Potential Selection Bias in Observational Studies Comparing Cervical Disc Arthroplasty to Anterior Cervical Discectomy and Fusion

Study Design. Retrospective study using a national administrative database. Objective. To define the cohort differences in patient characteristics between patients undergoing cervical disc arthroplasty (CDA) and anterior cervical discectomy and fusion (ACDF) in a large national sample, and to describe the impact of those baseline patient characteristics on analyses of costs and complications. Summary of Background Data. CDA was initially studied in high quality, randomized trials with strict inclusion criteria. Recently a number of non-randomized, observational studies have been published an attempt to expand CDA indications. These trials are predisposed to falsely attributing differences in outcomes to an intervention due to selection bias. Methods. Adults undergoing ACDF or CDA between 2004 and 2014 were identified using International Classification of Diseases, 9(th), Clinical Modification (ICD-9-CM) diagnosis and procedure codes. Perioperative demographics, comorbidities, complications, and costs were queried. Patient characteristics were compared via chi-square andttests. Cost, mortality, and complications were compared between ACDF and CDA cohorts using models that adjusted for demographics and comorbidities, as well as "naive" models that did not. Results. A total of 290,419 procedures, 98.2% ACDF and 1.8% CDA, were included in the sample. Compared with ACDF patients, CDA patients were younger, healthier as evidenced by number of comorbidities, and had an improved socioeconomic status as measured by income and insurance. The naive logistic regression model showed that hospital costs for CDA were $549 lower than ACDF. In the fully specified model, CDA was $574 more expensive. The naive model for medical complications suggests a protective advantage for CDA over ACDF, odds ratio of 0.627,P Conclusion. Patients undergoing CDA were younger and healthier with higher socioeconomic statuses compared with ACDF patients. Accounting for these baseline differences significantly attenuated the apparent benefit for CDR on costs and medical complications.