GEORGE®: A Pilot Study of a Smartphone Application for Huntington's Disease.

BACKGROUND: Current Huntington's disease (HD) measures are limited to subjective, episodic assessments conducted in clinic. Smartphones can enable the collection of objective, real-world data but their use has not been extensively evaluated in HD. OBJECTIVE: Develop and evaluate a smartphone application to assess feasibility of use and key features of HD in clinic and at home. METHODS: We developed GEORGE®, an Android smartphone application for HD which assesses voice, chorea, balance, gait, and finger tapping speed. We then conducted an observational pilot study of individuals with manifest HD, prodromal HD, and without a movement disorder. In clinic, participants performed standard clinical assessments and a battery of active tasks in GEORGE. At home, participants were instructed to complete the activities thrice daily for one month. Sensor data were used to measure chorea, tap rate, and step count. Audio data was not analyzed. RESULTS: Twenty-three participants (8 manifest HD, 5 prodromal HD, 10 controls) enrolled, and all but one completed the study. On average, participants used the application 2.1 times daily. We observed a significant difference in chorea score (HD: 19.5; prodromal HD: 4.5, p = 0.007; controls: 4.3, p = 0.001) and tap rate (HD: 2.5 taps/s; prodromal HD: 8.9 taps/s, p = 0.001; controls: 8.1 taps/s, p = 0.001) between individuals with and without manifest HD. Tap rate correlated strongly with the traditional UHDRS finger tapping score (left hand: r = -0.82, p = 0.022; right hand: r = -0.79, p = 0.03). CONCLUSION: GEORGE is an acceptable and effective tool to differentiate individuals with and without manifest HD and measure key disease features. Refinement of the application's interface and activities will improve its usability and sensitivity and, ideally, make it useful for clinical care and research.

An Electronic Medical Record-Based Discharge Disposition Tool Gets Bundle Busted: Decaying Relevance of Clinical Data Accuracy in Machine Learning.

BACKGROUND: Determining discharge disposition after total joint arthroplasty (TJA) has been a challenge. Advances in machine learning (ML) have produced computer models that learn by example to generate predictions on future events. We hypothesized a trained ML algorithm's diagnostic accuracy will be better than that of current predictive tools to predict discharge disposition after primary TJA. METHODS: This study was a retrospective cohort study from a single, tertiary referral center for primary TJA. We trained and validated an artificial neural network (ANN) based on 4368 distinct surgical encounters between 1/1/2013 and 6/28/2016. The ANN's ability to identify discharge disposition was then tested on 1452 distinct surgical encounters between 1/3/17 and 11/30/17. RESULTS: The area under the curve and accuracy achieved during model validation were 0.973 and 91.7%, respectively, with 25% of patients being discharged to skilled nursing facilities (SNFs). Within our testing data set, 6.7% of patients went to SNFs. The performance in the testing set included an area under the curve of 0.804, accuracy of 61.3%, sensitivity of 28.9%, and specificity of 93.8%. CONCLUSIONS: This is the first prediction tool using an electronic medical record-integrated ANN to predict discharge disposition after TJA based on locally generated data. Dramatically reduced numbers of patients discharged to SNFs due to implementation of a bundled payment model lead to poor recall in the testing model. This model serves as a proof of concept for developing an ML prediction tool using a relatively small data set and subsequent integration into the electronic medical record.

Convolutional Neural Network for Second Metacarpal Radiographic Osteoporosis Screening.

PURPOSE: Osteoporosis and osteopenia are extremely common and can lead to fragility fractures. The purpose of this study was to determine whether a computer learning system could classify whether a hand radiograph demonstrated osteoporosis based on the second metacarpal cortical percentage. METHODS: We used the second metacarpal cortical percentage as the osteoporosis predictor. A total of 4,000 posteroanterior (PA) radiographs of the hand were standardized through laterality correction, vertical alignment correction, segmentation, proxy osteoporosis predictor, and full pipeline. Laterality was classified using a LeNet convolutional neural network (CNN). Vertical alignment classification used 2,000 PA x-rays to determine vertical alignment of the second metacarpal. We employed segmentation to determine which pixels belong to the second metacarpal from 1,000 PA x-rays using the FSN-8 CNN. The full pipeline was tested on 265 previously unseen PA x-rays. RESULTS: Laterality classification accuracy was 99.62%, with a specificity of 100% and sensitivity of 99.3%. Rotation of the hand within 10° of vertical was accurate in 93.2% of films. Segmentation was 94.8% accurate. Proxy osteoporosis predictor was 88.4% accurate. Full pipeline accuracy was 93.9%. In the testing data set, the CNN had a sensitivity of 82.4% and specificity of 95.7%. In the balanced data set, 6 of 39 osteoporotic films were classified as nonosteoporotic; sensitivity was 82.4% and specificity, 94.3%. CONCLUSIONS: We have created a series of CNN that can accurately identify osteoporosis from non-osteoporosis. Furthermore, our CNN is able to make adjustments to images based on laterality and vertical alignment. CLINICAL RELEVANCE: Convolutional neural network and computer learning can be used as an adjunct to dual-energy x-ray absorptiometry scans or to screen and make appropriate referrals for further workup in patients with suspected osteoporosis.

Preoperative PROMIS Scores Predict Postoperative PROMIS Score Improvement for Patients Undergoing Hand Surgery.

Background: Patient-Reported Outcomes Measurement Information System (PROMIS) can be used alongside preoperative patient characteristics to set postsurgery expectations. This study aimed to analyze whether preoperative scores can predict significant postoperative PROMIS score improvement. Methods: Patients undergoing hand and wrist surgery with initial and greater than 6-month follow-up PROMIS scores were assigned to derivation or validation cohorts, separating trauma and nontrauma conditions. Receiver operating characteristic curves were calculated for the derivation cohort to determine whether preoperative PROMIS scores could predict postoperative PROMIS score improvement utilizing minimal clinically important difference principles. Results: In the nontrauma sample, patients with baseline Physical Function (PF) scores below 31.0 and Pain Interference (PI) and Depression scores above 68.2 and 62.2, respectively, improved their postoperative PROMIS scores with 95%, 96%, and 94% specificity. Patients with baseline PF scores above 52.1 and PI and Depression scores below 49.5 and 39.5, respectively, did not substantially improve their postoperative PROMIS scores with 94%, 93%, and 96% sensitivity. In the trauma sample, patients with baseline PF scores below 34.8 and PI and Depression scores above 69.2 and 62.2, respectively, each improved their postoperative PROMIS scores with 95% specificity. Patients with baseline PF scores above 52.1 and PI and Depression scores below 46.6 and 44.0, respectively, did not substantially improve their postoperative scores with 95%, 94%, and 95% sensitivity. Conclusions: Preoperative PROMIS PF, PI, and Depression scores can predict postoperative PROMIS score improvement for a select group of patients, which may help in setting expectations. Future work can help determine the level of true clinical improvement these findings represent.

Patient-reported outcomes use during orthopaedic surgery clinic visits improves the patient experience.

OBJECTIVES: The Patient-Reported Outcomes Measurement Information System (PROMIS) is growing in popularity as healthcare shifts towards a value-based system. However, it remains unclear if PROMIS use improves the patient experience. The aim of the present study was to determine if PROMIS use as part of routine orthopaedic clinical care is associated with improved patient experience, as measured by the Clinician and Group Consumer Assessment of Healthcare Providers and Systems (CGCAHPS) survey. METHODS: All patient visits to an orthopaedic surgery clinic at a single academic medical centre between February 2015 and September 2016 were reviewed. Accounting for known patient factors that have an impact on clinic visit satisfaction, CGCAHPS scores were compared between patients who had PROMIS used as part of their routine care and those who had not had PROMIS used as part of their routine care. A p-value of <0.05 was considered significant. RESULTS: A total of 8,607 patient visits fitted our inclusion criteria. Of these, surgeons elected not to use PROMIS in 8,422 patient encounters, leaving 185 patient visits in which PROMIS was actively used. When PROMIS was used, patients were significantly more likely to feel that the provider had spent enough time with them, to recommend this provider office to another patient and to rate the provider significantly higher on a scale from 0 to 10. Although not significant, a trend was found between use of PROMIS and whether a patient felt that a provider explained health information in way that the patient understood. CONCLUSIONS: PROMIS use in an orthopaedic clinic visit can have a positive impact on the patient experience, which is currently part of a number of alternative payment models.

Efficacy of PROMIS Pain Interference and Likert Pain Scores to Assess Physical Function.

PURPOSE: The Patient Reported Outcome Measurement Information System (PROMIS), developed by the National Institutes of Health, utilizes a health domain related to Pain Interference (PI). We evaluated this domain and its association with physical function (as determined by PROMIS Physical Function [PF]), administered as a computer adaptive test (CAT), and secondarily its association to a numerical 0 to 10 pain score. Our null hypothesis was that PI, as measured by CAT, has no correlation to PF and thus, there is no difference between comparisons of numerical pain scores and PROMIS PF. METHODS: Adult patients presenting to an upper extremity clinic from February to December 2015 completed PROMIS PF, PI, and numerical 0 to 10 pain score questionnaires. The PROMIS modules were completed electronically in their computer adaptive form. Mean population scoring on each module is defined at 50. Patients were also asked to rate their pain on a 0 to 10 scale. These data were collected as routine clinical care and were extracted from the electronic health record for cross-sectional evaluation. Bivariate Pearson correlation analysis defined the association between the PROMIS modules and the numerical pain scores. Correlations between PF and PI were compared with correlations between PF and pain scores. RESULTS: We recorded data from patients' 10,574 first, 5,210 second, 2,633 third, 1,382 fourth, and 722 fifth visits. The PROMIS PI was negatively correlated to the PROMIS PF. Numerical pain scores were also negatively correlated to PROMIS PF. Numerical pain scores were less correlated than PROMIS PI through time relative to PF. CONCLUSIONS: Both PROMIS PI and numerical pain scores had significant correlations with PF for each office visit. The PI had a larger correlation to PF than did numerical pain scores. The PI and numerical pain scale scores are also correlated. CLINICAL RELEVANCE: Patient-reported pain using a 0 to 10 pain score can be a predictor of patients' level of function, and although pain score does not replace other patient-reported outcomes, it can provide useful information, particularly when other patient-reported outcomes are not available.

Google Glass as an Alternative to Standard Fluoroscopic Visualization for Percutaneous Fixation of Hand Fractures: A Pilot Study.

This pilot study investigated the feasibility of Google Glass to assist visualization of fluoroscopic images during percutaneous pinning of hand fractures. Cadavers were used to compare total time to pin each fracture and total number of radiographs per fracture from a mini C-arm. A FluoroScan monitor was used for radiographic visualization compared to projecting the images in the Google Glass display. All outcome measures significantly improved for proximal phalanx fractures (127 versus 86 seconds, p = 0.017; 5.3 versus 2.2 images, p = 0.003), and fewer images were obtained during fixation of metacarpal fractures using Google Glass compared with traditional techniques (6.4 versus 3.6, p < 0.001). Typical FluoroScan monitor placement may require the surgeon to alter focus away from the operative field, whereas Google Glass allows constant attention directed toward the operative field.

The effect of informed consent on results of a standard upper extremity intake questionnaire.

PURPOSE: The trend toward requiring explicit consent from patients participating in observational research introduces the potential for consent bias, either through selection bias or through the Hawthorne effect. In the Hawthorne effect, patients may alter responses based on awareness of participation in a study, thus potentially limiting its applicability to a generalized orthopedic practice. We hypothesized that study subjects' awareness of participation in an observational study by informed consent would alter responses to a standard upper extremity questionnaire in a statistically and clinically meaningful way compared with patients who filled out the questionnaire as a quality control measure. METHODS: We retrospectively reviewed data on 39 patients who underwent isolated carpal tunnel release, who had completed preoperative and postoperative Quick Disabilities of the Arm, Shoulder, and Hand (Quick DASH) forms without providing consent for study participation. Next, we approached 35 patients scheduled to undergo isolated carpal tunnel release who completed the intake questionnaire on the day of surgery, for consent to participate in the study. After obtaining consent but before surgery, these patients completed a second questionnaire and then completed a postoperative questionnaire at a mean of 8 weeks postoperatively. RESULTS: There were no significant differences in age, sex, insurance status, symptom duration, nerve conduction study and electromyography results, or disease severity between groups. We identified no statistically significant difference in preoperative or postoperative Quick DASH score between the retrospective and prospective cohorts (40 ± 23 vs 40 ± 19 preoperatively; 27 ± 25 vs 19 ± 13 postoperatively) or within the prospective cohort before and after obtaining informed consent. CONCLUSIONS: Informed consent did not significantly alter patients' responses to the Quick DASH questionnaire. These results suggest that both opt-in and opt-out approaches to observational research in hand surgery provide results that may be applicable to a generalized orthopedic practice. CLINICAL RELEVANCE: This study provides evidence that will inform the interpretation of observational research findings in hand surgery.

Remodeling of murine intrasynovial tendon adhesions following injury: MMP and neotendon gene expression.

Tendon injury frequently results in the formation of adhesions that reduce joint range of motion. To study the cellular, molecular, and biomechanical events involved in intrasynovial tendon healing and adhesion formation, we developed a murine flexor tendon healing model in which the flexor digitorum longus (FDL) tendon of C57BL/6 mice was transected and repaired using suture. This model was used to test the hypothesis that murine flexor tendons heal with differential expression of matrix metalloproteases (MMPs), resulting in the formation of scar tissue as well as the subsequent remodeling of scar and adhesions. Healing tendons were evaluated by histology, gene expression via real-time RT-PCR, and in situ hybridization, as well as biomechanical testing to assess the metatarsophalangeal (MTP) joint flexion range of motion (ROM) and the tensile failure properties. Tendons healed with a highly disorganized fibroblastic tissue response that was progressively remodeled through day 35 resulting in a more organized pattern of collagen fibers. Initial repair involved elevated levels of Mmp-9 at day 7, which is associated with catabolism of damaged collagen fibers. High levels of Col3 are consistent with scar tissue, and gradually transition to the expression of Col1. Scleraxis expression peaked at day 7, but the expression was limited to the original tendon adjacent to the injury site, and no expression was present in granulation tissue involved in the repair response. The MTP joint ROM with standardized force on the tendon was decreased on days 14 and 21 compared to day 0, indicating the presence of adhesions. Peak expressions of Mmp-2 and Mmp-14 were observed at day 21, associated with tendon remodeling. At day 28, two genes associated with neotendon formation, Smad8 and Gdf-5, were elevated and an improvement in MTP ROM occurred. Tensile strength of the tendon progressively increased, but by 63 days the repaired tendons had not reached the tensile strength of normal tendon. The murine model of primary tendon repair, described here, provides a novel mechanism to study the tendon healing process, and further enhances the understanding of this process at the molecular, cellular, and biomechanical level.

Freeze-dried tendon allografts as tissue-engineering scaffolds for Gdf5 gene delivery.

Tendon reconstruction using grafts often results in adhesions that limit joint flexion. These adhesions are precipitated by inflammation, fibrosis, and the paucity of tendon differentiation signals during healing. In order to study this problem, we developed a mouse model in which the flexor digitorum longus (FDL) tendon is reconstructed using a live autograft or a freeze-dried allograft, and identified growth and differentiation factor 5 (Gdf5) as a therapeutic target. In this study we have investigated the potential of rAAV-Gdf5 -loaded freeze-dried tendon allografts as "therapeutically endowed" tissue-engineering scaffolds to reduce adhesions. In reporter gene studies we have demonstrated that recombinant adeno-associated virus (rAAV)-loaded tendon allografts mediate efficient transduction of adjacent soft tissues, with expression peaking at 7 days. We have also demonstrated that the rAAV-Gdf5 vector significantly accelerates wound healing in an in vitro fibroblast scratch model and, when loaded onto freeze-dried FDL tendon allografts, improves the metatarsophalangeal (MTP) joint flexion to a significantly greater extent than the rAAV-lacZ controls do. Collectively, our data demonstrate the feasibility and efficacy of therapeutic tendon allograft processing as a novel paradigm in tissue engineering in order to address difficult clinical problems such as tendon adhesions.

Adhesions in a murine flexor tendon graft model: autograft versus allograft reconstruction.

Reconstruction of flexor tendons often results in adhesions that compromise joint flexion. Little is known about the factors involved in the formation of flexor tendon graft adhesions. In this study, we developed and characterized a novel mouse model of flexor digitorum longus (FDL) tendon reconstruction with live autografts or reconstituted freeze-dried allografts. Grafted tendons were evaluated at multiple time points up to 84 days post-reconstruction. To assess the flexion range of the metatarsophalangeal joint, we developed a quantitative outcome measure proportional to the resistance to tendon gliding due to adhesions, which we termed the Gliding Coefficient. At 14 days post-grafting, the Gliding Coefficient was 29- and 26-fold greater than normal FDL tendon for both autografts and allografts, respectively (p < 0.001), and subsequently doubled for 28-day autografts. Interestingly, there were no significant differences in maximum tensile force or stiffness between live autograft and freeze-dried allograft repairs over time. Histologically, autograft healing was characterized by extensive remodeling and exuberant scarring around both the ends and the body of the graft, whereas allograft scarring was abundant only near the graft-host junctions. Gene expression of GDF-5 and VEGF were significantly increased in 28-day autografts compared to allografts and to normal tendons. These results suggest that the biomechanical advantages for tendon reconstruction using live autografts over devitalized allografts are minimal. This mouse model can be useful in elucidating the molecular mechanisms in tendon repair and can aid in preliminary screening of molecular treatments of flexor tendon adhesions.