The Anesthesia Patient Safety Foundation Stoelting Conference 2019: Perioperative Deterioration—Early Recognition, Rapid Intervention, and the End of Failure-to-Rescue

The Anesthesia Patient Safety Foundation (APSF) Annual Stoelting conference brings together diverse stakeholders to explore key perioperative care safety issues. The conference provides opportunities to collaborate and leverage expertise—catalyzing solutions to complex problems that continually harm patients. The 2019 theme: “preventing, detecting, and mitigating clinical deterioration in the perioperative period” heads the list of 12 perioperative patient safety priorities defined by APSF’s Board of Directors.1,2 Expert faculty (Supplemental Digital Content, Table 1, complete faculty list, https://links.lww.com/AA/D122) laid foundation for an interactive solution-oriented meeting, informed by deep understanding of the perioperative patient, caregiver, and clinician journeys. Personal narratives shared by Buist3 and Townsend eloquently described their own or loved ones’ suffering from failure to receive timely and appropriate medical care. These stories energized participants, giving a deep sense of the relevance and importance of the work we had gathered to do. The conference included presentation of fresh data, focused debates on controversial topics, and culminated in a human-centered design workshop to generate innovative, creative solutions. A total of 136 participants including patients, family members, clinicians, risk managers, industry representatives, and representatives from over 25 multiprofessional organizations took part in the September 4–5, 2019 meeting. LEARNING FROM THE ANESTHESIA QUALITY INSTITUTE DATA SOURCES Anesthesia Quality Institute (AQI) houses 2 important data sources: Anesthesia Closed Claims (ACC) and Anesthesia Incident ReportS (AIRS). Domino reviewed the ACC database for severe permanent injuries/death claims from 2005 to 2014, with a deterioration event occurring in postoperative phases (ie, postanesthesia care unit PACU >1 hour, ward, intensive care unit ICU). Failure-to-rescue (FTR) events occurred in 1 in 5 claims for severe permanent injuries or death. Compared to patients in other claims for severe outcomes, patients in FTR events were more likely to be American Society of Anesthesiologists (ASA) ≥III or have had orthopedic surgery. Damaging event FTR claims generally involved diagnosis and treatment of postoperative respiratory depression, cardiovascular, or respiratory issues due to patient comorbidities, postoperative hemorrhage, and spinal or epidural hematoma (Table). Domino found that communication and human factors4 (decision making, workload management, situational awareness, leadership, and teamwork) likely contributed to delayed diagnosis and treatment of these events. Table. - Failure-to-Rescue Compared to Other High-Severity Claims Injury Severity Score 6–9 (Permanent Disability to Death) Greater proportion ASA physical status III–V Greater proportion orthopedic surgery Types of adverse events in failure-to-rescue claims: • Postoperative respiratory depression • CV/respiratory issues related to comorbidities • Postoperative hemorrhage • Epidural/spinal hematoma Source: Anesthesia Closed Claims (2005–2014).Abbreviations: ASA, American Society of Anesthesiologists; CV, cardiovascular. Guffey reviewed the AIRS database of >1700 reports, where 63% of the FTR cases were felt to be preventable. FTR cases were categorized into (1) failure of recognition (eg, untimely recognition of hypotension), failure to monitor (eg, absence of blood pressure monitoring), delay in escalation (eg, untimely call for help), and making a definitive diagnosis (eg, diagnosing postoperative hemorrhage and definite return to surgery versus continued fluid resuscitation). Failure of recognition was the most frequent reason, followed by making a definitive diagnosis. Monitoring or delay in escalating issues were less frequent causes. Guffey noted that most AIRS-reported cases reflect occurrences in the operating room. He suggested escalation is more apt to be limited in the PACU—where situational awareness can be further compromised, and distractions more abundant than in the operating room. LEARNING FROM MODELS Dummett presented Kaiser Permanente’s implementation progress of an early warning system, which robustly scores integrated vital signs and laboratory values through their electronic record; develops predictive intelligence on the deteriorating patient; and uses a system of virtual rapid response team (RRT) nurses. Lessons learned over 5 years5 included the importance of being sensitive to alarm fatigue and integrating with existing workflows. Dummett emphasized that optimizing scores is not enough. It is critically important to create a culture of patient safety with a lens to rescue and recognize the possibility of diagnostic error. Blike shared Dartmouth-Hitchcock’s advancements with their Failure-to-Rescue Patient Safety Learning Laboratory focusing on metrics—distinguishing the difference in denominator between mortality rate (death/no. of surgeries) and FTR rate (death/no. patients with complications). He summarized work demonstrating hospital variations in surgical mortality are due far more to FTR rates than complication rates.6,7 Although the perfect metric is still elusive, Blike advocated actively sharing timely, outcome-based, complication-specific measures. He cautioned against the use of composite measures and measures insensitive to palliative-care cases. Buist, a key investigator in Australia’s early Medical Emergency Team (MET) studies,8 shared his personal story as patient, clinician, and researcher. Buist3 presented a provocative frame, coined “clinical futile cycles” (CFC), where a dangerous “spinning of wheels” occurs. Well-intentioned clinical activities—in keeping with traditional hierarchical referrals—fail to alter a deteriorating patient’s trajectory. He cited a cluster-randomized control trial9 where despite new MET activation procedures, rescue was not statistically improved. Buist asked that we emphasize the importance of microcultures within units. The CFC frame can educate both individuals and systems toward more timely solutions. FOCUSED DEBATES Who to Monitor and How? Peden argued that simply monitoring more does not make health care delivery better. She suggested that effective programs be developed by risk-stratifying patients, intermittently monitoring (frequency based on patient risk and use trends), and simultaneously utilizing early warning scoring systems.10–12 Devita countered with a vision of hospitals being high-reliability organizations and that not all patients will predictably deteriorate.13,14 He argued that intermittent monitoring results in sampling error, unreliability, and recognition errors and that recent monitoring systems (eg, electronic cardiac arrest risk triage e-CART score15) have evolved with better specificity and sensitivity. Augmented Intelligence: Hype or New Normal Mathur began his argument of augmented intelligence (AI) hype with Amara’s Law: “we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”16 He cautioned overgeneralization of AI’s capability and to be wary of machine learning from biased data sets.17 Buist18 countered that human fallibility cannot be ignored. AI has real-time potential to incorporate all patient data individualized to a patient with outcomes that are relevant to a particular clinical question. He shared data from an electronic alert system demonstrating significantly improved clinical rescue (eg, improved clinician attendance for warning scores above a threshold level from 29% to 78% and reduced hospital length of stay).19 Both agreed future collaboration between clinicians and machine-learning scientists with properly designed studies will determine the true role of AI in perioperative care. FTR: Afferent or Efferent Limb Issue? Hravnak argued that the afferent arm (all aspects of warning clinicians of failure) is the key issue. She shared a model of patient- and hospital-level factors that lead to FTR. Delayed calls aggregated into 2 categories: alarm fatigue and communication.20 She argued for improving specificity and sensitivity of alarms as essential for future success. Strengthening the afferent limb through education, resources, technology, and organizational structures are equally critical. Safavi defended the efferent limb (all aspects of responding to a warning) as key. As an analogy, he shared meta-analyses from beta-blocker therapy following myocardial infarction (MI) and tele-monitoring congestive heart failure (CHF),22 where the idealized technical solution (afferent limb) was insufficient to change outcomes. He conceded that, in contrast to technology inventions (eg, AI software platforms, biosensor wearables), focusing on organizational culture (eg, the sociotechnical relationships between alarming and responding teams)21 holds less glamour and rigor. Both debaters agreed that timely human response in both limbs is ultimately what is critically important.23 Human-Centered Design Workshop Sammann and team facilitated and led the human-centered design (HCD) workshop.24 They shared core principles of HCD as a methodology that empathizes with each stakeholder, understands their unique challenges and unmet needs, examines the problems from multiple perspectives, and designs creative solutions to meet the unmet needs of the stakeholders.25 Themes From Preconference Interviews As preconference work, qualitative research was conducted with diverse stakeholders to better understand FTR from their perspectives. Over 25 clinicians and patients across multiple specialties, clinical settings, and levels of experience were interviewed and observed. This work generated overarching themes, 9 key insights, and opportunities (Figure 1).Figure 1.: HCD failure-to-rescue insights and design challenges. Insights were obtained through qualitative interviews. Design challenges emerged from these insights to focus the design-thinking process toward solution-oriented prototypes. This list is in rank order of priority voting by conference participants. HCD indicates human-centered design.“There is no good time to raise the alarm” was one such insight. Interviewees discussed experiences of escalating care too early where they were perceived as overreacting, or instances of escalating care too late when the patient was already deteriorating. “You have to make sure your ducks are in a row, otherwise people will stop listening to you,” a PACU nurse shared. An intensivist recounted that “sinking gut feeling” when realizing that a patient could no longer be “rescued.” These insights exposed design opportunities for conference participants to brainstorm ideas on how we might build environments that support and encourage early escalation of care. In total, conference participants brainstormed 414 independent ideas, clustered into 117 themes. Ultimately 19 prototypes emerged (Figure 2), falling into 4 domains:Figure 2.: Example of insight to prototype progression. An example of a single insight/design challenge and the divergent and convergent process of the design-thinking process. Following divergent brainstorming, a workgroup team would select a theme and develop a prototype. EHR indicates electronic health record. Technology: improving data integration, user-interactive interfaces; exploring how technology contributes to workflow. Culture/recognition: deconstructing hierarchy, nurturing respectful behaviors, enabling cultures valuing all voices equally. Training: developing habits of reflection, flexibility, communication, and empathy System workflow: bringing wisdom/optimal decision-making to the bedside by anticipating and planning for deterioration. SUMMARY The Stoelting Conference 2019 convened multidisciplinary, professional experts to explore systems for patient rescue. Through a hybrid conference utilizing HCD, discussions and innovations anchored in the needs of clinicians and patients. It is essential that we build environments that support and encourage early escalation of care. We recommend that those designing solutions keep the 9 insights and challenge opportunities in mind. These solutions should enable 4 overlapping domains: technology, culture, training, and system/workflow. APSF will continue encouraging iterations of conference prototypes. We will foster dialog to deepen our understanding of the system dynamics, moving us reliably toward eliminating “Failure-to-Rescue.” ACKNOWLEDGMENTS We deeply thank Ben Alpers, George Blike, David Gaba, Devika Patel, Lynn Reede, Patty Reilly, Matthew Weinger, Bradley Winters, Stacey Maxwell, and Mark Warner for being essential members of the Stoelting Conference 2019 Steering Committee. This submission would not be possible without them. DISCLOSURES Name: Della M. Lin, MS, MD. Contribution: This author was critically involved with planning and executing the conference and drafted and revised the manuscript. Conflicts of Interest: None. Name: Carol J. Peden, MB ChB, MD, MPH. Contribution: This author helped with planning and executing the conference and with concept, drafting, and revision of the manuscript. Conflicts of Interest: None. Name: Simone M. Langness, MD. Contribution: This author helped with planning and executing the conference and with concept, drafting, and revision of the manuscript. Conflicts of Interest: None. Name: Amanda Sammann, MD, MPH. Contribution: This author was critically involved with planning and executing the conference and with drafting and revision of the manuscript. Conflicts of Interest: None. Name: Steven B. Greenberg, MD. Contribution: This author helped in planning and executing the conference and with critical revisions of the manuscript. Conflicts of Interest: S. B. Greenberg is the APSF Newsletter Editor-in-Chief and an Associate Editor, Safety Section, Anesthesia & Analgesia. Name: Meghan B. Lane-Fall, MD, MSHP. Contribution: This author helped in planning and executing the conference and with revisions of the manuscript. Conflicts of Interest: None. Name: Jeffrey B. Cooper, PhD. Contribution: This author helped in planning and executing the conference and with concept, drafting, and revision of the manuscript. Conflicts of Interest: None. This manuscript was handled by: Richard C. Prielipp, MD, MBA. Della M. Lin, MS, MDDepartment of SurgeryJohn A. Burns School of MedicineHonolulu, Hawaiiemail protected Carol J. Peden, MB ChB, MD, MPHDepartment of AnesthesiologyKeck School of Medicine at the University of Southern CaliforniaLos Angeles, California Simone M. Langness, MDAmanda Sammann, MD, MPHDepartment of SurgeryThe Better LabSan Francisco, California Steven B. Greenberg, MDDepartment of Anesthesiology Critical Care and Pain MedicineNorthShore University HealthSystemEvanston, IllinoisDepartment of Anesthesiology and Critical CareUniversity of Chicago, Pritzker School of MedicineChicago, Illinois Meghan B. Lane-Fall, MD, MSHPDepartment of Anesthesiology and Critical CarePerelman School of MedicineUniversity of PennsylvaniaPhiladelphia, Pennsylvania Jeffrey B. Cooper, PhDDepartment of Anesthesia, Critical Care & Pain MedicineMassachusetts General HospitalBoston, Massachusetts