Cuyahoga County Operation Stroke Speed of Emergency Department Evaluation and Compliance With National Institutes of Neurological Disorders and Stroke Time Targets
Background and Purpose— Although the National Institutes of Neurological Disorders and Stroke (NINDS) has established time targets for the evaluation of acute stroke patients, little is known about how often these targets are met in the community. To track compliance with NINDS guidelines, Cuyahoga County Operation Stroke implemented a countywide data collection process that provides comparative reporting of hospital performance.
Methods— Data on the emergency department evaluation of stroke patients have been retrospectively collected since January 2000 by hospitals participating in Cuyahoga County Operation Stroke. All hospitals used a 1-page form that includes NINDS time target information.
Results— Median time to emergency department arrival for the 1003 patients in this cohort was 115 minutes; 382 patients (38%) arrived in <3 hours; 506 (50%) arrived in <6 hours. After arrival, median time to physician contact was 12 minutes, time to CT was 65 minutes, and time to imaging results was 105 minutes. Earlier arrival intervals after symptom onset were independently associated with shorter evaluation times. All time targets were met in patients receiving intravenous tissue plasminogen activator.
Conclusions— A countywide data collection system for acute stroke evaluation is feasible. In the Cleveland metropolitan area, time to physician contact is close to the recommended NINDS target. Time to CT and time to imaging results are substantially longer than recommended. However, there was wide variation between hospitals. The association between time to arrival, speed of evaluation, and administration of intravenous tissue plasminogen activator suggests that community physicians selectively accelerate the evaluation and management of potential thrombolysis candidates.
Intravenous tissue plasminogen activator (tPA), the only approved treatment for acute ischemic stroke, must be administered within 3 hours of symptom onset.1 The short time window has necessitated institution-wide changes in hospitals throughout the United States so that eligible stroke patients can be evaluated and treated with intravenous tPA.
To assist hospitals in assessing institutional readiness to give intravenous tPA, time targets for the evaluation of acute stroke patients were developed at a National Institutes of Neurological Disorders and Stroke (NINDS) Symposium on the Rapid Identification and Treatment of Acute Stroke.2 Although widely accepted, little is known about how well community hospitals are able to meet the proposed time targets. The few available studies3–5 involved hospitals that participated in thrombolytic clinical trials and may not be representative of other US institutions. Adding impetus to the need to better understand the ability of community hospitals to comply with these targets, the Brain Attack Coalition recently recommended the establishment of primary stroke centers.6 One proposed criterion for primary stroke center designation is documentation that the NINDS time targets can be met.
Operation Stroke (OS), an initiative of the American Stroke Association, has been a major force in organizing healthcare delivery systems in the United States to better manage acute stroke patients. Cuyahoga County, located in northeastern Ohio and including the Cleveland metropolitan area, has had an OS task force in place since 1999. Of the 20 hospitals in Cuyahoga County, 18 are participating in OS. An additional affiliated hospital from neighboring Geauga County is also participating. A main goal of the Cuyahoga County OS is to increase the county rate of intravenous tPA use from 1.2% to 5%. To assess progress and to identify barriers to achieving this goal, we implemented a data collection process for all OS hospitals that provides comparative information on the speed of emergency department (ED) evaluation of acute stroke patients. Another goal was to assess the feasibility of establishing a community-wide acute stroke registry. We report our experience with the first 15 months of data collection, spanning January 2000 to March 2001.
Materials and Methods
Members of the OS medical subcommittee worked with The Stroke Group (Englewood, Colo) to develop a succinct and user-friendly data form for use at all Cuyahoga County hospitals. A 5-page “Stroke Stats” data form, which was used as the initial starting point, underwent significant modifications with extensive community physician input. The final 2-page data form, very similar to Ethos, the Stroke Group’s current Web-based stroke registry data form, was approved by the medical subcommittee. To be a member of Cuyahoga County OS, hospitals are required to participate in data collection, which has been approved by the Institutional Review Boards of all OS hospitals. Two pharmaceutical companies provided initial funding support to The Stroke Group to provide comparative hospital reporting to OS hospitals. No funding was provided for administrative support of the project, development of the final data form, or individual hospital data collection.
Hospitals in Cuyahoga County OS began retrospective data collection on patients with a primary discharge diagnosis of stroke (International Classifications of Diagnosis, ninth revision, clinical modification codes 431 through 436.) in January 2000. To minimize use of hospital resources while maximizing the usefulness of the data, OS required data collection only on patients arriving at the ED within 6 hours of symptom onset. Collection of information for stroke patients who arrived >6 hours from symptom onset, who were direct hospital admissions, or who had strokes during their hospital admission was optional.
Data Collection and Comparative Hospital Reporting
Data were collected by trained abstractors at each participating hospital using a 1-page data form that included information on demographics, method of hospital arrival, tPA use, and other acute interventions. No patient-identifying information such as name, clinic number, or birth date was collected. Each hospital was identified by a numeric code known only to that hospital. Time target information included time of symptom onset, hospital arrival time, time to initial physician contact, time to CT, and time to imaging results.
Hospital abstractors met in September 1999 to review definitions and patient inclusion and exclusion criteria. Symptom onset time was defined as the time the patient was last known to be well. If the time was not clearly specified but was known to be within 3 hours, then the time of 180 minutes was recorded. Similarly, if the time was not clearly specified but was known to be within 6 hours, abstractors used the time of 360 minutes. To allow more uniform data collection across hospitals, “time to CT” was defined as the time the patient was transported to CT rather than the time the initial CT was completed, which is the NINDS time target. No assessment of reliability of data abstraction was performed.
The completed paper forms were mailed in batches to the local American Heart Association office, where they were reviewed to ensure that they contained no patient- or hospital-identifying information. Data forms were then mailed to The Stroke Group in Colorado and entered into a relational database for analysis. Anonymous comparative hospital reporting was provided at the end of this 15-month period (March 2001). Each hospital was given individual results, along with an aggregate of Cleveland data divided into various time intervals from stroke onset to ED arrival (<3, 3 to 6, >6 hours).
Speed of Evaluation
Median values of the time elements were used for descriptive statistics because of their nonnormal distributions. Evaluations of time elements were stratified by time from symptom onset to ED arrival by use of the following intervals: <2, 2 to 3, 3 to 6, and >6 hours. Regression analyses were used to assess the significance of the arrival interval to the speed with which the evaluation was completed after log transformation of the time element data. To better define patients who were potentially eligible for intravenous tPA, patients with an ED diagnosis of transient ischemic attack were excluded from the speed of evaluation analyses for time to CT and time to imaging results. ED transfers were also excluded in assessments of these evaluation times because imaging was likely to have been done at the initial ED. Variation in speed of evaluation across hospitals was assessed with the Kruskal-Wallis test.
Method of Arrival
Logistic regression was performed to assess the effect of time from symptom onset (<2, 2 to 3, 3 to 6, >6 hours) with the use of Emergency Medical Services (EMS). It was also used to evaluate the use of EMS over time divided into 3-month intervals after adjustment for time from symptom onset to arrival.
Factors Affecting Speed of Evaluation
Three separate linear regression analyses were performed to evaluate the independent effect of the variables sex, race, time period (divided into 3-month periods), time interval to arrival (<2, 2 to 3, 3 to 6, >6 hours), and arrival by ambulance on the speed of 3 portions of the evaluation: time to physician contact (MD), time to CT, and time to imaging results. All covariates were entered as 1 block in each regression analysis. Log transformation of the time elements was done to normalize their distributions.
All analyses were performed with the SPSS 9.0 statistical package.
Of the 20 hospitals in Cuyahoga County, 14 collected data during the period of January 2000 to March 2001. Information was collected on 1003 patients, 50.4% of whom (506 of 1003) arrived within 6 hours of symptom onset (Table 1). The mean age of patients was 72.7 years, and 79.5% were white.
Speed of Evaluation
Median time to MD was 12 minutes, close to the NINDS time target of 10 minutes (Table 2). Patients arriving >6 hours after symptom onset had longer median times, although all other subgroups were fairly similar and near the national time target. There were differences in the median time to MD across hospitals (P=0.001), which ranged from 0 to 32 minutes. Data on time to MD were documented in 67% of cases.
Overall median time to CT was 65 minutes, with significantly longer times in patients arriving >6 hours from symptom onset. The only identified patient subgroup that met the NINDS time target were patients receiving intravenous thrombolysis. Median time to CT across hospitals ranged from 28 to 122 minutes (P=0.001). Data on time to CT were documented in 67% of cases.
Median time to imaging results was 102 minutes, much longer than the recommended 45-minute goal. Patients receiving intravenous tPA were the only subgroup that met the national time target. As with the other time measures, patients arriving >6 hours from symptom onset had longer median times to imaging results. There was wide variability in the median time to imaging results across hospitals, ranging from 43 to 193 minutes. Data were available in only 37.8% of patients.
Time to CT and time to imaging results were nonsignificantly longer for patients arriving within 2 hours of symptom onset who came to the ED between midnight and 4 am (Table 3). There was no change in any of the evaluation times over the 15-month period (data not shown).
Method of Arrival
Stroke patients arrived by ambulance in 63.9% of cases. Ambulance use was inversely associated with time to arrival (P<0.001). There was no significant difference in ambulance use according to race among any of the arrival intervals examined (<2, 2 to 3, 3 to 6, >6 hours; data not shown). Women had higher rates of ambulance use only among the 3-to-6-hour arrival interval (67.2% versus 48.9%, P=0.043). There was no change in the use of ambulance over time (P=0.476) with and without adjustment for time from symptom onset to arrival. Patients arriving by ambulance were evaluated more quickly, with significantly shorter unadjusted time to MD, time to CT, and time to imaging results (Figure 1).
Factors Affecting Speed of Evaluation
In regression analyses, factors independently associated with time to MD were arrival by ambulance (P=0.007) and time from symptom onset to arrival (inverse relationship, P=0.004). The only factor independently associated with time to CT and time to imaging results was initial time to arrival (P<0.001). Sex, race, and date of arrival (divided into 3-month periods) had no significant independent effect on speed of evaluation.
Intravenous tPA Use
The rate of intravenous tPA use was 6.8% in this select population of patients and 18.9% in patients arriving within 3 hours of symptom onset. The mean±SD age of patients receiving intravenous tPA was 69.1±13.4 years, 78% were white, and 66% were women, similar to the overall sample. There was no difference in rates of intravenous tPA use among patients according to sex (P=0.99) or race (P=0.18). However, there was wide variability across hospitals in the rates of intravenous tPA use for patients with no documented exclusions, ranging from 6.8% to 100% (P<0.001) for the 8 hospitals for which data were available (Figure 2). Notably, all evaluation times fell within the NINDS recommended time targets for patients receiving intravenous tPA.
Since the approval of intravenous tPA, the evaluation of acute stroke patients has become a priority nationwide.2,6–8 However, most metropolitan areas or regional health systems do not systematically collect performance data on patients with acute stroke. The Cuyahoga County OS initiative demonstrates that a voluntary countywide stroke data collection process is feasible. We believe important factors in the success of this process included the nonpunitive nature of the project and the ability for anonymous comparative reporting of individual hospital performance with the rest of Cuyahoga County in aggregate. In addition, to participate in OS, hospitals were required to collect data, which resulted in peer pressure among Cleveland hospitals. There was also stakeholder input with extensive community physician input into the data collection process. Finally, there was institutional support for the data collection process.
The NINDS stroke time targets were empirically established on the basis of the “golden hour” and the 1-hour door to needle time for acute myocardial infarction. The time targets were meant to be optimal guidelines for hospitals considering the administration of intravenous tPA for acute stroke; they were not based on any data to suggest that hospitals could in fact comply with the suggested targets. In the Cleveland region, time to initial MD contact is close to the recommended time target of 10 minutes. It took substantially longer, however, to obtain a CT than recommended by the NINDS. Our median time to initiation of a CT (time the patient was transported to CT) among patients arriving within 2 hours was 50 minutes, whereas the NINDS target for CT completion is 25 minutes. Our findings are in keeping with those of Morris et al,5 who found a median time to imaging of 1.1 hours in hospitals participating in a clinical tPA trial. Interestingly, time to CT was longer among patients arriving within 2 hours who came to the ED between midnight and 4 am, suggesting that speed of obtaining a CT depends on the time of day. The CT time during all periods of the day should be a focus of quality improvement efforts because it is a significant limiting factor in the ability of hospitals to complete acute stroke evaluations and start therapy in a timely manner. Neurological outcome after intravenous tPA is inversely related to the time from stroke onset to initiation of therapy.9,10
One potential explanation for the prolonged time to CT and time to imaging results, despite the prompt initial evaluation by physicians, is that only patients who may qualify for tPA or another acute intervention are entered into an accelerated stroke evaluation protocol. In fact, all time targets were met in patients who received intravenous tPA. Although we cannot exclude the possibility that their prompt evaluation enabled tPA patients to receive the therapy, earlier arrival intervals after symptom onset were independently associated with shorter evaluation times, which supports the presence of patient triaging. Thus, stratification according to the time interval to arrival is important in interpretation of individual hospital performance.
Caution should be used when interpreting data regarding the time imaging results were received. There was significant variability in times, possibly reflecting different and perhaps inconsistent methods of abstraction for this data element. The data element was present in only one third of the cases. For purposes of ongoing quality management, it may be useful to redefine the time targets in terms of data points that hospitals may reliably track.
The high percentage of missing data related to acute stroke management has been noted by others11 and highlights the importance of systematic approaches to improving documentation. The variability in available information raises the possibility of documentation bias; patients who were treated more swiftly according to protocol may be more likely to have times recorded in the chart, especially for poorly documented items such as time to interpretation. Cleveland-area hospitals are developing individual approaches to improve documentation. A strategy successful at 1 hospital has been placing the data form on the ED record, with those elements requiring ED physician input highlighted. Another potential strategy is the development of a standard ED documentation sheet for stroke patients that has reminders of inclusion and exclusion criteria for intravenous tPA, the elements necessary for billing, and the clinical and time elements important for the evaluation of institutional readiness. Prospective data collection is another option but would require extensive cooperation among physicians and nurses working all shifts in all EDs, reducing the chance that all hospitals in the county would be able to participate.
There is wide variation in speed of evaluation among hospitals, with median hospital times to initial MD contact ranging from 0 to 32 minutes and median hospital times to initiation of imaging ranging from 28 to 122 minutes. The rate of intravenous tPA use among potentially eligible patients ranged from 6.8% to 100% across hospitals. Although the percentage of tPA-eligible patients varies across institutions, these data suggest significant differences in Cleveland hospitals in either institutional preparedness or documentation, and we hope that the countywide data collection will alert poorly performing hospitals and rouse them to improve their system for managing acute stroke patients.
Only 64% of patients arrived to the ED via ambulance. Patients arriving to the ED via ambulance had shorter times to contact with a physician, independent of the time from symptom onset to arrival. Although this may be due in part to a higher acuity of illness in these patients, the data suggest that in addition to being transported more quickly, the evaluation is initiated more quickly in patients using EMS. This has been found in other studies5,12 and is another reason to educate potential stroke patients regarding the importance of calling 911 if they have strokelike symptoms.
There was no significant improvement in the speed of evaluation or patient use of EMS transport over the 15-month period that this information was collected, despite the active presence of OS in our county. However, there was no opportunity to use the data to effect change. Cleveland hospitals received only 1 report near the end of the 15-month period. We strongly recommend that community projects use a system that allows immediate reporting of the data such as that available from Web-based software. Cuyahoga County OS began using Ethos, the Internet version of the form that provides immediate comparative reports, in November 2001. Currently, 18 of 20 hospitals in Cuyahoga County and 1 affiliated hospital from neighboring Geauga County are actively entering data through the Internet.
Our regional data collection system has several limitations, including lack of information on the reliability of the data. In this initial effort, auditing of charts was not feasible. We hope that funding for this type of activity will become possible with the Stop Stroke Act currently before Congress. In addition, as discussed above, a significant percentage of specific time points was missing. We are actively working to improve the routine documentation of this information in the hospitals in Cleveland. Although a list of exclusion criteria for intravenous tPA was part of the data collection, the completeness of chart abstraction for these items is not known. Despite these limitations, this countywide quality improvement effort has demonstrated a community-wide commitment to optimize stroke care and provided important insights into the care of stroke patients in the Cleveland metropolitan region.
Cuyahoga County OS Hospitals
Bedford Hospital, Cleveland Clinic Foundation, Euclid Hospital, Fairview Hospital, Geauga Regional Medical Center, Hillcrest Hospital, Huron Hospital, Kaiser Permanente, Lakewood Hospital, Lutheran Hospital, Marymount Hospital, MetroHealth Medical Center, Parma Community Hospital, Richmond Heights Hospital, South Pointe Hospital, Southwest General Health Center, St John’s West Short Hospital, St Vincent’s Charity Hospital, and University Hospitals of Cleveland.
Genentech, Inc and AstraZeneca Inc provided funding support for the comparative hospital reporting done by The Stroke Group. Cuyahoga County OS gratefully acknowledges the work of the nurses and quality management personnel at each hospital whose dedication made this initiative possible. We also thank Annette DiRosa, Health Strategies Director at the American Heart Association, Northeastern Ohio Affiliate, for her tireless efforts throughout the 3 years of the Cuyahoga County OS Task Force.
Drs Furlan and Sila have received speaking honoraria from Genentech (tPA). Mr Houser is the president of The Stroke Group, which provided the comparative analysis for individual Operation Stroke Hospitals in March 2001 and operates Ethos, a Web-based stroke registry used by Cuyahoga County Operation Stroke.
Cuyahoga County Operation Stroke hospitals are listed in the Appendix, which is available online at http://stroke.ahajournals.org.
- Received July 3, 2002.
- Revision received October 4, 2002.
- Accepted October 11, 2002.
Marler J, Jones P, Emr Me. Rapid identification and treatment of acute stroke. In: Proceedings of a National Symposium. Bethesda, Md: National Institutes of Health; 1997. NIH publication 97–439.
Bratina P, Greenberg L, Pasteur W, Grotta JC. Current emergency department management of stroke in Houston, Texas. Stroke. 1995; 26: 409–414.
Morris DL, Rosamond W, Madden K, Schultz C, Hamilton S. Prehospital and emergency department delays after acute stroke: the Genentech Stroke Presentation Survey. Stroke. 2000; 31: 2585–2590.
Marler JR, Tilley BC, Lu M, Brott TG, Lyden PC, Grotta JC, et al. Early stroke treatment associated with better outcome: the NINDS rt-PA Stroke Study. Neurology. 2000; 55: 1649–1655.
Brott T. Analysis of the NINDS, ECASS I, ECASS II and ATLANTIS Data Sets. In: Proceedings of the 27th International Stroke Conference, February 7-9, 2002, San Antonio, Tex.
Nilasena D, Kresowik T, Wiblin R, Piskac A, Kresowik R, Brenton M, Wilwert JM, Hendel ME. Assessing patterns of tPA use in acute stroke. Stroke. 2002; 33: 354.Abstract.
Schroeder EB, Rosamond WD, Morris DL, Evenson KR, Hinn AR. Determinants of use of emergency medical services in a population with stroke symptoms: the Second Delay in Accessing Stroke Healthcare (DASH II) study. Stroke. 2000; 31: 2591–2596.