Acute pancreatitis is an increasingly common cause of hospital admission in the United States. The aim of this study was to evaluate independent predictors of mortality in adult patients with a diagnosis of acute pancreatitis and to assess the relationship of hospital length of stay (HLOS) with in-hospital mortality.
Adult (18–64 years) patients with acute pancreatitis as the primary diagnosis who required emergency admission were analyzed using the National Inpatient Sample (NIS) database from 2005–2014. A nonlinear relationship between HLOS and mortality was assessed using a multivariable generalized additive model (GAM). A multivariable logistic regression model was performed on the ascending part of the plot to assess the odds of mortality.
There were 344,120 patients studied. The mean (standard deviation) age of patients was 44.6 (11.55) years and 51% were females. The results are presented as EDF (effective degree of freedom), which is a summary statistic of GAM and reflects the degree of nonlinearity of a curve. Mean HLOS was 7 days. HLOS had a V-shaped relationship with mortality (EDF = 9.26, p < 0.001). The nadir of the V curve was at an HLOS of 3–6 days, with an increase in mortality for both a longer and a shorter stay. An HLOS longer than 6 days began to have a linear relationship with mortality. The associated logistic regression model had an odds ratio of 1.127 (95% confidence interval: 1.106–1.148) for HLOS. Age, sex, income quartile, and several comorbidities significantly increased in-hospital mortality.
Mortality for adults with acute pancreatitis was lowest for an HLOS of 3–6 days. After 6 days, there was a linear relationship with morality, with each day increasing the mortality by 12.7%.