Artificial intelligence for automated identification of total shoulder arthroplasty implants.

BACKGROUND: Accurate and rapid identification of implant manufacturer and model is critical in the evaluation and management of patients requiring revision total shoulder arthroplasty (TSA). Failure to correctly identify implant designs in these circumstances may lead to delay in care, unexpected intraoperative challenges, increased morbidity, and excess health care costs. Deep learning (DL) permits automated image processing and holds the potential to mitigate such challenges while improving the value of care rendered. The purpose of this study was to develop an automated DL algorithm to identify shoulder arthroplasty implants from plain radiographs. METHODS: A total of 3060 postoperative images from patients who underwent TSA between 2011 and 2021 performed by 26 fellowship-trained surgeons at 2 independent tertiary academic hospitals in the Pacific Northwest and Mid-Atlantic Northeast were included. A DL algorithm was trained using transfer learning and data augmentation to classify 22 different reverse TSA and anatomic TSA prostheses from 8 implant manufacturers. Images were split into training and testing cohorts (2448 training and 612 testing images). Optimized model performance was assessed using standardized metrics including the multiclass area under the receiver operating characteristic curve (AUROC) and compared with a reference standard of implant data from operative reports. RESULTS: The algorithm classified implants at a mean speed of 0.079 seconds (±0.002 seconds) per image. The optimized model discriminated between 8 manufacturers (22 unique implants) with AUROCs of 0.994-1.000, accuracy of 97.1%, and sensitivities between 0.80 and 1.00 on the independent testing set. In the subset of single-institution implant predictions, a DL model identified 6 specific implants with AUROCs of 0.999-1.000, accuracy of 99.4%, and sensitivity >0.97 for all implants. Saliency maps revealed key differentiating features across implant manufacturers and designs recognized by the algorithm for classification. CONCLUSION: A DL model demonstrated excellent accuracy in identifying 22 unique TSA implants from 8 manufacturers. This algorithm may provide a clinically meaningful adjunct in assisting with preoperative planning for the failed TSA and allows for scalable expansion with additional radiographic data and validation efforts.

Minimum 15-year follow-up for clinical outcomes of arthroscopic rotator cuff repair.

BACKGROUND AND HYPOTHESIS: Arthroscopic rotator cuff repair surgery is one of the most common shoulder procedures performed in the United States. Although several studies have shown considerable symptomatic relief in the short term following surgery, a relatively high rate of recurrent defects has led surgeons to question the long-term durability of this operation. We hypothesized that outcomes at a minimum of 15 years of follow-up in patients who underwent all-arthroscopic rotator cuff repair would be maintained and would remain significantly improved compared with the preoperative status. METHODS: All-arthroscopic rotator cuff repairs were performed in 193 patients from 2003 to 2005. Patient-reported outcomes were collected preoperatively and at 1, 2, 5, and ≥15 years postoperatively. The primary outcome was the American Shoulder and Elbow Surgeons (ASES) score. Secondary outcomes included Single Assessment Numeric Evaluation (SANE), Shoulder Activity Scale (SAS), visual analog scale, and Patient-Reported Outcomes Measurement Information System (PROMIS)-Upper Extremity (UE) scores. Patient demographic characteristics, revision surgical procedures, and complications were recorded. Generalized estimating equations were used to model scores over time, and multiple comparisons between time points were performed using Tukey adjustment. RESULTS: This study included 60 patients with a mean follow-up period of 16.5 years (range, 15.8-17.7 years). The mean ASES score improved from 60.2 ± 18.8 preoperatively to 93.0 ± 9.4 at ≥15 years (P < .0001). The mean visual analog scale pain score decreased from 4.1 ± 0.7 preoperatively to 0.7 ± 0.3 at ≥15 years (P < .0001). The average SANE, SAS, and PROMIS-UE scores at ≥15 years were 87.8 ± 14.8, 8.8 ± 4.3, and 49.6 ± 10.2, respectively. Of 60 patients, 7 underwent revision surgery. Older age and female sex were associated with lower SAS scores at 15 years, whereas female sex was associated with lower PROMIS-UE scores. There were no factors predictive of ASES or SANE scores. CONCLUSION: At long-term follow-up (≥15 years), the patient-reported outcomes of all-arthroscopic rotator cuff repair show significant improvement from baseline preoperative function and remain durable over a period of 15 years. This information is useful in counseling patients regarding the long-term results of this procedure.

Outcomes of shoulder arthroplasty by year of index procedure: are we getting better?

BACKGROUND: The purpose of this study was to determine whether postoperative patient-reported outcomes improved over time following anatomic total shoulder arthroplasty (TSA) and reverse total shoulder arthroplasty (RTSA). METHODS: We performed a retrospective analysis of prospectively collected patient-reported outcomes from our institution's registry between 2008 and 2018 (N = 1899). American Shoulder and Elbow Surgeons (ASES) scores at a minimum of 2 years postoperatively were required. Univariable linear models were used to test the association between year of surgery and improvement in ASES scores at 2- and 5-year follow-up, as well as any association with age, sex, primary or revision surgery, hand dominance, American Society of Anesthesiologists classification, rotator cuff status, primary diagnosis, and Walch classification. Multivariable models were created to analyze ASES score improvement by index year while controlling for significant factors. RESULTS: In the univariable analysis, 5-year ASES difference scores increased each year by a mean of 1.65 (P < .001; 95% confidence interval [CI], 0.75-2.55) for TSA, 2.50 (P = .014; 95% CI, 0.52-4.49) for RTSA, and 1.64 (P < .001; 95% CI, 0.81-2.47) for the overall population. Patient sex, American Society of Anesthesiologists classification, rotator cuff status, primary diagnosis, Walch classification, and revision procedures were also significant factors affecting ASES scores. On multivariable analysis controlling for these factors, 5-year ASES difference scores were still significantly associated with year of surgery, increasing each year by a mean of 2.20 (P < .001; 95% CI, 0.91-3.50) for TSA, 4.83 (P < .001; 95% CI, 1.17-8.49) for RTSA, and 1.66 (P < .001; 95% CI, 0.81-2.51) for the entire population. CONCLUSION: Both anatomic TSA and RTSA patients reported increasing ASES difference scores at 5-year follow-up as time passed. These findings may indicate that advances in shoulder arthroplasty have resulted in better patient outcomes over time. Further research is needed to clarify which factors influence improvements in outcomes, particularly for revision procedures.

What Factors Are Associated with Patient-reported Outcome Measure Questionnaire Completion for an Electronic Shoulder Arthroplasty Registry?

BACKGROUND: There has been an increase in the use of electronic systems to collect patient-reported outcome measures. There is limited data on the added value of electronic reporting on increasing patient response proportions and little knowledge of which patients are more likely to respond. QUESTIONS/PURPOSES: (1) What proportion of patients completed patient-reported outcome questionnaires at baseline and at 1 year and 2 years of follow-up after shoulder arthroplasty, and what methods did they use to complete these questionnaires (either automated or manual data collection)? (2) What factors were associated with questionnaire completion? METHODS: Our shoulder arthroplasty registry from a high-volume, tertiary care center implemented an electronic platform to collect patient-reported outcomes. A total of 2128 patients underwent shoulder arthroplasty between 2016 and 2019. Patients without an email address on file were excluded; 90% (1907 of 2128) of patients were included in the study. The population was 50% women (954 of 1907) with a mean age of 67 ± 9 years. A query was performed to determine whether patients completed questionnaires by either automated or manual data collection at baseline and 1 year and 2 years of follow-up after shoulder arthroplasty. In a logistic regression analysis, patient factors (such as demographics, education, and living arrangements) were evaluated for their association with whether patients completed these questionnaires. RESULTS: The proportion of questionnaire completion at baseline, 1 year, and 2 years were 72% (1369 of 1907), 47% (456 of 972), and 33% (128 of 393), respectively. Of the patients who completed their questionnaires, 63% (868 of 1369) did so through automated emails at baseline, 84% (381 of 456) did so at 1 year, and 81% (103 of 128) did so at 2 years. The remainder completed their questionnaires through manual data collection with a research assistant: 37% (501 of 1369) at baseline, 16% (75 of 456) at 1 year, and 19% (25 of 128) at 2 years. After controlling for potentially confounding variables like patient demographics, college education, and living arrangements, women were less likely to complete baseline questionnaires than men (odds ratio 0.78 [95% confidence interval 0.62 to 0.99]; p = 0.04), and white patients (OR 1.6 [95% CI 1.05 to 2.44]; p = 0.03) were more likely than nonwhite patients to have complete baseline questionnaires. At 2 years of follow-up, patients with a college education (OR 2.06 [95% CI 1.14 to 3.71]; p = 0.02), those who lived alone (OR 2.11 [95% CI 1.13 to 3.94]; p = 0.02), and those who had higher baseline Shoulder Activity Scale scores (OR 1.05 [95% CI 1.00 to 1.11]; p = 0.04) were more likely to have complete questionnaires than those without a college education, those who lived with other people, and those with lower SAS scores, respectively. CONCLUSION: The challenges of adopting an online platform include low follow-up proportions and the need for manual assistance by a research assistant to increase patient completion of questionnaires. CLINICAL RELEVANCE: The knowledge of which patient characteristics are associated with a higher likelihood of completing questionnaires has implications for targeted follow-up or representative sampling of the population in a registry. Populations that are less likely to respond may require more effort to reach to prevent exacerbating health outcome disparities. Random sampling with upweighting of hard-to-reach populations may also provide a solution to achieve a representative population of patients undergoing shoulder arthroplasty.

Clinical characteristics and patient-reported outcomes of total shoulder arthroplasty after anterior stabilization: a retrospective matched control study.

BACKGROUND: Instability arthropathy is a known cause of glenohumeral osteoarthritis (OA) among patients with and without prior shoulder stabilization. This study aims to compare the clinical, radiographic, and patient-reported outcome measure (PROM) scores among total shoulder arthroplasty (TSA) patients with and without a history of shoulder stabilization. METHODS: A case-control study was performed comparing 20 patients with a history of anterior shoulder stabilization (11 open, 9 arthroscopic) who underwent TSA to a matched cohort of 20 TSA patients without a history of shoulder surgery (mean follow-up = 2.8 years). Patients were matched by sex, age, and baseline American Shoulder and Elbow Surgeons Standardized Shoulder Assessment Form (ASES) score within 10 points (mean age 59.6 ± 9.6 years). Patient characteristics, operative findings, and preoperative and postoperative radiographic characteristics for both groups were reported. Comparisons were made regarding PROM scores (ASES, 12-Item Short Form Health Survey (SF-12), Shoulder Activity Scale [SAS], numeric rating scale for pain) at baseline, 2 years, and 5 years and patient satisfaction at 2 years. RESULTS: Intraoperative findings of subscapularis scarring or attenuation was common among patients with prior anterior stabilization. The instability cohort did have a higher percentage of B2/B3 glenoid types than the OA cohort (45% vs. 15%), but this was not significantly different possibly because of the small sample size. At 2 years, both instability and OA groups reported significant improvement in pain, function, and activity level. There was no difference between groups on any PROMs or patient satisfaction level. At 5 years, instability patients had significantly lower scores on the ASES and the SF-12 PCS than the OA group. CONCLUSION: There was notable alterations in both soft tissue and bony morphology among patients with prior anterior stabilization. After TSA, both instability and primary OA groups showed significant improvements at 2 years. However, PROMs for instability patients deteriorated at 5 years compared with the control group. Complex bony and soft tissue imbalances may contribute to more unpredictable long-term PROM scores. Thoughtful preoperative consideration of these factors should influence decision making regarding selection of TSA for management of OA in this complex patient cohort.