New American College of Surgeons risk calculator determines colorectal su...(University in Chicago. Using the AC...),Health

Using the ACS National Surgical Quality Improvement Program (NSQIP) database, data on 28,863 patients who underwent colorectal operations at 182 hospitals in 2006 to 2007 were analyzed to generate three logistic prediction models for 30-day morbidity, serious morbidity and mortality. The models considered more than 30 predictive variables such as patient age, gender, extent of disease, body mass index, shortness of breath and comorbidities such as chronic obstructive pulmonary disorder (COPD), high blood pressure, pneumonia, cardiovascular or neurologic diseases, diabetes and cancer. Outcomes were assessed at 30 days, regardless of whether the patient was discharged, remained hospitalized or was admitted to a different institution. The models were validated against 2005 data from 3,037 colorectal operations conducted at 37 hospitals, and similar model discrimination was shown. Results for these three models were used to construct a universal multivariable model to predict risk for all three outcomes.

Application of the variable selection process for the universal model yielded 13 variables that appeared in models for all three outcomes and two variables that appeared for two outcomes. Odds ratios for variables selected in the universal model showed findings generally consistent with clinical expectations.

The 15 predictive variables selected for the universal model were age; body mass index; extent of disease; sepsis (bloodstream infection); functional health status; preoperative laboratory values of albumin, creatinine and partial thromboplastin time (a measure of blood clotting); indication for operation (for example, cancer or obstruction); disseminated cancer; surgical extent for example, partial or total removal of the colon); whether the operation was associated with an emergent condition; shortness of breath; COPD; and type of wound (for example, clean versus infected).

The results reported here wer
'/>"/>