Effectiveness of regionalized systems for stroke and myocardial infarction

Abstract Background Acute ischemic stroke (AIS) and ST‐segment elevation myocardial infarction (STEMI) are ischemic emergencies. Guidelines recommend care delivery within formally regionalized systems of care at designated centers, with bypass of nearby centers of lesser or no designation. We review the evidence of the effectiveness of regionalized systems in AIS and STEMI. Methods Literature was searched using terms corresponding to designation of AIS and STEMI systems and from 2010 to the present. Inclusion criteria included report of an outcome on any dependent variable mentioned in the rationale for regionalization in the guidelines and an independent variable comparing care to a non‐ or pre‐regionalized system. Designation was defined in the AIS case as certification by the Joint Commission as either a primary (PSC) or comprehensive (CSC) stroke center. In the STEMI case, the search was conducted linking “regionalization” and “myocardial infarction” or citation as a model system by any American Heart Association statement. Results For AIS, 17 publications met these criteria and were selected for review. In the STEMI case, four publications met these criteria; the search was therefore expanded by relaxing the criteria to include any historical or anecdotal comparison to a pre‐ or nonregionalized state. The final yield was nine papers from six systems. Conclusion Although regionalized care results in enhanced process and reduced unadjusted rates of disparity in access and adverse outcomes, these differences tend to become nonsignificant when adjusted for delayed presentation and hospital arrival by means other than emergency medical services. The benefits of regionalized care occur along with a temporal trend of improvement due to uptake of quality initiatives and guideline recommendations by all systems regardless of designation. Further research is justified with a randomized registry or cluster randomized design to support or refute recommendations that regionalization should be the standard of care.


Introduction
Acute ischemic stroke (AIS) and ST-segment elevation myocardial infarction (STEMI) are common emergency conditions. Each year in the United States, approximately 87% of the 795,000 new or recurrent strokes are AIS and about a third of the 915,000 new or recurrent acute coronary syndromes (ACS) are STEMI. About three per thousand persons per year are affected, resulting in about 22% of all U.S. deaths and costing the U.S. economy about a quarter of a trillion dollars (Go et al. 2013). Emergency care providers have proposed treatment of these conditions within regionalized systems which (1) designate the level of disease-specific capability of each hospital in a region and (2) establish procedures for emergency medical services (EMS) to bypass facilities of no or lesser designation (Institute of Medicine, 2010). The rationale is to properly match patient need with hospital capability instead of defaulting to the nearest hospital which may lack the required capability.
The history of clinical practice guideline statements for AIS is depicted in Figure 1 (Adams et al. 1994(Adams et al. , 1996(Adams et al. , 2003(Adams et al. , 2005(Adams et al. , 2007 The National Institute of Neurological Disorders and Stroke rt-PA Stoke Study Group, 1995;Levine and Gorman 1999;Alberts et al. 2000Alberts et al. , 2005Alberts et al. , 2011Alberts et al. , 2013Albers et al. 2004;Schwamm et al. 2005;Acker et al. 2007;de Bustos et al. 2009;Del Zoppo et al. 2009;Schwamm et al., 2009a,b;Leifer et al. 2011;Lansberg et al. 2012;Higashida et al. 2013;Jauch et al. 2013;Lackland et al. 2014;Fargen et al. 2015;Powers et al. 2015). Soon after the determination that recombinant tissue plasminogen activator (rt-PA) was the first effective therapy (The National Institute of Neurological Disorders and Stroke rt-PA Stoke Study Group, 1995), stakeholders in stroke care convened the Brain Attack Coalition (BAC) which recommended criteria for primary stroke centers (PSC) (Alberts et al. 2000) to provide the care that most patients would need, and comprehensive stroke centers (CSC) (Alberts et al. 2005) to provide more advanced stroke care. In 2003, the Joint Commission (JC) began to certify PSC based on these criteria. The development of stroke systems of care (Schwamm et al. 2005), EMS systems of prehospital care (Acker et al. 2007), and protocols for EMS bypass of hospitals not capable of stroke care was recommended (Adams et al. 2007). In 2011, the EMS bypass recommendation was graded class I, level B (Alberts et al. 2011); it was upgraded to its current class I, level A in 2013 (Jauch et al. 2013), indicating adequate support by evidence from studies with a randomized design. The BAC has proposed criteria for acute strokeready hospitals to evaluate and triage patients to PSC and CSC as appropriate. Components of a modern stroke care system have been described, including public education outreach focused on symptom recognition and EMS use by patients, prompt recognition and proper triage by EMS personnel, appropriate transport protocols within and between hospitals to match the care with the patient's need, certification of stroke centers to improve outcomes, and implementation of telemedicine systems to support stroke care in rural areas. Endovascular therapy has Adams 1994 [3] NINDS 1995 [4] Adams 1996 [5] Levine 1999 [6] Alberts 2000  ACCP-9 reinforced rt-PA window 4.5 hrs EMS bypass of non-PSC class I, level A EMS bypass to nearest ASRH-PSC-CSC preferred common PSC/CSC data elements proposed for pooled analysis care elements and processes proposed for ASRH factors influencing decline in stroke mortality described meta-analysis: endovascular therapy yields outcomes superior to medical treatment regional systems of care should include endovascular-capable centers recently been associated with superior outcomes compared to medical therapy (Fargen et al. 2015). The current focused update of the AIS guideline of the American Heart Association/American Stroke Association recommends that regional systems of care include endovascularcapable centers for care to be delivered to selected patients (Powers et al. 2015).
The history of guideline statements for AMI and STEMI is depicted in Figure 2 (Gunnar et al. 1990;Ryan et al. 1996Ryan et al. , 1999Widimsky et al. 2000;Andersen et al. 2003;Topol and Kereiakes 2003;Antman et al. 2004Antman et al. , 2008Menon et al. 2004;Bassand et al. 2005;Nallamothu et al. 2005Nallamothu et al. , 2014Henry et al. 2006;Granger et al. 2007;Goodman et al. 2008;Kushner et al. 2009;White 2010;Brush 2012;El Khoury et al. 2012;Levine et al. 2012;O'Gara et al. 2012;Steg et al. 2012;Armstrong et al. 2013;Harold et al. 2013;Kontos et al. 2013;Bagai et al. 2014;Dehmer et al. 2014;Strom et al. 2014). The first guideline published in 1990 recommended thrombolytic therapy for patients with AMI symptoms and a pattern of ST-segment elevations (Gunnar et al. 1990). In 1996, thrombolysis indications were expanded and angioplasty was proposed as an alternative only if it could be implemented in a timely manner by experienced personnel at a high-volume facility (Ryan et al. 1996). In 1999, "timely" was defined as within 90 min of hospital arrival (Ryan et al. 1999). Randomized trials in the Czech Republic in 2000 (Widimsky et al. 2000) and Denmark in 2003(Andersen et al. 2003) demonstrated superior outcomes with transfer for angioplasty compared to thrombolysis. In 2004, the first guideline specifically addressing STEMI referred to primary percutaneous coronary intervention (PPCI) indicating that techniques such as stent implantation could result in superior outcomes compared to simple balloon angioplasty and recommended PPCI as the preferred reperfusion strategy (Antman et al. 2004). This guideline recommended development of EMS destination protocols to PPCI-capable centers and specified procedural volume criteria as a proxy for competence. Later guidelines and scientific papers have urged the provision of care within regionalized systems as "a matter of utmost importance" (Kushner et al. 2009)  angioplasty transfer preferred angioplasty transfer preferred regional and national networks encouraged low center volume associated with higher mortality and longer D2B time Mission: Lifeline regional system demonstration project description center (>36/yr) and operator (>11/yr) procedural volume criteria for PPCI reaffirmed PCI performance measures: center volume better quality indicator than operator volume meta-analysis: low operator volume associated with high MACE but not with mortality   (Brush 2012). The "Mission: Lifeline" regionalized system demonstration project is underway; comparisons of outcomes to accomplishments of nonparticipating systems have not yet been published (Bagai et al. 2014). The current guideline recommends regionalization of STEMI care as class I, level B, indicating a lack of support in studies with a randomized design (O'Gara et al. 2012). Volume criteria for centers specify at least 36 primary procedures per year (Harold et al. 2013); this threshold has been associated with lower mortality and shorter door-to-balloon times (Kontos et al. 2013). Center volume has been described as a better proxy for competence than operator volume (Nallamothu et al. 2014). Low operator volume has been associated with higher rates of major adverse cardiac events but not increased mortality (Strom et al. 2014). The purpose of this article is to provide a narrative review of the effectiveness of regionalized systems in AIS and STEMI sufficient to identify the gap between the adequacy of evidence in the AIS case and the deficiencies of the evidence in the STEMI case.

Search strategy
Literature searches were conducted with terms corresponding to designation of AIS and STEMI centers. Effectiveness was defined as outcome on a dependent variable corresponding to any rationale for regionalization specified in the references cited in Figures 1 and 2. In the AIS case, the terms "primary stroke center" OR "comprehensive stroke center" were searched in PubMed and Scopus, limited to the English language and publication within the trailing 5 years. Evidence regarding the effectiveness of JC-certified systems is provided in Table 1. In the STEMI case, since there is currently no standard means of designation agreed upon by national stakeholders, the terms "regionalization" AND "myocardial infarction" were searched in PubMed and Scopus and limited as above. The yield was augmented by additional searches in two stages. First, literature was searched for papers describing the effectiveness of model systems as described in any reference cited in Figure 2 (North Carolina; Minneapolis, Minnesota; Boston, Massachusetts; France; Denmark; the Czech Republic; and Vienna, Austria). Finally, literature was searched for any paper with a title including the phrase "A Report from the American Heart Association Mission: Lifeline Program." Criteria for a pre-or nondesignated comparison group were relaxed to include any anecdotal or statistical comparison. Papers without any pre-or nondesignated comparison were excluded.

Acute ischemic stroke
The rationale for regionalized systems in AIS is to optimize clinical and functional outcomes, decrease regional variability, reduce process times, and expand timely access. Lichtman and colleagues demonstrated that JCcertified status is associated in Medicare patients with decreased early mortality due to subarachnoid or intracerebral hemorrhage (Lichtman et al., 2011a) and with decreased early mortality, shorter hospitalization, and favorable disposition in ischemic stroke (Lichtman et al., 2011b). Neither study demonstrated reduced readmission rates.
A common finding is that certified systems can increase rates of rt-PA utilization. Panezai et al. (2014) showed that New Jersey CSC designation compared to PSC was associated with shorter process times and twice the proportion of eligible patients receiving thrombolysis. This analysis was limited by implementation of PSC designation after data collection had begun, whereas CSC designation was already established. Bhattacharya et al. (2013) found that JC certification mitigated racial disparities affecting African Americans and improved unadjusted rates of rt-PA utilization; but, when adjusted for delayed presentation and nonuse of EMS, these findings became nonsignificant. Rajamani et al. (2013) reported that JCcertified status in Michigan was associated with increased rates of rt-PA utilization in eligible patients but not favorable disposition or lower mortality. Mullen et al. (2013a) found that JC-certified status was associated with increased rates of rt-PA utilization. Their study design did not permit identification of hospitals participating in non-JC certification programs. Prabhakaran et al. (2012) reported that PSC status is associated with more than doubled rt-PA utilization that began several years before achievement of certified status and continued to steadily increase; however, there was no difference in mortality or complications (Prabhakaran et al. 2013). The authors acknowledged that not all PSC in the study area were sampled and that this might limit generalizability. McKinney et al. (2011) concluded that state CSC designation in New Jersey, but not PSC designation, mitigated the weekend effect of increased mortality and that state designation, but not JC designation, was associated with decreased 90-day mortality. State and JC designation were both associated with increased rates of rt-PA utilization.
Another common finding is that certified status is associated with improved adherence to other core measures. Johnson et al. (2014) demonstrated that JC PSC certification, as well as preparation for certification that would be achieved during the study period, was associated with Overall rate of rt-PA use was low and did not differ by status on certification AA race was associated with delayed presentation Self-presentation versus EMS was associated with nonreceipt of rt-PA Core measure compliance was better at certified hospitals, especially for Caucasian patients All disparities were present only, or more pronounced, at noncertified hospitals All disparities n.s. adjusted for delay The retrospective design and chart review methodology did not allow for detailed investigation of delayed presentation or nonuse of EMS JC certification mitigates disparity in access to care but does not increase rt-PA use adjusted for delayed presentation and nonuse of EMS sustained improvement in core measure adherence in North Carolina. Ballard et al. (2012) reported that PSC designation compared to predesignation performance in the Kaiser Permanente system is associated with reduced process times and improved utilization of timely imaging but not lower mortality, favorable disposition, or shorter hospitalization. Lewis et al. (2011) found that JC-certified compared to noncertified status was associated with better adherence to the anticoagulation performance measure for patients with atrial fibrillation. Certified status is associated with mitigation of disparity in access to stroke care. Bhattacharya et al. (2013) demonstrated that JC certification reduced racial disparities in receipt of rt-PA. Albright et al. (2012) showed that care at hospitals meeting CSC criteria mitigated the socalled "weekend effect" of increased mortality affecting off-hours presenters.
Despite the accomplishments reviewed above that have been associated with certified status, there is still room for improvement. Mullen et al. (2013b) reported that 10 years after JC certification became available, residents of the so-called "stroke belt" of the southeastern U.S. were still underserved. Fonarow et al. (2013) demonstrated that Performance Achievement Award recognition by the American Heart Association (AHA) was a more reliable indicator of performance than JC certification. Leira et al. (2012) concluded that expansion of access by the population of Iowa to care at certified centers would be superior if nomination for certified status were informed by a maximal coverage model compared to the current system of self-nomination. Lichtman et al. (2009) compared unadjusted and riskadjusted rates of 30-day mortality and readmission at centers which were early adopters of JC certification to corresponding rates at centers which did not achieve early certification. The authors demonstrated that JC certification identified centers with superior outcomes preceding the availability of the JC certification program by several years. In a recent review of factors associated with overall decreased stroke mortality conducted after 10 years of societal experience with the JC certification program, Lackland et al. (2014) concluded that evidence is inconclusive regarding the effect of certification of stroke systems of care on mortality. Taken as a group, these findings indicate that certified status is associated with superior performance; however, it is more nearly correct that certified status tends to be earned by centers already displaying superior performance and less nearly correct that certified status conveys superior performance capability. Certified status as a factor in superior performance is not well understood when other factors such as delayed presentation, nonuse of EMS, and improved performance by non-or precertified centers are also present.

ST-segment elevation myocardial infarction
The rationale for interventions specified in the first acute myocardial infarction (AMI) guideline was prevention of mortality (Gunnar et al. 1990), whereas the rationale for interventions specified in the first STEMI guideline was prevention of mortality and clinical and functional benefit from preserved left ventricular function (Antman et al. 2004). Recommendations for regionalization of STEMI care have been justified by expansion of access to reperfusion. Publications from six regionalized systems were found which included (1) at least one dependent variable corresponding to these rationales, (2) care provided by a PPCI-capable system, and (3) a non-or predesignated comparison. Clemmensen et al. (2013) in Denmark, Kalla et al. (2006) in Austria, Danchin et al. (2008) in France, and Le May et al. (2008) in Ontario, Canada reported decreased mortality with regionalized care. The Denmark report is an anecdotal assertion of an historic low; the other three reports are comparisons with baseline performance. Benedek et al. (2013) reported that implementation of the European Stent for Life quality initiative in Romania was followed by elimination of disparity in mortality affecting those initially presenting to a non-PPCI-capable hospital. Jollis et al. (2012) reported that statewide implementation of the Regional Approach to Cardiovascular Emergencies (RACE) in North Carolina resulted in an historic low proportion of patients clinically eligible but not reperfused. In an earlier report from the RACE system, Jollis et al. (2007) stressed that the study was not designed to examine mortality and that inferences based on their outcomes should be made with great caution. Henry (2012) described the RACE program as a model case of voluntary participation of all hospitals statewide. Glickman et al. (2012) concluded that despite the achievements of this system, mortality was not improved. Glickman et al. (2010) demonstrated that this system reduced all process times for female and elderly patients compared to baseline, significantly mitigating the baseline disparity for female patients, but that the disparity affecting elderly patients persisted. Forsyth et al. (2012) demonstrated that between 2001 and 2009, the proportion of patients admitted directly to a high-volume PCI center (then defined as performing >400 PCI procedures per year) in Florida increased from 62.4% at baseline to 89.7% at the study's end. This result was achieved despite the lack of a formal statewide protocol; the interpretation is that superior process can be achieved by "de facto" or self-organizing systems without formal governance. The system did not eliminate the disparity in admission to high-volume centers affecting female and elderly patients (Table 2).

Discussion
In the AIS case, the independent variable in the studies reviewed is status on an actual program of designation by JC certification and the comparison is to a non-or predesignation cohort. In the STEMI case, citation as a model system by a guideline statement or a scientific paper was chosen as proxy for designation and comparisons were less often supported by statistical comparisons. Taken as a group, these findings indicate that from a societal perspective, the goals of regionalization are incompletely realized. Most of the studies reviewed share several limitations.

Selection bias
Most of the studies are retrospective analyses of databases with voluntary participation. Several of the studies are comparisons of prospectively collected postimplementation data to historical preimplementation data reported without external validation. Where authors acknowledged this limitation, they asserted that audits had deemed the data valid. In the STEMI case, citation as a model system cannot occur unless system capabilities include not only clinical practice but also academic pursuits such as publication; this too can introduce selection bias.

Delayed and/or self-presentation
Most designs did not allow data collection regarding the proportion of patients who either delayed presentation past eligibility for reperfusion or presented by means other than EMS. The exception is Bhattacharya et al. (2013) who found that self-presentation compared to EMS use was associated with nonreceipt of rt-PA. This is likely an unmeasured confounder in other studies.

Hospital characteristics
The classification of hospitals as certified often included hospitals meeting criteria for designations which did not exist at the time data were collected. In several studies, the implementation of certification programs during data collection was acknowledged as a limitation along with other hospital characteristics.

Missing data
In several studies, authors acknowledged that data collection had been extended into the past before relevant data were routinely collected because the relevant variables were not yet defined. For example, it was generally not possible to control analyses for stroke severity because  NIH Stroke Scale data had not been routinely collected. Historical comparisons for rt-PA use were difficult because this therapy has not been consistently represented by reimbursable code; it has been shown that assignment of rt-PA treatment by allocation of procedure code results in an underrepresentation of actual numbers of patients treated Palazzo et al. (2014); Kleindorfer et al. (2008).

Secular trend
In the AIS case, the comparison group was either a preor a nondesignated cohort. Improvement in the comparison group was usually noted along with that which occurred in the certified group. In the STEMI case, the comparison was often a historical reference to baseline data without a statistical test. Authors usually acknowledged that their findings could not reliably be attributed to the effect of certification distinct from a secular trend.

Self-nomination
In several of the studies reviewed, the authors acknowledged that JC certification is a process for which hospitals self-nominate. This may not be a nuisance confounder but rather a requirement for a facility whose stakeholders seek to serve in a regionalized system. From a societal point of view, it makes sense to prefer organizations whose leadership is willing, if not eager, to fulfill this responsibility.
Achievement Fonarow et al. (2013) demonstrated that superior achievement in adherence to Get With the Guidelines (GWTG)-Stroke measures, and recognition by the AHA on that basis, is a better predictor of performance than JC certification. This may represent the best of both worlds in that voluntary participation in a quality initiative, followed by superior performance, identifies organizations not only willing but also capable of sustained implementation of best practices.

Recommendations
There is insufficient evidence from randomized designs to support or refute assertions that regionalization should be the standard for stroke care, although the weight of the observational evidence accumulated over the 10 years since JC certification became available indicates that the program is achieving its rationale. Since there is no standard definition of a regionalized system of STEMI care, it is not surprising that the proxy chosen, citation as a model system, yields even less evidence than in the AIS case.
It is important that conceptual clarity be achieved regarding what is, and what is not, a regionalized system so that stakeholders will know whether further efforts at formal regionalization are justified. There are several viable candidate definitions in the STEMI case. The Mission: Lifeline demonstration project is underway Bagai et al. (2014); when outcomes are published, there will be evidence regarding the effectiveness of systems so designated. The AHA sponsors an Accredited Heart Attack (STEMI) Receiving Center program (American Heart Association 2014); hospitals are designated by demonstrating that specific standards have been met. Since the first STEMI guideline (Antman et al. 2004), procedural volume has been proposed as proxy for evidence of competence (Levine et al. 2012;Kontos et al. 2013;Dehmer et al. 2014). Hospitals reporting volume to any database can be ranked from highest to lowest and an appropriate cutpoint can be selected for designation.
It is likely that there would be a great deal of overlap among hospitals with the above characteristics; in fact, the various candidate definitions for facility designation may be conceptually distinct ways of describing and measuring the same phenomenon. The difficulty in finding or generating evidence to either support or refute the STEMI regionalization recommendation is that there is not one explicit definition of the intended system attributes, but rather there several attractive candidate definitions. The accuracy of any research finding is dependent on the use of explicit definitions of variables (Wunsch et al. 2005). The best solution may be the one observed in Florida by Forsyth et al. (2012), who described what happened when participants are allowed to find their own place in the system. It may be that superior organizational ambition and superior outcomes will go hand in hand.
National registries such as Action-GWTG (American College of Cardiology, American Heart Association) are already in place enabling comparison of risk-adjusted mortality between a system and its national benchmark. Lauer and D'Agostino (2013) propose the randomized registry trial as a means of generating high-quality evidence by taking a random sample from such registries. This design would tend to avoid the limitations of the evidence reviewed above, confirm or refute causal links, and enable the pursuit of big data with small budgets.
Chew and Blows (2009) recommend a cluster randomized clinical trial design with the hospital as the unit of analysis, broad inclusion criteria, and data collection including a prespecified outcome of 30-day mortality and a means of tracking all the associated costs. A large universe of qualified candidate participating hospitals would be required for meaningful random selection to result in a sample suitable for meaningful random assignment to an experimental regionalized protocol or to usual care. If the effect of a universal single-payer health insurance program is an independent variable of interest, the universe of hospitals should ideally include many candidate hospitals from Canada and the U.S., for example. The complexity of such a design would be enormous; equipoise regarding the merits of networked versus un-networked care would be necessary and problematic for investigators and stakeholders at all candidate participating hospitals and both governments.
Until evidence from studies with a randomized design is available, the afferent flow of patients from the countryside to specialized centers should be matched with a brisk efferent flow of cognitive skill, decision support, and continuing medical education for rural providers. Decentralized grids of rural emergency response capability should be established or maintained and staffed with properly trained and equipped multidisciplinary teams so that guideline-adherent emergency care can begin where the patient lives. Regional systems that are in place should ensure that vigorous public outreach is undertaken or continued with a focus on symptom recognition and prompt EMS use.

Conclusion
Regionalization of emergency care for AIS and STEMI, defined as designation of hospital capabilities and EMS bypass of facilities with no or lesser designation, has been associated with expanded access to care, mitigation of disparity, and enhanced process. Improvement in outcomes due to regionalized care is difficult to distinguish from that which occurs due to quality initiatives and uptake of best practices by all facilities regardless of designation. Further research should include specific comparisons of outcomes achieved by designated systems to national benchmarks, nonregionalized systems, or pre-regionalized baseline states.