Background: Large administrative data sources such as the National Inpatient Sample are frequently used to estimate the incidence of surgery in the United States. These sources do not identify unique surgical cases (often containing multiple procedures). Investigators routinely assume that all surgical procedures on a given day comprise a single case, ignoring returns to the operating room (e.g., to address a complication). Methods: We estimated the inaccuracy of this assumption using 11 years of electronic data from 2 large academic hospitals. Results: An administrative database would have underestimated the actual case count by 0.22% (99% UCL = 0.25%) and 0.19% (99% UCL = 0.22%), respectively. Patients undergoing cardiac or vascular surgery had a much greater underestimation of case counts (99% UCL 1.72% and 1.60%, and 99% UCL 1.06% and 1.09%, respectively) than other specialties (99% UCL for each specialty <1.0%, including orthopedics, otolaryngology, and urology). The trauma surgery 99% UCL undercount was 1.7% at the one hospital where this could be measured. Conclusions: For most specialties, inferring overall surgical case frequency using administrative data based on an assumption of no returns to the surgical suite is reasonable. However, adjustment for cardiac, vascular, and trauma surgical caseloads using such sources should be considered.
- Medical informatics computing
- Organization and administration
ASJC Scopus subject areas
- Critical Care and Intensive Care Medicine
- Anesthesiology and Pain Medicine