To detect a possible selection bias, we compared the actual number of admissions with stroke in the two allocation groups, compared to expectations based on the allocation by date. The patients intended to be allocated to the SU, as included in our analysis, represented 49.7% of the patient population, compared with 51.1% of expected based on the proportion of birth days available for inclusion (p=0.35).
|
The Department of Neurology is completely separate from the Department of Medicine, with no rotation of physicians or nurses from the one to the other. The SU was allocated two physiotherapists, who did not rotate to the GMWs.
Assessment of baseline characteristics
To compare the presence of risk factors in the groups allocated by date of birth, we randomly selected two samples of 200 patients from the initial sample of 2321 patients for medical record review, each time sampling 100 patients from each of the two patient groups. The random selection was done using the random number function of our statistical software. All 400 medical records were available for review.
The first 200 medical records were assessed for eight key baseline characteristics. The second sample of 200 records was assessed for the same and a number of other baseline characteristics. In the second sample, we also validated the diagnosis of stroke (independently by two raters) and assessed co-existing conditions with the Charlson comorbidity index,10 a weighted aggregate of 19 possible co-existing conditions. Finally, we classified initial stroke severity using a retrospective algorithm for the Canadian Neurological scale.11 After exclusion of patients with low probability of stroke, as described above (n=9), we had 391 patients for analysis; 196 from the first sample of 200 and 195 from the second sample.
Stroke unit treatment
Patients allocated to the Department of Neurology were admitted to a 10-bed acute SU, operated as a short-stay unit with expected length of stay <10 days. This unit could expand or contract on demand. The medical treatment in this unit followed current practice guidelines for the management of patients with acute stroke,12,13 as previously reported.5
A standard examination was performed including neurological assessment, blood tests, electrocardiography, and a computer tomography (CT) scan of the brain within 2 h of admission. If an ischaemic stroke was suspected after the initial assessment and CT scan, aspirin 160–300 mg was administered orally or through a gastric tube, as soon as possible. As early as possible, often within the first hours after admission, both patients with ischaemic strokes and haemorrhages were mobilized. Patients with complete hemiplegia and patients who were impossible to mobilize for various reasons were given subcutaneous low-molecular-weight heparin. Parenteral isotonic fluid was administered routinely in the first 24 h. Hyperglycaemia was treated with insulin to keep serum glucose <12 mmol/l. Fever was treated with antipyretics (acetaminophen 500 mg tablet) when body temperature was 
38 °C. Antihypertensive treatment was not initiated in the first week except for markedly elevated blood pressure. If cardioembolic stroke was suspected, a cardiologist was consulted and eventually anticoagulation was initiated. The SU staff was multidisciplinary, comprising neurologists, trained nurses, physiotherapists, occupational therapists, and one speech therapist. The nurses were especially trained to detect and avoid complications. The physiotherapists followed the Bobath technique14 and instructed the staff to follow this approach over 24 h. A multidisciplinary team met weekly to evaluate progress and plan further treatment and discharge.
General medical ward treatment
In the Department of Medicine, patients were admitted to one of five wards, depending on available capacity. Patients in these wards received conventional medical treatment and care, without rigorous guidelines or standardization of treatment and care. Patients were immobilized until haemorrhage was excluded by CT scan, which most often would be performed within a few hours, sometimes the next day. Patients with ischaemic strokes were then mobilized, while patients with haemorrhages were often immobilized for up to 1 week. Aspirin was given if the CT scan did not reveal bleeding. Prophylactic administration of low-molecular-weight heparin was given to prevent venous thrombosis for immobilized patients. There was no routine of giving antipyretics or parental isotonic fluids. Anticoagulation was started when a possible cardiogenic embolic source was detected. Patients were offered supplementary oxygen, physiotherapy, occupational therapy, speech therapy, and evaluation by a neurologist when requested by the staff. Local department guidelines instructed personnel to pay attention to urine retention, bedsores, aspiration to the lungs and infections.
Follow-up after hospital discharge
After the initial care, subsequent rehabilitation was similar in the two groups. After acute medical treatment, stabilization and early rehabilitation, patients were discharged either to their homes, nursing homes, community-based long-term rehabilitation or hospital-based long-term rehabilitation. This treatment was given independently of their early treatment.
In a previous study of a subset of the patients included in this study with the same allocation procedure,5 after medical record review there was no difference between patients in the two arms of the study.
Outcome assessment
Information on death was collected through a computer link to the Norwegian National Population Register with only a few weeks updating lag. We included dates of death until 1 November 1999 in our analyses, resulting in 1198 deaths and 1071 censored observations in 2269 patients.
Statistical analysis
All the analyses were conducted according to the intention-to-treat principle, with survival for 30 and 365 days as the primary, prespecified outcome measures. In total, 9.0% of included patients (94 patients admitted to the SU, 111 to the GMWs) were admitted to the other treatment group than prescribed by the allocation rule, however this was adjusted for in the analyses by the intention-to-treat principle.
We used t-tests and 
2 tests for comparison of patient characteristics in the two study groups. Survival analysis was done using Kaplan-Meier survival curves and presentation of survival rates with approximate 95%CIs.15 We used Breslow's generalized Wilcoxon test for comparison of survival curves.16 Interrater agreement in the validity of the diagnosis of stroke was assessed using the kappa statistic.
The relationship between survival at 30 days and 1 year and its possible predictors was assessed using univariate and multivariate Cox proportional hazard regression analysis.17 These data were presented using risk ratios with 95%CIs and p values. In the Cox regression model, we included available predictors that we thought might be relevant for one-year survival without a computerized variable selection procedure. The proportional hazard assumption was checked with plots of the hazard function and log-log of the hazard function, and we found the assumptions to be satisfactorily fulfilled.
Apparent differences in risk factors between the groups were followed up with an assessment of the association between the risk factor and survival using univariate Cox regression of survival during the first 30 and 365 days.
We chose a 5% confidence level. SPSS for Windows 10.0 was used for all analyses.
|
Results |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionAppendixReferences |
|---|
Baseline characteristics
The two treatment groups were well balanced in terms of demographic and baseline characteristics (Table 1
). In the random sample of 391 patients, a diagnosis of stroke was confirmed in 184 patients (94%), while five were classified as TIAs and six as coding or punching errors. A second rater confirmed the diagnosis of stroke in 181 of the patients (93%). Agreement between the raters was substantial (kappa=0.79), using the terminology of Landis and Koch.18
|
Among the patients randomly selected for medical record review (n=391; 195 and 196 from the respective groups), there were 11.2% intracranial haemorrhages in the SU group and 6.2% in the GMW group (p=0.08). In this sample, we found no difference in the prevalence of previous stroke (p=0.76), diabetes (p=0.99), or cancer (ever) (p=0.21). At admission, there was no difference in the proportion of patients with current atrial fibrillation (p=0.20), reduced consciousness (p=0.95), use of warfarin (p=0.38), or use of acetylsalicylic acid (p=0.28) between the allocation groups. There was a difference in history of hypertension (p=0.02) and previous myocardial infarction (p=0.02), with more comorbidity in the group randomly sampled from the SU group. Further analysis, using univariate Cox proportional hazards regression analysis, revealed no association of previous myocardial infarction with 30- or 365-day survival in the sample of 391 patients. For previous hypertension there was an association with 365-day survival (RR=0.72, p=0.05), but no association with 30-day survival.
In the random sample of 195 patients, there were no group differences in seizures the first 24 h, history of atrial fibrillation, heart murmurs, systolic or diastolic blood pressure, pulse rate, body temperature, comorbidity assessed with the Charlson comorbidity index (p=0.93) or in stroke severity at admission as assessed with the Canadian Neurological scale (p=0.52).
Hospital use
The average length of stay was short in both patient groups; however, it was shorter in the GMWs than in the acute SU (Table 2). Assessment of hospital admissions and comorbidity after discharge from the hospital, showed no difference between the patients in the acute SU and patients in the GMWs (Table 2).
|
Survival
Unadjusted Kaplan-Meier survival curves for the two groups showed an early SU survival advantage (Figure 2). Survival for 30 days was achieved in 926 of the 1128 patients in the SU, as compared with 905 of the 1141 patients admitted to GMWs (p=0.08) (Table 3). Survival at 1 year was 732 patients in the SU and 731 in the GMWs (p=0.44) (Table 3). When analysis was restricted to patients surviving the first 30 days, there was no difference between the two Kaplan-Meier survival curves during the next 11 months (p=0.27).
|
|
The strongest predictors of death within 30 days were age and stroke type, with considerably better prognosis for ischaemic than haemorrhagic stroke (risk ratio of 0.24) (Table 4
). These variables and type of ward remained significant predictors of 30-day mortality after adjusting for the other variables in the model, while there was a weaker association with season of the year (winter vs. spring) (Table 4).
|
We constructed a similar Cox regression model to predict 1-year survival, achieving similar results. However the prediction by ward type was weaker and statistically not significant (Table 5
). In multivariate analysis, age, gender, season of the year (winter vs. spring), and stroke type were strong independent predictors of 1-year survival (Table 5).
|
|
Discussion |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionAppendixReferences |
|---|
This study is the largest single-centre study evaluating the survival benefit of a SU compared with conventional care in stroke. In multivariate Cox regression analysis, we found a 20% better survival during the first 30 days for patients admitted to a short stay, acute SU than for patients receiving conventional care in GMWs, although not reaching statistical significance in the Kaplan-Meier analysis. We found no survival benefit of SU treatment during the next 11 months.
The findings in our study are in accordance with the findings of reduced short-term mortality in SUs in prior trials including randomized trials,19–21 and give support to a previous report from a subset of the patients included in the present study.5 Hence, we show that the results from previous randomized trials can be reproduced over time in a routine setting. However, we could not confirm the long-term survival benefit of SU treatment reported in patients admitted within 24 h of stroke onset.5 Other trials offering rehabilitation for several weeks to patients who survived the first weeks after stroke onset also have shown a short-term survival benefit.22,23
A recent meta-analysis of organized in-patient (SU) care for stroke, combining the results of 20 SU studies, found a reduction in the odds of death (OR 0.83, 95%CI 0.71–0.97) recorded at final follow-up (median 1 year).2 This was in agreement with the findings of a previous review,3 but the conclusion about survival advantage was of marginal statistical significance and could be overturned by a relatively small number of unpublished randomized trials.2
The results using combined outcomes of death or requiring institutional care at the end of scheduled follow-up appears more robust.2 However, this outcome is probably more sensitive to local and national policies and practice, limiting generalization. Some of the limitations of meta-analyses have previously been highlighted: a possibility of publication bias, differences in study design, patient populations, heterogeneity in organization and treatment, or inappropriate analytic methods.24,25
Our results are also in agreement with the findings in a large multicentre observational study, also assessing routine care in stroke.4 Importantly, the patients in our study represent unselected stroke patients, in contrast to many of the randomized controlled trials.
In the present study, the finding of a higher mortality in the winter than in the spring for both 30-day and 1-year mortality, support the result of a previous study showing an association of stroke mortality with respiratory disease and temperature.26 We have no detailed data on functional status or other outcomes in this study, but there was an indication of a somewhat lower rate of hospital admission post stroke in the SU compared with the GMW patients.
We do not know which components of the SU care that are responsible for the early favourable effect on survival compared with GMWs. In the present study, care in the GMWs was neither very structured nor vigorously standardized. The early survival advantage in the SU might therefore be related to specialization and standardization of care in the SUs, concomitantly raising annual patient volume per unit. For other conditions, there is an association between hospital annual patient volume and survival.27–31
A previous study in our hospital with the same allocation procedure, in stroke patients admitted <24 h after onset of symptoms, showed some differences between care in our SU and in the medical wards:32 patients in the SU were mobilized earlier, more often received aspirin within 12 h after admission and more often received parenteral fluids, acetaminophen and were more liberally given antibiotics.
The findings of an observational study must be interpreted with caution. Our study was a large study in a single medical centre. A one-centre study limits the external validity of the study. However, it reduces the potential for differences in practices, registration and coding, although systematic differences in ICD-9 coding between the two departments might create bias. Neurologists might find more neurological deficits after 24 h in light strokes, classifying patients as having stroke, while internists might classify similar conditions as transitory ischaemic attacks (which were excluded in this study). The validation of the stroke diagnoses in a random sample of our cohort, confirming the stroke diagnoses in 93–94% of patients, compares favourably with the accuracy of coding in previous reports,9,33 although there were differences in sample selection and methodology between the studies. In a previous study of a subset (n=550) of the patients in this study, the diagnostic coding of the medical records was checked by one researcher and found to be accurate.32 In the present study, the equal proportion of haemorrhages in the two allocation groups also seemed reassuring, as did the distribution of most of the risk factors for survival in the random sample of medical records selected for review. We think the differences in history of hypertension and myocardial infarction in the random samples were probably due to chance, and we could only find an association between hypertension and one of the study outcomes, 365-day survival, in our sample of 391 patients. In addition, there was a difference in prevalence of haemorrhagic strokes in the allocation groups in this sample, in contrast with the situation in the total sample. This supports that the difference in prevalence of previous hypertension and myocardial infarction between the random samples might be due to chance. Additionally, these risk factors were more prevalent in the sample from the group with the most favorable outcome.
Our study was not randomized, as each individual patient did not have an equal chance of being allocated to each group. Thus, our allocation procedure was deterministic, by date of birth. However, the patients constituted two well-balanced groups comparable with the expected result of randomization. The large proportion of patients admitted to a different department than according to the allocation rule might be explained by unclear diagnosis at admission, reduced consciousness and several comorbidities. These patients would be admitted to the department most relevant for their other diseases rather than the frequently unrecognized acute stroke. However, we have accounted for this potential bias by adhering strictly to the intention-to-treat principle in the analysis.
In conclusion, our findings support previous findings of an initial survival advantage for patients admitted to a dedicated short-stay acute SU compared with conventional care in GMWs. This beneficial effect on survival cannot be confirmed beyond 30 days after hospital admission. We do not know which components of the SU care that are responsible for the initial survival benefit.
|
Appendix |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionAppendixReferences |
|---|
|
|
|
Acknowledgments |
|---|
Thanks to Jan Erikssen and Knut Nestvold for useful comments, and to Morten I. Lossius for assistance with the validation of the diagnoses.
|
Notes |
|---|
Address correspondence to Dr K. Stavem, Department of Medicine, Central Hospital of Akershus, N-1474 Nordbyhagen, Norway. e-mail: knut.stavem{at}klinmed.uio.no
|
References |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionAppendixReferences |
|---|
1. Garraway M. Stroke rehabilitation units: concepts, evaluation and unresolved issues. Stroke1985; 16:178–81.
2. Stroke Unit Trialist's Collaboration. Organised inpatient (stroke unit) care for stroke (Cochrane review). In: The Cochrane Library, Issue 4. Oxford, Update Software, 1999.
3.
Stroke Unit Trialists' Collaboration. Collaborative systematic review of the randomised trials of organised inpatient (stroke unit) care after stroke. Br Med J1997; 314:1151–9.
4.
Stegmayr B, Asplund K, Hulter-Asberg K, Norrving B, Peltonen M, Terent A, Wester PO. Stroke units in their natural habitat: can results of randomized trials be reproduced in routine clinical practice? Riks-Stroke Collaboration. Stroke1999; 30:709–14.
5.
Ronning OM, Guldvog B. Stroke units versus general medical wards, I: twelve- and eighteen-month survival—a randomized, controlled trial. Stroke1998; 29:58–62.
6.
Goldstein LB. Accuracy of ICD-9-CM coding for the identification of patients with acute ischemic stroke: effect of modifier codes. Stroke1998; 29:1602–4.
7.
Broderick J, Brott T, Kothari R, Miller R, Khoury J, Pancioli A, Gebel J, Mills D, Minneci L, Shukla R. The Greater Cincinnati/Northern Kentucky Stroke Study: preliminary first-ever and total incidence rates of stroke among blacks. Stroke1998; 29:415–21.
8.
Hankey GJ, Jamrozik K, Broadhurst RJ, Forbes S, Burvill PW, Anderson CS, Stewart-Wynne EG. Five-year survival after first-ever stroke and related prognostic factors in the Perth Community Stroke Study. Stroke2000; 31:2080–6.
9.
Ellekjaer H, Holmen J, Kruger O, Terent A. Identification of incident stroke in Norway: hospital discharge data compared with a population-based stroke register. Stroke1999; 30:56–60.
10. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis1987; 40:373–83.[Web of Science][Medline]
11.
Goldstein LB, Chilukuri V. Retrospective assessment of initial stroke severity with the Canadian Neurological Scale. Stroke1997; 28:1181–4.
12.
Adams HP, Jr., Brott TG, Crowell RM, Furlan AJ, Gomez CR, Grotta J, Helgason CM, Marler JR, Woolson RF, Zivin JA. Guidelines for the management of patients with acute ischemic stroke: a statement for healthcare professionals from a special writing group of the Stroke Council, American Heart Association. Circulation1994; 90:1588–601.
13. Anonymous. Treatment of stroke patients. Consensus Report No. 8. Oslo, Norwegian Research Council, 1995.
14. Bobath B. Adult hemiplegia: evaluation and treatment. London, William Heinemann Medical Books, 1978.
15. Altman DG. Practical statistics for medical research. London, Chapman & Hall, 1991.
16. Lang TA, Secic M. How to report statistics in medicine: annotated guidelines for authors, editors and reviewers. Philadelphia, American College of Physicians, 1997.
17. Cox DR. Regression models and life tables. J Roy Stat Soc (B)1972; 34:187–220.
18. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics1977; 33:159–74.[Web of Science][Medline]
19. Garraway WM, Akhtar AJ, Prescott RJ, Hockey L. Management of acute stroke in the elderly: preliminary results of a controlled trial. Br Med J1980; 280:1040–3.
20.
Stevens RS, Ambler NR, Warren MD. A randomized controlled trial of a stroke rehabilitation ward. Age Ageing1984; 13:65–75.
21.
Indredavik B, Bakke F, Solberg R, Rokseth R, Haaheim LL, Holme I. Benefit of a stroke unit: a randomized controlled trial. Stroke1991; 22:1026–31.
22.
Kalra L, Dale P, Crome P. Improving stroke rehabilitation: a controlled study. Stroke1993; 24:1462–7.
23.
Kalra L, Eade J. Role of stroke rehabilitation units in managing severe disability after stroke. Stroke1995; 26:2031–4.
24.
Sharp SJ, Thompson SG, Altman DG. The relation between treatment benefit and underlying risk in meta-analysis. Br Med J1996; 313:735–8.
25.
Sim I, Hlatky MA. Growing pains of meta-analysis. Br Med J1996; 313:702–3.
26.
Lanska DJ, Hoffmann RG. Seasonal variation in stroke mortality rates. Neurology1999; 52:984–90.
27.
Showstack JA, Rosenfeld KE, Garnick DW, Luft HS, Schaffarzick RW, Fowles J. Association of volume with outcome of coronary artery bypass graft surgery: scheduled vs nonscheduled operations. JAMA1987; 257:785–9.
28.
Thiemann DR, Coresh J, Oetgen WJ, Powe NR. The association between hospital volume and survival after acute myocardial infarction in elderly patients. N Engl J Med1999; 340:1640–8.
29. Cunningham WE, Tisnado DM, Lui HH, Nakazono TT, Carlisle DM. The effect of hospital experience on mortality among patients hospitalized with acquired immunodeficiency syndrome in California. Am J Med1999; 107:137–43.[Web of Science][Medline]
30. Hughes RG, Garnick DW, Luft HS, McPhee SJ, Hunt SS. Hospital volume and patient outcomes: the case of hip fracture patients. Med Care1988; 26:1057–67.[Web of Science][Medline]
31.
Jollis JG, Peterson ED, Nelson CL, Stafford JA, DeLong ER, Muhlbaier LH, Mark DB. Relationship between physician and hospital coronary angioplasty volume and outcome in elderly patients. Circulation1997; 95:2485–91.
32.
Ronning OM, Guldvog B. Stroke unit versus general medical wards, II: neurological deficits and activities of daily living—a quasi-randomized controlled trial. Stroke1998; 29:586–90.
33.
Benesch C, Witter DMJ, Wilder AL, Duncan PW, Samsa GP, Matchar DB. Inaccuracy of the International Classification of Diseases (ICD-9-CM) in identifying the diagnosis of ischemic cerebrovascular disease. Neurology1997; 49:660–4.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||