Stephen Ma

All Publications

• Recommendations for Clinicians, Technologists, and Healthcare Organizations on the Use of Generative Artificial Intelligence in Medicine: A Position Statement from the Society of General Internal Medicine. Journal of general internal medicine Crowe, B., Shah, S., Teng, D., Ma, S. P., DeCamp, M., Rosenberg, E. I., Rodriguez, J. A., Collins, B. X., Huber, K., Karches, K., Zucker, S., Kim, E. J., Rotenstein, L., Rodman, A., Jones, D., Richman, I. B., Henry, T. L., Somlo, D., Pitts, S. I., Chen, J. H., Mishuris, R. G. 2024

  Abstract

  Generative artificial intelligence (generative AI) is a new technology with potentially broad applications across important domains of healthcare, but serious questions remain about how to balance the promise of generative AI against unintended consequences from adoption of these tools. In this position statement, we provide recommendations on behalf of the Society of General Internal Medicine on how clinicians, technologists, and healthcare organizations can approach the use of these tools. We focus on three major domains of medical practice where clinicians and technology experts believe generative AI will have substantial immediate and long-term impacts: clinical decision-making, health systems optimization, and the patient-physician relationship. Additionally, we highlight our most important generative AI ethics and equity considerations for these stakeholders. For clinicians, we recommend approaching generative AI similarly to other important biomedical advancements, critically appraising its evidence and utility and incorporating it thoughtfully into practice. For technologists developing generative AI for healthcare applications, we recommend a major frameshift in thinking away from the expectation that clinicians will "supervise" generative AI. Rather, these organizations and individuals should hold themselves and their technologies to the same set of high standards expected of the clinical workforce and strive to design high-performing, well-studied tools that improve care and foster the therapeutic relationship, not simply those that improve efficiency or market share. We further recommend deep and ongoing partnerships with clinicians and patients as necessary collaborators in this work. And for healthcare organizations, we recommend pursuing a combination of both incremental and transformative change with generative AI, directing resources toward both endeavors, and avoiding the urge to rapidly displace the human clinical workforce with generative AI. We affirm that the practice of medicine remains a fundamentally human endeavor which should be enhanced by technology, not displaced by it.

  View details for DOI 10.1007/s11606-024-09102-0

  View details for PubMedID 39531100

  View details for PubMedCentralID 10714249

• "Covering provider": an effort to streamline clinical communication chaos. JAMIA open Joshi, M., Gokhale, A., Ma, S., Pendrey, A., Wozniak, L., Moturu, A., Schwartz, N. U., Wilson, A., Darmawan, K., Phillips, B., Cullum, S., Sharp, C., Brown, G., Shieh, L., Schmiesing, C. 2024; 7 (3): ooae057

  Abstract

  This report describes a root cause analysis of incorrect provider assignments and a standardized workflow developed to improve the clarity and accuracy of provider assignments.A multidisciplinary working group involving housestaff was assembled. Key drivers were identified using value stream mapping and fishbone analysis. A report was developed to allow for the analysis of correct provider assignments. A standardized workflow was created and piloted with a single service line. Pre- and post-pilot surveys were administered to nursing staff and participating housestaff on the unit.Four key drivers were identified. A standardized workflow was created with an exclusive treatment team role in Epic held by a single provider at any given time, with a corresponding patient list column displaying provider information for each patient. Pre- and post-survey responses report decreased confusion, decreased provider identification errors, and increased user satisfaction among RNs and residents with sustained uptake over time.This work demonstrates structured root cause analysis, notably engaging housestaff, to develop a standardized workflow for an understudied and growing problem. The development of tools and strategies to address the widespread burdens resulting from clinical communication failures is needed.

  View details for DOI 10.1093/jamiaopen/ooae057

  View details for PubMedID 38974405

  View details for PubMedCentralID PMC11226879

• Electronic Phenotyping of Urinary Tract Infections as a Silver Standard Label for Machine Learning. AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science Ma, S. P., Hosgur, E., Corbin, C. K., Lopez, I., Chang, A., Chen, J. H. 2024; 2024: 182-189

  Abstract

  This study explored the efficacy of electronic phenotyping in data labeling for machine learning with a focus on urinary tract infections (UTIs). We contrasted labels from electronic phenotyping against previously published labels such as urine culture positivity. In comparison, electronic phenotyping showed the potential to enhance specificity in UTI labeling while maintaining similar sensitivity and was easily scaled for application to a large dataset suitable for machine learning, which we used to train and validate a machine learning model. Electronic phenotyping offers a valuable method for machine learning label generation in healthcare, with potential benefits for patient care and antimicrobial stewardship. Further research will expand its application and optimize techniques for increased performance.

  View details for PubMedID 38827068

  View details for PubMedCentralID PMC11141812

• The promises and limitations of artificial intelligence for quality improvement, patient safety, and research in hospital medicine. Journal of hospital medicine Ma, S. P., Rohatgi, N., Chen, J. H. 2024

  View details for DOI 10.1002/jhm.13404

  View details for PubMedID 38751246

• Artificial Intelligence-Generated Draft Replies to Patient Inbox Messages. JAMA network open Garcia, P., Ma, S. P., Shah, S., Smith, M., Jeong, Y., Devon-Sand, A., Tai-Seale, M., Takazawa, K., Clutter, D., Vogt, K., Lugtu, C., Rojo, M., Lin, S., Shanafelt, T., Pfeffer, M. A., Sharp, C. 2024; 7 (3): e243201

  Abstract

  The emergence and promise of generative artificial intelligence (AI) represent a turning point for health care. Rigorous evaluation of generative AI deployment in clinical practice is needed to inform strategic decision-making.To evaluate the implementation of a large language model used to draft responses to patient messages in the electronic inbox.A 5-week, prospective, single-group quality improvement study was conducted from July 10 through August 13, 2023, at a single academic medical center (Stanford Health Care). All attending physicians, advanced practice practitioners, clinic nurses, and clinical pharmacists from the Divisions of Primary Care and Gastroenterology and Hepatology were enrolled in the pilot.Draft replies to patient portal messages generated by a Health Insurance Portability and Accountability Act-compliant electronic health record-integrated large language model.The primary outcome was AI-generated draft reply utilization as a percentage of total patient message replies. Secondary outcomes included changes in time measures and clinician experience as assessed by survey.A total of 197 clinicians were enrolled in the pilot; 35 clinicians who were prepilot beta users, out of office, or not tied to a specific ambulatory clinic were excluded, leaving 162 clinicians included in the analysis. The survey analysis cohort consisted of 73 participants (45.1%) who completed both the presurvey and postsurvey. In gastroenterology and hepatology, there were 58 physicians and APPs and 10 nurses. In primary care, there were 83 physicians and APPs, 4 nurses, and 8 clinical pharmacists. The mean AI-generated draft response utilization rate across clinicians was 20%. There was no change in reply action time, write time, or read time between the prepilot and pilot periods. There were statistically significant reductions in the 4-item physician task load score derivative (mean [SD], 61.31 [17.23] presurvey vs 47.26 [17.11] postsurvey; paired difference, -13.87; 95% CI, -17.38 to -9.50; P < .001) and work exhaustion scores (mean [SD], 1.95 [0.79] presurvey vs 1.62 [0.68] postsurvey; paired difference, -0.33; 95% CI, -0.50 to -0.17; P < .001).In this quality improvement study of an early implementation of generative AI, there was notable adoption, usability, and improvement in assessments of burden and burnout. There was no improvement in time. Further code-to-bedside testing is needed to guide future development and organizational strategy.

  View details for DOI 10.1001/jamanetworkopen.2024.3201

  View details for PubMedID 38506805

• Using Case Mix Index within Diagnosis-Related Groups to Evaluate Variation in Hospitalization Costs at a Large Academic Medical Center. AMIA ... Annual Symposium proceedings. AMIA Symposium Pi, S., Masterson, J., Ma, S. P., Corbin, C. K., Milstein, A., Chen, J. H. 2023; 2023: 1201-1208

  Abstract

  In analyzing direct hospitalization cost and clinical data from an academic medical center, commonly used metrics such as diagnosis-related group (DRG) weight explain approximately 37% of cost variability, but a substantial amount of variation remains unaccounted for by case mix index (CMI) alone. Using CMI as a benchmark, we isolate and target individual DRGs with higher than expected average costs for specific quality improvement efforts. While DRGs summarize hospitalization care after discharge, a predictive model using only information known before admission explained up to 60% of cost variability for two DRGs with a high excess cost burden. This level of variability likely reflects underlying patient factors that are not modifiable (e.g., age and prior comorbidities) and therefore less useful for health systems to target for intervention. However, the remaining unexplained variation can be inspected in further studies to discover operational factors that health systems can target to improve quality and value for their patients. Since DRG weights represent the expected resource consumption for a specific hospitalization type relative to the average hospitalization, the data-driven approach we demonstrate can be utilized by any health institution to quantify excess costs and potential savings among DRGs.

  View details for DOI 10.1089/pop.2013.0002

  View details for PubMedID 38222372

  View details for PubMedCentralID PMC10785921

• Targeting Repetitive Laboratory Testing with Electronic Health Records-Embedded Predictive Decision Support: A Pre-Implementation Study. Clinical biochemistry Rabbani, N., Ma, S. P., Li, R. C., Winget, M., Weber, S., Boosi, S., Pham, T. D., Svec, D., Shieh, L., Chen, J. H. 2023

  Abstract

  INTRODUCTION: Unnecessary laboratory testing contributes to patient morbidity and healthcare waste. Despite prior attempts at curbing such overutilization, there remains opportunity for improvement using novel data-driven approaches. This study presents the development and early evaluation of a clinical decision support tool that uses a predictive model to help providers reduce low-yield, repetitive laboratory testing in hospitalized patients.METHODS: We developed an EHR-embedded SMART on FHIR application that utilizes a laboratory test result prediction model based on historical laboratory data. A combination of semi-structured physician interviews, usability testing, and quantitative analysis on retrospective laboratory data were used to inform the tool's development and evaluate its acceptability and potential clinical impact.KEY RESULTS: Physicians identified culture and lack of awareness of repeat orders as key drivers for overuse of inpatient blood testing. Users expressed an openness to a lab prediction model and 13/15 physicians believed the tool would alter their ordering practices. The application received a median System Usability Scale score of 75, corresponding to the 75th percentile of software tools. On average, physicians desired a prediction certainty of 85% before discontinuing a routine recurring laboratory order and a higher certainty of 90% before being alerted. Simulation on historical lab data indicates that filtering based on accepted thresholds could have reduced 22% of repeat chemistry panels.CONCLUSIONS: The use of a predictive algorithm as a means to calculate the utility of a diagnostic test is a promising paradigm for curbing laboratory test overutilization. An EHR-embedded clinical decision support tool employing such a model is a novel and acceptable intervention with the potential to reduce low-yield, repetitive laboratory testing.

  View details for DOI 10.1016/j.clinbiochem.2023.01.002

  View details for PubMedID 36623759

• Bioengineered optogenetic model of human neuromuscular junction BIOMATERIALS Vila, O. F., Chavez, M., Ma, S. P., Yeager, K., Zholudeva, L., Colon-Mercado, J. M., Qu, Y., Nash, T. R., Lai, C., Feliciano, C. M., Carter, M., Kamm, R. D., Judge, L. M., Conklin, B. R., Ward, M. E., McDevitt, T. C., Vunjak-Novakovic, G. 2021; 276: 121033

  Abstract

  Functional human tissues engineered from patient-specific induced pluripotent stem cells (hiPSCs) hold great promise for investigating the progression, mechanisms, and treatment of musculoskeletal diseases in a controlled and systematic manner. For example, bioengineered models of innervated human skeletal muscle could be used to identify novel therapeutic targets and treatments for patients with complex central and peripheral nervous system disorders. There is a need to develop standardized and objective quantitative methods for engineering and using these complex tissues, in order increase their robustness, reproducibility, and predictiveness across users. Here we describe a standardized method for engineering an isogenic, patient specific human neuromuscular junction (NMJ) that allows for automated quantification of NMJ function to diagnose disease using a small sample of blood serum and evaluate new therapeutic modalities. By combining tissue engineering, optogenetics, microfabrication, optoelectronics and video processing, we created a novel platform for the precise investigation of the development and degeneration of human NMJ. We demonstrate the utility of this platform for the detection and diagnosis of myasthenia gravis, an antibody-mediated autoimmune disease that disrupts the NMJ function.

  View details for DOI 10.1016/j.biomaterials.2021.121033

  View details for Web of Science ID 000690381900001

  View details for PubMedID 34403849

  View details for PubMedCentralID PMC8439334

• Engineering of human cardiac muscle electromechanically matured to an adult-like phenotype NATURE PROTOCOLS Ronaldson-Bouchard, K., Yeager, K., Teles, D., Chen, T., Ma, S., Song, L., Morikawa, K., Wobma, H. M., Vasciaveo, A., Ruiz, E. C., Yazawa, M., Vunjak-Novakovic, G. 2019; 14 (10): 2781-2817

  Abstract

  The application of tissue-engineering approaches to human induced pluripotent stem (hiPS) cells enables the development of physiologically relevant human tissue models for in vitro studies of development, regeneration, and disease. However, the immature phenotype of hiPS-derived cardiomyocytes (hiPS-CMs) limits their utility. We have developed a protocol to generate engineered cardiac tissues from hiPS cells and electromechanically mature them toward an adult-like phenotype. This protocol also provides optimized methods for analyzing these tissues' functionality, ultrastructure, and cellular properties. The approach relies on biological adaptation of cultured tissues subjected to biomimetic cues, applied at an increasing intensity, to drive accelerated maturation. hiPS cells are differentiated into cardiomyocytes and used immediately after the first contractions are observed, when they still have developmental plasticity. This starting cell population is combined with human dermal fibroblasts, encapsulated in a fibrin hydrogel and allowed to compact under passive tension in a custom-designed bioreactor. After 7 d of tissue formation, the engineered tissues are matured for an additional 21 d by increasingly intense electromechanical stimulation. Tissue properties can be evaluated by measuring contractile function, responsiveness to electrical stimuli, ultrastructure properties (sarcomere length, mitochondrial density, networks of transverse tubules), force-frequency and force-length relationships, calcium handling, and responses to β-adrenergic agonists. Cell properties can be evaluated by monitoring gene/protein expression, oxidative metabolism, and electrophysiology. The protocol takes 4 weeks and requires experience in advanced cell culture and machining methods for bioreactor fabrication. We anticipate that this protocol will improve modeling of cardiac diseases and testing of drugs.

  View details for DOI 10.1038/s41596-019-0189-8

  View details for Web of Science ID 000487967700002

  View details for PubMedID 31492957

  View details for PubMedCentralID PMC7195192

• Quantification of human neuromuscular function through optogenetics THERANOSTICS Vila, O. F., Uzel, S. M., Ma, S. P., Williams, D., Pak, J., Kamm, R. D., Vunjak-Novakovic, G. 2019; 9 (5): 1232-1246

  Abstract

  The study of human neuromuscular diseases has traditionally been performed in animal models, due to the difficulty of performing studies in human subjects. Despite the unquestioned value of animal models, inter-species differences hamper the translation of these findings to clinical trials. Tissue-engineered models of the neuromuscular junction (NMJ) allow for the recapitulation of the human physiology in tightly controlled in vitro settings. Methods: Here we report the first human patient-specific tissue-engineered model of the neuromuscular junction (NMJ) that combines stem cell technology with tissue engineering, optogenetics, microfabrication and image processing. The combination of custom-made hardware and software allows for repeated, quantitative measurements of NMJ function in a user-independent manner. Results: We demonstrate the utility of this model for basic and translational research by characterizing in real time the functional changes during physiological and pathological processes. Principal Conclusions: This system holds great potential for the study of neuromuscular diseases and drug screening, allowing for the extraction of quantitative functional data from a human, patient-specific system.

  View details for DOI 10.7150/thno.25735

  View details for Web of Science ID 000459354100002

  View details for PubMedID 30867827

  View details for PubMedCentralID PMC6401498

• Advanced maturation of human cardiac tissue grown from pluripotent stem cells NATURE Ronaldson-Bouchard, K., Ma, S. P., Yeager, K., Chen, T., Song, L., Sirabella, D., Morikawa, K., Teles, D., Yazawa, M., Vunjak-Novakovic, G. 2018; 556 (7700): 239-+

  Abstract

  Cardiac tissues generated from human induced pluripotent stem cells (iPSCs) can serve as platforms for patient-specific studies of physiology and disease1-6. However, the predictive power of these models is presently limited by the immature state of the cells1, 2, 5, 6. Here we show that this fundamental limitation can be overcome if cardiac tissues are formed from early-stage iPSC-derived cardiomyocytes soon after the initiation of spontaneous contractions and are subjected to physical conditioning with increasing intensity over time. After only four weeks of culture, for all iPSC lines studied, such tissues displayed adult-like gene expression profiles, remarkably organized ultrastructure, physiological sarcomere length (2.2 µm) and density of mitochondria (30%), the presence of transverse tubules, oxidative metabolism, a positive force-frequency relationship and functional calcium handling. Electromechanical properties developed more slowly and did not achieve the stage of maturity seen in adult human myocardium. Tissue maturity was necessary for achieving physiological responses to isoproterenol and recapitulating pathological hypertrophy, supporting the utility of this tissue model for studies of cardiac development and disease.

  View details for DOI 10.1038/s41586-018-0016-3

  View details for Web of Science ID 000430082000047

  View details for PubMedID 29618819

  View details for PubMedCentralID PMC5895513

• Dual IFN-gamma/hypoxia priming enhances immunosuppression of mesenchymal stromal cells through regulatory proteins and metabolic mechanisms. Journal of immunology and regenerative medicine Wobma, H. M., Kanai, M., Ma, S. P., Shih, Y., Li, H. W., Duran-Struuck, R., Winchester, R., Goeta, S., Brown, L. M., Vunjak-Novakovic, G. 2018; 1: 45-56

  Abstract

  The immunosuppressive capacity of human mesenchymal stromal cells (MSCs) renders them promising candidates for treating diverse immune disorders. However, after hundreds of clinical trials, there are still no MSC therapies approved in the United States. MSCs require specific cues to adopt their immunosuppressive phenotype, and yet most clinical trials use cells expanded in basic culture medium and growth conditions. We propose that priming MSCs prior to administration will improve their therapeutic efficacy. Interferon-gamma (IFN-gamma) priming are cues common to situations of immune escape that have individually shown promise as MSC priming cues but have not been systematically compared. Using mixed lymphocyte reactions, we show that priming MSCs with either cue alone improves T-cell inhibition. However, combining the two cues results in additive effects and markedly enhances the immunosuppressive phenotype of MSCs. We demonstrate that IFN-gamma induces expression of numerous immunosuppressive proteins (IDO, PD-L1, HLA-E, HLA-G), whereas hypoxia switches MSCs to glycolysis, causing rapid glucose consumption and production of T-cell inhibitory lactate levels. Dual IFN-gamma/hypoxia primed MSCs display both attributes and have even higher induction of immunosuppressive proteins over IFN-gamma priming alone (IDO and HLA-G), which may reflect another benefit of metabolic reconfiguration.

  View details for DOI 10.1016/j.regen.2018.01.001

  View details for PubMedID 30364570

• Optogenetics for the Maturation of hiPS-CMs Shen, C. Y., Ma, S. P., White, E. C., Vila, O. F., Chen, T. H., Yeager, K., Vunjak-Novakovic, G. MARY ANN LIEBERT, INC. 2017: S156-S157

  View details for Web of Science ID 000416247300603

• Real-Time Bioluminescence Imaging of Cell Distribution, Growth, and Differentiation in a Three-Dimensional Scaffold Under Interstitial Perfusion for Tissue Engineering TISSUE ENGINEERING PART C-METHODS Vila, O. F., Garrido, C., Cano, I., Guerra-Rebollo, M., Navarro, M., Meca-Cortes, O., Ma, S. P., Engel, E., Rubio, N., Blanco, J. 2016; 22 (9): 864-872

  Abstract

  Bioreactor systems allow safe and reproducible production of tissue constructs and functional analysis of cell behavior in biomaterials. However, current procedures for the analysis of tissue generated in biomaterials are destructive. We describe a transparent perfusion system that allows real-time bioluminescence imaging of luciferase expressing cells seeded in scaffolds for the study of cell-biomaterial interactions and bioreactor performance. A prototype provided with a poly(lactic) acid scaffold was used for "proof of principle" studies to monitor cell survival in the scaffold (up to 22 days). Moreover, using cells expressing a luciferase reporter under the control of inducible tissue-specific promoters, it was possible to monitor changes in gene expression resulting from hypoxic state and endothelial cell differentiation. This system should be useful in numerous tissue engineering applications, the optimization of bioreactor operation conditions, and the analysis of cell behavior in three-dimensional scaffolds.

  View details for DOI 10.1089/ten.tec.2014.0421

  View details for Web of Science ID 000384100000005

  View details for PubMedID 27339005

• Protection of Organ Vasculature By Endothelial Overexpression of HLA-G Wobma, H. M., Ma, S. P., Fang, J., Vasavada, H., Duran-Struuck, R., Winchester, R., Hirschi, K., Vunjak-Novakovic, G. ELSEVIER SCIENCE INC. 2016: S362

  View details for DOI 10.1016/j.bbmt.2015.11.865

  View details for Web of Science ID 000370910300544

• Preconditioning Mesenchymal Stem Cells to Improve Transplant Tolerance Wobma, H. M., Kanai, M., Ma, S. P., Nakazawa, K. R., Duran-Struuck, R., Li, H., Vunjak-Novakovic, G. ELSEVIER SCIENCE INC. 2016: S149

  View details for DOI 10.1016/j.bbmt.2015.11.496

  View details for Web of Science ID 000370910300189

• Tissue-Engineering for the Study of Cardiac Biomechanics JOURNAL OF BIOMECHANICAL ENGINEERING-TRANSACTIONS OF THE ASME Ma, S. P., Vunjak-Novakovic, G. 2016; 138 (2): 021010

  Abstract

  The notion that both adaptive and maladaptive cardiac remodeling occurs in response to mechanical loading has informed recent progress in cardiac tissue engineering. Today, human cardiac tissues engineered in vitro offer complementary knowledge to that currently provided by animal models, with profound implications to personalized medicine. We review here recent advances in the understanding of the roles of mechanical signals in normal and pathological cardiac function, and their application in clinical translation of tissue engineering strategies to regenerative medicine and in vitro study of disease.

  View details for DOI 10.1115/1.4032355

  View details for Web of Science ID 000369441100011

  View details for PubMedID 26720588

  View details for PubMedCentralID PMC4845250