Trends in Practice Patterns of Conventional and Computer-Assisted Knee Arthroplasty

An Analysis of 570,671 Knee Arthroplasties Between 2010 and 2017

Abdalrahman G. Ahmed; Raymond Kang, MA; Mohamed Hasan, MD, MPH; Yao Tian, PhD, MS, MPH; Hassan M. Ghomrawi, PhD, MPH

Disclosures

J Am Acad Orthop Surg. 2021;29(22):e1117-e1125. 

In This Article

Methods

Data Sources

After obtaining institutional review board approval at our institution, we queried two statewide administrative data sets for the purposes of this study: the Statewide Planning and Research Cooperative System (SPARCS) and the Florida Administrative Data. Although these 2 states may not represent national trends, they are large states, representing 8% of the total US cohort, and a sizeable number of knee arthroplasties are performed in these states.[23] In addition, New York and Florida have very diversified environments that include big cities and small towns, a significant proportion of racial and ethnic minorities, and have different hospital types across those locations.[23] The New York and Florida statewide administrative data sets have also been used previously for examining utilization trends in total joint arthroplasty.[24–27] As such, analyzing trends in these two states may provide important insights into the CAKA practice trends occurring in the United States.

The SPARCS collects details on patient characteristics, diagnosis and treatments, outpatient services, ambulatory surgery, and emergency department visits from all nonfederal hospitals that are required to send data to the SPARCS database.[28] Florida administrative data include data on inpatient, outpatient, and ambulatory surgery in addition to emergency department discharge summaries for all procedures performed in Florida.[29] Hospital inpatient discharge data include data from short-term acute care psychiatric hospitals, long-term psychiatric hospitals, free-standing comprehensive rehabilitation facilities, and comprehensive rehabilitation distinct part units, which are all included in the Hospital Detailed Patient Database. We identified patients who underwent conventional and CAKAs between January 1, 2010, and September 30, 2017, in these 2 data sets.

Identification of Conventional and Computer-assisted Knee Arthroplasty

Using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes and International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) codes, patients who underwent primary KAs from 2010 to 2017 were identified in the databases (see Supplemental Digital Content, Table A.1, http://links.lww.com/JAAOS/A591, for full list of codes). KA patients whose data also included computer-assisted codes (ICD-9 codes 00.3, 00.31, 00.32, 00.33, 00.34, 00.35, and 00.39) (ICD-10 codes 8E0YXBZ, 8E0YXBF, 8E0YXBG, and 8E0YXBH) were identified as CAKA cases. Because the ICD-9-CM coding for KA does not differentiate total from unicompartmental KA, whereas the ICD-10-CM coding does, we included ICD-10-CM codes for unicompartmental for the period Q4 2015–end of 2017 to ensure continuity of trends.

Statistical Analysis

For each state, overall CAKA rate (ie, number of CAKA/[CAKA + conventional KA] across all years) and the proportion of CAKAs performed using each modality were calculated. To construct the utilization trend over time, quarterly CAKA rates (ie, number of CAKA/[CAKA + conventional KA] per quarter) were computed overall and then broken down by race and insurance status. Bivariate associations between patient demographics (ie, age, sex, race, insurance, and Charlson score) and those undergoing a CAKA were assessed using the t test and chi-square tests.[30–32] Multivariable logistic regression models were then used to determine the independent predictors of undergoing CAKA. The transition from ICD-9 to ICD-10 coding brought about a number of issues, most notably making sure that the ICD-9 and ICD-10 codes are capturing the same condition/procedure. Such an issue likely affects serial temporal assessments when using both ICD-9 and ICD-10 codes (Table 1).[33] To determine whether the ICD-9 to ICD-10 affected the regression results, separate models were run for each state and each ICD era (Q1 2010 to Q3 2015 and Q4 2015 to Q3 2017), a total of four regression models. All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC).

processing....