Abstract 20047: Nasa Model of "Threat and Error" in Paediatric Cardiac Surgery: Death Typically Results From Cycles of Error That Originate in the Operating Room and Are Amplified by Additional Error in Intensive Care

Introduction: We introduced the NASA “threat and error model” to our surgical unit; all admissions are considered “flights”, which should pass through stepwise de-escalations in risk. Hypothesis: Errors significantly influence risk de-escalation and contribute to poor outcomes. Methods: Patient flights (524) were tracked real-time for threats, errors and unintended states (figure). Expected risk de-escalation was: wean from mechanical support, sternal closure, extubation, ICU discharge and discharge home. Data were accrued via performance personnel, bedside data, reporting mechanisms and staff interviews. Infographics of flights were openly discussed weekly. Results: In 12% (64/524) of flights, the child failed to de-escalate sequentially through expected risk levels; unintended increments instead occurred. Failed de-escalations were highly associated with errors (426; 257 flights), however seemingly benign (P<.0001). Errors with clinical consequence (263; 173 flights) had 29% rate of failed de-escalations vs 4% (P<.0001). The most dangerous errors were “apical” errors typically (84%) occurring in the OR which led to cycles of propagating unintended states (n=110): these had 43% (47/110) rate of failed de-escalation (vs 4%, P<.0001). Apical errors were triggered by identifiable threats in 25% (28/110) (usually - 75% - morphology/ comorbidities); 75% were instead “unforced” errors. Cycles of unintended state were often (46%) amplified by additional (up to 7) errors in ICU that would worsen clinical deviation. Overall, failed de-escalations in risk were extremely closely linked to brain injury (N=13; P<.0001), or death (N=7; P<.0001). Conclusions: Deaths and brain injury almost always occur from propagating error cycles that originate in the OR and are often amplified by additional ICU errors. Improvements in threat management, error detection/rescue and vigilance at times of failed de-escalation will translate into improved outcomes.