Machine Learning Can Predict Level of Improvement in Shoulder Arthroplasty.

The ability to accurately predict postoperative outcomes is of considerable interest in the field of orthopaedic surgery. Machine learning has been used as a form of predictive modeling in multiple health-care settings. The purpose of the current study was to determine whether machine learning algorithms using preoperative data can predict improvement in American Shoulder and Elbow Surgeons (ASES) scores for patients with glenohumeral osteoarthritis (OA) at a minimum of 2 years after shoulder arthroplasty. METHODS: This was a retrospective cohort study that included 472 patients (472 shoulders) diagnosed with primary glenohumeral OA (mean age, 68 years; 56% male) treated with shoulder arthroplasty (431 anatomic total shoulder arthroplasty and 41 reverse total shoulder arthroplasty). Preoperative computed tomography (CT) scans were used to classify patients on the basis of glenoid and rotator cuff morphology. Preoperative and final postoperative ASES scores were used to assess the level of improvement. Patients were separated into 3 improvement ranges of approximately equal size. Machine learning methods that related patterns of these variables to outcome ranges were employed. Three modeling approaches were compared: a model with the use of all baseline variables (Model 1), a model omitting morphological variables (Model 2), and a model omitting ASES variables (Model 3). RESULTS: Improvement ranges of ≤28 points (class A), 29 to 55 points (class B), and >55 points (class C) were established. Using all follow-up time intervals, Model 1 gave the most accurate predictions, with probability values of 0.94, 0.95, and 0.94 for classes A, B, and C, respectively. This was followed by Model 2 (0.93, 0.80, and 0.73) and Model 3 (0.77, 0.72, and 0.71). CONCLUSIONS: Machine learning can accurately predict the level of improvement after shoulder arthroplasty for glenohumeral OA. This may allow physicians to improve patient satisfaction by better managing expectations. These predictions were most accurate when latent variables were combined with morphological variables, suggesting that both patients' perceptions and structural pathology are critical to optimizing outcomes in shoulder arthroplasty. LEVEL OF EVIDENCE: Therapeutic Level IV. See Instructions for Authors for a complete description of levels of evidence.

Improving Efficiency of Patient-Reported Outcome Collection: Application of Computerized Adaptive Testing to DASH and QuickDASH Outcome Scores.

PURPOSE: Patient-reported outcome measures assess health status and treatment outcomes in orthopedic care, but they may burden patients with lengthy questionnaires. Predictive models using machine learning, known as computerized adaptive testing (CAT), offer a potential solution. This study evaluates the ability of CAT to improve efficiency of the 30-item Disabilities of the Arm, Shoulder, and Hand (DASH) and 11-item QuickDASH questionnaires. METHODS: A total of 2,860 DASH and 27,355 QuickDASH respondents were included in the analysis. The CAT system was retrospectively applied to each set of patient responses stored on the instrument to calculate a CAT-specific score for all DASH and QuickDASH entries. The accuracy of the CAT scores, viewed in the context of the minimal clinically important difference for both patient-reported outcome measures (DASH, 12; QuickDASH, 9), was determined through descriptive statistics, Pearson correlation coefficient, intraclass correlation coefficient, and distribution of scores and score differences. RESULTS: The CAT model required an average of 15.3 questions to be answered for the DASH and 5.8 questions for the QuickDASH, representing a 49% and 47% decrease in question burden, respectively. Mean CAT score was the same for DASH and 0.1 points lower for QuickDASH with similar SDs (DASH, 12.9 ± 19.8 vs 12.9 ± 19.9; QuickDASH, 32.7 ± 24.7 vs 32.6 ± 24.6). Pearson coefficients (DASH, 0.99; QuickDASH, 0.98) and intraclass correlation coefficients (DASH, 1.0; QuickDASH, 0.98) indicated strong agreement between scores. The difference between the CAT and full score was less than the minimal clinically important difference in 99% of cases for DASH and approximately 95% of cases for QuickDASH. CONCLUSIONS: The application of CAT to DASH and QuickDASH surveys demonstrated an ability to lessen the response burden with negligible effect on score integrity. CLINICAL RELEVANCE: In the case of DASH and QuickDASH, CAT is an appropriate alternative to full questionnaire implementation for patient outcome score collection.

A concise shoulder outcome measure: application of computerized adaptive testing to the American Shoulder and Elbow Surgeons Shoulder Assessment.

BACKGROUND: Patient-reported outcome measures enable quantitative and patient-centric assessment of orthopedic interventions; however, increased use of these forms has an associated burden for patients and practices. We examined the utility of a computerized adaptive testing (CAT) method to reduce the number of questions on the American Shoulder and Elbow Surgeons (ASES) instrument. METHODS: A previously developed ASES CAT system was applied to the responses of 2763 patients who underwent shoulder evaluation and treatment and had answered all questions on the full ASES instrument. Analyses to assess the accuracy of the CAT score in replicating the full-form score included the mean and standard deviation of both groups of scores, frequency distributions of the 2 sets of scores and score differences, Pearson and intraclass correlation coefficients, and Bland-Altman assessment of patterns in score differences. RESULTS: By tailoring questions according to prior responses, CAT reduced the question burden by 40%. The mean difference between CAT and full ASES scores was -0.14, and the scores were within 5 points in 95% of cases (a 12-point difference is considered the threshold for clinical significance) and were clustered around zero. The correlation coefficients were 0.99, and the frequency distributions of the CAT and full ASES scores were nearly identical. The differences between scores were independent of the overall score, and no significant bias for CAT scores was found in either a positive or negative direction. CONCLUSION: The ASES CAT system lessens respondent burden with a negligible effect on score integrity.