Diagnostic Procedures, Treatments, and Outcomes in Stroke Patients Admitted to Different Types of Hospitals
Background and Purpose—In many countries, including Sweden, initiatives have been taken to reduce between-hospital differences in the quality of stroke services. We have explored to what extent hospital type (university, specialized nonuniversity, or community hospital) influences hospital performance.
Methods—Riksstroke collects clinical data during hospital stay (national coverage 94%). Follow-up data at 3 months were collected using administrative registers and a questionnaire completed by surviving patients (response rate 88%). Structural data were collected from a questionnaire completed by hospital staff (response rate 100%). Multivariate analyses with adjustment for clustering were used to test differences between types of hospitals.
Results—The proportion of patients admitted directly to a stroke unit was highest in community hospitals and lowest in university hospitals. Magnetic resonance, carotid imaging, and thrombectomy were more frequently performed in university hospitals, and the door-to-needle time for thrombolysis was shorter. Secondary prevention with antihypertensive drugs was used less often, and outpatient follow-up was less frequent in university hospitals. Fewer patients in community hospitals were dissatisfied with their rehabilitation. After adjusting for possible confounders, poor outcome (dead or activities of daily living dependency 3 months after stroke) was not significantly different between the 3 types of hospital.
Conclusions—In a setting with national stroke guidelines, stroke units in all hospitals, and measurement of hospital performance and benchmarking, outcome (after case-mix adjustment) is similar in university, specialized nonuniversity, and community hospitals. There seems to be fewer barriers to organizing well-functioning stroke services in community hospitals compared with university hospitals.
Hospital performance is often assessed at 3 levels: structure, processes, and outcomes,1 with case fatality being the most commonly available measure of stroke outcome.2 Hospitals with many stroke patients (high-volume hospitals) have been reported to have a lower 7-day case fatality3,4 or lower in-hospital stroke case fatality5–7 compared with hospitals with fewer stroke patients (low-volume hospitals). In the United States, greater spending on hospitals is also associated with lower risk-adjusted inpatient stroke case fatality.8,9 A recent national study from Denmark, however, failed to demonstrate any relationship between stroke case volume and 30-day or 1-year case fatality after adjusting for differences in case-mix between hospitals.10
Between-hospital differences in early stroke case fatality are usually attributed to differences in the quality of acute stroke care provided (although the scientific evidence for a direct link between hospital performance and stroke outcome is not robust2). Another possible explanation for between-hospital differences in outcome is differences in case-mix. In some previous studies, the use of routine administrative registers has limited the ability to adjust for differences in key prognostic variables, such as stroke severity at onset.
Compared with low-volume hospitals, high-volume hospitals provide, on average, more evidence-based services to acute stroke patients. These services include diagnostic procedures, acute treatment, such as thrombolysis, aspects of nursing, early rehabilitation, and secondary prevention.10–12 In the United States, the “Get With The Guidelines Stroke Program,” hospital certification programs, and recognition programs are associated with higher conformity with care measures for acute stroke patients.13
With the exception of the Danish study,10 previous studies on hospital volume have been performed in settings where stroke units have not followed a national standard. In Sweden, all hospitals admitting acute stroke patients have a dedicated stroke unit (Table I in the online-only Data Supplement). National stroke guidelines under the auspices of the National Board of Health and Welfare are available. Riksstroke, a hospital performance register, includes benchmarking of stroke processes and outcomes in all hospitals.
In the dissemination of new medical technology, the prevailing view is that university hospitals are, in general, early adopters and community hospitals are, on average, later adopters. However, the complexity of university hospitals may negatively affect factors, such as patient satisfaction and follow-up visits. In this study, we explored to what extent differences in stroke care procedures and outcomes among university, large nonuniversity, and community hospitals exist.
Patients presenting with acute stroke between January 1, 2012, and December 31, 2013, and who were recorded in The Swedish Stroke Register (Riksstroke), were included in the study. The primary aim of this national register was to monitor and support improvement in quality of stroke care in Sweden. The register, established in 1994, covers all hospitals in the country admitting acute stroke patients (72 hospitals in 2012 and 2013). The Riksstroke follow-up procedures include linkage to the national Cause of Death Register and a questionnaire to surviving patients 3 months after stroke (response rate 87.6% in 2012–2013). Details on the organization, funding, data collection, and reporting and description of variables are available at the Riksstroke Web site (http://www.riks-stroke.org/index.php?content=&lang=eng&text=). This study has been approved by the Regional Ethical Review Board at Umeå University (Dnr 2013/353-31).
Of all patients discharged from Swedish hospitals in 2012 and 2013 with a diagnosis of acute stroke in routine administrative registers, 88.2% were recorded in Riksstroke. Allowing for the estimated 6% false-positive diagnosis of acute stroke in Swedish routine administrative registers,14 the actual coverage of the register is estimated to be 94%.
Basic characteristics of acute stroke services in Sweden with emphasis on patient allocation to different types of hospital are described in Table I in the online-only Data Supplement. In the present study, hospitals were categorized as university hospitals (n=9, mean 562 admissions per year during 2012–2013, range 291–879), specialized nonuniversity hospitals (n=22; mean 507 admissions, range 209–1139), and community hospitals (n=41; mean 204 admissions, range 64–389). The delineation between specialized nonuniversity hospitals and community hospitals was determined by their degree of specialization; community hospitals have only basic inpatient specialities (typically internal medicine, surgery, anesthesiology, x-ray department, and a laboratory). Large nonuniversity hospitals have a wider range of specialities and provide more advanced diagnostic procedures (eg, various magnetic resonance diagnostics) and interventions (eg, carotid surgery).
Stroke patient volume was categorized by quartiles with 18 hospitals in each category; Q1, 64 to 170 admissions; Q2, 201 to 263 admissions; Q3, 291 to 436 admissions; Q4, 439 to 1139 admissions per year.
In Sweden, acute stroke care is centralized only to a limited extent. For example, thrombolysis is performed in 69 of the 72 hospitals, and patients are seldom referred to a tertiary (university) hospital except for neurosurgery or endovascular procedures. Between 2012 and 2013, 2621 patients (5.3%) were treated in >1 hospital. In the Riksstroke register, the patient is assigned to the hospital where he/she spends most of the acute hospital stay.
In April 2013, a 34-item questionnaire addressing structural aspects of stroke care was distributed to the 72 hospitals admitting acute stroke patients. All hospitals responded. We used this information to describe stroke unit care in the 3 types of hospitals.
Riksstroke uses definitions agreed upon by the Stroke Unit Trialists’ Collaboration15,16 and the European Stroke Initiative17 to define a dedicated stroke unit. In Riksstroke, the level of consciousness is used as a proxy for stroke severity. Based on the Reaction Level Scale (RLS 85),18 Riksstroke classifies patients as alert (RLS 1), drowsy (RLS 2–3), or unconscious (RLS 4–8). Recordings of patient-reported outcomes (mood, self-assessed general health, and satisfaction with care) have been described in previous publications.19,20
In univariate analyses of hospital type (university hospitals, specialized nonuniversity hospitals, and community hospitals), χ2 tests for 2×3 cross-tables were used for binary variables. Proportions and means are presented with corresponding 95% confidence intervals within different hospital type. In 2 multiple logistic regression models (with and without stroke patient volume), independent predictors of the probability of poor outcome (dead or activities of daily living [ADL] dependency 3 months after stroke) were analyzed. To adjust for the fact that observations on individual patients within one and the same hospital may not be entirely independent on each other (clustering), we used generalized estimating equations with an exchangeable correlation structure. It was used in univariate analyses, as well as in multiple logistic regression models. Possible presence of multicollinearity in the models was assessed by calculating variance inflation factors. The analyses were performed with the statistical software IBM SPSS Statistics 22.
Role of the Funding Source
The funders of Riksstroke had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
Hospital and Patient Characteristics
In 2012 to 2013, the Riksstroke register recorded 49 144 admissions for acute stroke from 72 hospitals. Of these, 9 were university hospitals, 22 were specialized nonuniversity hospitals, and 41 were community hospitals. Table 1 shows some basic characteristics of the 3 types of hospitals. According to self-assessments, the great majority of hospitals of all types fulfilled the basic set of stroke unit criteria. The number of stroke unit beds per population in the hospital catchment area was twice as high in community hospitals compared with that in university hospitals. A dedicated stroke care coordinator was less common in community hospitals, whereas a routine for early multidisciplinary rehabilitation was less common in university hospital stroke units.
Patient characteristics in the 3 types of hospital are shown in the Table II in the online-only Data Supplement. Patients admitted to university hospitals were on average 2 years younger than patients admitted to nonuniversity hospitals. In university hospitals, a higher proportion were ADL independent before the index stroke, and a lower proportion were recorded to have hypertension. The proportions of patients with intracerebral hemorrhage and a lowered level of consciousness (as a proxy for stroke severity) were highest in university hospitals.
For acute stroke patients admitted to community hospitals, it was significantly more common that they were admitted directly to specialized stroke care (stroke unit, intensive care unit, or transferred to a neurosurgical unit) compared with patients admitted to university or specialized nonuniversity hospitals. When instead stroke unit care during more than half of the hospital was tested, the differences between hospital types were not significant (Table 2).
The proportion of stroke patients examined by computed tomography scan was high (>98%) in all types of hospital. In contrast, the proportion examined by magnetic resonance imaging (MRI) was considerably higher in university hospitals than that in nonuniversity hospitals, whether specialized or not (Table 2). In addition, the proportion of patients with ischemic stroke examined by any type of carotid artery imaging was highest in university hospitals.
The proportion of patients with ischemic stroke treated by thrombolysis was somewhat higher in university hospitals than that in nonuniversity hospitals (Table 2). Median door-to-needle time for patients treated with thrombolysis was on average 12 minutes shorter in university hospitals and 8 minutes shorter in specialized nonuniversity hospitals compared with that in community hospitals. In Sweden, thrombectomy is performed only in university hospitals, and some patients from other hospitals are referred to university hospitals for the procedure (Table 2).
Because there are large local variations in the organization of stroke services, we report the total length of hospital stay (ie, the time spent in acute care hospitals plus time spent in rehabilitation or geriatric units). Median length of total hospital stay was identical in the 3 types of hospital, whereas mean length of stay was 1.6 days longer in university hospitals than that in community hospitals. There were no clinically meaningful differences between the 3 types of hospitals with respect to prescription of secondary prevention drugs at discharge from hospital, except for antihypertensive drugs that were prescribed significantly more often in patients discharged from nonuniversity than those from university hospitals (Table 2). The proportion with an outpatient follow-up visit to a physician or a stroke nurse was lower in university hospitals than that in the 2 other hospital types (Table 2; nonoverlapping 95% confidence intervals but nonsignificant P value).
Death and ADL Dependency
A somewhat larger proportion of surviving patients had poor outcome (death from ADL dependency) at 3 months after stroke if they had initially been admitted to a university hospital rather than a nonuniversity hospital, but after adjustment for clustering the differences were not statistically significant (Table IV in the online-only Data Supplement).
In a multivariate statistical model that included case-mix variables and hospital type and was adjusted for clustering, poor outcome (death or ADL dependency 3 months after stroke) was similar in the 3 types of hospital (Table 3).Other factors associated with increased risk of poor outcome in the multivariate model were high age, being ADL dependent before the index stroke, living alone, and a history of stroke, diabetes mellitus, and atrial fibrillation (Table 3). The risk of poor outcome was strongly associated with a diagnosis of intracerebral hemorrhage and lowered consciousness at admission.
In an additional multivariate model, hospitals were subdivided by stroke patient volume into quartiles (Table III in the online-only Data Supplement). After adjustment for clustering, there were no significant differences in poor outcome between high volume (439–1139 admissions per year; reference), medium-high volume (291–436 admissions per year; odds ratio, 0.87; 95% confidence interval, 0.74–1.03), medium-low (201–263 admissions per year; odds ratio, 0.89; 95% confidence interval, 0.74–1.06), and low-volume hospitals (64–170 admissions per year; odds ratio, 0.97; 95% confidence interval, 0.79–1.18). In this model, the variance inflation factor was 2.18 for both hospital type and stroke patient volume, indicating some degree of multicollinearity. For all other variables, the variance inflation factor was ≤1.36 (absence of multicollinearity).
The great majority of patients in all types of hospitals reported that they were satisfied with the care they had received in hospital and the rehabilitation they had received after discharge. The proportion of patients who were dissatisfied with the acute in-hospital care was similar in the 3 types of hospitals, whereas dissatisfaction with rehabilitation after discharge from hospital was significantly more common in patients treated in university and specialized university hospitals than in those treated in community hospitals (Table 4). A lower proportion of community hospital patients reported their general health as poor 3 months after stroke, whereas similar proportions reported some degree of low mood in the 3 types of hospital. The rates of smoking cessation 3 months after stroke did not differ substantially by type of hospital.
Our results show important differences in stroke services and management among university, nonuniversity, and community hospitals, and these differences are not always in the favor of university hospitals. Whereas nonintensive stroke units are established in all acute hospitals in Sweden, patient access to stroke units is better in community hospitals than that in university hospitals. However, patients treated in university hospitals are more often investigated using MRI, are more often having their carotid arteries examined, have shorter door-to-needle times for thrombolysis, and are more often treated by thrombectomy. After adjusting for differences in case-mix, hospital type is not associated with the risk of poor outcome (dead or ADL dependency 3 months after stroke). Further adjustment for stroke patient volume did not affect the main results.
A strength of this study is that it is nationwide with up-to-date information and that all hospitals in the country admitting acute stroke patients are covered. As acute care hospitals in Sweden have reasonably well-defined catchment areas and are all publicly financed, there is minimal risk of active patient selection by ability to pay. The coverage of all acute stroke patients admitted to hospital is high (see Methods). Extensive validations of the data submitted to Riksstroke have not indicated any systematic differences in data quality between different types of hospitals (Riksstroke, unpublished). The case-mix, including prevalence of risk factors, is similar to what has been reported in other large-scale studies, for example, the US Get With The Guidelines project21 and the UK Oxfordshire Community project,22 except that atrial fibrillation is a more prominent risk factor in the Swedish population. Most previous studies have only reported on case fatality as outcome. Our study, however, also included functional outcome. The Riksstroke database has permitted more complete case-mix adjustments than in studies based on routine administrative data only.
A limitation of the present study is that data on stroke care structure (Table I in the online-only Data Supplement) were self-reported by the hospitals and not validated. A further limitation is that there may be some residual confounding although major factors known to be prognostic were included in the statistical models of outcome. Socioeconomic data, except being married/cohabitant versus single, were not collected. It should also be noted that most Swedish community hospitals have higher stroke patient volumes than reported from other countries (eg, ref. 3–5).
To some extent, a different patient mix in university hospitals may explain the differences in procedures by type of hospital. The lower mean age and more pronounced stroke severity at onset may have contributed to a higher proportion of patients undergoing MRI scanning and carotid artery imaging and receiving thrombolysis. The longer mean (but not median) stay in university hospitals could possibly indicate higher proportions of patients with severe stroke. An alternative interpretation is that it reflects less-than-optimal interactions between hospitals and community services in large cities with university hospitals. Despite adjustment for the most common prognostic factors in the regression models, some residual unmeasured confounding cannot be excluded.
Remarkably, there is a 6% unit difference in favor of community compared with university hospitals in patients’ access to direct stroke unit admission. It seems that community hospitals give higher priority to stroke patients than university hospitals do, at least in terms of number of stroke unit beds per population in the hospital catchment area. The large proportion of patients in university and specialized university hospitals not admitted directly to specialized stroke care (stroke unit, intensive care unit, or transferred to a neurosurgical unit) indicates organizational deficits in large hospitals. Patient dissatisfaction with rehabilitation is also more common in university and specialized nonuniversity hospitals than that in community hospitals, and patients treated in university hospitals are less often followed up after discharge. These observations indicate that there are more barriers when organizing well-functioning stroke services in complex university hospital settings. It also seems that stroke care suffers because of the strong internal competition for resources in university hospitals.
The moderate differences in stroke care procedures and outcomes between different types of hospitals should be seen in the context of how the quality improvement in stroke care has developed in Sweden during the past 2 decades. The first version of national guidelines for stroke care was issued by a governmental agency in 1990s, and these guidelines have been regularly updated. A national system for benchmarking of hospital performance (Riksstroke) has been in operation since the mid-1990s, with all hospitals admitting acute stroke patients participating since 1998.23 A large set of performance indicators is publicly available.
One of the aims of the national guidelines and the benchmarking of hospital performance has been to reduce regional and between-hospital differences in stroke care quality. A previous study that used Riksstroke data has shown a considerable delay in adopting new technologies, such as thrombolysis in community hospitals.11 The present data show that differences still exist, but they are modest. Major differences now exist for MRI scanning and thrombectomy. Only a minor fraction of patients first admitted to a community hospital are referred to a larger hospital for these procedures.
After adjusting for case-mix and clustering, the risk of poor outcome was similar in the 3 types of hospital. The similarities in outcome after case-mix adjustment suggest that the various differences in quality of stroke management counterbalance each other. It seems reasonable that this counterbalance is, in part, the result of the emphasis on national guidelines and measurements of hospital performance with publicly available benchmarking.
We observed, however, that, independent of hospital type and case-mix and after adjustment for clustering, the outcome was similar in high-volume and low-volume hospitals. This finding agrees with findings from a nationwide Danish study10 but not with most other studies on stroke patient volume.3–9 It seems that, in some contexts, it is possible to uphold the same quality of stroke care in small as in large hospitals.
Summary and Conclusions
Patients admitted to university hospitals, compared with community hospitals, are less often admitted directly to specialized stroke services and are more often dissatisfied with rehabilitation after discharge from hospital. Patients in university hospitals have better access to diagnostic procedures, and the in-hospital delay to thrombolysis is shorter. After adjusting for differences in case-mix, stroke patient outcome is similar in university and community hospitals. The Swedish setting with national stroke guidelines, stroke units in all hospitals, measurement of hospital performance, and publicly available benchmarking may have helped eliminate differences in outcome between types of hospitals. There seems to be fewer barriers to organizing well-functioning stroke services in community hospitals compared with those in university hospitals.
Sources of Funding
Riksstroke is funded by the Ministry of Health and the Swedish Association of Local Authorities and Regions.
* See the online-only Data Supplement for details.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.007212/-/DC1.
- Received September 15, 2014.
- Revision received December 12, 2014.
- Accepted December 26, 2014.
- © 2015 American Heart Association, Inc.
- Katzan IL,
- Spertus J,
- Bettger JP,
- Bravata DM,
- Reeves MJ,
- Smith EE,
- et al
- Heuschmann PU,
- Kolominsky-Rabas PL,
- Roether J,
- Misselwitz B,
- Lowitzsch K,
- Heidrich J,
- et al
- Carretta HJ,
- Chukmaitov A,
- Tang A,
- Shin J
- Svendsen ML,
- Ehlers LH,
- Ingeman A,
- Johnsen SP
- Eriksson M,
- Jonsson F,
- Appelros P,
- Asberg KH,
- Norrving B,
- Stegmayr B,
- et al
- Reed SD,
- Cramer SC,
- Blough DK,
- Meyer K,
- Jarvik JG
- Fonarow GC,
- Liang L,
- Smith EE,
- Reeves MJ,
- Saver JL,
- Xian Y,
- et al
- Langhorne P,
- Pollock A
- Brainin M,
- Olsen TS,
- Chamorro A,
- Diener HC,
- Ferro J,
- Hennerici MG,
- et al
- Asplund K,
- Jonsson F,
- Eriksson M,
- Stegmayr B,
- Appelros P,
- Norrving B,
- et al
- Eriksson M,
- Asplund K,
- Glader EL,
- Norrving B,
- Stegmayr B,
- Terént A,
- et al
- Schwamm LH,
- Fonarow GC,
- Reeves MJ,
- Pan W,
- Frankel MR,
- Smith EE,
- et al
- Sandercock PA,
- Warlow CP,
- Jones LN,
- Starkey IR