L. E. Kuo1, G. C. Karakousis1, K. D. Simmons1, D. N. Holena1, R. R. Kelz1 1Hospital Of The University Of Pennsylvania,Department Of Surgery,Philadelphia, PA, USA
Introduction: Surgeons struggle to counsel families on the role of surgery and the likelihood of survival in the moribund patient. A recent study demonstrated a nearly 50% 30-day survival rate for the moribund surgical patient, but without information on what factors are associated with survival, it is difficult to provide patients and their family members with information on the best course of action for a specific patient. We sought to develop a risk prediction model for postoperative inpatient death for the moribund surgical candidate.
Methods: Using ACS NSQIP data from 2007-2012, we identified ASA class 5 (moribund) patients who underwent an operation by a general surgeon. The sample was randomly divided into development and validation cohorts. In the development cohort, patient characteristics which could be readily discerned on preoperative evaluation were evaluated for inclusion in the predictive model. The primary outcome measure was in-hospital mortality, and factors found to be significant in univariate logistic regression were entered into a multivariable model. Points were assigned to these factors based on beta coefficients. This model was used to generate a simple scoring system to predict inpatient mortality. Models were developed separately for operations performed within 24 hours of admission and operations performed at least one day after admission as a means of differentiating between patients who presented to the hospital in the moribund state, and those whose condition reflected deterioration over their hospital course. Each model was tested on the validation cohort.
Results: 3,130 patients were included in the study. In-hospital mortality was 50.5% in the overall sample. In multivariable regression modeling, patient characteristics associated with in-hospital mortality were age, functional status (odds ratio 2.11, confidence interval 1.39-3.19), dialysis within the previous 30 days (1.63, 1.22-2.32), recent myocardial infarction (1.52, 1.04-2.22), and ventilator dependence (2.17, 1.43-3.30). For patients undergoing surgery within 24 hours of admission, body mass index was also associated with inpatient death. The scoring system generated from this model accurately predicted in-hospital mortality in both the development and validation cohorts for patients undergoing surgery within and after 24 hours (Table 1).
Conclusion: A simple risk prediction model using readily available preoperative patient characteristics can be used to accurately predict postoperative mortality in the moribund patient undergoing surgery. This scoring system can easily be applied in the clinical setting to assist in counseling and decision-making.
