Pancreatic cancer symptom trajectories from Danish registry data and free text in electronic health records

Abstract

AbstractPancreatic cancer is one of the deadliest cancer types with poor treatment options. Better detection of early symptoms and relevant disease correlations could improve pancreatic cancer prognosis. In this retrospective study, we used symptom and disease codes (ICD-10) from the Danish National Patient Registry (NPR) encompassing 8.1 million patients from 1977 to 2018, of whom 22,727 were diagnosed with pancreatic cancer. To complement and compare these diagnosis codes with deeper clinical data, we used a text mining approach to extract symptoms from free text clinical notes in electronic health records (4,418 pancreatic cancer patients and 44,180 controls). We used both data sources to generate and compare symptom disease trajectories to uncover temporal patterns of symptoms prior to pancreatic cancer diagnosis for the same patients. We show that the text mining of the clinical notes was able to capture richer statistically significant symptom patterns, in particular general pain, abdominal pain, and liver-related conditions. We also detected haemorrhages (p-value =4.80·10-08) and headache (p-value =2.12·10-06) to be linked as early symptoms of pancreatic cancer. Chaining symptoms together in trajectories identified patients with jaundice conditions having higher median survival (>90 days) compared to patients following trajectories that included haemorrhage, oedema or anaemia (≤90 days). Additionally, we discovered a group of cardiovascular patients that developed pancreatic cancer with a lower median survival (≤90 days). These results provide an overview of two types of pancreatic cancer symptom trajectories. The two approaches and data types complement each other to provide a fuller picture of the early risk factors for pancreatic cancer.