Prediction of in-hospital cardiopulmonary arrest outcome.

Clinical and demographic characteristics of 122 patients undergoing cardiopulmonary resuscitation were retrospectively collected to develop a predictive model for immediate success of resuscitation (restoration of pulse and blood pressure). The project focused on objective measurement of parameters available before resuscitation was performed. Variables included age, diagnoses, objective severity of illness, laboratory data, and clinical course variables. A four-variable model was developed using logistic regression to predict resuscitation success immediately after resuscitation. The four predictive before arrest factors were age between 40 and 70 years, scheduled for surgery, location of arrest in an intensive care unit, and before arrest PO2 greater than 8 mm Hg. The model had an accuracy of 69%, sensitivity of 76%, and specificity of 61%.