The use of technology in TKA may allow surgeons to achieve more accurate and precise implant placement. However, whether this translates to improved clinical outcomes, if certain patient populations benefit from intraoperative technology over unassisted techniques, or if advanced technologies confer improved long-term implant survivorship remains an area of active research.
Our findings generally support prior literature findings, which demonstrated that TA-TKA has not been universally adopted in the USA [16, 19, 20, 25]. We found that only 3.1% of patients underwent primary TKA with the use of intraoperative technology in the year 2018. In contrast, a National Inpatient Sample (NIS) database study found utilization rates for primary TKA to be 7.0% in 2014, a rate which has likely further increased in recent years [19]. Notably, the NSQIP and NIS databases include data from different subsets of hospitals. Geographic region and socioeconomic status influence TA-TKA utilization rates, whereas TA-TKA is more likely to be performed at high-volume, urban, and teaching hospitals [16, 19]. Furthermore, patients with private insurance are more likely to undergo surgery with technology as opposed to conventional methods [16, 26]. Another NSQIP database study conducted between 2012 and 2018 found that 2.6% of primary TKAs were performed with the assistance of computer navigation [25]. On the other hand, in an analysis of the New York Statewide Planning and Research Cooperative System (SPARCS) database, robotic TKA utilization rates were lower than 1.5% [19, 20]. Although database studies are well suited to determine national trends, they are limited by the number of hospitals that participate in data collection and often take multiple years to release datasets. Since the last year included in this study, we would expect the utilization rates of TA-TKA to have increased each year.
Surprisingly, the annual proportion of TKAs performed using technology within the NSQIP database decreased from 2010 to 2018. A recent study found that the utilization of computer-assisted TKA increased from 4.9% to 9.5% in New York and from 4.0% to 5.7% in Florida between the years 2010 and 2017 [14]. An NIS database analysis found that the proportion of TA-TKA increased steadily from 1.2% in 2005 to 7.0% in 2014 [19]. Our contradictory results may be explained by the increase in the number of participating centers in the NSQIP database during the study period, which coincides with the decrease in technology use. These additional centers may be performing TA-TKA at lower rates than the NSQIP-participating institutions at the start of the study period.
We found significant differences in race between the unmatched TA-TKA and U-TKA groups. In the TA-TKA cohort as compared with the U-TKA cohort, Asian, Black/African Americans, and Native American/Pacific Islanders were underrepresented and Whites overrepresented. This highlights the potential for racial disparities in TA-TKA utilization across the US health system. In an NIS database analysis, African American patients were more likely to undergo TA-TKA as compared with Whites, Hispanics, Asians or Pacific Islanders, and Native Americans [19]. Several authors have suggested that patient income, geographic region, and socioeconomic status may significantly influence the probability of receiving TA-TKA [16, 19]. Since the NSQIP database lacks hospital and geographic data, we were unable to investigate these geographic or socioeconomic trends.
To our knowledge, this is the largest study to compare operative times for patients undergoing TA-TKA and U-TKA. Prior studies have shown that TA-TKA and U-TKA have similar operative times, and not surprisingly, we found that mean operative times did not significantly differ between the matched TA-TKA and U-TKA cohorts [8, 10, 25, 27, 28]. Although these data support existing findings, operators using technology must register bony landmarks, interpret intraoperative data, and adjust component positions accordingly. These additional steps have the potential to increase operative time, but high-volume adult reconstructive knee surgeons may perform these swiftly such that no clinically meaningful increase in operative time is observed. During the study period, TA-TKA operative times decreased, suggesting that hospital systems and surgeons have become more efficient over time as they emerged from their initial learning curve and as the technology itself has substantially improved over time. Notably, operative times in the TA-TKA cohort increased in the first few study years before decreasing by a greater amount in the latter study years. The number of participating centers in the NSQIP database increased during the first few years, and the increased operative time may be due to these additional centers reporting patients with longer procedure times.
In examining inpatient outcomes, we found that patients undergoing TA-TKA had significantly shorter LOS as compared with that of U-TKA. In the existing literature, the association between TA-TKA and LOS is unclear. Several studies have found either similar [27, 28] or longer LOS [8, 29] in patients undergoing TA-TKA, but our analysis is one of the first to report shorter LOS. Additionally, patients in the TA-TKA cohort were more likely to be discharged home and less likely to be discharged to rehabilitation or skilled nursing facilities as compared with the U-TKA cohort. Since the NSQIP database lacks hospital data, we were unable to explain these trends, though varying institutional discharge policies across individual hospital systems may account for these differences. It is important to note that institutions supporting robotic and navigation technology may also have more robust perioperative protocols, same-day surgery programs, and physical therapy resources, possibly leading to improved in-patient outcomes as well.
Patients undergoing TA-TKA had lower complication rates than those undergoing U-TKA. A recent meta-analysis comparing robotic and unassisted TKA observed similar complication rates between groups [30], though some analyses comparing computer-assisted and unassisted TKA have demonstrated lower complication rates in those undergoing computer-assisted TKA [8, 13]. In our study, TA-TKA patients had significantly lower postoperative transfusion rates compared with U-TKA patients, which is consistent with prior findings [8, 13, 28] and associated with improved surgical outcomes [31]. Postoperative readmission rates did not significantly differ between groups. In prior national database analyses, computer-assisted TKA has yielded similar [25] and decreased [13] readmission rates. Further analysis revealed that patients in the TA-TKA cohort were readmitted primarily for superficial surgical site infections and hematological issues in addition to unspecified and other medical diagnoses. Manually reviewing the diagnosis codes of readmitted patients, it was difficult to determine whether certain readmissions were caused by periprosthetic fracture or hardware complications as well, possibly causing under- or overreporting of these conditions in our analysis. Nonetheless, the trend toward lower readmission rates for TA-TKA patients may be due to reduced need for bone and periarticular soft tissue manipulation intraoperatively due to improved component positioning and tracking [32]. Moreover, fewer TA-TKA patients were discharged to skilled nursing facilities, which is associated with increased readmission rates as well [33]. Lastly, reoperation and mortality rates did not differ between groups, which is consistent with previous findings [27, 28].
Surgeons who perform TA-TKA and U-TKA may work in different regions of the country or different practice environments; this provides one possible explanation for the perioperative differences observed between TA-TKA and U-TKA groups. Prior studies assessing TA-TKA utilization have demonstrated that it is more likely to be performed in urban, teaching, and high-volume hospitals [16, 19]. Given this, we hypothesize that TA-TKA cases in the NSQIP database were predominantly performed by operators at urban, tertiary referral centers, orthopedic specialty hospitals, or ambulatory surgery centers. This may explain the observed differences in perioperative outcomes.
For certain complications, our analysis produced slightly different incidence rates on subgroup analysis of major complications, readmissions, and reoperations. This discrepancy is due to coding differences within the database for each category of complication. Since the standard of care for a deep hardware infection involves operative management, we believe that infection rates are best approximated in the reoperation subgroup analysis.The NSQIP database lacks the granularity in ICD diagnosis coding needed to capture orthopedic-specific complication rates, and therefore, the reoperation subgroup analysis provides the most accurate measure of infection rates. Additionally, the NSQIP database does not record complications treated on an outpatient basis or through emergency department care, and this causes underreporting of certain postoperative complications—notably superficial wound complications, suboptimal functional outcomes, and other low-acuity complications related to TKA.
This study had several limitations. Since this was a retrospective database study, selection bias likely exists between study groups. We performed propensity score matching to limit potential confounders between patients, but since the dataset was completely de-identified, we were unable to control for the hospital the surgery was performed at. This likely influenced our results and explains the paradoxical trends in TA-TKA utilization within the dataset. This also likely influenced our analysis of short-term outcomes because there is significant heterogeneity between hospital systems in terms of demographics, patients’ social support, perioperative rehabilitation protocols, clinical care coordination, postoperative inpatient care, and access to multidisciplinary home service. All of these factors have been shown to reduce short-term complications and readmission rates [34,35,36]. Since the NSQIP database utilized CPT codes to record operative details, we were also unable to differentiate between computer navigation and robotic TKA.
Furthermore, NSQIP lacks specific ICD coding for readmission and reoperation diagnoses and does not record emergency department visits in which the patient is discharged. This likely caused our study to underrepresent the true complication rates, specifically superficial site infections in which the patient can receive antibiotics and then undergo follow-up with their primary care provider. Additionally, although technology use likely increased from 2018 to 2021, we were unable to capture this because, at the time of this analysis, the latest year of released NSQIP data was 2018. We also did not assess which patient characteristics are associated with technology utilization, though prior studies have investigated this issue [16, 19]. The NSQIP database only provided 30-day follow-up, which does not capture the true complication rate within the 90-day bundled period. Lastly, since our study was highly powered, our analysis detected miniscule differences as statistically significant. Given this, although certain clinical outcomes differed significantly, these differences may not be clinically significant.