Abstract
Purpose of Review
With the unprecedented advancement of data aggregation and deep learning algorithms, artificial intelligence (AI) and machine learning (ML) are poised to transform the practice of medicine. The field of orthopedics, in particular, is uniquely suited to harness the power of big data, and in doing so provide critical insight into elevating the many facets of care provided by orthopedic surgeons. The purpose of this review is to critically evaluate the recent and novel literature regarding ML in the field of orthopedics and to address its potential impact on the future of musculoskeletal care.
Recent Findings
Recent literature demonstrates that the incorporation of ML into orthopedics has the potential to elevate patient care through alternative patient-specific payment models, rapidly analyze imaging modalities, and remotely monitor patients.
Summary
Just as the business of medicine was once considered outside the domain of the orthopedic surgeon, we report evidence that demonstrates these emerging applications of AI warrant ownership, leverage, and application by the orthopedic surgeon to better serve their patients and deliver optimal, value-based care.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
References
Papers of particular interest, published recently, have been highlighted as: • Of importance •• Of major importance
• Bini SA. Artificial intelligence, machine learning, deep learning, and cognitive computing: what do these terms mean and how will they impact health care? J Arthroplast. 2018;33(8):2358–61. https://doi.org/10.1016/j.arth.2018.02.067The purpose of this article was to discuss the fundamentals of artificial intelligence so that practicing surgeons may better understand how and where to apply it. The article touches on the origins of artificial intelligence and its progression into machine learning, as well as the development of artificial neural networks and deep learning algorithms. It provides a foundation for knowledge of machine learning and its use in healthcare.
Maxmen JS. The post-physician era : medicine in the twenty-first century. Hoboken: Wiley; 1976.
Naylor CD. On the prospects for a (deep) learning health care system. JAMA. 2018;320(11):1099–100. https://doi.org/10.1001/jama.2018.11103.
Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nat Med. https://doi.org/10.1038/s41591-018-0300-7.
•• Haeberle HS, Helm JM, Navarro SM, et al. Artificial intelligence and machine learning in lower extremity arthroplasty: a review. J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.05.055The purpose of this review was to (1) summarize and review the most recent applications of artificial intelligence and machine learning–specific to lower extremity arthroplasty, (2) discuss the origins and model-specifics of machine learning, and (3) examine the progression of machine learning into healthcare. This review specifically examines osteoarthritis gait models, joint-specific imaging, and value-based payment models. It is one of the few reviews to look at the advancement and application of machine learning within the field of orthopaedics.
•• Ramkumar PN, Karnuta JM, Navarro SM, et al. Preoperative prediction of value metrics and a patient-specific payment model for primary total hip arthroplasty: development and validation of a deep learning model. J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.04.055The purpose of this study was to (1) test an artificial neural network capable of predicting variables such as length of stay, inpatient charges, and discharge disposition for total hip arthroplasty, and (2) create a patient-specific payment model using this artificial neural network to account for patient complexity. Using 78,335 primary total hip arthroplasty cases, the model demonstrated area under the curves of 82.0%, 83.4%, and 79.4% for the three variables. The patient-specific payment model established risk increases of 2.5%, 8.9%, and 17.3% for moderate, major, and severe comorbidities. This study validated the use of machine learning in its prediction of patient-centered outcomes and tiering of payments based on case complexity.
•• Karnuta JM, Navarro SM, Haeberle HS, et al. Predicting inpatient payments prior to lower extremity arthroplasty using deep learning: which model architecture is best? J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.05.048The purpose of this study was to compare two deep learning models with different regularization techniques in predicting outcomes for primary total hip arthroplasty and total knee arthroplasty. Using 295,605 patients, the study demonstrated an improvement in performance when the DenseNet model was used with regularization. This study established an important set of features when using deep learning to more accurately predict inpatient surgical costs, and provided a foundation for other studies seeking to use machine learning.
Milinovich A, Kattan MW. Extracting and utilizing electronic health data from epic for research. Ann Transl Med. 2018;6(3):42. https://doi.org/10.21037/atm.2018.01.13.
•• Ramkumar PN, Haeberle HS, Bloomfield MR, et al. Artificial intelligence and arthroplasty at a single institution: real-world applications of machine learning to big data, value-based care, mobile health, and remote patient monitoring. J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.06.018The purpose of this review was to discuss the objectives of Cleveland Clinic’s newly created Machine Learning Arthroplasty Laboratory: patient-specific, value-based care, and human movement. This review highlights an important step in the application of machine learning in order to improve patient outcomes in the field of orthopaedics.
•• Ramkumar PN, Navarro SM, Haeberle HS, et al. Development and validation of a machine learning algorithm after primary total hip arthroplasty: applications to length of stay and payment models. J Arthroplast. 2019;34(4):632–7. https://doi.org/10.1016/j.arth.2018.12.030The purpose of this study was to (1) develop and validate a machine learning algorithm to predict length of stay and patient-specific inpatient payments after primary total hip arthroplasty, and (2) to propose a risk-adjusted patient-specific payment model taking patient comorbidity into account. Applying 122,334 patients to the model, the algorithm used race, age, gender, and comobidity scores to demonstrate area under the curve’s of 0.87 and 0.71 for length of stay and payment. As patient complexity increased, predicting payment increased in tiers of 3%, 12%, and 32% for moderate, major, and extreme comorbidities, respectively. This study demonstrated the potential of using a risk-based patient-specific payment model that offers reimbursement commensurate with case complexity.
•• Navarro SM, Wang EY, Haeberle HS, et al. Machine learning and primary total knee arthroplasty: patient forecasting for a patient-specific payment Model. J Arthroplast. 2018;(12):33, 3617–3623. https://doi.org/10.1016/j.arth.2018.08.028The purpose of this study was to (1) develop a machine learning algorithm to predict length of stay and inpatient costs after total knee arthroplasty, and (2) to propose a tiered patient-specific payment model commensurate with case complexity for reimbursement. Using 141,446 patients, a Bayesian model was created and trained to demonstrate an area under the curve of 0.7822 and 0.7382 for length of stay and cost. Cost ad-ons increased in teirs of 3%, 10%, and 15% for moderate, major, and extreme morality risks, respectively. This study supported the use of machine learning in predicting length of stay and cost in total knee arthroplasty, and demonstrated the poential of this application in the creation of a patient-specific payment model.
•• Karnuta JM, Navarro SM, Haeberle HS, Billow DG, Krebs VE, Ramkumar PN. Bundled care for hip fractures: a machine-learning approach to an untenable patient-specific payment model. J Orthop Trauma. 2019;33(7):324–30. https://doi.org/10.1097/BOT.0000000000001454The purpose of this study was to (1) develop a supervised naïve Bayes machine learning algorithm to pedict length of stay and cost after hip fracture, and (2) propose a patient-specific payment model to adjust reimbursement based on patient comborbidities. Studying 98,562 Medicare patients who underwent operative management for hip fracture, a model was built that used age, sex, ethnicity, race, type of admission, risk of mortality, and severity of illness to demonstrate an accuracy of 76.5% and 79.0% for length of stay and cost. Performance was 88% for length of stay and 89% for cost. Analysis showed increasing model error with increasing risk of mortality, thus lending validity to increases in risk-adjusted payment for each risk of mortality. This study concluded that bundled care is an implausible payment model for hip fractures due to the cost of delivery of hip fracture care being dependent on non-modifiable patient-specific factors.
• Bini SA, Shah R, Bendich I, Patterson J, Hwang K, Zaid M. Machine learning algorithms can use wearable sensor data to accurately predict 6-week patient reported outcome scores following joint replacement in a prospective trial. J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.07.024The aim of this study was to demonstrate the feasibility of combining machine learning with patient-generated health data in total joint arthroplasty to predict patient-reported outcome measures. The pilot study used 22 patients and used 3 activity trackers to collect 35 features from 4 weeks before to 6 weeks following surgery. Of the 22, 15 patients completed the study and 3 million data points were collected. The machine learning algorithm grouped patients into 3 clusters predictive of 6-week patient-reported outcome measure results. This study served to prove the concept that machine learning can be used in combination with patient-generated health data to predict patient-reported outcome measures following surgery.
•• Cabitza F, Locoro A, Banfi G. Machine learning in orthopedics: a literature review. Front Bioeng Biotechnol. 2018;6. https://doi.org/10.3389/fbioe.2018.00075The purpose of this review was to provide a systematic literature review of articles published in the last two decades in which the application of machine learning was described in relation to an orthopaedic problem or pupose. The content of 70 articles was screened and analyzed in order to outline the articles’ content in terms of main machine learning technique used, the orthopaedic application domain, and the source of data and quality of their predictive performance. This review is critical as it serves as one of the first to examine the use of artificial intelligence and machine learning in the field of orthopaedics specifically.
Thong W, Parent S, Wu J, Aubin C-E, Labelle H, Kadoury S. Three-dimensional morphology study of surgical adolescent idiopathic scoliosis patient from encoded geometric models. Eur Spine J. 2016;25(10):3104–13. https://doi.org/10.1007/s00586-016-4426-3.
Olczak J, Fahlberg N, Maki A, Razavian AS, Jilert A, Stark A, et al. Artificial intelligence for analyzing orthopedic trauma radiographs. Acta Orthop. 2017;88(6):581–6. https://doi.org/10.1080/17453674.2017.1344459.
Artificial Intelligence and Machine Learning Across the Care Continuum in Lower Extremity Arthroplasty. Cleveland Clinic. https://consultqd.clevelandclinic.org/artificial-intelligence-and-machine-learning-across-the-care-continuum-in-lower-extremity-arthroplasty/. Accessed 16 Aug 2019.
Shah RF, Martinez AM, Pedoia V, Majumdar S, Vail T, Bini S. Variation in the thickness of knee cartilage. The use of a novel machine learning algorithm for cartilage segmentation of MRI images. J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.07.022.
• Kruse C, Eiken P, Vestergaard P. Machine learning principles can improve hip fracture prediction. Calcif Tissue Int. 2017;100(4):348–60. https://doi.org/10.1007/s00223-017-0238-7The purpose of this study was to apply machine learning principles in the prediction of hip fractures and estimate the predictive importance of dual-energy X-ray absorptiometry. Data from 4722 women and 717 men with 5 years of follow-up was collected and 24 statistical models were built, In women, the bootstrap aggregated flexible discriminant analysis performed best with an area under the curve of 0.91. For men, eXtreme Gradient Boosting performed best with an area under the curve of 0.89. This study’s importance is highlighted by its conclusion that machine learning can improve fracture prediction beyond logistic regression using ensemble methods.
• Kuo CY, Yu LC, Chen HC, Chan CL. Comparison of models for the prediction of medical costs of spinal fusion in Taiwan diagnosis-related groups by machine learning algorithms. Healthc Inform Res. 2018;24(1):29–37. https://doi.org/10.4258/hir.2018.24.1.29The aims of this study were (1) to compare the performance of machine learning methods for the prediction of medical costs associated with spinal fusion in terms of profit or loss, and (2) to apply these methods to explore the factors associated with the medical costs of spinal fusion. Through the categorization of 532 cases, the random forest method was found to be the most accurate predictive model, achieving an accuracy of 84.30%, a sensitivity of 71.4%, and a specificity of 92.2%, and an area under the curve of 0.904. This study demonstrated the utility of machine learning in the prediction of medical costs, a potential tool to inform hospital strategy in terms of increasing the efficiency of financial management.
•• Ramkumar PN, Karnuta JM, Navarro SM, et al. Deep learning preoperatively predicts value metrics for primary total knee arthroplasty: development and validation of an artificial neural network model. J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.05.034The purpose of this study was to (1) develop and validate an artificial neural network to learn and predict length of stay, inpatient charges, and discharge disposition in patients undergoing total knee arthroplasty, and (2) to apply the artificial neural network to propose a risk-based, patient-specific payment model commensurate with case complexity. Using 175,042 cases, a neural network was developed that demonstrated an area under the curve of 74.8%, 82.8%, and 76.1% for length of stay, charges, and discharge disposition. The subsequent patient-specific payment model demonstrated an increase in risk of 2.0%, 21.8%, and 82.6% as patient comorbidity increased from moderate to major to severe. This study demonstrated a model that is able to be applied to patient-specific payment models when tiering reimbursement based upon case complexity.
AAHKS. 2019 Medicare physician fee schedule rule includes potentially misvalued codes for THA/TKA. http://www.aahks.org/newsroom/press-releases/2019-medicare-physician-fee-schedule-rule-includes-potentially-misvalued-codes-for-tha-tka/. Accessed 25 Nov 2018.
Learmonth ID, Young C, Rorabeck C. The operation of the century: total hip replacement. Lancet. 2007;370(9597):1508–19. https://doi.org/10.1016/S0140-6736(07)60457-7.
• Karhade AV, Schwab JH, Bedair HS. Development of machine learning algorithms for prediction of sustained postoperative opioid prescriptions after total hip arthroplasty. J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.06.013The purpose of this study was to develop machine learning algorithms for the preoperative prediction of prolonged opioid prescriptions after total hip arthroplasty. Reviewing the records of 5507 patients, five machine learning algorithms were developed and determined the factors for prolonged addiction to be age, duration of opioid exposure, preoperative hemoglobin, and preoperative medications. The importance of this study is that it demontrates a utility of machine learning that can be used to improve preoperative screening and prediction of patients at risk for prolonged postoperative opioid prescriptions.
•• Ramkumar PN, Haeberle HS, Ramanathan D, et al. Remote patient monitoring using mobile health for total knee arthroplasty: validation of a wearable and machine learning–based surveillance platform. J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.05.021The purpose of this study was to validate the feasibility of remote patient monitoring systems in terms of the frequency of data interruptions and patient acceptance. Using a pilot cohort of 25 patients undergoing total knee arthroplasty, a mobile application and wearable knee sleeve was used to report data for mobility, knee range of motion, patient-reported outcome measures, opioid usage, home exercise program, and compliance. Of the 25 patients, 100% had uninterrupted passive data collection, establishing the ability to remotely acquire continuous data in postoperative patients.
Bloomfield RA, Williams HA, Broberg JS, Lanting BA, McIsaac KA, Teeter MG. Machine learning groups patients by early functional improvement likelihood based on wearable sensor instrumented preoperative timed-up-and-go tests. J Arthroplast. 2019. https://doi.org/10.1016/j.arth.2019.05.061.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of Interest
J. Matthew Helm, Andrew M. Sweirgosz, Heather S. Haeberle, and Jaret M. Karnuta report no conflicts of interest.
Viktor E. Krebs reports royalties and consultancy fees from Stryker outside the submitted work.
Prem N. Ramkumar reports royalties and consultancy fees from Focus Ventures outside the submitted work.
Jonathan L. Schaffer reports royalties and consultancy fees from Zimmer Biomet, outside the submitted work.
Andrew I. Spitzer reports consultancy fees from Flexion Therapeutics Inc., Medical Device Business Services Inc., FIDIA Pharma USA Inc., and Sanofi-Aventis USA LLC, outside the submitted work.
Human and Animal Rights and Informed Consent
This article does not contain any studies with human or animal subjects performed by the authors.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article is part of the Topical Collection on The Use of Technology in Orthopaedic Surgery—Intraoperative and Post-Operative Management
Rights and permissions
About this article
Cite this article
Helm, J.M., Swiergosz, A.M., Haeberle, H.S. et al. Machine Learning and Artificial Intelligence: Definitions, Applications, and Future Directions. Curr Rev Musculoskelet Med 13, 69–76 (2020). https://doi.org/10.1007/s12178-020-09600-8
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12178-020-09600-8