Inaccuracies in ICD Coding for Obesity Would Be Expected to Bias Administrative Database Spine Studies Toward Overestimating the Impact of Obesity on Perioperative Adverse Outcomes
Retrospective cohort study.Objective.
To determine if International Classification of Diseases (ICD) coding for obesity is biased toward certain subgroups and how potential bias may influence the outcomes of database research in spine.Summary of Background Data.
There has been increased use of national databases using administrative data in the spine surgery literature. Past research demonstrates that sensitivity of ICD codes for obesity is poor, but it is unknown if such inaccuracies are systematically biased and if they may bias studies utilizing such data.Methods.
Patients who underwent elective posterior lumbar fusion, 2013 to 2016, at a large academic hospital were identified. All ICD codes assigned to the encounter were obtained. Body mass index (BMI) was calculated based on height and weight. The sensitivity of ICD coding for obesity was calculated. Sensitivity was compared for subgroups defined by demographic, comorbidity, intraoperative, and postoperative factors. The association of obesity (as defined by BMI≥30 and ICD coding) with 30-day postoperative adverse events was tested with multivariate regression.Results.
The study included 796 patients. The overall sensitivity of ICD coding for obesity was 42.5%. The sensitivity of ICD coding for obesity was significantly higher in patients with greater BMI, diabetes, American Society of Anesthesiologists class≥III, increased length of stay, venous thromboembolism, any adverse event, and major adverse event. Multivariate analysis for determining outcomes of increased risk with obesity as defined by ICD coding included venous thromboembolism, major adverse events, and any adverse events. However, multivariate analysis for determining outcomes of increased risk with obesity defined by BMI did not yield any positive associations.Conclusion.
ICD codes for obesity are more commonly assigned to patients with other comorbidities or postoperative complications. Further, use of such nonrandomly assigned ICD codes for obesity has the potential to skew studies to suggest greater associated adverse events than calculated BMI would demonstrate.Conclusion.
Level of Evidence: 3