In-Patient Code Stroke
A Quality Improvement Strategy to Overcome Knowledge-to-Action Gaps in Response Time
Background and Purpose—Stroke is a relatively common and challenging condition in hospitalized patients. Previous studies have shown delays in recognition and assessment of inpatient strokes leading to poor outcomes. The goal of this quality improvement initiative was to evaluate an in-hospital code stroke algorithm and educational program aimed at reducing the response times for inpatient stroke.
Methods—An inpatient code stroke algorithm was developed, and an educational intervention was implemented over 5 months. Data were recorded and compared between the 36-month period before and the 15-month period after the intervention was implemented. Outcome measures included time from last seen normal to initial assessment and from last seen normal to brain imaging.
Results—During the study period, there were 218 inpatient strokes (131 before the intervention and 87 after the intervention). Inpatient strokes were more common on cardiovascular wards (45% of cases) and occurred mainly during the perioperative period (60% of cases). After implementation of an inpatient code stroke intervention and educational initiative, there were consistent reductions in all timed outcome measures (median time to initial assessment fell from 600 [109–1460] to 160 [35–630] minutes and time to computed tomographic scan fell from 925 [213–1965] to 348.5 [128–1587] minutes).
Conclusions—Our study reveals the efficacy of an inpatient code stroke algorithm and educational intervention directed at nurses and allied health personnel to optimize the prompt management of inpatient strokes. Prompt assessment may lead to faster stroke interventions, which are associated with better outcomes.
In-hospital acute strokes account for 7% to 15% of all acute cerebrovascular events and represent a challenge for healthcare systems.1–4 Special characteristics make this group more susceptible to a higher incident risk of stroke and poorer outcomes compared with patients arriving from outside the hospital. For example, in-hospital stroke patients are usually older, have higher prevalence of comorbid conditions,2,4–6 and many occur during the perioperative period, often among patients undergoing cardiac procedures.1–3,5 Others occur on cardiology, general medicine, or surgical wards while receiving care for other medical conditions requiring hospitalization.5,7
More recently, there has been a focus on educating the general public to recognize the cardinal features of stroke and to create rapid triage systems so that patients with suspected acute stroke in the community setting can be routed to the nearest stroke center.8–10 The same degree of effort has not been spent on in-hospital strokes.
On first consideration, patients who have a stroke while admitted to hospital should be at an advantage compared with out-of-hospital strokes. Inpatients already are in a monitored environment, have rapid access to nurses and physicians, are in close proximity to imaging facilities, and often have recent laboratory testing. Together, these factors should facilitate prompt recognition, investigation, and management of acute stroke in hospital. However, numerous studies revealed greater delays in the care of in-hospital stroke patients compared with out-of-hospital stroke patients.5,7,9,11,12 Some identified causes include lack of education about identifying stroke on inpatient wards, delayed notification of the most appropriate personnel, and poor communication about the need for urgent medical evaluation.
Consequently, there is an opportunity for improvement in the care of patients who have an acute stroke while admitted to hospital. We hypothesized that an educational intervention targeting first responders (eg, nurses and physicians) would improve initial response time for inpatient strokes, a critical outcome determining access to new interventional treatment for acute stroke.
Our aim was to develop an algorithm for the management of inpatient strokes, implement an educational program as a quality improvement intervention to be disseminated to inpatient wards, and assess if this reduced delays in the management of inpatient stroke.
St. Michael’s Hospital is a 500-bed tertiary care academic and teaching hospital located in downtown Toronto, Canada. The hospital has active general neurology, general internal medicine, cardiology, and cardiovascular surgery wards, in addition to other medical and surgery wards and intensive care units. There are ≈2 to 4 in-hospital strokes per month, and many are identified beyond the time window for intervention. No protocol existed for managing in-hospital strokes. Approval was obtained from our institutional research ethics board for this study.
Development of the Intervention
Our initial step was to collect baseline data for inpatient strokes on the time from last seen normal (LSN) to initial assessment and time from initial assessment to brain imaging (eg, computed tomographic [CT] scan or magnetic resonance imaging), in order to confirm that these times were not ideal and present these data to stakeholders. We met with key stakeholders to determine perceived barriers to timely and effective care of in-hospital stroke patients, including neurologists, the stroke medical director, stroke clinical nurse specialists, regional education coordinators, nurse managers, nursing staff, and allied health staff (physiotherapists, occupational therapists, and speech-language pathologists). These interviews revealed that there was a lack of knowledge about identifying strokes, the need for timely evaluation of an acute stroke, and the availability of a stroke team.
The medical education team discussed possible interventions aimed at achieving a tangible change in behavior (eg, rapid in-hospital stroke assessment and management). After the discussion with stakeholders, an algorithm was developed for the identification and initial management of in-hospital stroke (Figure 1). On the basis of the algorithm, a 13-slide electronic presentation was developed, which was presented at in-service education sessions lasting 60 minutes, on each of the targeted inpatient units (this presentation is available in the online-only Data Supplement). The wards receiving this intervention included the following: cardiovascular surgery, cardiology, the cardiovascular intensive care unit, the cardiac catheterization laboratory, general internal medicine, respirology, neurology, nephrology, orthopedic surgery, neurosurgery, trauma surgery, vascular surgery, and general surgery. The in-service education sessions contained information on stroke symptoms and signs, the importance of speed because of a tight time window for thrombolysis, and the process of activating a code stroke and were delivered to the nursing staff on each unit, as well as the unit managers and the allied health staff (eg, physiotherapists, speech-language pathologists). These groups were targeted because they represent stable staff members on each ward that most commonly identify stroke symptoms. The learning objectives of the educational intervention included the following: to be able to describe the different types of strokes, to be able to recognize the common signs and symptoms of an acute stroke, to understand why acute stroke is an emergency, to be aware of the treatment options for an acute stroke, to describe the role of different medical team members in activating a code stroke, and to describe the procedure for activating a code stroke for an inpatient. During each in-service educational session, a written log was kept of questions that were asked or needed clarification, allowing the presentation to adapt if there were gaps or portions that were unclear. Laminated posters of the inpatient code stroke algorithm were placed throughout the inpatient units, and pocket cards were provided to nursing staff.
The algorithm was reviewed with key stakeholders (stroke nurse practitioner and case manager, medical education team, neurology, internal medicine, and stroke leaders) responsible for the management of inpatients with an acute stroke. Because of the high turnover of medical residents, our intervention was targeted at nurses, unit managers, and allied health. However, the inpatient code stroke algorithm was distributed to all residents starting rotations at St. Michael’s Hospital and reviewed during their new staff orientation sessions.
The primary outcome was time from LSN to initial assessment. Secondary outcomes included the following: time from LSN to brain imaging, time from initial assessment to brain imaging, poststroke complications, neurological deficits, number of patients receiving intravenous thrombolysis, or vascular interventional procedures. Time to acute stroke treatment was not chosen as a primary outcome as most inpatient strokes would not be likely to receive thrombolysis or interventional procedures. Timed variables were collected prospectively for both the pre- and postintervention periods. Chart reviews were performed of all patients who had a stroke while admitted to hospital for the study period, identified by reviewing discharge summaries, consults to neurology, or activation of the inpatient code stroke protocol. Data abstractors were trained on how to collect the necessary information using a standardized data collection form. Demographic features, presenting symptoms, stroke severity, vascular risk factors, and complications after the stroke were collected.
The preimplementation (baseline) study period was from April 2006 to April 2009. The implementation period was May 2009 to October 2009 and then a postimplementation evaluation period went from November 2009 to February 2011.
Adjusted and unadjusted primary and secondary outcome measures were compared pre- and post-implementation of the intervention (comparing the April 2006 to April 2009 time period with the November 2009 to February 2011 time period). For the multivariate analysis, data were adjusted for demographics, stroke severity as defined by the number of neurological deficit at onset of stroke,13 and baseline cardiovascular risk factors (eg, hypertension, coronary artery disease).
Unadjusted comparisons between the preimplementation and postimplementation groups with respect to categorical indicators were performed using the χ2 test or the Fisher exact test. Continuous variables were compared between groups using the 2-sample t test or the nonparametric Wilcoxon rank-sum test. We considered the exponential, Weibull, generalized γ, log-normal, and log-logistic regression models to estimate time ratios and 95% confidence intervals to measure the effect of post–Inpatient Code Stroke intervention on time from LSN to initial assessment, time from LSN to brain imaging, and time from initial assessment to brain imaging, adjusted for any previous surgery, indicator of at least 2 neurological symptoms, and coronary artery disease. The model that best fit the data was the one with smaller Akaike information criterion. For times recorded as zero minutes, we added 0.5 minutes to include them in the statistical models. We also included run charts, commonly used in quality improvement studies, to display process performance over time.
All analyses were conducted using SAS 9.4 (SAS Institute, Inc, Cary, NC), and statistical significance was defined as 2-sided P values <0.05.
Overall, 245 inpatient strokes were identified during the 2 study periods. Twenty-seven patients were excluded as outliers because they were identified >72 hours from LSN and could have represented preexisting or nonacute deficits. The final sample size was 218 inpatient strokes (131 in the preintervention period and 87 in the postintervention period). Demographic characteristics of the cohort are shown in Table 1. Coronary artery disease was more common in the postintervention group (P=0.028), but other demographic features and comorbidities did not differ (Table 1). Overall, 60% of inpatient strokes occurred in the perioperative period (defined as a stroke in which symptoms or signs were first noted on the patient waking in the postanesthesia care unit), and this was significantly more common in the preintervention group (P=0.012).
The most commonly identified symptoms were unilateral weakness (86%), speech disturbance (46%), decreased level of consciousness (30%), and facial droop (29%). A larger proportion of milder strokes were observed in the postintervention group compared with preintervention (78.1% versus 48.1%; P=0.001), and more patients were identified with speech disturbance after the intervention. The distribution of inpatient strokes by type of ward is shown in Table 2, and these did not differ statistically comparing the pre- and postintervention periods. The cardiovascular service accounted for the largest proportion of inpatient strokes (42.9% overall).
After implementation of the inpatient code stroke protocol, there were a total of 35 inpatient code stroke activations out of 87 inpatient strokes. All of these were appropriate uses of the code stroke protocol, since those 35 patients were within 4 hours of LSN and had signs or symptoms of stroke. The most common reason for not activating the code stroke was that the time from LSN was either unknown or >4 hours.
Table 3 compares the outcomes for the preintervention and postintervention groups. For the entire cohort of in-hospital strokes, the median time from LSN to initial assessment fell from 600 minutes (109–1460 minutes) before implementation of the educational code stroke intervention to 160 minutes (35–630 minutes) after its implementation (P=0.0065; Figure 2). Similarly, median time from LSN to brain imaging fell from 925 minutes (213–1965 minutes) to 348 minutes (128–1587 minutes; P=0.0288) and from initial assessment to brain imaging scan fell from 135 minutes (43–480 minutes) to 110 minutes (51–331 minutes), although this difference did not reach significance (P=0.5088; Figure 2). For the 35 patients in whom a code stroke was activated, the median time from LSN to initial assessment was 75 minutes and to CT scan was 125 minutes, both significantly shorter than that during the preintervention phase (P<0.0001 for both outcomes). Run charts were created for the year before the implementation of the intervention and the postintervention period, documenting the time from LSN to initial assessment (Figure 3A) and from LSN to brain imaging (Figure 3B) over time. Although there were fluctuations in these times on a case-by-case basis, the overall trend for lower time to initial physician assessment and brain imaging was sustained throughout the postintervention period.
Reduction in times was also estimated after adjusting for the presence of cardiovascular risk factors, >2 neurological deficits, and whether the stroke was perioperative. Using the generalized γ regression model, the time from patient LSN to initial assessment was significantly reduced by 35.4% (time ratio=0.646; 95% confidence interval, 0.455–0.918; P=0.0147). Likewise, the time from patient LSN to brain imaging was significantly reduced by 38.4% (time ratio, 0.616; 95% confidence interval, 0.412–0.921; P=0.0182). However, reduction in time from initial assessment to brain imaging was not significant (log-logistic regression model, P=0.1894; time ratio=0.729; 95% confidence interval, 0.454–1.169; reduction=27.1%).
Few patients received intravenous thrombolysis (n=12) or endovascular intervention (n=2) to conduct an analysis (Table 3).
Inpatient strokes are medical emergencies and should be afforded the same high-quality care as strokes that occur out of hospital. Numerous studies have documented delays in the evaluation and management of strokes in hospitalized patients.3,12,14 This may lead to poor outcomes for patients experiencing inpatient strokes, including long hospitalization and greater disability.4,5 Although many institutions have protocols for stroke patients arriving through the emergency department, such protocols do not always exist for in-hospital strokes.
In this study, we have shown that the implementation of an inpatient code stroke algorithm combined with stroke education targeting key stakeholders can improve the response times (eg, LSN to initial assessment and brain imaging). These time periods are 2 key measures of access to class I-level evidence treatment options for acute stroke (eg, thrombolysis or endovascular thrombectomy). The target of our intervention was the dissemination of knowledge and translation of knowledge-to-action gaps among key stakeholders who care for inpatients, in order to change their conception of stroke as an emergency and to alter behavior (management of a suspected inpatient stroke). The fact that there were significantly more mild strokes (with fewer neurological deficits) identified in the postintervention phase suggests that the educational intervention may have improved the recognition of early or mild strokes. In addition, a larger percentage of inpatient strokes were perioperative in the preintervention phase, suggesting that the intervention may have had a larger impact on nonsurgical wards. This finding may be explained by improvements in the detection of subtle signs of stroke; in the postoperative period, patients are often closely monitored, whereas on a medical floor where patients may have multiple comorbidities, subtle signs of stroke may be more easily overlooked.
There was strong consensus among stakeholders that the inpatient code stroke protocol was an important step in improving the quality of inpatient stroke care, as allied health felt better equipped to respond to a suspected stroke, and found that physicians were more responsive to their request for an urgent assessment for a suspected inpatient stroke (as evidenced by the large number of code strokes activated by physicians). This concept became more relevant with the introduction of endovascular thrombectomy as part of the standard of acute stroke care. In this study, the small number of patients treated with intravenous thrombolysis is attributable mainly to the presence of contraindications, including recent surgery and medical comorbidities (eg, gastrointestinal bleeding). Few patients were treated with endovascular thrombectomy because this procedure was not yet standard of care at the time of our study.
One recent study demonstrated that compared with out-of-hospital strokes, patients with in-hospital strokes showed longer times to neuroimaging, lower rates of thrombolysis, and left with greater poststroke disability.5 Similarly, in-hospital strokes are less likely to meet various quality-of-care metrics, such as the Get-With-the-Guidelines-Stroke achievement and quality measures.4
Many factors lead to delayed recognition or assessment of the hospitalized patient with an acute stroke, which may result in delayed or missed treatment opportunities.7,15 Factors suggested as leading to delays include the fact that neurological deficits may be attributed to other general medical conditions; medication effects; or delirium, lack of education on stroke signs, the short time window for thrombolysis, and the lack of a dedicated protocol for triaging acute stroke in hospitalized patients.1,7,9,14–16
The disparity in the care of inpatient versus out-of-hospital stroke should be amenable to improvement because inpatients have the potential to be diligently monitored and are close to resources such as CT scanners and stroke teams. In a study by Cumbler et al,17 quality improvement methodology was applied to reduce time to evaluation for inpatient strokes. In that study, the quality improvement methodology focused on creating a care pathway that included a checklist of ideal practices and were able to reduce time to CT scan by 57%.17 A unique feature of our study is that our primary intervention addressed both gaps in knowledge and action gaps when a stroke was identified, with emphasis on prompt action. This was addressed through the education of ward personnel and the creation of a formalized inpatient stroke protocol.
Our study has several limitations that deserve comment. First, this is a single-center quality improvement study. As such, we caution about the generalizability of our results. On the contrary, our protocol was embedded into standard practice targeting relevant time benchmarks reflecting care in most stroke centers. Second, we cannot completely exclude the effect of residual confounding from other variables. Third, even after the intervention, the median time from LSN to brain imaging was still 110 minutes, higher than recommended by the current guidelines.18 There is, therefore, still room for further systems-based quality improvement strategies to drive this value lower. Fourth, this study was performed before the publication of seminal trials demonstrating the effectiveness of interventional thrombectomy for acute stroke, and thus many of the patients ineligible for intravenous thrombolysis may have been good candidates for endovascular therapy had the study been performed more recently. Lastly, greater identification of perioperative strokes could have the unintended consequences of increasing the reported complication rates for surgical procedures at a given institution. However, this is vastly outweighed by the potential benefits of identifying and treating acute strokes. Despite these limitations, our study demonstrates that a brief educational intervention consisting of an in-service training conducted by members of the stroke team, combined with a formalized pathway for managing acute in-hospital stroke, can be successful at improving the appropriate assessment and access to care (for which there is class I evidence) for inpatient strokes.
Future improvements should focus on adapting this intervention to more directly target medical and surgical residents who spend time on the wards. Further emphasis of the urgency of an acute stroke is also needed because some healthcare workers continue to consider stroke a lesser emergency compared with other hospital codes. In addition, it would be important to have basic education around concepts of risk management and decision making.19,20 Sustaining the improvement over the long term will require continued commitment to educating new staff members and commitment from ward managers to emphasize the importance of the inpatient code stroke algorithm. Other targets to further improve time to CT scanning would include specific protocols for transportation of patients to the scanner from the ward.
The timely assessment and imaging of inpatients with suspected strokes is essential to provide parity in the care of in-hospital and out-of-hospital stroke patients. An inpatient code stroke algorithm and quality improvement educational strategy aimed at overcoming knowledge-to-action gaps can be successful in reducing most relevant initial times in acute stroke care (LSN to assessment and LSN to brain imaging), which have been associated with improved clinical outcomes.18
We appreciate the participation of nurses, allied health members, and physicians involved in stroke care.
Sources of Funding
This project was founded by the Ontario Stroke Network and the Heart and Stroke Foundation of Canada after an open peer-reviewed competition. Dr Saposnik is supported by the Distinguished Clinician Scientist and Mid-Career Awards from Heart and Stroke Foundation of Canada after an open peer-reviewed competition.
Guest Editor for this article was Eric Smith, MD, MPH.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.017622/-/DC1.
- Received February 20, 2017.
- Revision received May 9, 2017.
- Accepted May 23, 2017.
- © 2017 American Heart Association, Inc.
- Kelley RE,
- Kovacs AG
- Cumbler E,
- Murphy P,
- Jones WJ,
- Wald HL,
- Kutner JS,
- Smith DB
- Saltman AP,
- Silver FL,
- Fang J,
- Stamplecoski M,
- Kapral MK
- Alberts MJ,
- Brass LM,
- Perry A,
- Webb D,
- Dawson DV
- Alberts MJ,
- Perry A,
- Dawson DV,
- Bertels C
- Nolan S,
- Naylor G,
- Burns M
- Kapral MK,
- Fang J,
- Silver FL,
- Hall R,
- Stamplecoski M,
- O’Callaghan C,
- et al
- Aly N,
- McDonald K,
- Leathley M,
- Sharma A,
- Watkins C
- Sposato LA,
- Esnaola MM,
- Zamora R,
- Zurrú MC,
- Fustinoni O,
- Saposnik G
- Cumbler E,
- Wald H,
- Bhatt DL,
- Cox M,
- Xian Y,
- Reeves M,
- et al
- Jauch EC,
- Saver JL,
- Adams HP,
- Bruno A,
- Connors JJB,
- Demaerschalk BM,
- et al
- Saposnik G,
- Redelmeier D,
- Ruff CC,
- Tobler PN
- Saposnik G,
- Johnston SC