Acute Versus Delayed Reverse Shoulder Arthroplasty for the Primary Treatment of Proximal Humeral Fractures

Henry D. Seidel, BS; Sarah Bhattacharjee, BS; Jason L. Koh, MD; Jason A. Strelzow, MD; Lewis L. Shi, MD


J Am Acad Orthop Surg. 2021;29(19):832-839. 

In This Article



This retrospective study used the Medicare Standard Analytic File data set from PearlDiver (PearlDiver), a national insurance claims database consisting of 51 million deidentified Medicare patient records for individuals with orthopaedic diagnoses. This database contains records from 2005 to 2014 and is searchable by Current Procedural Terminology (CPT) codes and International Classification of Diseases, Ninth Revision (ICD-9) diagnostic and procedural codes. Because of the anonymized records of PearlDiver, this study was determined exempt from the Institutional Review Board under IRB18-0215.

Study Cohort

Patients aged 65 years and older with proximal humeral fractures were identified in the database using the first instance of a recorded ICD-9 diagnostic code for proximal humeral fracture (Figure 1). Among this group, we identified patients who underwent rTSA (ICD-9 81.88) within 1 year after the proximal humeral fracture diagnosis. Because the ICD-9 code 81.88 for rTSA was established in the fourth quarter of 2010, only patients with rTSA procedures done after October 1, 2010, were included in this study. Patients were separated into two cohorts based on the treatment. The acute primary rTSA cohort included patients with an rTSA code recorded within 4 weeks of fracture. The delayed primary rTSA cohort included patients who had the first instance of an rTSA code recorded greater than 4 weeks after fracture diagnosis and within 1 year of fracture diagnosis. These intervals were outlined by previous studies.[32,34,37] To account for patients who lost or changed insurance, only patients who were continuously active in the database for 1 year after the rTSA code were included in the study. Therefore, all included patients effectively had a follow-up of 1 year, and outcomes were only measured within that 1-year time frame. To ensure that none of the rTSA surgeries in the delayed treatment were salvage procedures for other failed operative treatments, patients who underwent additional proximal humeral fracture treatment before rTSA, including ORIF and hemiarthroplasty, were excluded from this study.

Figure 1.

Flow diagram illustrating the cohort selection process and treatment breakdown of proximal humeral fractures in elderly patients from the PearlDiver database between 2010 and 2014. The acute primary rTSA cohort underwent rTSA <4 weeks following fracture. The delayed primary rTSA cohort underwent rTSA >4 weeks following fracture and within 1 year. rTSA = reverse total shoulder arthroplasty

We examined age, sex, tobacco use, obesity, diabetes, and Elixhauser Comorbidity Index of the two groups to determine any differences in the demographics and comorbidities of the patient populations. The 1-year revision rate (ICD-9 81.97, CPT-23473, and CPT-23474), 1-year postoperative surgical complication rates, surgery day cost of treatment, and length of stay were assessed for the acute and delayed primary rTSA treatment cohorts. Postoperative surgical complications included mechanical complications (including but not limited to aseptic loosening, periprosthetic fracture, and periprosthetic osteolysis), infection, and dislocation. The surgical complication rates were defined using CPT and ICD-9 codes. For patients who underwent revision surgery, surgical complications proceeding the revision surgery that were recorded on the patient's record were identified.

Statistical Analysis

We used Pearson chi-squared analysis to assess the univariate differences in rates of revision and surgical complication between the acute and delayed treatment cohorts. The Student t-test was used to compare the surgery day cost and length of stay between the cohorts. All tests were conducted at an alpha level of 0.05. For revision and complication outcomes that were statistically significantly, a multivariate logistic regression was used to account for potential confounding from the comorbidities and demographic factors of age, sex, tobacco use, diabetes, and obesity. The adjusted odds ratios and confidence intervals (CIs) were determined from the multivariate analysis. Statistical analysis was done using the R statistical package available through PearlDiver.