About

An-Kwok Ian Wong is an Assistant Professor of Medicine and an Assistant Professor in Biostatistics & Bioinformatics at Duke University. His professional role involves contributing to the Duke Department of Biostatistics and Bioinformatics, where he is engaged in research and teaching activities. His work is situated within the broader context of Duke's initiatives in biostatistics, computational biology, and bioinformatics, supporting the university's mission to advance health sciences through innovative research and education.

Research topics

• Artificial Intelligence
• Medicine
• Machine Learning
• Data Mining
• Computer Science
• Mathematics
• Engineering
• Anesthesia
• Emergency medicine
• Statistics
• Algorithm
• Internal medicine

Selected publications

• The Association of Skin Pigmentation With Pulse Oximetry Bias in Adult Patients in the ICU

  CHEST Critical Care · 2026-04-01

  articleOpen access
  PublisherDOI

• Skin tone and clinical dataset from a prospective trial on acute care patients

  Scientific Data · 2026-01-28

  articleOpen accessSenior authorCorresponding

  Although hypothesized to be the root cause of the pulse oximetry disparities, skin tone and its use for improving medical therapies have yet to be extensively studied. Studies that previously used self-reported race as a proxy variable for skin tone cannot account for skin tone variabilities within race groups. This study aimed to create a unique baseline dataset that included skin tone and electronic health record (EHR) data to better evaluate health disparities associated with pulse oximetry. We collected skin tone data at 16 different body locations using multiple devices, including administered visual scales, colorimetric, spectrophotometric, and photography via mobile phone cameras. All patients' data were converted into a common data model and de-identified before publication in PhysioNet. We assessed 167 features per skin location on 128 patients linked with their EHR data, such as laboratory data, vital sign recordings, and demographic information. We also include 2,438 images from mobile phones to assist in developing artificial intelligence tools to combat health disparities.

  PublisherOA PDFDOI

• Strengthening the Evidence on Intraoperative Occult Hypoxemia

  Anesthesia & Analgesia · 2026-03-02

  article
  PublisherDOI

• Evaluating Large Language Models for Automated Clinical Abstraction in Pulmonary Embolism Registries: Performance Across Model Sizes, Versions, and Parameters

  Lecture notes in computer science · 2026-01-01

  book-chapterSenior author
  PublisherDOI

• Evaluation of SOFA-2 Score Performance Across Demographic Subgroups: An External Validation Study Using MIMIC-IV

  medRxiv · 2026-03-11

  articleOpen access

  The Sequential Organ Failure Assessment (SOFA)-2 score was recently validated for ICU mortality prediction across more than 3 million admissions but was not evaluated across demographic subgroups. We assessed the discrimination and calibration of the SOFA-2 score for ICU mortality across subgroups defined by age, sex, race and ethnicity, primary language, and insurance status. We conducted a retrospective cohort study of adult patients (aged 18 years or older) admitted to ICUs at Beth Israel Deaconess Medical Center between 2008 and 2022 (MIMIC-IV, version 3.1), selecting the first ICU admission per patient. First-day SOFA-2 scores (range, 0-24) were calculated using worst recorded values across 6 organ systems. Discrimination was assessed using AUROC, calibration using intercepts and slopes, and subgroup differences using bootstrap resampling. Among 64,015 ICU admissions (median age, 66 years [IQR, 54-78]; 56.1% male; 66.1% White), overall ICU mortality was 7.2% (n=4,596). Overall AUROC was acceptable at 0.77 (95% CI, 0.76-0.77). Notably, discrimination declined significantly with age: AUROC was 0.85 (95% CI, 0.83-0.87) for ages 18-44 and 0.72 (95% CI, 0.70-0.73) for ages 75 and older (difference in AUROC, -0.14; 95% CI, -0.16 to -0.11), with systematic underprediction of mortality in older patients (calibration intercept, 0.39). Discrimination was also significantly lower among non-English speakers (difference in AUROC, -0.04; 95% CI, -0.07 to -0.01) but did not differ significantly across documented racial and ethnic groups. Patients with unknown race/ethnicity (14.3% of the cohort) had nearly double the overall mortality rate and poor calibration. SOFA-2 demonstrated good overall performance for ICU mortality prediction but with clinically meaningful variation across demographic subgroups, particularly a substantial decline in discrimination with advancing age. These findings underscore the need for routine equity evaluation of clinical prediction tools before widespread implementation.

  PublisherOA PDFDOI

• FAIR-EC: A Global Research Network for Fair, Accountable, Interpretable, and Responsible AI in Emergency Care (Preprint)

  2025-03-22

  preprintOpen access

  <sec> <title>BACKGROUND</title> The current landscape of Emergency Care (EC) is marked by high demand leading to issues such as Emergency Department boarding, overcrowding and subsequent delays that impact the quality and safety of patient care. Integrating data science into EC can enhance decision-making with predictive, preventative, personalized, and participatory approaches. However, gaps in adherence to fairness, accountability, interpretability, and responsibility are evident, particularly due to barriers in data-sharing, which often result in a lack of transparency and robust oversight in these applications. </sec> <sec> <title>OBJECTIVE</title> The Fair, Accountable, Interpretable and Responsible (FAIR)-EC collaboration adapts the existing FAIR principles to address emerging challenges as data science integrates with EC. This initiative aims to transform EC by establishing ethical artificial intelligence (AI) standards specifically tailored for this integration. By bridging the gap between EC professionals, data scientists and other stakeholders, the collaboration promotes international cooperation that leverages advanced data science techniques to enhance EC outcomes across different care settings. </sec> <sec> <title>METHODS</title> We propose a federated research design that enables analyses of extensive datasets from various global institutions without compromising patient privacy. This approach transforms epidemiological research with advanced data science techniques, emphasizing the harmonization of data for comprehensive analyses across different healthcare systems. </sec> <sec> <title>RESULTS</title> The FAIR-EC initiative has facilitated the collection and analysis of datasets from diverse geographical regions, enabling the examination of regional variations in EC practices. Initial projects have demonstrated promising outcomes, including the successful development of a federated scoring system and the adaptation of association studies and predictive models across various regions. These efforts highlight the feasibility of leveraging advanced data science techniques to address the complexities of EC while preserving patient privacy. </sec> <sec> <title>CONCLUSIONS</title> FAIR-EC integrates data science ethically and effectively into EC, addressing challenges like fragmented data, real-time handoffs, and public health crises. Its federated design harmonizes diverse data streams while preserving privacy, and its emphasis on ethical AI aligns with the dynamic nature of EC. Despite challenges in data variability and system complexity, FAIR-EC establishes a strong foundation for innovation in global EC. </sec>

  PublisherOA PDFDOI

• EHRmonize: A Framework for Medical Concept Abstraction from Electronic Health Records using Large Language Models

  Lecture notes in computer science · 2025-01-01 · 5 citations

  book-chapterSenior author
  PublisherDOI

• 1122: AUTOMATED ABSTRACTION OF PULMONARY EMBOLISM CONCEPTS FROM CT REPORTS WITH LARGE LANGUAGE MODELS

  Critical Care Medicine · 2025-01-01

  articleSenior author
  PublisherDOI

• Consultation and Echocardiography Timing for Duke Health Patients With Pulmonary Embolism Diagnosed Via Computed Tomography

  American Journal of Respiratory and Critical Care Medicine · 2025-05-01

  articleSenior author

  Abstract INTRODUCTION Effective treatment of pulmonary embolism (PE), a major cause of mortality, depends on a timely sequence of care. This includes diagnosis via computed tomography (CT) with a protocol-timed contrast agent (CTPE) and possibly right ventricular assessment via echocardiograms (echo). PE Response Teams (PERT) can expedite treatment decisions and interventions. While previous registries have characterized established PE patients or those receiving interventions, the sequence and timing of care for all patients—including CTPE, echocardiograms, and PERT consultations—are poorly characterized. This study aims to characterize the demographics and the sequence of timing of care for all patients diagnosed with PE via CT in a quaternary care hospital system. METHODS CTPE reports were obtained from Duke Health's electronic health record using Epic Clarity. Patient IDs were used to extract relevant Echos within 48 hours of the CTPE. Pulmonary vascular disease (PVD) consultations were identified as PERT consultations if ordered within 48 hours of the CTPE event. RESULTS From 2014 to 2024, Duke Health recorded 82,004 CTPE studies involving 60,567 patients across 80,682 admissions. The mean patient age was 58.27 ± 17.22 years. The majority were Black (38.98%) or White (54.31%), with a significant proportion not Latino (92.98%). The median time from the CTPE order to scan was 1.71 hours [0.93, 3.20]. Among these patients, 19,281 CTPE episodes needed urgent echo, ordered a median 0.13 hours [-0.05, 0.36] after the CTPE, with results documented 0.56 hours [0.27, 0.89] after performing the echo. The total time from CTPE to echo results was 0.77 hours [0.33, 1.10]. Additionally, 479 PVD consults were initiated for 470 patients, with a median time of 0.12 hours [0.03, 0.55] post-CTPE, closely correlating with echo orders at 0.03 hours [-0.04, 0.42]. CONCLUSIONS With these findings, we describe patient demographics and care timelines for CTPE, TTE, and PERT consults at an academic center. This emphasizes the need for coordination for timely PE diagnosis and suggests opportunities for process optimization.

  PublisherDOI

• 1665: IMPACT OF VENTILATOR MECHANICAL POWER ON OUTCOMES IN PATIENTS WITH CARDIOGENIC SHOCK

  Critical Care Medicine · 2025-01-01 · 1 citations

  article
  PublisherDOI

Frequent coauthors

• Leo Anthony Celi

  Harvard University

  29 shared

• João Matos

  Massachusetts Institute of Technology

  16 shared

• Mary E. Lough

  Stanford University

  14 shared

• Jack Gallifant

  Massachusetts Institute of Technology

  14 shared

• Marie‐Laure Charpignon

  Massachusetts Institute of Technology

  12 shared

• Andre L. Holder

  Emory University

  8 shared

• Judy Wawira Gichoya

  Emory University

  8 shared

• Azade Tabaie

  Georgetown University

  7 shared