Coverdell National Acute Stroke Registry Michigan

Objective: To assess the use of IV recombinant tissue plasminogen activator (rt-PA) in a statewide hospitalbased stroke registry and to identify factors associated with its use among eligible patients. Methods: A modified stratified sampling scheme was used to obtain a representative sample of 16 hospitals. Prospective case ascertainment and data collection were used to identify all acute stroke admissions over a 6-month period. Subjects eligible for IV rt-PA were defined as those who arrived within 3 hours of onset, who had no evidence of hemorrhage on initial brain image, and who had no physician-documented reasons for non-treatment with IV rt-PA. Multivariate logistic regression was used to identify factors associated with IV rt-PA use. Results: Of 2,566 stroke admissions, 330 (12.9%) met the eligibility criteria for rt-PA treatment, and of these 43 (13%) received IV rt-PA treatment. Among 2,236 admissions excluded from consideration, 21% had evidence of hemorrhage on initial imaging, 35% had unknown stroke onset times, 38% had an onset to arrival time 3 hours, and 6% had physician documented contraindications. Among eligible patients, being male, use of emergency medical services, and rapid presentation were associated with increased IV rt-PA use. Conclusions: Treatment with IV rt-PA was underutilized in this hospital-based stroke registry. The primary reason for nontreatment was delayed presentation. Reducing prehospital and in-hospital response times would help increase IV rt-PA use, as would greater emergency medical services use. Improving the documentation of onset times would help clarify the underlying causes of delayed presentation. NEUROLOGY 2006;66:306–312 Since the approval of IV recombinant tissue plasminogen activator (rt-PA) in June 1996, its use among patients with acute stroke has been disappointingly low; several studies have reported treatment rates of between only 1.2% and 9% among acute ischemic stroke admissions.1-6 A recent study summarizing the results of four state prototypes participating in the Paul Coverdell National Acute Stroke Registry (PCNASR) reported that between 3% and 8.5% of patients with ischemic stroke were treated with rt-PA.7 Major reasons cited for the underutilization of rt-PA include the short 3-hour treatment window,8,9 the lack of hospital infrastructure to facilitate the administration of rt-PA,10 and physicians’ concern regarding the efficacy and safety of rt-PA treatment, especially in more community-based settings.8,11,12 Delay in presentation is the most common exclusion reported, however, with most studies finding that only about 20 to 25% of patients with acute stroke arrive to the hospital within 3 hours.3,7,13 However, even among this group, the majority of patients still do not receive IV rt-PA treatment; a recent study found that only 29% of subjects who presented within 3 hours of stroke onset were eligible for IV rt-PA treatment.13 Other common reasons for exclusion from IV rt-PA therapy include mild or rapid resolution of stroke symptoms, and medical contraindications such as hypertension, seizure, and bleeding disorders.3,9,13,14 Commentary, see page 297 From the Departments of Epidemiology (Y.Z.D., M.J.R., A.J.M.) and Neurology and Ophthalmology (G.L.B., S.W., A.M.), Michigan State University, East Lansing; Department of Neurology (B.S.J.), Wayne State University, Detroit; Borgess Medical Center (R.U.K.), Kalamazoo; and Department of Neurology (S.L.H., K.M.), University of Michigan, Ann Arbor. Supported by US Centers for Disease Control and Prevention Collaborative Agreement No. U50/CCU520272-01. Disclosure: The authors report no conflicts of interest. Received June 16, 2005. Accepted in final form October 3, 2005. Address correspondence and reprint requests to Dr. Mathew J. Reeves, Department of Epidemiology, B 601 West Fee Hall, Michigan State University, East Lansing, MI 48824; e-mail: [email protected] 306 Copyright © 2006 by AAN Enterprises, Inc. at McGill University on March 9, 2006 www.neurology.org Downloaded from The objectives of this study were to use data from a statewide hospital-based stroke registry that prospectively identified acute stroke admissions to describe the use of IV rt-PA treatment, and to identify factors associated with its use among eligible patients. Methods. Study design and hospital selection. The Michigan Acute Stroke Care Overview & Treatment Surveillance System (MASCOTS) was a statewide, hospital-based, stroke registry prototype that was part of the initial development phase of the PCNASR.7,15 A single-stage cluster design was used, whereby a sample of 16 Michigan hospitals was obtained from a modified stratified sampling regime. First, in the initial pilot phase of the project, we selected eight hospitals with certainty from four urban communities that were participating in a community-based stroke awareness project. These hospitals represented larger urban and academic institutions. Second, the remaining 114 statewide acute care hospitals that had at least 30 stroke discharges in 2000 were ranked according to their total number of stroke discharges, and four equal sized strata were created. Two hospitals were then randomly selected, with probability proportional to size, from each of the four strata, which in conjunction with the eight pilot phase hospitals resulted in a final sample of 16 hospitals. These 16 hospitals treated 26% of the 36,000 annual stroke discharges statewide in 2000. Subsequent examination of the characteristics of these 16 selected hospitals showed that the sample was broadly representative of Michigan in terms of its urban, suburban, and rural communities, and key demographic characteristics of its population. Case ascertainment and acute stroke case definition. Following an initial pilot phase, data were collected at the 16 hospitals over a 6-month period from May to November 2002. To identify all acute stroke admissions, prospective case ascertainment was carried out by trained hospital personnel who regularly reviewed the emergency department (ED) admission log, ward logs, intensive care unit log, and neurology consults. All consecutive hospital admissions older than 18 years of age that had a chief complaint or clinical signs and symptoms consistent with acute stroke onset were considered as suspect stroke cases and were documented in a logbook. To be included in the final registry of acute stroke admissions, subjects then had to meet one of seven acute stroke subtype definitions: ischemic stroke (IS), intracerebral hemorrhage (ICH), subarachnoid hemorrhage (SAH), TIA, ischemic stroke of uncertain duration (ISUD), hemorrhagic stroke of uncertain type (HSUT), or stroke of uncertain type. To be included in the registry TIA cases had to exhibit symptoms at the time of initial ED presentation. The final stroke subtype diagnosis was determined at the time of discharge and may have been different from that suggested by the initial image results. These acute stroke subtype definitions were based on previous work,16 and were used by all of the participating prototype sites.7,15 Data collection. A standardized data collection instrument was developed to obtain information on demographics; use of emergency medical services (EMS); ED arrival and evaluation; initial brain imaging; stroke onset time; NIH Stroke Scale (NIHSS); rt-PA treatment; in-hospital diagnosis, procedures, and treatment; prior medical history; discharge status; and secondary prevention interventions. Information on the initial ED evaluation, including that relevant to the decision to treat with IV rt-PA, was collected at the time the patient arrived at the hospital. Other data, such as in-hospital procedures, treatments, discharge instructions, modified Rankin Scale (mRS) at discharge, and inhospital mortality, were abstracted from the charts after discharge. In order to ensure the accuracy and uniformity of data collection across the sites, the data coordinators attended a training session prior to the onset of data collection and had access to ongoing training and technical support throughout the study. All data submitted to the main study center underwent an extensive series of quality and logic checks that allowed the hospital staff to update the information through an iterative process of data editing. Prior to the start of the study, a survey of the 16 participating hospitals was conducted to obtain hospital-level information relevant to acute stroke care. This included basic hospital characteristics (i.e., the number of admissions, bed size of the hospital, rural or urban location), the presence of an acute stroke team, the availability of stroke specialists, the use of written clinical guidelines and care pathways, access to imaging technology, use of in-hospital rehabilitation therapy services, the presence of stroke quality improvement programs, and participation in databank or other registry-like activities. Human subjects’ approval was obtained from each hospital’s Institutional Review Board (IRB) before the start of the study. All IRBs granted the project exempt status because the registry’s primary purpose was quality improvement, and no direct patient contact was involved. It was therefore not possible to collect outcomes data post-discharge. Definition of eligible subjects for IV rt-PA therapy. Subjects eligible for IV rt-PA treatment were defined as all acute stroke admissions seen in the ED within 3 hours of stroke onset, who had no evidence of hemorrhage on the initial brain image, and who had no physician-documented reasons for non-treatment with IV rt-PA. Outcome measures and statistical analysis. The primary outcomes of interest were the proportion of eligible patients who received IV rt-PA, and the occurrence of symptomatic ICH within 36 hours of IV rt-PA treatment or other life-threatening systemic hemorrhage requiring blood transfusions within 7 days of treatment. Two other secondary outcomes measures, in-hospital mortality and mRS at discharge (i.e., 4 vs 3), were compared between IV rt-PA treated and non-treated patients who had a final discharge diagnosis of IS.17 For descriptive analysis, categorical data are presented as percents, and continuous data are presented as means with SD. The t test was used to test for differences in continuous variables, and the 2 or Fisher exact test was used to compare categorical data (i.e., proportions). To identify the demographic or clinical factors associated with the use of IV rt-PA therapy, univariate logistic regression analysis was first conducted followed by a backwards stepwise multiple logistic regression analysis.18 Age, sex, and race were considered a priori confounding variables and were therefore included in all models regardless of significance. Other candidate explanatory variables included nursing home residence, ED arrival mode (i.e., EMS vs other), stroke onset to ED arrival time, ED arrival day (i.e., weekday vs weekend), ED arrival time (i.e., morning, afternoon, evening, or night), CT image results (i.e., normal, old infarct, acute/subacute infarct, other), pre-stroke ambulatory status, and previous medical history. These variables were considered in the multivariable modeling phase only if they had a univariate significance level of p 0.3. Variables that were not significant in the multivariate model (i.e., p 0.05) were excluded in a hierarchical backward elimination fashion. Relevant interaction terms involving main effect variables were tested in the final multivariable model.18 Adjusted ORs and 95% CI were reported for variables in the final model. Statistical analyses were performed using SAS software, version 8.2 (SAS Institute Inc., Cary, NC). Results. One of the original 16 hospitals closed down during the early phase of the study. A total of 2,566 acute stroke patients were admitted to the 15 remaining hospitals during the 6-month period. Of these 2,566 stroke admissions, 469 (18.3%) had hemorrhagic signs on their initial brain image and were excluded from further consideration (figure). Among the remaining 2,097 nonhemorrhagic acute stroke admissions, 793 (37.8%) were excluded because their stroke onset times were unknown, 851 (40.6%) were excluded because they arrived 3 or more hours after onset, and 123 (5.9%) had physiciandocumented contraindications to IV rt-PA treatment. The remaining 330 patients, representing 12.9% of the 2,566 admissions, were therefore considered eligible for rt-PA therapy. Among these, a total of 47 patients (14.2%) received rt-PA therapy, of which 43 (13%) received IV rt-PA (see figure). These 43 patients represented 2.1% of the 2,097 acute non-hemorrhagic stroke admissions, or 1.7% of the 2,566 admissions. The two major reasons documented by the treating phyFebruary (1 of 2) 2006 NEUROLOGY 66 307 at McGill University on March 9, 2006 www.neurology.org Downloaded from sician for not giving IV rt-PA to the 123 subjects were that the patient had undergone a significant improvement (n 44, 35.8%), and that the stroke severity was considered too mild (n 22, 17.9%). Other physician-documented reasons for exclusion included abnormal activated partial thromboplastin time (n 6), life expectancy 1 year or severe comorbidity (n 4), stroke too severe (n 4), history of ICH, or brain aneurysm, or tumor (n 4), CT suggested an alternative diagnosis (n 3), advanced age (n 2), seizure at onset (n 2), recent surgery/trauma (n 2), recent intracranial (IC) surgery (3 months) or head trauma (n 2), uncontrolled hypertension (n 1), or other miscellaneous reasons (n 17). Finally, consent was not available or the family refused treatment in 10 patients. Univariate logistic analyses that compared the 43 IV tPA-treated patients to the 283 non-treated patients found differences (p 0.05) by sex, arrival mode, and onset to ED arrival time (table 1). The past medical histories did not differ significantly between treated and non-treated groups (data not shown). The results of the final multivariate logistic regression analysis are shown in table 2. Sex, ED arrival mode, and stroke onset to ED arrival time remained significant in the final model. Female patients were significantly less likely to get IV rt-PA treatment than male patients (OR 0.4, 95% CI 0.2 to 0.8), and patients who arrived via EMS were seven times (OR 7.3, 95% CI 2.9 to 18.2) more likely to receive IV rt-PA treatment, compared to patients who arrived by non-EMS. Stroke onset to ED arrival time was strongly associated with likelihood of IV rt-PA treatment. Patients who arrived during the second hour after stroke onset were less than half as likely to receive IV rt-PA compared to those who arrived within an hour of onset (OR 0.4, 95% CI 0.2 to 0.9), while those who arrived between 2 and 3 hours after onset were 33 times less likely to be receive treatment (OR 0.03, 95% CI 0.003 to 0.2). After adjustment, neither age (p 0.5) nor race (p 0.8) had an effect on the likelihood of IV rt-PA use. Finally, two-way interaction terms between arrival mode and age, sex and race, Figure. Flow diagram defining the potentially eligible patients with acute stroke for rt-PA therapy. *51 of the 851 patients (6%) had missing information on the emergency department arrival time, hence an onset to arrival time could not be calculated. Table 1 Demographic and clinical characteristics of IV rt-PAtreated vs non-treated subjects among eligible patients (n 326)*