Given the large-scale deployment of Electronic Health Records (EHR), secondary use of EHR data will be increasingly needed in all kinds of health services or clinical research. This paper reports some data quality issues we encountered in a survival analysis of pancreatic cancer patients. Using the clinical data warehouse at Columbia University Medical Center in the City of New York, we mined EHR data elements collected between 1999 and 2009 for a cohort of pancreatic cancer patients. Of the 3068 patients who had ICD-9-CM diagnoses for pancreatic cancer, only 1589 had corresponding disease documentation in pathology reports. Incompleteness was the leading data quality issue; many study variables had missing values to various degrees. Inaccuracy and inconsistency were the next common problems. In this paper, we present the manifestations of these data quality issues and discuss some strategies for using emerging informatics technologies to solve these problems.