C. L. Scaife1, K. C. Hewitt1, X. Sheng2, K. W. Russell1, M. C. Mone1 1University Of Utah,General Surgery / Surgery,Salt Lake City, UT, USA 2University Of Utah,Biostatistics / Pediatrics,Salt Lake City, UTAH, USA
Introduction:
Patients with an advanced cancer diagnosis, who develop an intra-abdominal surgical emergency, pose a medical decision making dilemma. While patient survival is already limited due to advanced malignancy, a surgical intervention may be futile and costly. Patients, families, and medical providers need more accurate data to assist with these difficult treatment decisions. This study was designed to establish morbidity and mortality rates, and to attempt to identify preoperative risk factors which may predict a poorer outcome.
Methods:
The national NSQIP database was queried for patients with disseminated cancer undergoing emergent abdominal surgery (2005-2012). Preoperative NSQIP variables were used for prediction models for 30-day major morbidity and mortality. Example NSQIP variables that were used included patient functional status, ASA class, weight loss, renal function, age, albumin, sepsis, HCT, and pre-op cardiopulmonary comorbidities. A tree model and logistic regression were employed to find factors associated with these outcomes. The analysis was carried out on a training dataset; model performance (misclassification rate) was then evaluated on a validation dataset.
Results:
In patients with an abdominal surgical emergency and disseminated or Stage IV cancer, there was overall a major surgical morbidity rate of 47% and a surgical mortality rate of 26%. The tree model for morbidity showed that sepsis or a hematocrit <29.6 was predictive for a major morbidity (error rate 36%). The tree model for mortality showed an ASA score of 4 or 5 with a totally dependent functional status to be predictive of mortality (error rate 24%).
Conclusion:
The decision to operate for an intra-abdominal emergency in the setting of disseminated cancer is difficult when considering the morbidity to the patient, the overall expected cancer specific and post-operative expected survival, and cost. Our study confirms that the risk for surgical morbidity and peri-operative death in this population is very high. Preoperative patient factors, including sepsis, ASA class, anemia, and patient functional dependence, all strongly predict poor patient outcomes. We have also developed further logistic regression models, derived from this database, to provide detail to help with decisions related to care.
