I. An Interdisciplinary Initiative to Reduce Unplanned Extubation in the Pediatric Critical Care Units at CHCO
Project Leader(s): Jonathan Kaufman, MD; Eduardo Da Cruz, MD, Emily Dobyns, MD, Chris Peyton, MS,CPNP; Mike Rannie RN,MS; Beth Wathen, MSN, CCRN, Jerrold Judd RRT, Matt Vitaska ND,RN, and Michael Kahn MD, PhD
Rationale: Unplanned extubations in PICUs and pediatric cardiac ICUs (CICU) can result in increased mortality, morbidity, and length of stay. We sought to reduce the incidence of these events by reliably measuring occurrences and instituting a series of coordinated interdisciplinary interventions.
Relationship to external efforts: While this was a project developed locally at CHCO, as opposed to as part of a national improvement effort, there is within the critical care literature, many examples of initiatives directed at reducing this unintended occurrence of preventable harm. Initiatives to reduce these events are increasingly common in ICUs. Guidelines and recommendations have recently been published for reducing unplanned extubations in pediatric critical care units.
Goals and Objectives
Aim: The project goal was to significantly reduce (by ~25-50% from baseline) the frequency of unplanned extubation for children with endotracheal tubes within the cardiac and pediatric intensive care units from a baseline rate of 0.8 per 100 ventilator days over a two-year period.
Population: Intubated patients of any age in the PICU and CICU at Children's Hospital Colorado (CHCO).
IOM Quality Dimensions: Safety, effectiveness, efficiency, patient/family centered
Primary Outcome Measures: the rate for unplanned extubation in this patient population. The numerator was the number of unplanned extubations; the denominator was ventilator days. Other data collected in the root cause analysis of every event included time of day, patient age, patient diagnosis, and disposition (e.g. re-intubated or remained extubated), and, if re-intubated for unplanned extubation, how many experienced CPR peri-re-intubation.
Calculation of Measure: Calculation Rate = # of unplanned extubations/# of ventilator days expressed as a rate per 100 ventilator days for each critical care unit. Average rates for each of the 3 periods (baseline, intervention, and post-intervention) were calculated by dividing the sum of all unplanned extubation events by the total patient ventilator days. Differences in rates across all 3 intervals were examined by using the nonparametric Kruskal-Wallis 1-way analysis of variance.
Target for Measure: A goal of significant reduction was established by the project team, initially estimated to be 25-50%.
Sampling Strategy: All patients were included in the data collection.
Data Submission over course of QI effort: Occurrence rate data was reported on a monthly basis to project team members and leadership committees. In addition, a white board was placed in the ICUs publicly showing “Days Since last unplanned extubations” as a strategy for building will and engaging staff and families in this improvement activity.
Improvement and Feedback: The data collected and experience in these two ICUs was reported within the Critical Care practice council and the work subsequently spread to the Neonatal ICU.
Intervention: First, all patients admitted to either ICU from the operating room or an outside facility, including emergency departments, underwent re-taping of the endotracheal tubes in a uniform manner. Second, in the CICU, handoffs from the cardiovascular operating room were standardized and included discussing the endotracheal tube position by the anesthesiologist and visibly recording it at the bedside. Third, all unplanned extubations were reviewed by a short root cause analysis that specifically addressed the causal and contributing factors surrounding the event to identify improvement opportunities. Fourth, the level of sedation was specifically addressed during daily rounds as it related to extubation readiness. Fifth, both units began publically displaying a large white-board that recorded “days since the last adverse event.”
The results of the project are demonstrated on the below u charts (Figure 4) showing reduction in mean occurrence rates in both ICUs during the three project periods; the reduction in rates reached statistical significance in the CICU.
Figure 1 & 2: Statistical Process Control chart of monthly unplanned extubation rates for CICU (top) and PICU (bottom). Average rates are calculated separately for baseline, intervention, and post-intervention periods.
Improvement: Reducing unplanned extubation in critically ill infants is important because these events may result in increased mortality and morbidity in already vulnerable infants and children. Severe hypoxia leading to end organ injury, aspiration, and airway trauma are among the serious adverse consequences of emergent unplanned re-intubation. Patients experiencing an unplanned extubation may be at higher risk for pneumonia, prolonged ventilator requirement, and increased length of stay.
Resources for Implementation of Intervention: Team members were trained in the PDSA method and utilized it to plan tests of change within the intervention period.
Sustainability of Improvement: The CICU has sustained a mean rate of 0.2 unplanned events through the end of 2011 and into early 2012. The PICU has sustained a mean rate of 0.4 unplanned events through this time period. The success of this quality improvement initiative required the leadership of identified unit specific and discipline specific champions. These individuals were the key drivers of change during the project and maintain leadership roles, but the sustainability of the effort depends upon a cultural change in the unit that enables continuation of the practices that resulted in the demonstrated improvements. The units have been able to maintain these process improvements including having ongoing, accurate data collection and reporting, consistent taping and re-taping practices, structured handoffs between the OR and CICU and reviews of occurrences. The team has been able to maintain its rates and address variation in practice that may occur from time to time.
II. Reducing Codes Outside the ICU
Project Leader(s): Core Code Reduction Team-Emily Dobyns, MD, Jenny Reese, MD, Jennifer Roth, RN, BSN, Chris Peyton, MS,CPNP; Beth Wathen, MSN, Patrick Guffey, MD, Daniel Hyman MD, Jodi Thrasher RN MSN. (Other teams included an SBAR Communication Training Taskforce, an RN Resident Council Initiative and the Code Simulation Taskforce).
Rationale: What is the gap in quality that the QI effort is to address? Why is this QI effort being undertaken? Compare the current state of care within your organization relative to this QI effort with the state of care in other settings.
In-hospital survival for patients who experience a cardiopulmonary arrest outside the ICU is less than 25%. Many of these codes, upon review, are preventable in that the child has demonstrated signs of clinical deterioration in the 6-12 hours prior to the arrest. We sought to reduce the incidence of these events through a series of planned and measured educational and system interventions.
Relationship to external efforts: The initial work done on code reduction in 2008 was as part of a national collaborative to reduce codes sponsored by Child Health Corporation of America (CHCA.) While the work at CHCO continued following the formal collaborative, the core project team at CHCO led an ongoing virtual national collaboration across CHCA that continued to focus on code reduction, and defined the national measurement strategy to permit benchmarking for the ~40 participating hospitals. The code measure is a part of a larger benchmarking effort called Whole System Measures (WSM).
Goals and Objectives
Aim: Following initial reductions in codes on the medical unit alone, in 2011 the hospital set an organization wide “pillar goal” to reduce codes outside the ICU by 25% within one year.
Population: Inpatients of any age admitted to the medical-surgical units. (ICU and surgical patients were excluded from the denominator.)
IOM Quality Dimensions: Safety, effectiveness, efficiency, and patient/family centered.
Primary Outcome Measures: The rate of non-ICU codes in this patient population. The numerator was the number of codes; the denominator was inpatient days outside the ICU. Other data collected related to specific interventions, e.g. % of nurses and physicians trained in SBAR and CUS communication as well as measures related to changes in the reasons for codes that did occur, all of which were reviewed to identify opportunities for improvement (e.g. codes with communication failure, failure to diagnose, etc.)
Calculation of Measure: Calculation Rate = # of codes/# of adjusted days expressed as a rate per 1000 adjusted patient days for the hospital.
Target for Measure: A goal of a 25% reduction was established by the project team as a pillar goal in 2011.
Sampling Strategy: All patients were included in the data collection.
Data submission over course of QI effort: Occurrence rate data was reported on a monthly basis to project team members, a code committee that was formed, to leadership committees, to the hospital generally with all other pillar goals. It was also ultimately reported on the hospital’s external website.
Improvement and Feedback: The data collected and more importantly, the analyses of the codes that occurred were shared at a multidisciplinary code committee each month, and periodically at quality leadership committee reviews of the project.
Intervention: We implemented a range of interventions as part of this effort over the past few years, including:
-Implementation of PEWS (Pediatric Early Warning Scores) with vital signs for all inpatients in order to detect clinical deterioration earlier.
-We also implemented the use of PEWS at the time of admission from the ED or network of care/ambulatory sites in order to reduce the likelihood of placing a child on the inpatient unit who would be better served in the ICU.
-Training of all clinical staff in the use of structured communication tools, SBAR (Situation/Background/Assessment/Recommendation) and CUS (an assertion model utilizing words to more effectively communicate risk- Concerned/Uncomfortable/Safety problem).
-Review of all codes to identify trends and learning opportunities that were shared in a regular publication called “Code Comments”.
-Implementation of “radar rounds”- twice daily discussions between supervising residents and nurses to identify patients of concern to either group.
-Regular review of cases with codes at resident M and M conferences to promote earlier use of the RRT program.
-Family activation of RRT was instituted in late 2008
-Changes in how oncology patients with fever and neutropenia were managed so as to assure hemodynamic stability prior to determination of placement upon admission. Similarly, patients being loaded with antiepileptic medications are now observed for longer periods prior to transfer out of the Emergency Department.
-Establishment of a council for nurses and residents to identify and pursue opportunities for better communication and teamwork in the care of shared patients
Establishing code reduction as an organizational pillar goal was an important strategy in that it created tremendous visibility as well as promoted engagement in the project by all clinical staff and hospital leadership teams
The results of the project are demonstrated in the statistical process control chart showing reduction in mean occurrence rates in for codes over the past two years (Fig. 5). The chart demonstrates a 50% reduction in codes outside the ICUs in 2011. The code rate fell from 0.51 codes/1000 inpatient days to 0.24 codes/1000 inpatient days. In raw numbers, we estimate that approximately 15 fewer codes occurred in 2011 as compared with 2010. The project was nominated for and received the Medical Staff’s 2011 Excellence in Quality and Safety Award.
Figure 3. Statistical process control chart of monthly codes outside the ICU.
Improvement: Reducing the rate of codes outside the ICU is important because the risk of mortality in children who arrest outside the ICU is poor. In our review of prior codes we identified that a majority of them were preventable and would potentially reduce mortality if interventions and intensive care were instituted sooner in their course.
Resources for Implementation of Intervention: A multidisciplinary code team reviewed cases monthly. The codes database was created by the PICU advanced practice nurse and utilized for data tracking and case analysis. A specific team was created for the purpose of training staff in SBAR/CUS. Staff members (including attending physicians and residents) were provided with lanyard cards to remind them of the meanings of these acronyms and to encourage their use.
Sustainability of Improvement: Interventions were implemented over time and standard strategies are being used to assure their sustainability, including ongoing data reporting, regular case reviews, and monitoring of process changes including the use of PEWS, SBAR, and the RRT program.
Back to top
III. In-Hospital Stroke Response Quality Improvement Initiative University of Colorado Hospital
Project Leader(s): Ethan Cumbler, MD: Alex Graves, NP; Tracey Anderson, NP
Rationale: Between 35,000 and 75,000 patients suffer stroke during hospitalization. Similar to out-of-hospital ischemic stroke, strokes that occur to patients in the hospital have a limited time window for therapy and require prompt evaluation to allow for thrombolysis. The literature suggests greater delays in evaluation of strokes on hospital wards compared to those in the ED. American Stroke Association guidelines recommend that brain imaging be obtained within 25 minutes of symptom onset, yet this benchmark is rarely achieved for the in-hospital stroke. Analysis of a Michigan stroke registry found that only 3.1% of patients with in-hospital strokes met this goal and our own research on a Colorado state-wide stroke registry found time-to-evaluation to be more than twice the recommended benchmark. Data from another multi-center stroke registry in Spain showed that half of all thrombolysis-eligible, in-hospital stroke patients could not be treated due to delays in evaluation.
Relationship to external efforts: At conception there were no external relationships. However, after project success was seen a partnership with the National Stroke Association was commenced.
Goals and Objectives
Aim: The project goal was to reduce the time to inpatient stroke alert imaging to less than or equal to 25 minutes.
Population: Hospitalized patients at University of Colorado Hospital.
IOM Quality Dimensions: Safety, effectiveness, efficiency, patient/family centered.
Primary Outcome Measures: The time to stroke imaging after an inpatient stroke alert was called. Later measures included improving the in-hospital stroke alert criteria, that is improving the specificity of stroke alerts to reduce the waste created through responding to stroke mimics. Final measures included a better understanding of the processes in place at nationally successful stroke programs and implementing those processes at our institution.
Calculation of Measure: For the overall goal of time to stroke imaging the measure included the median time from in-hospital stroke alert to first slice of CT imaging.
Target for Measure: The goal was set to achieve all head CTs in in-hospital stroke in 25 minutes or less. We also used median time to head CT as a measure.
Sampling Strategy: All patients were included in the data collection.
Data submission over course of QI effort: Occurrence rate data was reported on a monthly basis to project team members and leadership committees.
Improvement and Feedback: The data collected was reported to the institution’s Stroke Council team at least monthly. The feedback was provided both orally, in written summary and graphically. The data was also shared with the institutional leadership.
Intervention: Implementation and Continuous Improvement of an In-hospital Stroke Response Team (2006-2012)
Ten in-hospital stroke alerts including seven ischemic strokes occurred in the year prior to initiating the inpatient stroke alert program in late 2006. .Administrative coding data followed by chart review identified six additional in-hospital ischemic strokes. Forty four in-hospital stroke alerts including twenty two ischemic strokes occurred in the year following the inpatient stroke alert program initiation (through Q4, 2007). Administrative coding data followed by chart review identified three additional in-hospital ischemic strokes.
The percentage of true ischemic strokes with stroke alert activation for inpatients increased from 54% to 88% after program implementation although the percentage of alerts for non-ischemic symptoms rose from 30% to 50%. Median time from symptom recognition to CT for in-hospital ischemic strokes decreased from 271 minutes (IQR 127, 363) in the year prior to the inpatient stroke program to 74 minutes in the year after (IQR 51, 178) (p=0.02).
In an effort to reduce the frequency of “false alarms,” we performed a focused review of the characteristics and symptoms of 64 true in-hospital strokes between January and September 2009. In-hospital stroke alerts comprised 64 of 402 (15.9%) of all stroke alerts during this study interval with the remaining 338 occurring in the ED. Of the 64 in-hospital stroke alerts, 22.7% had a documented ischemic stroke, 15.2% had a TIA, 15.2% had metabolic derangement, 19.7% were due to medication, and 27.3% were due to sepsis or delirium.
Through review we discovered that altered mental status was never the lone presentation for true stroke, resulting in a change in the stroke alert criteria to remove confusion as a sole criterion for alerts thereby improving the utilization of stroke team resources.
Improving hospital systems for processes such as stroke involves multiple departments and professionals interacting as teams for a time-critical event. In order to understand the features that characterize high performing hospitals and gain understanding of the methods and institutional culture needed to create and maintain excellence we conducted structured interviews with 12 best-performing hospitals in the American Heart Association’s Target Stroke program, which recognizes hospitals capable of treating ischemic stroke with thrombolysis in under an hour. We identified the following key features:
-Identification of strong stroke champions in multiple departments
-Empowerment of staff to find department-specific solutions to solve common challenges
-Setting of aggressive improvement targets and broad sharing of data to foster accountability and expectation of improvement
Over the past several years, we have continued to pursue improvements in the efficiency and overall quality of the stroke alert program. Four key changes were implemented to address the fact that in-hospital stroke alerts were taking more than twice as long as stroke alerts in the ED. Time from stroke alert to initiation of CT scan was prospectively tracked for the 6-month period prior to intervention (9/08 to 2/09). After a 3-month interval for intervention roll-out from March to May 2009, the response times for the pre-intervention period were compared to a 6-month post-intervention evaluation period June 2009-November 2009.
Four essential interventions that we believe resulted in improved efficiency were:
1. Process mapping and direct observation to identify unreliable and/or steps causing delays (e.g. IV access, appropriate lab evaluation, timely interdisciplinary communication)
2. System redesign with input from all stakeholders, resulting in several process changes (e.g. EKG after CT, not before; rapid transport by residents/nurses not transport staff; pagers for CT techs)
3. In-hospital Stroke Alert Checklist developed and disseminated on a laminated pocket card including key steps and expected timelines
4. Real-Time Feedback
During the intervention rollout and post-intervention periods, feedback was provided from the stroke program to the front line providers following each in-hospital stroke alert. The clinicians involved were notified of the final diagnosis and patient outcome and were provided with feedback about how the patient’s evaluation times compared with benchmark goals. The feedback process was designed to be bi-directional, with requests for input from staff on barriers experienced to rapid evaluation and suggestions for future process improvement.
As a result of these interventions, Pre-intervention median inpatient stroke alert-to-CT time was 69.0 minutes with 19% meeting goal of 25 minutes. Post-intervention median inpatient stroke alert-to-CT time was reduced to 29.5 minutes with 32% at goal (p<0.0001).
In the 12 months, extending to November 2010, following the conclusion of the formal study of the second cycle of interventions the median response time to in-hospital strokes continues to be maintained at 30 minutes.
Improvement: See above.
Resources for Implementation of Intervention: Team members were able to utilize the inpatient Stroke Council and their personnel resources as well as a modest (< $200) budget for needed supplies. Additionally, our residency training program within the internal medicine residency training program had developed a program for QI project work by housestaff. Many of these housestaff played a role and were an added resource.
Sustainability of Improvement: The project is an ongoing initiative that has the support of the institution, especially the inter-disciplinary Stroke Council. Additionally, new projects aimed at improving stroke care in the Emergency Department have commenced in the wake of these successes.
Back to top