Abstract Background/objectives Hyperlipasemia is frequently encountered in patients in the intensive care unit (ICU). The degree to which it should be valued in making the diagnosis of acute pancreatitis (AP) in critically ill patients remains uncertain. We sought to determine the diagnostic accuracy of hyperlipasemia and the optimal lipase cutoff for diagnosing AP in critically ill patients. Methods Four hundred and seventeen ICU patients with hyperlipasemia, defined as lipase greater than three times the upper limit of normal from 2009 to 2012 were retrospectively identified. A diagnosis of AP was confirmed by the additional presence of either characteristic abdominal pain or cross-sectional imaging. Results The overall positive predictive value (PPV) of hyperlipasemia was 38.1%. Median initial lipase levels were 1164 IU/L in patients with AP and 284.5 IU/L in patients without AP (p < 0.001). The optimal diagnostic lipase cutoff of 532 IU/L correlated with a sensitivity, specificity, negative predictive value and PPV of 77.4%, 78.0%, 84.9%, and 67.0% respectively. The most common primary diagnoses in non-AP patients with elevated lipase included shock, cardiac arrest and malignancy. Conclusions Physicians should maintain caution when interpreting hyperlipasemia in the critically ill due its relatively low PPV. However, a greater lipase cutoff improves its diagnostic value in AP and helps to reduce unnecessary imaging in these patients.