Back to 2011 Program

Objective: To compare a novel institutional database linked with a hospital information system (HIS), to a traditional prospective surgical database. Design: A newly established institutional data warehouse autopopulated by HIS was queried for 9 ICD-9-CM diagnosis codes for pancreatic neoplasms. Another query was performed using a prospective surgical oncology database maintained by manual entry. Duplicated patients and those unique to either dataset were identified. A subset of the identified HIS-warehouse patients was subjected to manual review to determine accuracy of diagnosis. Setting: Single-institution academic medical center. Patients: Patients with ICD-9-CM diagnosis code of pancreatic neoplasm. Interventions: Database query and medical record review. Main Outcome Measure: Pancreatic neoplasm patient capture by each database. Results: 1107 patients were identified from the HIS-linked dataset with pancreatic neoplasm from 1999-2009. Of these, 254 (22.9%) were also captured by surgical oncology, while 853 (77.1%) were only in the HIS-linked dataset. Manual review of a randomized 10% subset of the HIS-only group demonstrated 52.4% of patients were without identifiable pancreatic pathology, suggesting miscoding, while 34.9% had diagnoses consistent with pancreatic neoplasm, and 12.7% with other pancreatic pathology. Of 394 patients tracked by surgical oncology, 254 (64.5%) were captured in both datasets, while 140 (35.5%) were not captured in the HIS-linked dataset. Manual review of the non-captured patients demonstrated 85.9% with pancreatic neoplasm and 14.1% with other pancreatic pathology. Lastly, a 10% review of the 254 patient overlap demonstrated 87.5% of patients had pancreatic neoplasm and 12.5% with other pancreatic pathology. Conclusions: These results suggest cautious interpretation of national-level administrative data utilizing ICD-9-CM diagnosis codes. Our findings suggest that the current state-of-the-art data warehouses require clinical correlation and validation through retrospective mechanisms.