Acute pancreatitis has been known as a distinct clinical entity since the end of the 19th century, but its management is still characterised by an absence of specific treatments, despite extensive clinical and experimental research. In part this has stemmed from the limitations of the classifications for acute pancreatitis severity used so far. It may well be that the perpetual use of classifications which were solely based on empirical description of severity has retarded progress in the field due to systematic misclassification in selecting patients for clinical trials, evaluating the effects of treatment, and comparing the inter-institutional data. This article considers several important questions with regard to classifying the severity of acute pancreatitis. These relate to how the classification is best used, what determines the severity of acute pancreatitis, how determinants relate to each other and the time course of determinants. Further, the concept of causal inference is applied to classifying the severity of acute pancreatitis, and the best available evidence suggests that classification of acute pancreatitis severity should be based on local and systemic determinants. It is anticipated that the determinants-based classification of acute pancreatitis severity will underpin the ability of future research to discover a specific treatment for acute pancreatitis and make routine clinical management of patients with acute pancreatitis more tailored and evidence-based.