Building large-scale registries from unstructured clinical notes using a low-resource natural language processing pipeline.

Building clinical registries is an important step in clinical research and improvement of patient care quality. Natural Language Processing (NLP) methods have shown promising results in extracting valuable information from unstructured clinical notes. However, the structure and nature of clinical notes are very different from regular text that state-of-the-art NLP models are trained and tested on, and they have their own set of challenges. In this study, we propose Sentence Extractor with Keywords (SE-K), an efficient and interpretable classification approach for extracting information from clinical notes and show that it outperforms more computationally expensive methods in text classification. Following the Institutional Review Board (IRB) approval, we used SE-K and two embedding based NLP approaches (Sentence Extractor with Embeddings (SE-E) and Bidirectional Encoder Representations from Transformers (BERT)) to develop comprehensive registry of anterior cruciate ligament surgeries from 20 years of unstructured clinical data at a multi-site tertiary-care regional children's hospital. The low-resource approach (SE-K) had better performance (average AUROC of 0.94 ± 0.04) than the embedding-based approaches (SE-E: 0.93 ± 0.04 and BERT: 0.87 ± 0.09) for out of sample validation, in addition to minimum performance drop between test and out-of-sample validation. Moreover, the SE-K approach was at least six times faster (on CPU) than SE-E (on CPU) and BERT (on GPU) and provides interpretability. Our proposed approach, SE-K, can be effectively used to extract relevant variables from clinic notes to build large-scale registries, with consistently better performance compared to the more resource-intensive approaches (e.g., BERT). Such approaches can facilitate information extraction from unstructured notes for registry building, quality improvement and adverse event monitoring.

Systematic evaluation of common natural language processing techniques to codify clinical notes.

Proper codification of medical diagnoses and procedures is essential for optimized health care management, quality improvement, research, and reimbursement tasks within large healthcare systems. Assignment of diagnostic or procedure codes is a tedious manual process, often prone to human error. Natural Language Processing (NLP) has been suggested to facilitate this manual codification process. Yet, little is known on best practices to utilize NLP for such applications. With Large Language Models (LLMs) becoming more ubiquitous in daily life, it is critical to remember, not every task requires that level of resource and effort. Here we comprehensively assessed the performance of common NLP techniques to predict current procedural terminology (CPT) from operative notes. CPT codes are commonly used to track surgical procedures and interventions and are the primary means for reimbursement. Our analysis of 100 most common musculoskeletal CPT codes suggest that traditional approaches can outperform more resource intensive approaches like BERT significantly (P-value = 4.4e-17) with average AUROC of 0.96 and accuracy of 0.97, in addition to providing interpretability which can be very helpful and even crucial in the clinical domain. We also proposed a complexity measure to quantify the complexity of a classification task and how this measure could influence the effect of dataset size on model's performance. Finally, we provide preliminary evidence that NLP can help minimize the codification error, including mislabeling due to human error.

Prevalence and Predictors of Concomitant Meniscal Surgery During Pediatric and Adolescent ACL Reconstruction: Analysis of 4729 Patients Over 20 Years at a Tertiary-Care Regional Children's Hospital.

Background: The rate of concomitant meniscal procedures performed in conjunction with anterior cruciate ligament (ACL) reconstruction is increasing. Few studies have examined these procedures in high-risk pediatric cohorts. Hypotheses: That (1) the rates of meniscal repair compared with meniscectomy would increase throughout the study period and (2) patient-related factors would be able to predict the type of meniscal operation, which would differ according to age. Study Design: Cohort study (prevalence); Level of evidence, 2. Methods: Natural language processing was used to extract clinical variables from notes of patients who underwent ACL reconstruction between 2000 and 2020 at a single institution. Patients were stratified to pediatric (5-13 years) and adolescent (14-19 years) cohorts. Linear regression was used to evaluate changes in the prevalence of concomitant meniscal surgery during the study period. Logistic regression was used to determine predictors of the need for and type of meniscal procedure. Results: Of 4729 patients (mean age, 16 ± 2 years; 54.7% female) identified, 2458 patients (52%) underwent concomitant meniscal procedures (55% repair rate). The prevalence of lateral meniscal (LM) procedures increased in both pediatric and adolescent cohorts, whereas the prevalence of medial meniscal (MM) repair increased in the adolescent cohort (P = .02). In the adolescent cohort, older age was predictive of concomitant medial meniscectomy (P = .031). In the pediatric cohort, female sex was predictive of concomitant MM surgery and of undergoing lateral meniscectomy versus repair (P≤ .029). Female sex was associated with decreased odds of concomitant LM surgery in both cohorts (P≤ .018). Revision ACLR was predictive of concomitant MM surgery and of meniscectomy (medial and lateral) in the adolescent cohort (P < .001). Higher body mass index was associated with increased odds of undergoing medial meniscectomy versus repair in the pediatric cohort (P = .03). Conclusion: More than half of the young patients who underwent ACLR had meniscal pathology warranting surgical intervention. The prevalence of MM repair compared with meniscectomy in adolescents increased throughout the study period. Patients who underwent revision ACLR were more likely to undergo concomitant meniscal surgeries, which were more often meniscectomy. Female sex had mixed effects in both the pediatric and adolescent cohorts.

Editorial Commentary: Artificial Intelligence Models Using Machine Learning Can Improve Preoperative Identification of Subscapularis Pathology.

Subscapularis pathology is difficult to diagnose, in part because of decreased sensitivity and accuracy in identifying tears with magnetic resonance imaging (MRI) when compared to other cuff tendons. Artificial intelligence evaluation of patient physical examination and MRI data using a machine learning model shows that arthroscopically confirmed partial- or full-thickness subscapularis tears are highly associated with abnormal subscapularis tendon length, long head of the biceps tears, and subscapularis fatty atrophy, and on physical examination, with weakness with internal rotation and positive lift-off, belly press, and bear hug tests. Today, physicians may use machine learning as a tool, but this model may not currently be sufficient to drastically change practice. However, with continued research and development, which is occurring rapidly, similar models could aid physicians in timely identification of pathology and optimization of preoperative planning, as well as physician training and education.

Prevalence and Predictors of Concomitant Meniscal and Ligamentous Injuries Associated With ACL Surgery: An Analysis of 20 Years of ACL Reconstruction at a Tertiary Care Children's Hospital.

BACKGROUND: There is an increasing rate of procedures being performed for concomitant injuries during anterior cruciate ligament (ACL) surgery. Few studies have examined risk factors for these associated injuries in young patients. HYPOTHESIS: There are patient-related factors predictive of concomitant knee pathology that differ between age-based cohorts. STUDY DESIGN: Cross-sectional study; Level of evidence, 3. METHODS: Natural language processing was used to extract clinical variables from available notes of patients undergoing ACL surgery between 2000 and 2020 at a single institution (5174 ACL surgeries; mean age, 17 ± 4 years; 53.1% female; accuracy, >98%). Patients were stratified to pediatric (5-13 years), adolescent (14-19 years), and young adult (20-35 years) cohorts. Logistic regression was used to determine predictors of concomitant injury to the menisci, medial collateral ligament (MCL), posterolateral corner (PLC), and posterior cruciate ligament (PCL). RESULTS: Between 2000 and 2020, 54% of pediatric, 71% of adolescent, and 70% of adult patients had ≥1 concomitant soft tissue injury. In children and adolescents, increased age was consistently predictive of sustaining a concomitant injury (P < .02). Female children had increased odds of concomitant medial meniscal injury, while female adults had decreased odds (P≤ .046). Adolescent and adult female patients had decreased odds of concomitant lateral meniscal injury (P≤ .027). Female children had increased odds of injury to the MCL (P = .015), whereas female children and adolescents had decreased odds of PCL injury (P≤ .044). Adolescents undergoing revision ACL surgery had increased odds of meniscal injury (P≤ .001) and decreased odds of concomitant MCL injury (P = .028). Increased body mass index (BMI) was associated with increased odds of concomitant medial meniscal injury in all cohorts (P≤ .041), lateral meniscal injury in adults (P = .045), and PLC injury in children (P = .016). Contact injuries were associated with increased odds of MCL injury in adolescents (P = .017) and PLC injury in adolescents and adults (P < .014). CONCLUSION: These findings support the hypothesis, as there were multiple factors that significantly affected the risk of concomitant injuries that differed between cohorts. Increased age, BMI, and contact injury history were generally associated with increased odds of sustaining a concomitant injury, whereas female sex and revision ACL surgery had mixed effects. Further studies are essential to investigate the sex-based differences in risk for concomitant injuries and to develop tailored treatment plans that minimize the risk of secondary ACL injury.

Autologous Osteochondral Transfer of the Knee Demonstrates Continued High Rates of Return to Sport and Low Rates of Conversion to Arthroplasty at Long-Term Follow-Up: A Systematic Review.

PURPOSE: To perform a systematic review of the literature to evaluate (1) activity level and knee function, (2) reoperation and failure rates, and (3) risk factors for reoperation and failure of autologous osteochondral transfer (AOT) at long-term follow-up. METHODS: A comprehensive review of the long-term outcomes of AOT was performed. Studies reported on activity-based outcomes (Tegner Activity Scale) and clinical outcomes (Lysholm score and International Knee Documentation Committee score). Reoperation and failure rates as defined by the publishing authors were recorded for each study. Modified Coleman Methodology Scores were calculated to assess study methodological quality. RESULTS: Twelve studies with a total of 495 patients and an average age of 32.5 years at the time of surgery and a mean follow-up of 15.1 years (range, 10.4-18.0 years) were included. The mean defect size was 3.2 cm2 (range, 1.9-6.9 cm2). The mean duration of symptoms before surgery was 5.1 years. Return to sport rates ranged from 86% to 100%. Conversion to arthroplasty rates ranged from 0% to 16%. The average preoperative International Knee Documentation Committee scores ranged from 32.9 to 36.8, and the average postoperative International Knee Documentation Committee scores at final follow-up ranged from 66.3 to 77.3. The average preoperative Lysholm scores ranged from 44.5 to 56.0 and the average postoperative Lysholm scores ranged from 70.0 to 96.5. The average preoperative Tegner scores ranged from 2.5 to 3.0, and the average postoperative scores ranged from 4.1 to 7.0. CONCLUSIONS: AOT of the knee resulted in high rates of return to sport with correspondingly low rates of conversion to arthroplasty at long-term follow-up. In addition, AOT demonstrated significant improvements in long-term patient-reported outcomes from baseline. LEVEL OF EVIDENCE: Level IV, systematic review of Level I-IV studies.

Acetabular Rim Disorders: Epidemiology, Etiology, Management, and Outcomes.

¼ Anatomic disorders of the acetabular rim are a common, correctable source of hip pain in younger patients.¼ Some common conditions of involving abnormal acetabular rim morphology include developmental dysplasia of the hip, pincer-type femoroacetabular impingement, acetabular protrusion, and acetabular retroversion.¼ Treatment option for these conditions were historically limited to open osteotomy and osteoplasty procedures; however, there is increasing use of arthroscopic intervention for these patients.¼ Arthroscopic intervention has demonstrated short-term success in a variety of focal acetabular rim disorders; however, further research is needed to determine the long-term outcomes of these procedures and their utility in more global pathology.

American Orthopaedic Foot and Ankle Society Annual Meeting All-in-person Attendance Results in Immense Carbon Expenditure.

BACKGROUND: Professional society conferences are integral to the medical profession. However, airline travel is a major contributor to greenhouse gas production, and the environmental impact of in-person attendance at an orthopaedic conference has yet to be described. With growing concern about the climate crisis, we sought to quantify the carbon footprint of in-person attendance to help potential attendees more consciously consider in-person attendance, inform strategies to minimize greenhouse gas emissions during travel to annual meetings, and increase awareness about and momentum for efforts in orthopaedic surgery to reduce the carbon footprint of society conferences. QUESTIONS/PURPOSES: (1) What was the magnitude of greenhouse gas production resulting from all-in-person 2019 American Orthopaedic Foot and Ankle Society (AOFAS) annual meeting attendance in Chicago, IL, USA? (2) What was the magnitude of greenhouse gas production resulting from the all-virtual 2020 AOFAS annual meeting, and how does it compare with the 2019 AOFAS annual meeting carbon footprint? (3) To what extent could an alternative in-person meeting model with four or seven hubs decrease greenhouse gas production resulting from round-trip air travel compared with the 2019 AOFAS annual meeting? METHODS: A list of the postal codes and countries of all 1271 registered participants attending the four-day 2019 AOFAS annual meeting in Chicago, IL, USA, was obtained from AOFAS headquarters. The 2019 conference was chosen because it was the last pre-COVID meeting and thus attendance was more likely to resemble that at prepandemic in-person conferences than more recent meetings because of pandemic travel restrictions. We estimated carbon dioxide-equivalent (CO 2 e) production from round-trip air travel using a publicly available internet-based calculator (Myclimate: https://co2.myclimate.org/en/flight_calculators/new ). Emissions produced by the conference venue, car travel, and hotel stays were estimated using published Environmental Protection Agency emission factors. To estimate emissions produced by the all-virtual 2020 AOFAS annual meeting (assuming an equal number of attendees as in 2019), we used the framework published by Faber and summed estimated network data transfer emissions, personal computer and monitor emissions, and server-related emissions. Using the 2019 registrant list, we modeled four-hub and seven-hub in-person meeting alternatives to determine potential decreased round-trip air travel greenhouse gas production. Meeting hub locations were selected by visualizing the geographic distribution of the 2019 registrants and selecting reasonable meeting locations that would minimize air travel for the greatest number of attendees. Registrants were assigned to the nearest hub location. Myclimate was again used to estimate CO 2 e production for round-trip air travel for the hub meeting models. RESULTS: The total estimated emissions of the all-in-person 2019 AOFAS annual meeting (when accounting for travel, conference space, and hotel stays) was 1565 tons CO 2 e (median 0.61 tons per attendee, range 0.02 to 7.7 tons). The total estimated emissions of the all-virtual 2020 meeting (when accounting for network data transfer emissions, personal computer and monitor emissions, and server-related emissions) was 34 tons CO 2 e (median 0.03 tons per attendee). This corresponds to a 97.8% decrease in CO 2 e emissions compared with the in-person conference. The model of a four-hub in-person meeting alternative with meetings in Chicago, Santiago, London, and Tokyo predicted an estimated 54% decrease in CO 2 e emissions from round-trip air travel. The seven-hub meeting model with meetings in Chicago; Washington, DC; Dallas; Los Angeles; Santiago; London; and Tokyo was predicted to diminish the CO 2 e emissions of round-trip air travel by an estimated 71%. CONCLUSION: The 2019 AOFAS annual meeting had an enormous carbon footprint and resulted in many individuals exceeding their annual allotted carbon budget (2.5 tons) according to the Paris Agreement. Hosting the meeting virtually greatly reduced the annual meeting carbon footprint, and our hub-based meeting models identified potential in-person alternatives for reducing the carbon footprint of conference attendance. CLINICAL RELEVANCE: Professional societies must consider our responsibility to decarbonizing the healthcare sector by considering innovative approaches-perhaps such as our multihub proposals-to decarbonize carbon-intensive annual meetings without stalling academic progress.

Survival analyses and their applications in orthopaedics.

Survival analyses are a powerful statistical tool used to analyse data when the outcome of interest involves the time until an event. There is an array of models fit for this goal; however, there are subtle differences in assumptions, as well as a number of pitfalls, that can lead to biased results if researchers are unaware of the subtleties. As larger amounts of data become available, and more survival analyses are published every year, it is important that healthcare professionals understand how to evaluate these models and apply them into their practice. Therefore, the purpose of this study was to present an overview of survival analyses, including required assumptions and important pitfalls, as well as examples of their use within orthopaedic surgery.

Natural language processing in radiology: Clinical applications and future directions.

Natural language processing (NLP) is a wide range of techniques that allows computers to interact with human text. Applications of NLP in everyday life include language translation aids, chat bots, and text prediction. It has been increasingly utilized in the medical field with increased reliance on electronic health records. As findings in radiology are primarily communicated via text, the field is particularly suited to benefit from NLP based applications. Furthermore, rapidly increasing imaging volume will continue to increase burden on clinicians, emphasizing the need for improvements in workflow. In this article, we highlight the numerous non-clinical, provider focused, and patient focused applications of NLP in radiology. We also comment on challenges associated with development and incorporation of NLP based applications in radiology as well as potential future directions.

Supervised machine learning and associated algorithms: applications in orthopedic surgery.

Supervised learning is the most common form of machine learning utilized in medical research. It is used to predict outcomes of interest or classify positive and/or negative cases with a known ground truth. Supervised learning describes a spectrum of techniques, ranging from traditional regression modeling to more complex tree boosting, which are becoming increasingly prevalent as the focus on "big data" develops. While these tools are becoming increasingly popular and powerful, there is a paucity of literature available that describe the strengths and limitations of these different modeling techniques. Typically, there is no formal training for health care professionals in the use of machine learning models. As machine learning applications throughout medicine increase, it is important that physicians and other health care professionals better understand the processes underlying application of these techniques. The purpose of this study is to provide an overview of commonly used supervised learning techniques with recent case examples within the orthopedic literature. An additional goal is to address disparities in the understanding of these methods to improve communication within and between research teams.

Natural language processing: using artificial intelligence to understand human language in orthopedics.

Natural language processing (NLP) describes the broad field of artificial intelligence by which computers are trained to understand and generate human language. Within healthcare research, NLP is commonly used for variable extraction and classification/cohort identification tasks. While these tools are becoming increasingly popular and available as both open-source and commercial products, there is a paucity of the literature within the orthopedic space describing the key tasks within these powerful pipelines. Curation and navigation of the electronic medical record are becoming increasingly onerous, and it is important for physicians and other healthcare professionals to understand potential methods of harnessing this large data resource. The purpose of this study is to provide an overview of the tasks required to develop an NLP pipeline for orthopedic research and present recent examples of successful implementations.

The development and deployment of machine learning models.

Applications of artificial intelligence, specifically machine learning, are becoming increasingly popular in Orthopaedic Surgery, and medicine as a whole. This growing interest is shared by data scientists and physicians alike. However, there is an asymmetry of understanding of the developmental process and potential applications of machine learning. As new technology will undoubtedly affect clinical practice in the coming years, it is important for physicians to understand how these processes work. The purpose of this paper is to provide clarity and a general framework for building and assessing machine learning models.

Changes in Anatomic Risk Factors for Patellar Instability During Skeletal Growth and Maturation.

BACKGROUND: Several anatomic risk factors have been identified in the pathogenesis of patellofemoral instability. The literature is sparce regarding how these anatomic risk factors change during skeletal growth and development. HYPOTHESIS: The anatomic risk factors associated with patellar instability change significantly during skeletal growth and maturation with different patterns in male versus female patients. STUDY DESIGN: Cross-sectional study; Level of evidence, 3. METHODS: Magnetic resonance imaging data from 240 unique, asymptomatic knees (7-18 years of age; 50% female) were used to measure patellar height (Caton-Deschamps index), lateral patellar tilt angle, trochlear height, trochlear groove depth, trochlear sulcus angle, and tibial tubercle-trochlear groove (TT-TG) distance. Linear regression was used to test the associations between age and anatomic findings. Two-way analysis of variance with Holm-Sídák post hoc test was used to compare anatomic characteristics between sexes in 3 age groups: prepubertal school-aged children (7-10 years old), early adolescents (11-14 years old), and late adolescents (15-18 years old). RESULTS: Patellar height (female sex), lateral patellar tilt angle (male sex), and trochlear sulcus angle (both sexes) decreased with age (P < .001). Trochlear height, depth, and TT-TG distance increased with age in both male and female participants (P < .02). Male participants had a larger sulcus angle (by 5.3°± 1.6° at age 11-14 years) and greater trochlear height (by >5 mm across medial, central, and lateral regions at age 15-18 years) than age-matched female participants (P < .01). We found no other sex-related differences in quantified anatomic features (P > .1). CONCLUSION: The findings partially support our hypothesis indicating significant age-related changes in all quantified features, which were not different between male and female participants except for trochlear sulcus angle in early adolescence and trochlear height in late adolescence. In general, the majority of anatomic risk factors for patellar instability change with maturity in a direction that assists in reducing the risk of patellar instability and/or dislocation. The only outlier is the TT-TG distance, which increased by age, and in our oldest cohort of patients, the mean fell below the normal adult range. The current observations highlight the importance of age in the interpretation of risk for injury as well as the need for further studies to identify intrinsic and extrinsic factors that may result in abnormal development of these anatomic features during skeletal growth and maturation.

The readability of online patient education materials for slipped capital femoral epiphysis.

Given the long-term complications of undiagnosed slipped capital femoral epiphysis (SCFE) and the importance of readable health information materials on positive, equitable health outcomes, the objective of this study was to determine if the online patient education materials regarding SCFE are written at or below accepted recommendations. The secondary objective was to determine whether the readability of these materials varied when stratified by the type of website. 'Slipped capital femoral epiphysis', 'SCFE', and 'slipped femoral head' were used as search queries in three common search engines. The readability of each website was evaluated using five established metrics, and the scores were compared by website type and by the complexity of the search query. In this study of 53 unique websites about SCFE, we demonstrated that only one of the web pages was written at the recommended sixth-grade level, and the mean reading level of the online material was above the 10th-grade level. Post hoc testing showed that only websites associated with pediatric academic institutions were written at a significantly lower grade level than general health websites [P < 0.05 for all, range (0.003, 0.04)]. The materials about SCFE that are available to patients and their families continue to be written at an inappropriate level. To increase accessibility and allow for equitable long-term health outcomes, physicians, universities, hospitals and medical societies must ensure that they produce readable education to increase patients' understanding of SCFE, its symptoms and available treatment options. Future studies evaluating progress regarding these metrics are warranted.