Abstract
Introduction
In 2015, the healthcare system transitioned from International Classification of Diseases, Ninth Revision (ICD-9) coding to the Tenth Revision (ICD-10). Given that administrative claims are used for quality initiatives, risk adjustment models and clinical research, we sought to determine the effect of new, more detailed coding on the incidence of complications following primary total knee arthroplasty (TKA).
Methods
The Humana administrative claims database was queried from 2-years prior to October 1, 2015 (ICD-9 cohort) and for 1-year after this date (ICD-10 cohort) to identify all primary TKA procedures. Each TKA was then tracked for occurrence of an arthroplasty specific post-operative complication within 6 months of surgery using the respective coding systems. Laterality and joint specific codes were utilized for the ICD-10 cohort to ensure complications occurred on the same side and joint as the index procedure. Incidence of each complication was compared between cohorts using risk ratios (RR) and 95% confidence intervals.
Results
There were 19,009 TKAs in the ICD-10 cohort and 38,172 TKAs in the ICD-9 cohort. The incidence of each post-operative complication analyzed was significantly higher in the ICD-9 cohort relative to the ICD-10 cohort (Figure 1). Comparing ICD-9 cohort to ICD-10 cohort, PJI occurred in 1.9% vs 1.3% (RR 1.5), loosening in 0.3% vs 0.1% (RR 2.7), periprosthetic fracture in 0.3% vs 0.1% (RR 3.0) and other mechanical complications in 0.7% vs 0.4% (RR 2.0), respectively (p < 0.05 for all). These findings remained significant when subgroup analyses were performed to control for seasonal variation between groups.
Conclusion
The transition from ICD-9 to ICD-10 coding has dramatically altered the reported incidence of complications following TKA. These results are likely due to the added complexity of ICD-10 which is joint and laterality specific. It is important to quantify and understand the differences between coding systems as this data is used for quality initiatives, risk adjustment models and clinical research. Thoughtful methodology will be necessary for future research if both ICD-9 and ICD-10 data are being analyzed simultaneously. Furthermore, additional studies are needed to compare the hospital billing data for complex diagnoses or revision procedures to coding performed by skilled researchers in the field.
For figures, tables, or references, please contact authors directly.
