Accuracy and reliability of extubation decisions by intensivists.

BACKGROUND Mortality increases when extubations fail. Although predictors of extubation failure have been evaluated, physicians' reasoning to extubate a patient has received minimal attention. We hypothesized that the accuracy and reliability of physicians' extubation decisions are low. METHODS We sent surveys to 55 physicians in the divisions of pulmonary and critical care medicine of 3 teaching hospitals in Chicago, Illinois. The survey comprised 32 clinical vignettes of real patients who were extubated after they tolerated a spontaneous breathing trial (16 failed extubations). Unaware of the outcomes of extubation, the physicians were asked if they would extubate each patient, and to give reasons if they opted not to. We quantified the agreement between and accuracy of the physicians' decisions, determined the patient characteristics that influence the extubation decision, and described the tradeoffs leading to that decision. RESULTS Completed surveys were obtained from 45 physicians (82%). The physicians postponed extubation in 37% of the cases. Agreement between any 2 physicians was fair (mean ± SD phi 0.34 ± 0.15) and was highest between attending physicians from the same institution (0.37 ± 0.15). In deciding to extubate a patient, 33% of the physicians relied on the breathing pattern on pressure support ventilation, 49% relied on the acid-base status, 13% relied on the mental status, and 8% relied on the amount of secretions. The accuracy of the physicians' extubation decisions was low (area under the receiver operating characteristic curve 0.35). The sensitivity of the physicians identifying the patients who were successfully extubated was 57%, and the specificity was 31%. A model that comprises the same variables that influenced the physicians was more accurate in predicting extubation outcome (area under the receiver operating characteristic curve 0.88). CONCLUSIONS For a decision made on an almost daily basis in intensive care units, physicians' extubation decisions are inaccurate and only fairly reliable.