Acute stroke alert activation, emergency service use, and reperfusion therapy in Sweden

Abstract Objectives Ambulance services and stroke alerts reduce the time from stroke onset to acute stroke diagnosis. We describe the use of stroke alerts and ambulance services in different hospitals and patient groups and their relationship with reperfusion therapy. Methods This nationwide study included 49,907 patients admitted with acute stroke who were registered in The Swedish Stroke Register (Riksstroke) in 2011–2012. Results The proportions of patients admitted as stroke alerts out of all acute stroke admissions varied from 12.2% to 45.7% in university hospitals (n = 9), 0.5% to 38.7% in specialized nonuniversity hospitals (n = 22), and 4.2% to 40.3% in community hospitals (n = 41). Younger age, atrial fibrillation (AF), living in an institution, reduced consciousness upon admission, and hemorrhagic stroke were factors associated with a higher probability of stroke alerts. Living alone, primary school education, non‐European origin, previous stroke, diabetes, smoking, and dependency in activities of daily living (ADL) were associated with a lower probability of stroke alert. The proportion of patients arriving at the hospital by ambulance varied from 60.3% to 94.5%. Older age, living alone, primary school education, being born in a European country, previous stroke, AF, dependency in ADL, living in an institution, reduced consciousness upon admission, and hemorrhagic stroke were associated with ambulance services. Hospital stroke alert frequencies correlated strongly with reperfusion rates (r = .75). Conclusion Acute stroke alerts have a significant potential to improve stroke reperfusion rates. Prehospital stroke management varies conspicuously between hospitals and patient groups, and the elderly and patients living alone have a markedly reduced likelihood of stroke alerts.


| INTRODUCTION
High-priority ambulance service and stroke prenotification are key components in reducing prehospital delay and time to diagnosis in patients with suspected acute stroke (Fassbender et al., 2013;Meretoja et al., 2012Meretoja et al., , 2013Oostema, Nasiri, Chassee, & Reeves, 2014). Stroke alerts (prenotification by ambulance staff members or initiated in the emergency department) have been shown to be strongly associated with shorter in-hospital delay to the onset of treatment (door-toneedle time [DNT]) and higher proportions of patients being treated with thrombolysis (Binning et al., 2014;Lin et al., 2012a;McKinney et al., 2013;Patel, Rose, O'Brien, & Rosamond, 2011;Prabhakaran, O'Neill, Stein-Spencer, Walter, & Alberts, 2013;Ragoschke-Schumm et al., 2014). Additional benefits of prenotification have been reported in patients with warfarin-associated intracerebral hemorrhage, in whom the time to reversal of anticoagulation was shortened (Dowlatshahi et al., 2013). Rapid triage and immediate stroke team activation (stroke alert) are important components of Target: Stroke, which is a nationwide quality-improvement initiative of the American Heart Association and the American Stroke Association to improve the care of stroke in the US (Fonarow et al., 2011(Fonarow et al., , 2014, and they are recommended in the guidelines issued by the European Stroke Organisation (2008).
In Sweden, national guidelines include stroke alerts as one of the indicators of high-quality acute stroke care (National Board of Health and Welfare, 2015). However, annual reports from the Swedish Stroke Register (Riksstroke, 2014) show marked regional and betweenhospital variations in the use of ambulance services and stroke alerts (Riksstroke) (as has also been reported in the US; Lin et al., 2012b).
We aimed to describe (1) between-hospital variations in the use of stroke alerts and ambulance services in different types of hospitals; (2) longitudinal changes in stroke alerts and the use of stroke alerts and ambulance services in different patient groups; and (3) time from onset to hospital admission and the association between stroke alert frequency, ambulance services, and reperfusion therapy rates. Riksstroke was established in 1994 to monitor, support was included and modeled as a fixed effect (Varewyck, Goetghebeur, Eriksson, & Vansteelandt, 2014). The effect of hospital type was analyzed in a separate model not including individual hospital effects.

| Material
Pearson correlation was used to assess the association between hospital stroke alert frequency, ambulance service, and hospital reperfusion rate. We performed a separate analysis of in-hospital stroke alerts in the subgroup of patients who did not use ambulance.
Education level and smoking status were missing for >2% of the patients, and these were included in the analysis by adding a separate category ("unknown") in the multiple regression. Patients with other missing variables were omitted from the multiple regression analysis. A difference with a p-value <.05 was considered statistically significant.
We used SAS 9.4 for statistical analyses. After adjustment for other patient factors, the probability of stroke alert was lower in patients older than 74 years than in younger patients. Living alone, primary school education, non-European origin, previous stroke, diabetes, smoking, and dependency in ADL were other factors associated with a lower probability of stroke alert in the multiple regression model. The probability of stroke alert was higher in patients with AF, those who were living in an institution, those who were drowsy at hospital admission, and those who had a hemorrhagic stroke ( Table 1). The observed socioeconomic differences in stroke alert frequencies persisted throughout the study period ( Figure 1).

| Stroke alerts
The frequency of in-hospital stroke alerts was 6.7% in the 11,372 patients who did not use ambulance transport to hospital 2011-2012.
After adjustment for other factors, older patients, patients born outside the Nordic countries, and patients living alone had lower probability of in-hospital stroke alerts (Table S1). In-hospital stroke alert frequencies were higher in university hospitals than in community hospitals.

| Ambulance services
The proportion of patients arriving at the hospital by ambulance in Older age, living alone, primary school education, European origin, previous stroke, AF, dependency in ADL, living in an institution, being drowsy or unconscious at admission, and hemorrhagic stroke were factors associated with use of ambulance services in the multiple regression (Table 2). Patients arriving at the hospital by ambulance also had a higher probability of stroke alert (29.6% vs. 6.7%, p < .001).  There was a strong correlation between hospital stroke alert frequency and reperfusion rate (r = .75, p < .001, Figure 2), but not between hospital reperfusion rate and ambulance service frequency (r = −.02, p = .84).

| DISCUSSION
This study shows substantial between-hospital variation in the use of stroke alert and ambulance services, also within hospitals of the same type. There are important factors at the individual patient level that are associated with the likelihood of having a stroke alert. Older age, less severe stroke, and living alone are factors that are independently associated with a low chance of having a stroke alert. Even after adjustment for individual patient characteristics, stroke alerts are more common in university hospitals than in nonuniversity hospitals.
Hospitals with frequent use of stroke alerts have a high proportion of patients with ischemic stroke being treated with acute reperfusion (thrombolysis or thrombectomy). The use of ambulance services follows a partly different pattern with similar average proportions in university and nonuniversity hospitals, and a predominance of elderly patients with a low level of education. Common to ambulance services and stroke alerts is that they are used more infrequently with immigrants born outside Europe and, as expected, more often with patients with severe stroke.
Most previous studies on stroke alerts have been performed in single centers (Binning et al., 2014;McKinney et al., 2013) or at the regional level (Patel et al., 2011;Prabhakaran et al., 2013), but a very large US study involving 1,585 hospitals has been performed within the Get-With-The-Guidelines-Stroke (GWTG-Stroke) project (Lin et al., 2012a(Lin et al., , 2012b. Our results agree with those from the US in that there is a marked between-hospital variation in the use of stroke alerts (Lin et al., 2012b). Our results also agree with previous observations that a high proportion of acute stroke patients with stroke alerts is associated with frequent use of reperfusion therapy such as thrombolysis (Lin et al., 2012a;McKinney et al., 2013;Prabhakaran et al., 2013).
Contrary to the US findings (Lin et al., 2012b), we observed that stroke alerts were more common in university than in nonuniversity hospitals.
The use of stroke alerts is low in Sweden compared to the US (23% in this study vs. 64% in US hospitals with >50 stroke admissions per year; Lin et al., 2012b), highlighting a potential to further improve the access to acute reperfusion. To the best of our knowledge, only one previous study has analyzed how patient-level factors relate to stroke alerts.
In the GTWG-Stroke study, factors associated with lower use of stroke alerts included older age, having diabetes mellitus, and having peripheral vascular disease. Stroke alerts were less likely among black patients than white patients (Lin et al., 2012b). We found socioeconomic factors to be independently associated with the frequency of stroke alerts, and having a low level of education reduced the odds of stroke alert by 5%, stroke, diabetes, and being dependent in ADL before the index stroke or unconscious on arrival at the hospital. Because stroke alerts are associated with shorter time from admission to diagnosis and reversal of anticoagulation in warfarin-associated intracerebral hemorrhage (Dowlatshahi et al., 2013), it is worth noting that patients who were diagnosed with an intracerebral hemorrhage were more likely than average to have a stroke alert.
At first glance, it would seem paradoxical that ambulance services are used less frequently and stroke alerts more frequently in people with a high level of education compared to those with a low level of education. However, it has previously been reasonably well established that even after adjustment for multiple possible determinants, emergency medical transports are used more often by socioeconomically disadvantaged people (Rucker, Edwards, Burstin, O'Neil, & Brennan, 1997;Svenson, 2000), whereas people with a high level of education have higher awareness of stroke warning symptoms (Hickey et al., 2009).
The present observations help to explain the socioeconomic differences in access to thrombolytic treatment that we have previously described. Compared with stroke patients with primary school education, patients with university-level education are more often treated with thrombolysis (Stecksen et al., 2014). Our results now show that stroke alerts are more frequent and that OAT times are shorter in patients with a high level of education. The use of ambulance services per se does not seem to follow the same socioeconomic pattern.
Living alone versus cohabiting does not adversely affect the access to ambulance services, but it markedly reduces the frequency of stroke alerts, most likely because of delays in calling 112 (911 in the US) and because of more uncertainties about onset of stroke symptoms. As a result, fewer patients living alone were admitted to hospital within 3 hr of onset of stroke symptoms (present results), and 44% fewer received thrombolytic treatment (Stecksen et al., 2014) compared to patients who were cohabiting. It seems that elderly single people should be especially targeted when campaigns to improve stroke awareness in the population are designed (Desai, Herter, Riccardi, Rorden, & Fridriksson, 2015).

| Strengths and limitations
Strength of this study is that it is nationwide and covers all hospitals admitting acute stroke patients in the country and that it has high coverage of all patients admitted to hospital for acute stroke in the country. This study was based on individual-level socioeconomic information, like the GWTG-Stroke study (Lin et al., 2012b). Most studies on socioeconomic status (SES) and different aspects of stroke have used SES data based on states, counties, or postal numbers.
We acknowledge that geographical-level SES data may add another dimension of information to the more specific individual-level SES data; individual-level and geographical-regional data should be seen as complementary.
A limitation of this register-based study is that it was not possible to distinguish between stroke alerts elicited by ambulance services (also called prehospital triage [Prabhakaran et al., 2013] or prehospital notification [Blomberg, Lundstrom, Toss, Gedeborg, & Johansson, 2014]), emergency room prenotification, or other modes of stroke alert. Hospital effects were accounted for in the statistical models, but the associations between patient characteristics and stroke alert frequency remained. This indicates that stroke alerts in themselves, irrespective of how they were applied in each hospital, account for the associations, but it does not exclude that some modes of application might have a stronger relationship than others. Another limitation is that we had no information on negative stroke alerts (i.e., stroke alerts in patients who were subsequently not diagnosed with acute stroke).
This includes the relatively small proportion of patients who fully recovered after thrombolysis and were diagnosed with transient ischemic attack. A limitation was also that we did not have information on whether a patient was living in an urban or a rural setting (that could partly explain differences in stroke alerts between different types of hospital).
Results from observational studies like the present one raise the question of possible residual confounding variables that might have influenced some or all of the findings. There are, for instance, many aspects of living conditions that might affect the use of stroke alerts and ambulance services, for example, distance to hospital, that might affect the use of stroke alerts and ambulance services, but these have not been measured. Riksstroke did not include a stroke scale during the study period, and hence we could not specify which stroke signs influenced stroke alerts. There was no information on stroke size or location. Level of consciousness at hospital arrival was used as a surrogate for stroke severity.
F I G U R E 2 Hospital reperfusion rate (%) in ischemic stroke patients ≤80 years versus hospital stroke alert frequency in all admitted stroke patients, 2011-2012. The size of the bubble corresponds to hospital patient volume

| CONCLUSION
Stroke alerts are low-technology interventions with a significant potential to improve acute stroke outcome. Because every second stroke alert in Sweden actually leads to reperfusion treatment (Figure 2), it is notable that 50% of stroke patients are living alone in Sweden (Table 1) and that they have a 37% reduction in the likelihood of having a stroke alert. Inequity in other socioeconomic factors, such as level of education and country of birth, also contribute considerably to reduce the use of stroke alerts.
Public stroke warning campaigns have been reported to have varying degrees of success (Mellon, Doyle, Rohde, Williams, & Hickey, 2015;Rasura et al., 2014). The present findings may help to define population groups that should be targeted to make early ambulance calls. Such a specific target group would be elderly people living alone.