Multicenter Comparison of Processes of Care Between Stroke Units and Conventional Care Wards in Australia
Background and Purpose— Approximately 23% of Australian hospitals provide Stroke Units (SUs). Evidence suggests that clinical outcomes are better in SUs than with conventional care. Reasons may include greater adherence to processes of care (PoC). The primary hypothesis was that adherence to selected PoC is greater in SUs than in other acute care models.
Methods— Prospective, multicenter, single-blinded design. Models of care investigated: SUs, mobile services, and conventional care. Selected PoC were related to care models and participant outcomes. Data were collected at acute hospitalization (median 9 days) and at medians of 8 and 28 weeks after stroke.
Results— 1701 patients were screened from 8 hospitals, 823 were eligible, and 468 participated. Response rate was 96% at final follow-up. Mean age was 73 years (SD 14). Overall PoC adherence rates for individual care models were SU 75%, mobile service 65%, and conventional care 52% (P<0.001). The adjusted odds of participants being alive at discharge if adhering to all or all but 1 PoC was significant (aOR 3.63; 95% CI: 1.04 to 12.66; P=0.043). Important trends at 28 weeks were found for being at home (aOR 3.09; 95% CI: 0.96 to 9.87; P=0.058) and independent (aOR 2.61; 95% CI: 0.96 to 7.10; P=0.061), with complete PoC adherence.
Conclusion— Adherence to key PoC was higher in SUs than in other models. For all patients, adherence to PoC was associated with improved mortality at discharge and trends found with independence at home, providing support for the need to increase access to stroke units.
Stroke affects >44 000 Australians each year, with a 28-day case fatality rate of 20% for a first-ever stroke.1 Approximately 89% of people who have a stroke in Australia are admitted to hospital,1 with care provided mainly within the public system.2 Approximately 23% of Australian hospitals provide a dedicated service for stroke,2 with no formal prioritization of patients who have stroke to be transported to a hospital with a stroke unit (SU) raising equity issues.
Randomized, controlled trials of SUs have shown consistent and significant trends toward improved patient outcome compared with conventional care.3–6 Reductions in mortality have been demonstrated at 3 and 12 months, and even at 10 years after stroke.7 The aspects of care delivery responsible for better outcomes in organized services remain unclear.8 Important practices include early mobilization, physiological homeostasis, early initiation of aspirin, and, when appropriate, thrombolysis, anticoagulation in patients with atrial fibrillation, measures to avoid aspiration, early nutrition, frequent monitoring, and management of co-morbidity to avoid complications.6,9
The SCOPES (Stroke Care Outcomes: Providing Effective Services) study was undertaken to provide evidence to support greater uptake of SUs by delineating the factors that make them more effective. This was achieved by using a quality assessment framework underpinned by the work of Donabedian.10 Structure (model of care), process (clinical practices), and outcomes of care delivery were examined to better understand the differences between alternate stroke care models.11 The primary hypothesis was that adherence to selected processes of care (PoC) is greater in SUs than in other acute care models.
Materials and Methods
This was a prospective, observational, multicenter cohort design. Researchers collecting follow-up data were blinded to the model of care.
All public hospitals in metropolitan Melbourne (Victoria, Australia) were screened. Eligibility criteria included facilities admitting >100 stroke patients per year, willingness to provide patient and hospital level data, a stroke care model operating >12 months, and were not participating in research that could influence data. All eligible hospitals (n=8) participated and provided ethics committee approval. All nonparticipating hospitals were sent a questionnaire regarding stroke care provision at their site.
All consecutive admissions were screened over 12 months. Inclusion criteria were first ever or recurrent stroke (ischemic or intracranial hemorrhage only), hospital presentation within 3 days of onset, age older than 18, and written consent provided. Exclusion criteria were patients transferred after admission to another hospital or an inpatient stroke.
Sample sizes were estimated at 64 patients per model of care, based on the detection of a clinically important difference in adherence rates to a single PoC for SU 95%, mobile service 85%, and conventional care 75%, with power of 0.8 and alpha of 0.05. We aimed to recruit 80 patients per hospital to allow between-hospital comparisons and to account for loss to follow-up.
SU—a designated (geographically localized) acute ward area where a multidisciplinary stroke team focuses its expertise. Patients were classified as being treated in SU if admitted there directly or transferred there during the acute admission, irrespective of duration of stay.
Mobile service—a dedicated, hospital-based, multidisciplinary team who review stroke patients located throughout a hospital.
Conventional care—no specific service or health professional team dedicated to hospital stroke management.
Stroke—vascular lesion of the brain resulting in a neurological deficit persisting for at least 24 hours or resulting in the death of the individual.12
Overall PoC adherence—number of PoC completed/number applicable for patients within each care model.
Thorough adherence—adherence to all or all except 1 applicable PoC.
Complete adherence—adherence to all applicable PoCs.
Outcomes—alive at discharge, alive and independent at final follow-up, and home at final follow-up.
Independence—modified Rankin score of 0 to 2, dependence score is 3 to 5, and death score is 6.13
Institutionalization—placement in an aged care or other medical facility.
Data related to hospitalization (PoC, case-mix, and outcome) were extracted retrospectively from case notes. When possible, case-mix information was obtained directly from respondents to augment information in their medical records and limit missing data. Long-term outcome data were prospectively collected.
Process of Care Data
We undertook literature reviews and expert consensus meetings (as robust evidence of improved outcomes for some PoCs was limited) to derive a set of clinically important PoC, which could be abstracted from medical records. The final 21 PoCs reflected aspects of care within 24 hours of admission, documentation, and general management.
Additional data were collected to further explain differences between the models of care, including use of clinical management plans, managing doctor, number and type of clinical investigations, the receipt of palliative care, and intrahospital transfers.
Clinical Outcome Measures
Outcome data were collected at acute hospitalization (median 9 days; quartiles 5, 16), median 8 weeks (quartiles 6, 10), and median 28 weeks (quartiles 26, 33). Most patient and informal career follow-up data were collected using validated measures via telephone interview and postal survey, unless face-to-face or proxy interviews were required. The measures qualified stroke type and severity, quality of life, disability, handicap, satisfaction with services, and career strain.
Data were double-entered and a computer program was used to identify discrepancies. An alternative researcher audited a random selection of 10% of all acute hospital data. Variables identified as having a low level of interobserver agreement were subjected to auditing in all cases.
A multidisciplinary panel of local and international experts agreed on the analysis framework.11 Fifteen of the 21 PoCs were used in the analysis because they were deemed high-priority (selection criteria included consideration of reliability, robustness, and importance regarding outcome). In brief, these were: (1) activities within 24 hours, which included CT scan, swallowing assessment, allied health assessment, neurological observations; (2) documentation of premorbid function and discharge needs; and (3) management practices: enteric feeding if nil by mouth >48 hours, measures to avoid aspiration, deep vein thrombosis prophylaxis, fever management, and use of antiplatelet agents at discharge (protocol developed before published guidelines for acute therapy) (Table 1).
Patient Applicability, Adherence to Processes of Care, and Stroke Outcomes
PoCs were examined for their applicability to participants (number of cases applicable/total cases) and adherence rates (number of cases adhered/total applicable cases) calculated for each (Table 1).
Participants were then categorized into those with thorough and complete adherence, and the effect of adherence levels on long-term outcomes was investigated.
Statistical analysis was performed with SPSS for Windows, Version 10.0.5. Categorical variables were analyzed using the χ2 test. ANOVA was used to compare age across care models. Logistic regression was used to investigate the effect of model of care and adherence to PoC on outcome measures, adjusted for case-mix. Level of significance was P<0.05.
Adjustment for patient case-mix was undertaken using variables identified as predictive of stroke outcome:14,15 age, premorbid function, living alone, normal verbal Glasgow Coma Score, ability to lift both arms, ability to walk alone, and urinary incontinence within 72 hours of stroke.14 Discrimination of this statistical model for predicting independence at 28 weeks was greater than a model using age, gender, and variables that differed significantly (Table 2) between care modalities (areas under receiver operating characteristic curves 0.887 and 0.841, respectively).
Inter-rater reliability was assessed using the kappa (κ) statistic for dichotomous variables and the weighted κ statistic for ordinal variables.
The 3 SU hospitals were tertiary teaching hospitals. The 3 mobile services and 2 conventional care hospitals were smaller suburban centers. Various models of care could be received within SU or mobile service hospitals, such as mobile service care in SU hospital when beds were unavailable; hence, patients were categorized according to treatment actually received.
Of 1701 patients screened between September 1998 and October 1999, 878 were ineligible and 468 patients provided informed consent to participate (57%). The main causes of ineligibility were TIA and nonstroke (45%). The SU hospitals recruited significantly greater participants (SU 63%, mobile service 56%, conventional care 43%, P<0.01). No statistically significant differences between participants and nonresponders for age and gender were detected. Further, the proportions of stroke patients admitted with ischemic and hemorrhagic stroke did not vary significantly across the hospitals, suggesting similar stroke type presentations.
Key variables used for case-mix adjustment, subgroup classification, or as PoC measures, were assessed for inter-rater reliability and demonstrated excellent agreement (κ >0.8).
Outcome data were available on 96% of patients (survivors n=357, deceased n=91). Nine participants (2%) withdrew from the study and 11 (2%) were lost. Mean age was 73 years (SD 14, range 19 to 102 years) (Table 2).
Of the 230 participants presenting to an SU hospital, 138 (60%) were admitted directly to the SU, 175 (76%) spent some time in the SU, and 55 (24%) did not receive any SU care. Of these, the SU team saw 52 as a mobile service. Of the 169 participants from mobile service hospitals, 12 received conventional care. Conventional care hospitals recruited 69 participants.
Age, prestroke level of independence, history of atrial fibrillation, Oxfordshire stroke subtype classification,16 and admission muscle strength varied significantly across the care models (Table 2). There were no significant differences in occupational classification (used as a surrogate for socio-economic status and education level) or length of stay for patients experiencing no discharge delays (SU median 7 days, quartiles 4, 9; mobile service median 7 days, quartiles 5, 9; and conventional care 6 days, quartiles 4, 8). The variances observed in the baseline characteristics atrial fibrillation and stroke subtype when included in a multivariate model were found to not influence the effect of model of care for outcomes evaluated (discharged P=0.528; at 28 weeks: home P=0.467, independent P=0.729, and alive P=0.563).
Most participants were discharged to a rehabilitation facility (43%) or home (32%); 10% went to an aged care facility and 3% were transferred to another hospital for interim or palliative care. At final follow-up, 62% were home, 16% were in an aged care facility, 3% were in hospital, and 20% had died.
Quality of Acute Care
More than half of the PoC were considered applicable in all cases, with applicability ranging from 11% to 100%. Overall adherence rates to PoC varied from 27% for “occupational therapist within 24 hours” to 97% for “documentation of premorbid function” (Table 1).
Completeness of Care and Stroke Service Model
Irrespective of the total number of applicable PoC, which ranged from a minimum requirement of 9 to all 15, a consistent level of adherence was observed in each model (Figure). The overall adherence rate was significantly greater in SUs, and a significantly greater proportion of SU participants had thorough and complete adherence than the other models (Table 3).
Completeness of Care and Patient Outcome
Adjusting for case-mix, the odds of being alive at discharge for participants with thorough adherence (n=89, 19%) was significantly higher. No significant differences in longer-term outcomes between the 2 groups for either adherence category was found, although there were trends suggesting improved outcomes. The absolute differences in outcomes for those that did or did not have thorough or complete adherence ranged from 8.6% to 19% (Table 4).
Model of Care and Explanatory Variables
Clinical management plans detail the activities of care over time. Of the 8 hospitals, 2 had established use of clinical management plans (SU=1, MS=1), 3 hospitals were in various trial phases (MS=1, CC=2), and 2 hospitals did not use them (SU=1, MS=1). Clinical management plans were more likely to be used (SU 81 [46%], mobile service 87 [42%], conventional care 9 [11%]; P<0.001) and completed in SU participants (SU 55 [68%], mobile service 43 [49%], conventional care 1 [11%]; P=0.001). The managing physician was more likely to be a stroke specialist (neurologist) in the SU (P<0.001). There was no significant difference between patients receiving palliative care (SU 8%, mobile service 12%, and conventional care 6%; P=0.161). Patients treated by a mobile service were less likely to be transferred to another ward during their acute hospitalization. The main reasons for ward transfers were to receive care in a SU, coronary care ward, intensive care unit, general medical ward, or an on-site rehabilitation ward.
Univariate analysis demonstrated significant differences in types of diagnostic tests undertaken by participants treated in SU and mobile service compared with conventional care. Conventional care patients were less likely to have a transesophageal echocardiogram (P=0.043), undergo a carotid ultrasound (P=0.001), cerebral or carotid angiogram (P<0.001), transcranial Doppler ultrasound (P<0.001), or magnetic resonance imaging (P<0.001).
We have demonstrated that adherence to high-priority PoC was associated with improved mortality at discharge and trends found toward independence at home. These management practices are more consistently delivered in a geographically localized SU setting. Overlapping care models existed within some participant hospitals. An intention to treat analysis in this evaluation would be expected to underestimate the true effect of the intervention (model of care). The models of care compared were representative of those in nonparticipating metropolitan hospitals.
Important trends emerged, suggesting improved patient outcomes at 28 weeks with complete adherence to PoC. A >3-fold increase in odds of being discharged from acute care was found if all applicable PoC or all except 1 (thorough adherence) were undertaken by health professionals.
Because PoCs were completed more often in SUs, these findings may explain why SUs achieve better outcomes compared with other models. Other studies have also shown that SUs have greater adherence to aspects of evidence-based investigation and treatment.9,17,18 We acknowledge that the overall effectiveness of SU care may be influenced by intangible factors such as greater enthusiasm and specialization of staff, which cannot be attributed completely by the sum of individual PoC. The categorization of SU care was based on a conservative premise that any care in the SU should result in patient benefits,11 potentially underestimating benefits observed.
Seven of the 15 PoC had adherence rates <65%, indicating the capacity for improvement. Some of the parameters were more complex because they required the availability of allied health staff or had time restrictions. Improvement in adherence is possible regardless of the model of care. A systematic method for measuring clinical practice for stroke is needed to ensure that optimal care is delivered.17
The finding that 24% of patients admitted to SU hospitals received no care in the SU raises issues with respect to equity of care and capacity to meet demand. Limited diagnostic services at conventional care hospitals may have influenced the reduced investigations performed at these hospitals. Because proportions of patients receiving palliative care were similar across models, this could not be attributed to the PoC adherence rates.
Differences in patient case-mix between models were appropriately adjusted for in the analysis and can be explained, in part, by referral and response bias. Our response rate is in line with similar studies examining PoC.19 The SCOPES cohort may not be representative of the entire stroke population but contained sufficient breadth to be able to examine quality of care, particularly because process indicators are often universally applicable irrespective of stroke type or severity. The higher refusal rate in conventional care hospitals was possibly influenced by these patients tending to be older and more dependent before their stroke. Studies that tend to include mild/moderate strokes have higher response rates.5 A randomized, controlled trial was not possible because of the potential for contamination and crossovers of the control group within an SU or mobile service hospital, although perhaps studies of particular components of care could be undertaken. In addition, the size of SU hospitals may have influenced PoC adherence because of the availability of resources and, coincident with higher recruitment rates, thorough and complete care rates may have been overestimated.
Several potential sources of reporting bias arise in this study, such as use of telephone follow-up surveys or abstracting PoC from medical records, which may have been undertaken but not routinely recorded or required subjective judgements. Criteria to select reliable PoC for the analysis were used. In addition, SUs may be more adept at documenting PoC given their specialized focus and better use of clinical management tools, although the spread of use of these tools across models should have limited this form of bias. Abstracting PoC from medical records may underestimate rates up to 10%.19
Although not a primary hypothesis, we did not demonstrate significant differences in long-term outcomes associated with PoC adherence. This may reflect a lack of statistical power, because the number of participants receiving thorough or complete adherence was small. Nonetheless, important trends were demonstrated and a 50% improvement in independence at 28 weeks with complete adherence compared with thorough adherence was observed.
This study demonstrates aspects of care that may be important in determining outcome and further highlights the better clinical practices of SUs. Establishing key factors that contribute to optimal management in SUs provides additional incentive to increase equitable access to dedicated SUs in Australia and elsewhere.
The Department of Human Services Victoria, the Ian Potter Foundation, and the National Stroke Foundation of Australia supported this study. We thank Maria Di Pietro, Michelle Fox, Tamara Clements, Jason Faux, and Karen Martin for their contribution as research officers. We also thank Sonia Morrissey and Penny Bisset for their contribution to database management. In addition, we thank Franca Smarrelli, previously of the National Stroke Foundation of Australia, Professor Shah Ebrahim (UK), and Professor Peter Langhorne (UK) for their support and contribution to this research. Finally, we thank our consumer representative Gillian Simons, President of Stroke Association of Victoria.
Authors Donnan and Davis are the heads of dedicated stroke units in their respective hospitals. They were not directly involved in the collection of data or analysis of the results.
- Received January 21, 2004.
- Accepted January 29, 2004.
Thrift A, Dewey H, Macdonell R, McNeil J, Donnan G. Stroke incidence on the East Coast of Australia: The North East Melbourne Stroke Incidence Study (NEMESIS). Stroke. 2000; 31: 2087–2092.
Smarelli F, Cadilhac D, Pearce D. Australian survey of hospital services for stroke. Cerebrovasc Dis. 2001; 11. Abstract 93.
Stroke Unit Trialists’ Collaboration. Collaborative systematic review of the randomised trials of organised inpatient (stroke unit) care after stroke. BMJ. 1997; 314: 1151–1159.
Stroke Unit Trialists’ Collaboration. Organised inpatient (stroke unit) care for stroke (Cochrane Review). In: The Cochrane Library, Issue 1, 2003. Oxford: Update Software.
Indredavik B, Bakke F, Slordahl SA, Rokseth R, Haheim LL. Stroke unit treatment. 10-year follow-up. Stroke. 1999; 30: 1524–1527.
Duncan PD, Williams L, Brass L, Horner R, Hinchey J, Wolf PA. Quality Assessment of Stroke Care. American Heart Association: Council Newsletter, Winter 1999.
National Stroke Foundation. Stroke Care Outcomes: Providing Effective Services (SCOPES). An Evaluation of Victorian Stroke Services: Final Report. October 2002.
World Health Organisation. Report of the World Health Organisation on stroke and other cerebrovascular disorders. Stroke. 1989; 20: 1407–1431.
Counsell CE. The prediction of outcome in patients with acute stroke. (MD thesis). Cambridge, UK: Cambridge University; 1998.
Counsell C, Dennis M, McDowall M, Warlow C. Predicting outcome after acute and subacute stroke: development and validation of new prognostic models. Stroke. 2002; 33: 1041–1047.
Dennis MS, Burn JP, Sandercock PA, Bamford JM, Wade DT, Warlow CP. Long-term survival after first-ever stroke: the Oxfordshire Community Stroke Project. Stroke. 1993; 24: 796–800.
Weir N, Dennis MS, Scottish Stroke Outcomes Study Group. Towards a national system for monitoring the quality of hospital-based stroke services. Stroke. 2001; 32: 1415–1421.