Surgeon Variation Is Likely the Primary Confounding Factor in Observational Studies of Technology Use in Total Knee...

Large observational datasets are increasingly being used to study the impact of technology on the outcomes of total knee arthroplasty (TKA). The validity of these studies is unclear, and they may be susceptible to bias. We aimed to uncover potential confounding variables for patients receiving technology during TKA.

A retrospective review of 1,001 consecutive patients who underwent primary TKA from 2021 to 2023 at an academic medical center was performed. All four surgeons at this institution selectively used technology (computer navigation, robotics, or augmented reality). Patient charts and radiographs were manually assessed for multiple medical and orthopedic variables. Univariate analyses were performed with Chi-square tests, t-tests/Kruskal-Wallis tests, and analysis of variance tests to compare those patients who had technology used versus those who did not. Binary logistic regression analyses were used to determine the relationship between variables from the univariate analysis with P < 0.1.

There were 45% of the patients who had technology used during TKA. For univariate analyses, age, range of motion, preoperative Knee Injury and Osteoarthritis Outcome Score Joint Replacement (KOOS JR) score, surgeon, payer, diabetes, liver disease, limb alignment, peripheral vascular disease, retained hardware, excessive bone loss, and obesity were significant variables associated with technology use (P < 0.05). For multivariate analyses, surgeon, excessive bone loss, retained hardware, and peripheral vascular disease remained significant (P < 0.05). The surgeon performing the procedure explained the greatest amount of variation in the model for whether technology was used (P < 0.001).

The surgeon performing the operation appears to be a strong confounding variable when assessing the use of technology during TKA. Administrative and registry data studies that do not adjust for the surgeon are likely confounded and may lead to misleading conclusions. Several additional variables were also identified that may further confound observational research on this topic.