Predicting adverse outcomes in a diagnosis-based protocol system for rapid sequence intubation

Our ED at Louisiana State University developed a unique approach to airway management by having four diagnosis-based protocols for rapid sequence intubation (RSI). This study examines protocol use and outcome from RSI in an academic ED. The study objective was to identify variables that are predictive of adverse outcomes in patients requiring RSI. This was a 4-year prospective, observational, data-gathering study of all intubations in an academic ED setting with >250,000 patient visits per year. Four protocols were established for 1) children <10 years of age, 2) adults with increased intracranial pressure, 3) adults with chronic obstructive pulmonary disease/asthma, and 4) other adults not fitting B or C. A special continuing quality improvement (CQI) committee was established to examine each case of RSI. Prospective data were collected, including age, race, gender, protocol, diagnostic group, intubation indication, and preintubation oxygen saturation. Diagnostic group was categorized as medical, blunt trauma, or penetrating trauma. Adverse outcome was defined as any case with hemodynamic changes, those requiring surgical or bronchoscopic intervention, and those requiring more than three attempts at intubation. Data were analyzed using univariate analysis, logistic regression, and a binomial regression tree analysis with SPSS 9.0 (Chicago, IL) and Answer Tree (SPSS). A total of 1,320 consecutive intubated patients were included. Protocol A was used in 4%, B in 43%, C in 15%, and D in 38%. Significant differences in number of cases with adverse outcome were based on protocol (P =.03) and final diagnosis (P <.03). Protocol C was less likely to be associated with adverse outcome than protocol D (odds ratio [OR] = 0.2, 95% confidence interval [CI] = 0.1-0.7). Penetrating trauma was more likely to be associated with adverse outcome (OR = 1.8, 95%, CI = 1.1-3.2) than blunt trauma. A regression tree analysis yielded the following, all cases using protocol A or C or medical cases using B had an adverse event in 11 of 458 (2.4%), whereas nonmedical cases using protocols B or D and medical cases using D had adverse outcomes in 73 of 862 cases (8.5%). The decision rules lead to a better classification of cases with adverse outcomes (2.4 vs 8.5%, of = 6.1%, 95% CI = 3.7-8.4). Adult trauma patients who fit the protocols B or D or adult medical patients who fit protocol B were at higher risk for adverse outcomes with RSI. This could alert the physician to a population at higher risk for adverse outcomes. Variables available in a diagnosis-based protocol RSI system can be used to predict adverse outcome among patients requiring RSI.