Retrospective study of prospectively collected data.Objective.
To explore interdatabase reliability between National Inpatient Sample (NIS) and National Surgical Quality Improvement Program (NSQIP) for anterior cervical discectomy and fusion (ACDF) in data collection and its impact on subsequent statistical analyses.Summary of Background Data.
Clinical studies in orthopedics using national databases are ubiquitous, but analytical differences across databases are largely unexplored.Methods.
A retrospective cohort study of patients undergoing ACDF surgery was performed in NIS and NSQIP. Key demographic variables, comorbidities, intraoperative characteristics, and postoperative complications were analyzed via bivariate and multivariate analyses.Results.
A total of 112,162 patients were identified from NIS and 10,617 from NSQIP. Bivariate analysis revealed small, but significant, differences between patient demographics, whereas patient comorbidities and ACDF intraoperative variables were largely much more distinct across the two databases. Multivariate analysis identified independent risk factors between NIS and NSQIP for mortality, cardiac complications, and postoperative sepsis, some of which were identified in both but most of which were unique to one database. Identification of independent risk factors from both databases specifically highlights their greater validity and importance in stratifying patient risks. In addition, NSQIP was found to be a more accurate predictor for complications based on the average areas under the receiver-operating curve (CNSQIP = 0.83 vs. CNIS = 0.81) across the multivariate models. Complication rate analysis between inpatient and outpatient settings in NSQIP showed the importance of at least 30-day patient follow up, which was devoid in NIS data tabulation and further marked its weakness compared with NSQIP.Conclusion.
Despite having largely similar patient demographics, this study highlights critical risk factors for ACDF and demonstrates how different patient profiles can be across NIS and NSQIP, the impact of such differences on identification of independent risk factors, and how NSQIP is ultimately better suited for adverse-event studies.Conclusion.
Level of Evidence: 3