Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Curr Rev Musculoskelet Med . 2026 Apr 13;19(1):33. doi: 10.1007/s12178-026-10019-w Search in PMC Search in PubMed View in NLM Catalog Add to search Artificial Intelligence and its Current Role in Clinical Outcome Prediction, Musculoskeletal Imaging, and Economic and Ethical Considerations within Orthopedics and Sports Medicine Emmett O’Malley Emmett O’Malley 1 Sports Medicine Institute, Hospital for Special Surgery, New York, NY USA Find articles by Emmett O’Malley 1 , Bryan Soth Bryan Soth 1 Sports Medicine Institute, Hospital for Special Surgery, New York, NY USA Find articles by Bryan Soth 1 , Alex Capitano Alex Capitano 1 Sports Medicine Institute, Hospital for Special Surgery, New York, NY USA Find articles by Alex Capitano 1 , Alessandro Bensa Alessandro Bensa 1 Sports Medicine Institute, Hospital for Special Surgery, New York, NY USA 2 Service of Orthopaedics and Traumatology, Department of Surgery, EOC, Lugano, Switzerland 3 Università della Svizzera Italiana, Faculty of Biomedical Sciences, Lugano, Switzerland Find articles by Alessandro Bensa 1, 2, 3 , Joshua Eskew Joshua Eskew 1 Sports Medicine Institute, Hospital for Special Surgery, New York, NY USA Find articles by Joshua Eskew 1 , Malik Dancy Malik Dancy 1 Sports Medicine Institute, Hospital for Special Surgery, New York, NY USA Find articles by Malik Dancy 1 , Benedict Nwachukwu Benedict Nwachukwu 1 Sports Medicine Institute, Hospital for Special Surgery, New York, NY USA Find articles by Benedict Nwachukwu 1, ✉ Author information Article notes Copyright and License information 1 Sports Medicine Institute, Hospital for Special Surgery, New York, NY USA 2 Service of Orthopaedics and Traumatology, Department of Surgery, EOC, Lugano, Switzerland 3 Università della Svizzera Italiana, Faculty of Biomedical Sciences, Lugano, Switzerland ✉ Corresponding author. Received 2025 Oct 23; Accepted 2026 Feb 20; Collection date 2026 Dec. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2026. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. PMC Copyright notice PMCID: PMC13076830 PMID: 41973309 Abstract Purpose of Review Artificial intelligence (AI) has emerged as a useful tool across the field of orthopedic surgery. This review highlights recent literature on AI’s role in surgical outcome prediction, musculoskeletal imaging, economic and ethical considerations, with a focus on its integration in sports medicine workflow and procedures. Recent Findings Machine learning AI models have demonstrated superior accuracy in predicting orthopedic related patient-reported outcomes, surgical complications, and the utilization of healthcare compared to traditional, non-AI methods. Within imaging, AI applications now produce automated measurements for clinical and presurgical planning with precision equivalent to expert-level measurements. Large language AI models are increasingly used for clinical documentation, research workflows, and administrative support for healthcare delivery and effectiveness. Despite increasing integration of AI into orthopedics and its subspecialties, challenges in validation, accessibility due to cost, and ethical considerations remain. Summary Orthopedic surgery and sports medicine are particularly well suited for AI applications due to their well-defined, measurable clinical outcomes. Emerging AI tools and models show promise in enhancing patient outcomes, surgical planning, and healthcare efficiency. Continued AI research must prioritize external validation, ethical implementation, and educational integration to ensure responsible, effective, and reproducible use. Keywords: Artificial intelligence, Machine learning, Orthopedic surgery, Clinical outcome prediction, Imaging, Sports medicine Introduction Orthopedic surgery is undergoing a technological revolution driven by the integration of artificial intelligence (AI) into diagnostic, surgical, and rehabilitative processes. AI has become a tool capable of enhancing clinical decision making, streamlining workflow efficiency, and personalizing patient care. This transformation parallels the broader digital evolution of orthopedic surgery, as data-driven approaches and robotics are increasingly guiding clinical decision-making and patient outcomes [ 1 – 3 ]. Clinical and preoperative care are among the most active domains for AI integration, with demonstrated success in imaging interpretation and automation [ 4 , 5 ]. For example, in anterior cruciate ligament (ACL) reconstruction, deep learning (DL) AI models can quantify posterior tibial slope from radiographs with surgeon-level accuracy, enabling anatomic risk assessment for the potential risks of graft failure [ 6 ]. Similarly, machine learning (ML) AI models have enhanced prediction of ACL reconstruction revision risk based on patient specificities supporting individualized preoperative planning [ 7 ]. As AI adoption expands within orthopedic surgery, careful evaluation of its capabilities and limitations is essential. The following subsections summarize core applications of AI across orthopedics and the subspeciality of sports medicine by providing in-depth discussion of outcome prediction in upper and lower extremity procedures, imaging automation and quantitative analysis, as well as economic and ethical considerations. This current concept review aims to provide a broad synthesis of contemporary applications and future directions for the clinically meaningful implementation of AI in musculoskeletal care. Core Applications of AI in Sports Medicine Outcome Prediction and Patient Selection AI outcome prediction has been applied across upper and lower extremity sports medicine procedures, enabling joint-specific risk stratification and functional outcome prioritization. ML algorithms trained on clinical and imaging datasets enable predictive models to estimate a patient’s likelihood of achieving functional milestones or experiencing surgical complications in individualized and selected cohorts. These models are particularly valuable for high-volume procedures such as rotator cuff repair, where identifying patients at risk for suboptimal recovery can enhance preoperative counseling and postoperative planning [ 8 , 9 ]. Efforts to standardize AI applications in outcome prediction underscore the need for transparent reporting of AI model training, comprehensive dataset composition, and AI performance metrics to ensure meaningful clinical utility [ 10 ]. Recent AI models enable quantification of individual clinical contributors to injury risk. In professional baseball athletes, these models demonstrated strong predictive performance (AUC > 0.80), with prior injury history and workload-related performance metrics emerging as the dominant predictors compared with traditional single-variable analyses (Fig. 1 ) [ 11 ]. Such insights reinforce the validity of AI outputs and support their use in prehabilitation planning, workload monitoring, and risk stratification. Fig. 1. Open in a new tab Variable importance plot for predicting injury risk in professional athletes. A machine learning (ML) model’s assessment of the most influential factors contributing to future injury risk in Major League Baseball players. Prior injury history and advanced performance metrics, such as weighted cutter runs per 100 pitches and wins above replacement, ranked highest in predictive value. Visualizations like this enhance model interpretability and support the integration of artificial intelligence (AI)-based tools into personalized prehabilitation strategies and athlete risk profiling. Adapted from Karnuta et al. [ 11 ] In the lower extremity, Kunze et al. validated ML algorithms capable of predicting clinically meaningful improvement following hip arthroscopy for femoracetabular impingement (FAI) [ 12 , 13 ]. The algorithms incorporated preoperative variables such as demographics, baseline clinical scores, and interoperative findings to generate individualized outcome estimates [ 14 , 15 ]. Subsequent external validation confirmed the model’s reliability, reinforcing that well-designed algorithms can retain predictive accuracy across institutions and patient demographics [ 15 ]. ML models demonstrated fair to good discrimination for predicting subsequent hip procedures, including need for revision arthroscopy and progression to total hip arthroplasty (Fig. 2 ). Model performance was strongest for predicting progression to total hip arthroplasty, with an area under the curve (AUC) of approximately 0.80, and demonstrated moderate discrimination for revision arthroscopy and periacetabular osteotomy (AUCs ≈ 0.75–0.77) [ 16 ]. This reproducibility is critical for advancing AI predictive tools from single-center studies to large-scale clinical use, where variability in patient populations and practice settings can otherwise compromise validity and reliability. Fig. 2. Open in a new tab Receiver operating characteristic (ROC) curves for predicting subsequent hip surgeries using machine learning (ML). ROC curves display the predictive performance of random forest models developed to forecast the risk of subsequent hip procedures following primary hip arthroscopy for femoroacetabular impingement (FAI). Area under the curve (AUC) values were 0.77 for revision hip arthroscopy, 0.80 for total hip arthroplasty (THA), 0.62 for hip resurfacing arthroplasty (HRA), and 0.76 for periacetabular osteotomy (PAO), reflecting fair to good discriminative accuracy. These models integrate preoperative demographic, radiographic, and functional data to assist in surgical risk stratification and patient counseling Despite this progress, widespread clinical translation has remained limited. Most models demonstrate strong internal validation but lack external validation across diverse populations and practice settings [ 17 ]. The absence of replicability and widespread access constrains their universal utility. Integration into clinical workflows will require emphasis on interpretability and evaluation of bias among clinicians. In sports medicine, where return-to-sport (RTS) and long-term functional success define treatment goals, AI prediction platforms have the potential to become staple components of individualized surgical and non-surgical treatment. As healthcare increasingly transitions towards value-based care, AI-based tools capable of estimating patient satisfaction, functional recovery, and treatment efficacy will become central to optimizing patient care delivery [ 18 ]. These upper and lower extremity applications highlight AI’s capacity to enhance care from outcome prediction to improved patient experience through more accurate patient assessment and treatment. As models continue to evolve, their integration into clinical pathways will expand, resulting in a new era of data-driven, patient-centered care. Imaging and Biometric Analysis AI is transforming the interpretation of orthopedic imaging. DL algorithms now demonstrate high accuracy in recognizing radiographic parameters and alignment abnormalities [ 19 ]. In ligamentous knee injuries, ML can support more informed decision making by standardizing the assessment of imaging-derived risk factors known to influence graft failure risk [ 7 ]. Tasks such as posterior tibial slope quantification, once prone to interobserver variability, can now be standardized through automated algorithms [ 6 ]. In patients with ACL injury, DL models have demonstrated close agreement with expert human measurements, achieving a mean absolute error of approximately 2° while reducing processing time to approximately 5 seconds per image compared with an estimated 180 minutes for manual segmentation and measurement [ 6 ]. The utility of AI also extends to subtle injury detection. In hip preservation surgery, radiographic indices of native anatomy and pathology derived from computed tomography (CT) can be paired with random forest classifiers to predict long-term functional outcomes, illustrating a typical imaging-driven approach to outcome prediction in hip arthroscopy (Fig. 3 ). These methods underscore AI’s impact on musculoskeletal radiology, particularly in high volume or resource-limited environments where automation can reduce time and error. Fig. 3. Open in a new tab Radiographic indices and ML pipeline for predicting hip arthroscopy outcomes. This figure outlines the ML workflow used to evaluate the prognostic value of preoperative CT-based radiographic indices in patients undergoing hip arthroscopy. Radiologic angles were extracted by board-certified radiologists and analyzed using a random forest classifier to predict one- and two-year patient-reported outcomes, including Modified Harris Hip Score (mHHS), Hip Outcome Score – Sport Specific (HOS-SS), Hip Outcome Score - Activities of Daily Living (HOS-ADL), and International Hip Outcome Tool (iHOT-33). The model’s performance was assessed using the AUC, as shown for both timepoints. Adapted from Ramkumar et al. [ 14 ] However, as reliance on algorithmic outputs increases, concerns are warranted regarding clinical oversight and algorithmic bias, particularly when models trained on limited or non-representative datasets are applied across diverse patient populations [ 20 ]. Optimizing widespread deployment will require standardized and protocol-driven approaches for integrating multimodal imaging inputs including magnetic resonance imaging (MRI), CT, and radiographs into unified predictive models [ 21 , 22 ]. For example, a study by Ramkumar et al. used a ML model to predict functional recovery after hip arthroscopy by incorporating imaging and radiographic morphology as potential predictors. The cohort reflected a wide range of morphology across femoral and acetabular parameters, including alpha angle, center-edge angles, version profiles, and combined indices such as the McKibbin and hip impingement indices (Table 1 ). Tables 2 and 3 demonstrate that isolated radiographic indices show limited utility for functional recovery at 1–2 years [ 14 ]. Across sport-specific functions and activities of daily living, odds ratios remain close to null with non-significant associations, and discrimination is consistently modest, with AUC values clustering near 0.50–0.57 across most parameters [ 14 ]. Specifically, alpha angle and coronal center-edge angle both demonstrated odds ratios near 1.0 and AUCs near 0.50–0.55 at one- and two-year follow-up [ 14 ]. Table 1. Baseline patient characteristics for machine learning (ML) cohort undergoing hip arthroscopy. This table outlines demographic, laterality, and radiographic characteristics for 1,735 patients who underwent hip arthroscopy. Radiographic angles include alpha, beta, neck shaft, acetabular version, and femoral version. These data were used as features in ML models predicting post-operative outcomes. Adapted from Ramkumar et al. [ 14 ] Patient characteristics α Ethnicity, n (%) American Indian/Alaska Native 1 (0.1) Asian 28 (1.6) Black/African American 30 (1.7) Hispanic/Latino 7 (0.4) Mixed/Undetermined 311 (17.9) White 1358 (78.3) Laterality, n (%) Left 792 (45.6) Right 943 (54.4) Sex, n (%) Female 864 (49.8) Male 871 (50.2) Age, Y 31.1 ± 10.1 Body mass index, kg/m 2 24.3 ± 4.0 Alpha angle, deg 63.6 ± 11.7 Beta angle, deg 50.6 ± 15.5 Sagittal center-edge angle, deg 57.1 ± 10.1 Coronal center-edge angle, deg 32.8 ± 9.2 Neck shaft angle, deg 132 ± 9.3 Acetabular version angle, deg 1 o’clock 2.7 ± 8.6 2 o’clock 10.9 ± 8.6 3 o’clock 16.5 ± 6.6 Femoral version angle, deg 13.9 ± 11.1 McKibbin index 30.3 ± 13.1 Hip impingement index −49.7 ± 16.0 Ethnicity, n (%) 1 (0.1) Open in a new tab α Data are shown as mean ± standard deviation unless otherwise indicated Table 2. Change in sport function (HOS-SS) by radiographic subgroup. Mean improvement in HOS-ADL scores is displayed across various anatomical groupings (e.g., alpha angle ≥ 55° vs. < 55°). No statistically significant differences were observed, suggesting radiographic measures do not strongly correlate with recovery in daily activities. Adapted from Ramkumar et al. [ 14 ] Relationship between radiographic indices and the HOS-SS Score at 1- and 2-Years’ Follow-up α Radiographic parameter Odds ratio (95% CI) P Value AUC (95% CI) Alpha Angle 1 year 0.9944 (0.9803–1.0084) 0.433 0.515 (0.468–0.562) 2 years 1.0077 (0.9853–1.0304) 0.5 0.540 (0.463–0.617) Coronal center-edge angle 1 year 0.9981 (0.9798–1.0185) 0.842 0.500 (0.453–0.548) 2 years 0.9650 (0.9285–1.0005) 0.068 0.569 (0.490–0.648) Femoral Version Angle 1 year 1.004 (0.989–1.020) 0.595 0.510 (0.464–0.557) 2 years 1.016 (0.996–1.037) 0.126 0.544 (0.485–0.602) McKibbin index 1 year 1.004 (0.991–1.017) 0.51 0.517 (0.471–0.563) 2 years 1.008 (0.993–1.025) 0.322 0.546 (0.486–0.607) Hip impingement index 1 year 1.006 (0.996–1.016) 0.246 0.520 (0.473–0.567) 2 years 1.003 (0.991–1.016) 0.628 0.516 (0.455–0.577) Open in a new tab α AUC, area under the curve; CI, confidence interval; HOS-SS, Hip Outcomes Score-Sport Specific Table 3. Change in daily living function (HOS-ADL) by radiographic subgroup. This table presents mean changes in HOS-S scores following hip arthroscopy, stratified by radiographic parameters. As with HOS-ADL, minimal variation between groups highlights the limited prognostic value of static anatomical indices. Adapted from Ramkumar et al. [ 14 ] Relationship between radiographic indices and the HOS-ADL Score at 1- and 2-Years’ Follow-up α Radiographic parameter Odds Ratio (95% CI) P Value AUC (95% CI) Alpha Angle 1 year 0.9945 (0.9807–1.0083) 0.435 0.487 (0.441–0.533) 2 years 1.0070 (0.9849–1.0293) 0.535 0.538 (0.462–0.613) Coronal center-edge angle 1 year 0.9990 (0.9810–1.0192) 0.919 0.502 (0.456–0.548) 2 years 0.9645 (0.9283–1.0005) 0.063 0.568 (0.490–0.646) Femoral Version Angle 1 year 1.001 (0.985–1.013) 0.873 0.504 (0.461–0.546) 2 years 1.007 (0.991–1.025) 0.391 0.489 (0.435–0.543) McKibbin index 1 year 1.003 (0.991–1.015) 0.65 0.509 (0.466–0.551) 2 years 1.012 (0.998–1.027) 0.119 0.538 (0.484–0.591) Hip impingement index 1 year 1.009 (0.999–1.018) 0.078 0.540 (0.497–0.583) 2 years 1.002 (0.991–1.014) 0.702 0.516 (0.462–0.570) Open in a new tab α AUC, area under the curve; CI, confidence interval; HOS-ADL, hip outcomes score-sport specific These findings suggest that commonly cited anatomic indices alone do not meaningfully stratify postoperative functional improvement and should not be used as sole inputs for patient selection or counseling. Instead, they support a more clinically useful paradigm where radiographic morphology is used within multivariable models that incorporate baseline function, intraoperative findings, and patient-level factors of recovery. This shifts AI-driven patient selection from morphology-only prediction on imaging to multidimensional prognostication [ 14 ]. Machine Learning and Time- and Cost-Effective Care Delivery AI also offers a powerful means to enhance value-based care through cost prediction and improved operational efficiency. ML models consistently outperform traditional heuristics and clinician judgment in forecasting surgical costs, discharge disposition, and inpatient resource utilization [ 9 ]. These predictive tools can identify high-cost drivers preoperatively. This allows providers and payers to anticipate factors such as prolonged operative time, inpatient admission, or post-acute care needs and allocate resources accordingly [ 9 ]. A simplified decision tree (Fig. 4 ) illustrates how individual clinical inputs are processed to generate a cost estimate. Fig. 4. Open in a new tab Decision tree example for predicting cost following anterior cruciate ligament (ACL) reconstruction A decision tree used by a ML model to predict average total charges for ACL reconstruction based on key perioperative variables. The model first stratifies cases by anesthesia type and operative time, with green nodes representing lower-cost predictions. Such tools can facilitate real-time decision-making and cost forecasting in surgical planning. Adapted from Lu et al. [ 23 ]. Models predicting clinically meaningful improvement also support value-driven care pathways. For example, the Orthopedic Sports Medicine and Shoulder Outcomes (OSSO) algorithm predicted attainment of the minimal clinically important difference (MCID) following hip arthroscopy with similar performance in the development and external validation cohorts, with c-statistics of 0.77 and 0.80, respectively (Table 4 ). Such reproducibility enhances clinical confidence and supports health system-wide implementation for consistent outcome tracking. Table 4. Comparative performance of the OSSO machine learning (ML) algorithm in development and external validation cohorts. This table presents a side-by-side comparison of performance metrics for the OSSO minimal clinically important difference (MCID) prediction model in patients undergoing hip arthroscopy. The model demonstrated strong generalizability, with similar c-statistics, calibration slopes, and Brier scores across both the original internal validation and the external validation cohorts. Decision curve analysis revealed greater net clinical benefit compared with other management strategies. Adapted from Kunze et al. [ 12 ] Comparative performance of the OSSO ML minimal clinically important difference prediction algorithm between original development study and current external validation study α Performance measure Development/Internal validation cohort External validation cohort c-statistic 0.77 (0.69–0.84) 0.80 (0.71–0.87) Calibration slope 1.22 (0.79–1.65) 1.16 (0.74–1.61) Calibration intercept 0.07 (− 0.27–0.41) 0.13 (− 0.26–0.53) Brier Score 0.14 (0.12–0.17) 0.15 (0.12–0.18) Null model Brier Score 0.18 0.2 Decision curve analysis Greater net benefit than other management strategies Greater net benefit than other management strategies Open in a new tab Values are presented as mean (95% confidence interval). c-statistic, concordance statistic; ML, machine learning; OSSO, Orthopedic Sports Medicine and Shoulder Outcomes Procedure-specific economic modeling has likewise gained traction. AI-based frameworks can integrate clinical, radiographic, and demographic data to predict the financial burden of surgeries such as ACL reconstruction [ 23 ]. This capability is particularly relevant as bundled payments and outcome-linked reimbursement models become standard across the practices of orthopedic surgery. Emerging Technologies and Large Language Models Large language models (LLMs) have recently emerged as tools capable of advancing musculoskeletal workflow. These systems, such as chat-based LLMs, can draft clinical notes, simplify medical instructions, and clarify highly complex information into patient-accessible language, thereby improving patient adherence and satisfaction [ 24 – 26 ]. However, while LLMs offer a promise of streamlined communication, recent reviews indicate that their performance remains inconsistent. Issues such as hallucinated responses, lack of source transparency, and poor reproducibility raise concerns about reliability in clinical environments [ 27 – 29 ]. As a result, an integral component to ensure their safe and effective utilization is developing solid study designs, openly using the training data, and providing domain-specific fine-tuning approaches. By processing natural language inputs at scale, these systems can automate complex tasks such as literature review and synthesis, and protocol drafting. Different LLM model types have diverse input–output modalities and functional capabilities. Appropriate alignment between model capabilities and intended use may help reduce clinician burden and improve consistency in data-driven care pathways (Table 5 ) [ 29 ]. In practice, inadequate alignment between model capability and intended use has contributed to variable methodological quality in the literature, with systematic reviews identifying redundancy, limited rigor, and insufficient customization across healthcare applications [ 28 ]. These findings emphasize the need for purposeful model design, transparent evaluation, and orthopedic specific modeling prior to clinical implementation. Table 5. Model concepts and functions of basic large language models (LLMs). Input-output modalities and core functions of LLM types, highlighting distinctions between task-based models and foundational systems. Adapted from Kunze et al. [ 29 ] Patient characteristics α Model type Model functions Basic Text-text Use natural language processing to make text-prediction based on language input. Example: asking to finish a sentence. Text-image Use training from large sets of images captioned with short text descriptions to generate new image from text input. One such method of accomplishing this task is diffusion. Example: asking model to generate picture of a hip replacement Text-Video Generate video outputs that correspond to text input. Example: asking models to generate videos of driving a car. Text-3D output Generate 3D objects that correspond to text input description. Example: Asking models to find and navigate to kitchen within 3D representation of a house. Text-Task Models trained to perform specific tasks or actions based on text input. Examples: answering question, fix non-functioning code, or generate a grocery list. Advanced Foundational LLM Expanded output capabilities, including question answering, sentiment analysis, object recognition, information extraction, image generation, and instruction following. Both inputs and outputs are a combination of the basic models. Open in a new tab 3D, 3-dimensional; LLM, large language model Effective use of LLMs in orthopedics depends not only on algorithmic performance, but also on clinician literacy and judgment. As AI platforms become more accessible, orthopedic surgeons must learn to interpret, validate and deploy output responsibly to ensure alignment with the most responsible patient care. Educational reform is recognized as essential to improving AI fluency among trainees and practitioners [ 25 , 30 ]. Retrieval augmented generation (RAG) models have demonstrated superior evidence synthesis compared with general purpose LLMs, underscoring the value of tailored customization [ 27 ]. In parallel, natural language processing (NLP) has become an important modality for extracting structured insights from unstructured clinical text. By mining electronic medical records (EMR), NLP algorithms can identify trends, stratify patient risk, and facilitate registry creations. These functions are otherwise limited by human bandwidth and variability [ 31 ]. The implications for research efficiency, quality assurance, and personalized care delivery are established, particularly as orthopedic institutions seek to build scalable data infrastructure. When viewed collectively, the implementation of AI across imaging, prediction, and administration provides clear support for its general utility in orthopedics. Whether by improving diagnostic precision, optimizing resource allocation, or supporting communication, AI has the potential to drive a new era of efficient, personalized musculoskeletal care. In both academic and community healthcare settings, the collective impact of AI suggests a unifying benefit: higher-quality musculoskeletal care delivered more efficiently and at a lower cost. Ethical, Educational, and Global Considerations The integration of AI into orthopedic surgery introduces complex ethical, educational, and equity-based challenges that need to be addressed to ensure responsible implementation. One of the central concerns is premature clinical adoption of unvalidated AI tools, a phenomenon termed the “valley of despair”, where enthusiasm surpasses its actionable benefit [ 32 ]. To mitigate this, bioethical frameworks are being developed to guide AI deployment with principles centered on transparency, equity, and safety [ 33 , 34 ]. These frameworks advocate for complete model testing, clear documentation of limitations, and mechanisms to ensure human oversight remains central to the decision-making process. From an educational standpoint, there is growing recognition that AI competency should be integrated into orthopedic training at every level. Recent surveys of orthopedic trainees reveal significant variability in AI knowledge, highlighting the need for structured exposure to these technologies early in surgical education [ 35 ]. Without a foundational understanding, clinicians risk overreliance on or misinterpretation of AI-generated outputs. Editorial voices have underscored the danger of deploying AI tools without sufficient user competency, equating these actions to the introduction of unregulated instruments into the operating room [ 36 – 38 ]. Multimodal LLMs can process and summarize radiologic findings, a capability that underscores both their potential utility and the need for user training and oversight (Fig. 5 ). Fig. 5. Open in a new tab Multimodal large language model (LLM) for medical imaging interpretation. This diagram illustrates how a multimodal LLM integrates radiographic imaging with text-based prompts to generate clinical findings. A pretrained vision transformer extracts features from the image (in this case, a chest radiograph), which are then processed through a linear transformation layer and interpreted by a medically trained LLM. The model outputs a structured diagnostic report, showcasing the potential of LLMs to enhance decision-making through image-text alignment. Adapted from Kunze et al. [ 29 ] Global equity remains another important consideration. Most existing AI models are trained and validated on data from high-income countries limiting their applicability in resource constrained regions risking digital health disparities. Ensuring fairness will require inclusion of diverse patient populations and open access international datasets to promote equitable outcomes [ 39 , 40 ]. Ultimately, the promise of AI in orthopedics depends not only on technological advancement, but also on adherence to ethical principles, comprehensive education, and global inclusivity. These pillars are essential for transforming AI from a theoretical innovation to a trusted partner in musculoskeletal care delivery. Future Directions and Conclusion AI continues to evolve in musculoskeletal medicine offering opportunities to enhance diagnostic accuracy, improve outcome prediction, and streamline care delivery and costs. As adoption progresses, the field must transition from proof-of-concept models to sustained clinical integration. The importance of robust model design, external validation, and ethical deployment across a broad range of orthopedic subspecialties has been clearly demonstrated by the contributions of physician-scientists thus far [ 24 , 29 , 31 , 41 , 42 ]. Ongoing translational scholarship will be essential in bridging the gap between data science and bedside application. Despite significant progress, many existing algorithms remain limited by small training cohorts and inconsistent reporting standards. To achieve clinical utility, future efforts must prioritize interpretability and workflow compatibility, ensuring the AI tools enhance rather than complicate clinical decision making. As with any paradigm-shifting technology, enthusiasm should be balanced with critical appraisal and a commitment to patient centered care. Priority areas for future development include: Prospective, multicenter validation of ML and DL algorithms across diverse patient populations to ensure external validity and reduce algorithmic bias. Standardized transparent reporting of model architecture and performance metrics, including calibration curves, confusion matrices, and AUC values, to support reproducibility and clinician confidence. Development of formal orthopedic-specific educational initiatives and programs to improve AI fluency among trainees and practicing surgeons, ensuring responsible clinical use. Integration of validated AI tools into real-time EMR systems and surgical planning platforms, facilitating decision support and reducing the cognitive burden of data management. The next chapter of orthopedic innovation will depend on the ability to validate, educate, and integrate AI responsibly. If these priorities are met through multi-center collaboration, policy support, and academic rigor, AI will be well positioned to lead into a new era of orthopedics. This new era will be defined by clinical precision, personalization, and measurable improvements in healthcare and patient outcomes. Key References Ramkumar PN, Karnuta JM, Haeberle HS, Sullivan SW, Nawabi DH, Ranawat AS, et al. Radiographic Indices Are Not Predictive of Clinical Outcomes Among 1735 Patients Indicated for Hip Arthroscopic Surgery: A Machine Learning Analysis. Am J Sports Med. 2020;48(12):2910–8. https://doi.org/10.1177/0363546520950743 . ⚬ This large multicenter study of 1735 hip arthroscopy patients was among the first to apply machine learning to outcome prediction, highlighting both the promise of AI-driven prognostic models and the limitations of traditional radiographic indices in hip preservation surgery. Kunze KN, Polce EM, Clapp I, Nwachukwu BU, Chahla J, Nho SJ. Machine Learning Algorithms Predict Functional Improvement After Hip Arthroscopy for Femoroacetabular Impingement Syndrome in Athletes. J Bone Joint Surg Am. 2021;103(12):1055–62. https://doi.org/10.2106/jbjs.20.01640 . ⚬ This landmark study developed and internally validated one of the first supervised ML models to predict clinically significant improvement after hip arthroscopy, establishing a reproducible framework for AI-based outcome prediction in hip preservation. Karnuta JM, Luu BC, Haeberle HS, Saluan PM, Frangiamore SJ, Stearns KL, et al. Machine Learning Outperforms Regression Analysis to Predict Next-Season Major League Baseball Player Injuries: Epidemiology and Validation of 13,982 Player-Years From Performance and Injury Profile Trends, 2000-2017. Orthop J Sports Med. 2020;8(11):2325967120963046. https://doi.org/10.1177/2325967120963046 . ⚬ This pivotal study applied a machine learning model to a large dataset of professional baseball players (13,982 player-seasons), demonstrating that AI significantly outperforms traditional regression in predicting athlete injuries. It underscores the real-world predictive capacity of AI in sports medicine. Lu Y, Pareek A, Yang L, Rouzrokh P, Khosravi B, Okoroha KR, et al. Deep Learning Artificial Intelligence Tool for Automated Radiographic Determination of Posterior Tibial Slope in Patients With ACL Injury. Orthop J Sports Med. 2023;11(12):23259671231215820. https://doi.org/10.1177/23259671231215820 . ⚬ This work introduced a deep learning AI tool that automatically measures posterior tibial slope on knee radiographs, achieving high accuracy. It exemplifies how AI can enhance orthopedic imaging by enabling efficient, automated extraction of quantitative anatomical metrics relevant to ACL injury risk assessment. Woo JJ, Yang AJ, Olsen RJ, Hasan SS, Nawabi DH, Nwachukwu BU, et al. Custom Large Language Models Improve Accuracy: Comparing Retrieval Augmented Generation and Artificial Intelligence Agents to Noncustom Models for Evidence-Based Medicine. Arthroscopy. 2025;41(3):565–73.e6. https://doi.org/10.1016/j.arthro.2024.10.042 . ⚬ This recent investigation evaluated custom large language models (LLMs) for evidence-based medicine in orthopedics, showing that domain-specific, retrieval-augmented LLMs significantly improve accuracy in literature synthesis compared to standard models. The study highlights the transformative potential of tailored AI agents to enhance clinical decision support and research in musculoskeletal care. Author contributions E.O. and B.S. wrote the main manuscript text. E.O. prepared all figures.All authors reviewed and edited the manuscript. Data Availability No datasets were generated or analysed during the current study. Declarations Competing interests The authors declare no competing interests. Disclosure statement Statement: Below are the healthcare industry relationships reported by Dr. Nwachukwu as of July 8, 2025. BICMD - Co-founder, Ownership Interest Figur8 - Advisory Board HSS West Side ASC - Ownership Interest Medbridge- Advisory Board Quantum - Advisory Board Stryker Corporation - Consultant Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References 1. Myers TG, Ramkumar PN, Ricciardi BF, Urish KL, Kipper J, Ketonis C. Artificial intelligence and orthopaedics: an introduction for clinicians. J Bone Joint Surg Am. 2020;102(9):830–40. 10.2106/jbjs.19.01128. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Padash S, Mickley JP, Vera Garcia DV, Nugen F, Khosravi B, Erickson BJ, et al. An overview of machine learning in orthopedic surgery: an educational paper. J Arthroplasty. 2023;38(10):1938–42. 10.1016/j.arth.2023.08.043. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Girod MM, Saniei S, Ulrich MN, Bukowiec LG, Mulford KL, Taunton MJ, et al. Artificial intelligence in the diagnosis and prognostication of the musculoskeletal patient. HSS Journal®: The Musculoskeletal Journal of Hospital for Special Surgery. 2025;15563316251339660. 10.1177/15563316251339660. [ Google Scholar ] 4. Atik O. Artificial intelligence, machine learning, and deep learning in orthopedic surgery. Jt Dis Relat Surg. 2022;33(2):484–5. 10.52312/jdrs.2022.57906. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Wang VM, Cheung CA, Kozar AJ, Huang B. Machine learning applications in orthopaedic imaging. J Am Acad Orthop Surg. 2020;28(10):e415-7. 10.5435/jaaos-d-19-00688. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Lu Y, Pareek A, Yang L, Rouzrokh P, Khosravi B, Okoroha KR, et al. Deep learning artificial intelligence tool for automated radiographic determination of posterior tibial slope in patients with ACL injury. Orthop J Sports Med. 2023;11(12):23259671231215820. 10.1177/23259671231215820. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Vasavada K, Vasavada V, Moran J, Devana S, Lee C, Hame SL, et al. A novel machine learning model to predict revision ACL reconstruction failure in the MARS cohort. Orthop J Sports Med. 2024;12(11):23259671241291920. 10.1177/23259671241291920. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Alaiti RK, Vallio CS, Assunção JH, de Andra ESFB, Gracitelli MEC, Neto AAF, et al. Using machine learning to predict nonachievement of clinically significant outcomes after rotator cuff repair. Orthop J Sports Med. 2023;11(10):23259671231206180. 10.1177/23259671231206180. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Kumar V, Roche C, Overman S, Simovitch R, Flurin PH, Wright T, et al. What is the accuracy of three different machine learning techniques to predict clinical outcomes after shoulder arthroplasty? Clin Orthop Relat Res. 2020;478(10):2351–63. 10.1097/corr.0000000000001263. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Jang SJ, Rosenstadt J, Lee E, Kunze KN. Artificial intelligence for clinically meaningful outcome prediction in orthopedic research: current applications and limitations. Curr Rev Musculoskelet Med. 2024;17(6):185–206. 10.1007/s12178-024-09893-z. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 11. Karnuta JM, Luu BC, Haeberle HS, Saluan PM, Frangiamore SJ, Stearns KL, et al. Machine learning outperforms regression analysis to predict next-season Major League Baseball player injuries: epidemiology and validation of 13,982 player-years from performance and injury profile trends, 2000–2017. Orthop J Sports Med. 2020;8(11):2325967120963046. 10.1177/2325967120963046. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Kunze KN, Kaidi A, Madjarova S, Polce EM, Ranawat AS, Nawabi DH, et al. External validation of a machine learning algorithm for predicting clinically meaningful functional improvement after arthroscopic hip preservation surgery. Am J Sports Med. 2022;50(13):3593–9. 10.1177/03635465221124275. [ DOI ] [ PubMed ] [ Google Scholar ] 13. Kunze KN, Polce EM, Clapp I, Nwachukwu BU, Chahla J, Nho SJ. Machine learning algorithms predict functional improvement after hip arthroscopy for femoroacetabular impingement syndrome in athletes. J Bone Joint Surg Am. 2021;103(12):1055–62. 10.2106/jbjs.20.01640. [ DOI ] [ PubMed ] [ Google Scholar ] 14. Ramkumar PN, Karnuta JM, Haeberle HS, Sullivan SW, Nawabi DH, Ranawat AS, et al. Radiographic indices are not predictive of clinical outcomes among 1735 patients indicated for hip arthroscopic surgery: a machine learning analysis. Am J Sports Med. 2020;48(12):2910–8. 10.1177/0363546520950743. [ DOI ] [ PubMed ] [ Google Scholar ] 15. Kunze KN, Polce EM, Nwachukwu BU, Chahla J, Nho SJ. Development and internal validation of supervised machine learning algorithms for predicting clinically significant functional improvement in a mixed population of primary hip arthroscopy. Arthroscopy. 2021;37(5):1488–97. 10.1016/j.arthro.2021.01.005. [ DOI ] [ PubMed ] [ Google Scholar ] 16. Haeberle HS, Ramkumar PN, Karnuta JM, Sullivan S, Sink EL, Kelly BT, et al. Predicting the risk of subsequent hip surgery before primary hip arthroscopy for femoroacetabular impingement syndrome: a machine learning analysis of preoperative risk factors in hip preservation. Am J Sports Med. 2021;49(10):2668–76. 10.1177/03635465211024964. [ DOI ] [ PubMed ] [ Google Scholar ] 17. Gupta P, Haeberle HS, Zimmer ZR, Levine WN, Williams RJ, Ramkumar PN. Artificial intelligence-based applications in shoulder surgery leaves much to be desired: a systematic review. JSES Rev Rep Tech. 2023;3(2):189–200. 10.1016/j.xrrt.2022.12.006. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 18. Balch JA, Chatham AH, Hong PKW, Manganiello L, Baskaran N, Bihorac A, et al. Predicting patient reported outcome measures: a scoping review for the artificial intelligence-guided patient preference predictor. Front Artif Intell. 2024;7:1477447. 10.3389/frai.2024.1477447. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Velasquez Garcia A, Bukowiec LG, Yang L, Nishikawa H, Fitzsimmons JS, Larson AN, et al. Artificial intelligence-based three-dimensional templating for total joint arthroplasty planning: a scoping review. Int Orthop. 2024;48(4):997–1010. 10.1007/s00264-024-06088-6. [ DOI ] [ PubMed ] [ Google Scholar ] 20. De Simone B, Deeken G, Catena F. Balancing ethics and innovation: can artificial intelligence safely transform emergency surgery? A narrative perspective. J Clin Med. 2025. 10.3390/jcm14093111. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Ko S, Pareek A, Ro DH, Lu Y, Camp CL, Martin RK, et al. Artificial intelligence in orthopedics: three strategies for deep learning with orthopedic specific imaging. Knee Surg Sports Traumatol Arthrosc. 2022;30(3):758–61. 10.1007/s00167-021-06838-8. [ DOI ] [ PubMed ] [ Google Scholar ] 22. Bozzo A, Tsui JMG, Bhatnagar S, Forsberg J. Deep learning and multimodal artificial intelligence in orthopaedic surgery. J Am Acad Orthop Surg. 2024;32(11):e523-32. 10.5435/jaaos-d-23-00831. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Lu Y, Kunze K, Cohn MR, Lavoie-Gagne O, Polce E, Nwachukwu BU, et al. Artificial intelligence predicts cost after ambulatory anterior cruciate ligament reconstruction. Arthrosc Sports Med Rehabil. 2021;3(6):e2033–45. 10.1016/j.asmr.2021.10.013. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Lavoie-Gagne O, Woo JJ, Williams RJ 3rd, Nwachukwu BU, Kunze KN, Ramkumar PN. Artificial intelligence as a tool to mitigate administrative burden, optimize billing, reduce insurance- and credentialing-related expenses, and improve quality assurance within health care systems. Arthroscopy. 2025;41(8):3270–5. 10.1016/j.arthro.2025.02.038. 25. Giorgino R, Alessandri-Bonetti M, Luca A, Migliorini F, Rossi N, Peretti GM, et al. ChatGPT in orthopedics: a narrative review exploring the potential of artificial intelligence in orthopedic practice. Front Surg. 2023;10:1284015. 10.3389/fsurg.2023.1284015. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Kunze KN. Artificial intelligence and digital applications in musculoskeletal healthcare: ready or not, here it comes! HSS Journal®: The Musculoskeletal Journal of Hospital for Special Surgery. 2025;15563316251341314. 10.1177/15563316251341314. [ Google Scholar ] 27. Woo JJ, Yang AJ, Olsen RJ, Hasan SS, Nawabi DH, Nwachukwu BU, et al. Custom large language models improve accuracy: comparing retrieval augmented generation and artificial intelligence agents to noncustom models for evidence-based medicine. Arthroscopy. 2025;41(3):565-73.e6. 10.1016/j.arthro.2024.10.042. [ DOI ] [ PubMed ] [ Google Scholar ] 28. Kunze KN, Gerhold C, Dave U, Abunnur N, Mamonov A, Nwachukwu BU, et al. Large language model use cases in health care research are redundant and often lack appropriate methodological conduct: a scoping review and call for improved practices. Arthroscopy. 2025;41(11):4928-45.e2. 10.1016/j.arthro.2025.03.066. [ DOI ] [ PubMed ] [ Google Scholar ] 29. Kunze KN, Nwachukwu BU, Cote MP, Ramkumar PN. Large language models applied to health care tasks may improve clinical efficiency, value of care rendered, research, and medical education. Arthroscopy. 2025;41(3):547–56. 10.1016/j.arthro.2024.12.010. [ DOI ] [ PubMed ] [ Google Scholar ] 30. Lisacek-Kiosoglous AB, Powling AS, Fontalis A, Gabr A, Mazomenos E, Haddad FS. Artificial intelligence in orthopaedic surgery. Bone Joint Res. 2023;12(7):447–54. 10.1302/2046-3758.127.Bjr-2023-0111.R1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Pruneski JA, Pareek A, Nwachukwu BU, Martin RK, Kelly BT, Karlsson J, et al. Natural language processing: using artificial intelligence to understand human language in orthopedics. Knee Surg Sports Traumatol Arthrosc. 2023;31(4):1203–11. 10.1007/s00167-022-07272-0. [ DOI ] [ PubMed ] [ Google Scholar ] 32. Wellington IJ, Cote MP. Editorial commentary: machine learning in orthopaedics: venturing into the valley of despair. Arthroscopy. 2022;38(9):2767–8. 10.1016/j.arthro.2022.05.010. [ DOI ] [ PubMed ] [ Google Scholar ] 33. Kiwinda LV, Kocher SD, Bryniarski AR, Pean CA. Bioethical considerations of deploying artificial intelligence in clinical orthopedic settings: a narrative review. HSS Journal®: The Musculoskeletal Journal of Hospital for Special Surgery. 2025;15563316251340303. 10.1177/15563316251340303. [ Google Scholar ] 34. Karaibrahimoglu A, İnce F, Hassanzadeh G, Alizadeh A, Bagheri K, Yucel I, et al. Ethical considerations in telehealth and artificial intelligence for work related musculoskeletal disorders: a scoping review. Work. 2024;79(3):1577–88. 10.3233/wor-240187. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Pawelczyk J, Kraus M, Eckl L, Nehrer S, Aurich M, Izadpanah K, et al. Attitude of aspiring orthopaedic surgeons towards artificial intelligence: a multinational cross-sectional survey study. Arch Orthop Trauma Surg. 2024;144(8):3541–52. 10.1007/s00402-024-05408-0. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Shang Z, Chauhan V, Devi K, Patil S. Artificial intelligence, the digital surgeon: unravelling its emerging footprint in healthcare - the narrative review. J Multidiscip Healthc. 2024;17:4011–22. 10.2147/jmdh.S482757. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Mavrogenis AF, Hernigou P, Scarlat MM. Artificial intelligence, natural stupidity or artificial stupidity: who is today the winner in orthopaedics? What is true and what is fraud? What legal barriers exist for scientific writing? Int Orthop. 2024;48(3):617–23. 10.1007/s00264-024-06102-x. [ DOI ] [ PubMed ] [ Google Scholar ] 38. Leopold SS, Haddad FS, Sandell LJ, Swiontkowski M. Artificial intelligence applications and scholarly publication in orthopedic surgery. J Orthop Res. 2023;41(6):1137–8. 10.1002/jor.25566. [ DOI ] [ PubMed ] [ Google Scholar ] 39. Banatwala U, Ibrahim MT, Shaikh RH, Shahzad H, Hoodbhoy Z, Noordin S. A comprehensive exploration of artificial intelligence in orthopaedics within lower-middle-income countries: a narrative review. J Pak Med Assoc. 2024;74(4):S90–6. 10.47391/jpma.Aku-9s-14. [ DOI ] [ PubMed ] [ Google Scholar ] 40. Murphy MP, Brown NM. Corr synthesis: when should the orthopaedic surgeon use artificial intelligence, machine learning, and deep learning? Clin Orthop Relat Res. 2021;479(7):1497–505. 10.1097/corr.0000000000001679. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Pruneski JA, Williams RJ 3rd, Nwachukwu BU, Ramkumar PN, Kiapour AM, Martin RK, et al. The development and deployment of machine learning models. Knee Surg Sports Traumatol Arthrosc. 2022;30(12):3917–23. 10.1007/s00167-022-07155-4. 42. Ramkumar PN, Luu BC, Haeberle HS, Karnuta JM, Nwachukwu BU, Williams RJ. Sports medicine and artificial intelligence: a primer. Am J Sports Med. 2022;50(4):1166–74. 10.1177/03635465211008648. [ DOI ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement No datasets were generated or analysed during the current study. Articles from Current Reviews in Musculoskeletal Medicine are provided here courtesy of Humana Press ACTIONS View on publisher site PDF (1.4 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top