This retrospective cohort study used de-identified electronic health record (EHR) data from the TriNetX Research Network, which aggregates data from 92 healthcare organizations (HCOs), including hospitals, primary care, and specialty providers across diverse patient populations. The study specifically queried TriNetX’s Natural Language Processing dataset, which includes both insured and uninsured patients.

TriNetX has been reviewed by the Western Institutional Review Board, and since only de-identified data were used, additional IRB approval and informed consent was not required. TriNetX offers access to continuously updated, de-identified aggregate EHR data, including demographics, diagnoses, procedures, medications, laboratory values, genomics, and cohort analysis tools. Detailed information regarding the TriNetX database and software has been previously published [14,15].

Patients aged 65 and older undergoing TSA or RSA on or prior to September 23, 2024, with a previous diagnosis of shoulder OA were first queried. All patients with diagnoses of rotator cuff pathologies prior to their procedure were excluded using the following codes: M75.1A, M75.5A, and S46.0A. The diagnosis of OA in both cohorts was identified using the following International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) codes: M19.019, M19.012, M19.011, M19.21, M12.211, M19.212, and M19.219. RSA patients were identified with the ICD-10 Procedure Coding System (ICD-10-PCS) codes 0RRK00Z and 0RRJ00Z, whereas TSA patients were identified with the ICD-10-PCS codes 0RRJ0JZ, 0RRJ07Z, 0RRK0JZ, 0RRK07Z, 0RRJ0KZ, and 0RRK0KZ. Each cohort excluded patients identified by the ICD-10-PCS codes of the other cohort.

Demographic and patient characteristic data were collected and compared using TriNetX’s cohort comparison tools. Propensity score matching was conducted using the platform, which applies logistic regression to user-defined covariates to generate propensity scores for individual subjects. A 1:1 matching approach was then performed using a greedy nearest neighbor algorithm with a caliper width of 0.1 pooled standard deviations. To mitigate bias associated with the nearest neighbor algorithm, TriNetX randomizes the row order during the matching process. The matched characteristics and comorbidities included average age at index, sex, race, body mass index (BMI), and the incidence of hypertension, diabetes, nicotine dependence, heart failure, chronic obstructive pulmonary disease (COPD), end-stage renal disease (ESRD), and ascites. Standardized mean differences (SMDs) were used to ascertain the distribution balance among our baseline variables. All variables had SMD values of less than 0.1 and were considered well-matched.

Following matching, infectious and orthopedic complication rates were tracked in each cohort from the first day after surgery (index event) for 90 days. Pertinent codes used to query these outcomes are found in Supplementary Table 1. An additional category named “Any Orthopedic Complication” was created to ascertain an overview of musculoskeletal adverse events including dislocation, periprosthetic fracture, scapula fracture, implant loosening, stiffness, rotator cuff injury, and hematoma. Risk differences were utilized to compare outcomes across groups, and significant differences were calculated with t-tests using the TriNetX platform. Significance was set at P<0.05.

Additionally, multivariable risk factor analysis was conducted independently for both the RSA and TSA cohorts using TriNetX’s logistic regression tool. This analysis examined the risk for orthopedic and infectious complications concerning patient comorbidities. Infections as an outcome of logistic regression were defined by the presence of any of the following: musculoskeletal infection, deep and superficial surgical site infection, wound dehiscence, unspecified infection, or sepsis (Supplementary Table 1). P-values less than 0.05 were considered significant.

Patients treated with RSA experienced a significantly higher number of musculoskeletal infections compared to TSA. In contrast to Ross et al. [16], who reported a 0.5% rate of prosthetic joint infections 90 days following RSA, our rate of musculoskeletal infections was two-fold that rate at 1.39%. Variation in diagnosis codes across studies when defining infection may yield differences in results, as our musculoskeletal infection category consisted of multiple diagnosis codes. Nonetheless, our findings are compatible with prior literature, which generally reported higher rates of infectious complications following RSA compared to TSA [16,17]. Previously suggested mechanisms to explain this disparity ranged from variation in implant type to surgical techniques [18]. Overall, the rate of musculoskeletal infections following shoulder arthroplasty in our study fell within the published range of 1% to 10% of prosthetic joint infections [19].

Although RSA may be performed in patients with existing rotator cuff deficiencies, patients with a previous diagnosis of rotator cuff arthropathy were excluded from this study to identify new postoperative deficiencies. A significantly lower number of postoperative rotator cuff injuries were found in patients treated with TSA than RSA. The higher rate of rotator cuff injuries following RSA in our study may reflect intraoperative variation between RSA and TSA, with undiagnosed rotator cuff deficiencies more frequently discovered during RSA. This warrants further investigation [20].

Postoperative dislocation was observed at a higher rate following RSA, at 0.87% compared to0.27% in TSA in the present study. Rates of dislocation in RSA vary widely from 2.4% to 31% in the existing literature, whereas our study found a dislocation rate under 1% [21]. We hypothesize that an outcome window of 90 days may have been insufficient to capture the true rate, accounting for this difference. Instability in RSA may vary from TSA due to differences related to implant type, lack of soft tissue tension, impingement, and a deltopectoral approach leading to eventual dislocation [19,22]. Furthermore, RSA places greater reliance on the deltoid muscle, and inadequate tension may create instability or future dislocations. Finally, greater BMI, as seen with the RSA cohort, as well as a history of prior surgery (not accounted for in this study) constitute potential nonsurgical risk factors for dislocation [18].

Stiffness was observed at greater rates in our TSA cohort (5.51%) compared to RSA (4.23%). Although few studies have examined this relationship in the context of OA, the higher rate of early stiffness following TSA compared to RSA is logical considering their respective postoperative rehabilitation protocols. TSA involves relatively stricter postoperative immobilization, whereas early range of motion is more strongly encouraged in RSA [23,24]. Although early stiffness during the 90-day postoperative period is not necessarily indicative of a surgical complication, it can persist beyond 1-year postoperative in up to 25% of patients following RSA, limiting shoulder function, decreasing patient satisfaction, and necessitating reoperation [17,25]. Still, despite increased stiffness, TSA generally yields better postoperative range of motion compared to RSA [26,27]. The contradictory findings in TSA may stem from yet-undetermined biomechanical mechanisms. For instance, one study linked early stiffness to improved tendon healing after arthroscopic rotator cuff repair [28]. While rotator cuff pathology was excluded from our study, a similar mechanism may underlie our results. Further research is needed to determine whether such biomechanical factors apply to shoulder arthroplasty for OA.

After matching to account for comorbidity and demographic differences, patients experienced a periprosthetic fracture rate of 0.85% following treatment with RSA compared to 0.61% in patients treated with TSA. Published literature has demonstrated similarly increased rates of periprosthetic fractures in patients undergoing RSA compared to TSA [16-18]. In RSA, acromial and scapular spine fractures as well as humeral fractures are observed, with rates documented to be between 4.5% and 3.5% respectively [29]. Intraoperative glenoid and humeral fractures are reported to increase the risk of subsequent periprosthetic fracture and may be more likely depending on existing patient characteristics such as glenoid arthritis or osteopenia [17,18]. Furthermore, screw placement, implant design, and deltoid lengthening are risk factors implicated in periprosthetic fracture following RSA.

When comparing nonorthopedic complications, we found greater rates of renal complications as well as deep vein thrombosis in patients undergoing RSA, which is consistent with the existing literature [16,30]. Prior research established older age (>70 years), higher comorbidity, and history of cardiac or thromboembolic events as risk factors for 90-day cardiac and thromboembolic complications following primary TSA [31]. Higher rates of UTIs were also observed among the RSA group. Although previous reports comparing UTI in the setting of RSA versus TSA are sparse, existing research on total joint arthroplasty has identified increasing age, female sex, steroid use, operative time greater than 130 minutes, and preoperative creatinine as risk factors for UTI [32].Taken together, the literature regarding the risk for medical complications in total joint arthroplasty overwhelmingly attributes differences to age, comorbidity, and sex. Further research is required to identify risk factors for these medical complications based on additional comorbidities not explored in the present study as well as intraoperative factors, which are not available on TriNetX.

A secondary aim of this study was to determine risk factors for orthopedic complications following shoulder arthroplasty, with a greater number of significant risk factors identified in the TSA group (Table 3). Hypertension was a significant risk factor for orthopedic complications in both the RSA and TSA groups. Prior research found hypertension to be a risk factor for fracture as well as extended hospital stay, which may contribute to postoperative stiffness [33,34]. Increasing age was associated with a lower risk of orthopedic complications in both the RSA and TSA groups, suggesting that older patients may experience fewer postoperative complications despite often having greater comorbidity burdens. This protective effect may stem from lower functional demands, reduced mechanical stress on implants, and more conservative rehabilitation protocols in older patients, highlighting the need for further investigation.

Risk factors that were significant exclusively in the RSA group included a personal history of nicotine dependence. Prior research identified active tobacco use as a risk factor for numerous complications in the 90-day window following TSA including fractures, dislocation, and other forms of instability; the underlying mechanisms are multifaceted and have been attributed to the deleterious effects of nicotine on bone density, peripheral blood flow, and gene expression in healing tendons [35].

In TSA, risk factors included male sex, COPD, and obesity. Male sex was associated with a lower risk of orthopedic complications, but other literature attributed this to male TSA patients often being younger and more active, which could influence long-term mechanical failure rates. This lack of consensus regarding sex-specific TSA outcomes warrants further study [36-40]. COPD was a significant risk factor, consistent with literature linking it to periprosthetic fracture, implant failure, instability, and stiffness, likely due to chronic inflammation and physical inactivity [41]. Obesity was the strongest TSA-specific risk factor. Prior studies have demonstrated that obesity contributes to higher joint loading, altered biomechanics, increased intraoperative complexity, and, consequently, greater risk of implant loosening, stiffness, periprosthetic fractures [42-45].

This study also analyzed risk factors for infectious complications within the 90-day postoperative period following RSA and TSA performed for OA (Table 4). Diabetes mellitus was a significant risk factor for infections in both the RSA and TSA groups, consistent with previous literature in which diabetes mellitus is well-documented as a risk factor for postoperative infection due to its detrimental effects on immune response and wound healing [46].

In the RSA group, male sex and heart failure were identified as significant risk factors for infectious complications. Heart failure, which is associated with systemic inflammation and poor circulation, also significantly raises the risk of infection following RSA [47]. The higher risk for infection in male RSA patients is well-documented and likely multifactorial, involving a greater comorbidity burden, weaker immune response compared to females, and perhaps social factors such as lower adherence to postoperative care and higher rates of smoking and alcohol use [36,48,49]. There were no significant risk factors for infection exclusive to the TSA group.

This study is not without its limitations. First, while TriNetX offers the advantage of analyzing a large sample size, this same feature limits data analysis due to its reliance on medical billing codes (e.g., ICD-10, Current Procedural Terminology) to establish cohorts and compare outcomes. Although the selection of codes was informed by a thorough literature review, the coding process is inherently subjective, and as such any given combination of codes may conceivably fail to capture some members of the target cohort (in this case, readmitted TSA or RSA recipients) due to discrepancies in coding across institutions or even inaccuracies in coding. Similarly, not all adverse events identified in this analysis necessarily represent unintended complications of the procedure itself. For example, pre-existing rotator cuff pathology is frequently managed with RSA, and patients who had such diagnoses prior to surgery may continue to experience symptoms in the short term. As a result, follow-up encounters coded for rotator cuff-related issues may be captured as complications in this dataset, even though they may reflect the natural progression of a pre-existing condition rather than a new postoperative complication. Several differences in baseline comorbidities persisted despite propensity matching, including white ethnicity, nicotine dependence, COPD, and ascites; thus, some differences such as the postoperative occurrence of respiratory complications could be attributed in part to baseline group differences in COPD rather than the procedures alone. Additionally, the present study investigated complications confined to the 90-day postoperative window, which may not be sufficient to fully capture procedure-related adverse events that occur later in the postoperative period. One final limitation of the TriNetX database is the possibility that one or more HCO is unresponsive when querying the database, potentially yielding inconsistent results when attempting to analyze outcomes for the same cohort across different time periods; this limitation was mitigated in this our study by replicating the present results multiple times to ensure any significant findings were consistently demonstrated across numerous queries. Furthermore, the retrospective nature of this study may also hinder the ability to establish causality. Variation in surgical techniques and patient management strategies, which are not accounted for in this study, may influence outcomes and should be considered when interpreting the results.

Future research should examine similar orthopedic and infectious outcomes over a longer postoperative course, beyond the 90-day postoperative window. Additional research encompassing a broader array of complications in various organ systems and/or indications for revision surgery, would likewise contribute to a more complete understanding of the different outcomes observed following TSA and RSA.