Back to Annual Meeting Posters

BACKGROUND: The aim of this study was to develop a preoperative normogram to predict thedischarge status following pancreatic resection for malignancy.
METHODS: The Nationwide Inpatient Sample (NIS) database (2000-2009) was used. Discharge disposition was categorized as routine (home without any assistance) or non-routine (discharge to home with assistance or to skilled facility). Multivariate logistic regression model was used to identify variables influencing discharge disposition and a normogram was created. The training set (2000-2005) was used to develop the model which was further validated using the validation set (2006-2009).
RESULTS A weighted total of 21250 patients (2000-2005) were used to create a predictive model and 20390 patients (2006-2009) were used to validate it. The mean total points for the 2000-2005 dataset was 134.7 (SE=1.25), which correspond to approximately a 42% non-routine discharge which is similar to the actual observed non-routine discharge rate of 43%. The normogram was validated using the NIS 2006-2009 dataset. The mean total points for the 2006-2009 sample is 128.5 (SE=1.70) with an observed non-routine discharge rate of 46%. The concordance index was found to be 0.67 (95% confidence interval of 0.65 to 069). Calibration plots of the normogram revealed agreement between the observed non-routine discharge probabilities versus the model predicted non-routine discharge probability.(Figure: 1)
CONCLUSION This preoperative normogram may accurately predict the chance of a non routine discharge following pancreatic resection for malignancy and may be used as an adjunctive clinical tool in the preoperative counseling of these patients.

Calibration plots for Normogram dataset (2000-2005) and Validation Dataset (2006-2009).