About

Jean Crowell Beckham is a Professor in the Department of Psychiatry and Behavioral Sciences at Duke University. She serves as the Co-Division Director of Behavioral Medicine and Neurosciences within the department. Her role involves leadership in clinical and research initiatives related to behavioral medicine and neuroscience, contributing to the academic and clinical missions of Duke University. Her contact information is provided through her email beckham@duke.edu, and her office is located at Box 3625, Durham, NC 27710.

Research topics

• Medicine
• Biology
• Genetics
• Internal medicine
• Psychology
• Clinical psychology
• Computational biology
• Bioinformatics
• Psychiatry
• Computer Science
• Medical emergency
• Pharmacology
• Cardiology
• Neuroscience
• Surgery
• Demography
• Pediatrics
• Environmental health
• Endocrinology

Selected publications

• A Longitudinal Clinical Foundation Model on Nationwide Veteran Health Trajectories

  medRxiv · 2026-05-17

  articleOpen access

  ABSTRACT We present VA-LLM, a 1.62-billion-parameter autoregressive transformer pre-trained from scratch on 1.74 trillion tokens of clinical text spanning 22 years of care for 13.8 million patients in the Veterans Health Administration, with mortality outcomes confirmed through the National Death Index for 7.8 million patients. In a retrospective–prospective evaluation on 107,555 withheld patients, VA-LLM achieved higher 5-year AUPRC than Llama-2 (7 billion parameters), BioGPT_large (1.57 billion parameters), and GatorTron (3.91 billion parameters), matching GatorTron’s 100,000-patient performance with only 10,000 labeled patients. In a clinical validation against the VA’s operational Care Assessment Need (CAN) score on 5.5 million patients one year beyond the pre-training corpus, VA-LLM achieved a 90-day mortality AUROC of 90.00% versus 87.74% ( p < 0.001) and a 45% relative improvement in AUPRC; post-hoc recalibration recovered calibration comparable to CAN (Brier 0.0091 versus 0.0093) without sacrificing discrimination. Across 21 pre-training checkpoints, discriminative performance correlated more strongly with cumulative mortality experience (CME), the total person-years contributed by patients with confirmed deaths, than with token count (Δ R 2 = 0.15; Williams p < 10 −6 ). Performance plateaued once marginal cohorts added fewer confirmed deaths, even as pre-training loss continued to decrease. These findings suggest that the clinical composition of pre-training data, particularly the completeness of documented patient trajectories, correlates with predictive performance more closely than corpus size alone.

  PublisherOA PDFDOI

• Toxic exposure and rates of suicidal thoughts and behaviors among U.S. military veterans

  Psychiatry Research · 2026-04-15

  article
  PublisherDOI

• Utility of Candidate Genes From an Algorithm Designed to Predict Genetic Risk for Opioid Use Disorder

  JAMA Network Open · 2025-01-09 · 5 citations

  articleOpen access

  Importance: Recently, the US Food and Drug Administration gave premarketing approval to an algorithm based on its purported ability to identify individuals at genetic risk for opioid use disorder (OUD). However, the clinical utility of the candidate genetic variants included in the algorithm has not been independently demonstrated. Objective: To assess the utility of 15 genetic variants from an algorithm intended to predict OUD risk. Design, Setting, and Participants: This case-control study examined the association of 15 candidate genetic variants with risk of OUD using electronic health record data from December 20, 1992, to September 30, 2022. Electronic health record data, including pharmacy records, were accrued from participants in the Million Veteran Program across the US with opioid exposure (n = 452 664). Cases with OUD were identified using International Classification of Diseases, Ninth Revision, or International Classification of Diseases, Tenth Revision, diagnostic codes, and controls were individuals with no OUD diagnosis. Exposures: Number of risk alleles present across 15 candidate genetic variants. Main Outcome and Measures: Performance of 15 genetic variants for identifying OUD risk assessed via logistic regression and machine learning models. Results: A total of 452 664 individuals with opioid exposure (including 33 669 with OUD) had a mean (SD) age of 61.15 (13.37) years, and 90.46% were male; the sample was ancestrally diverse (with individuals of genetically inferred European, African, and admixed American ancestries). Using Nagelkerke R2, collectively, the 15 candidate genes accounted for 0.40% of variation in OUD risk. In comparison, age and sex alone accounted for 3.27% of the variation. The ensemble machine learning. The ensemble machine learning model using the 15 variants as predictive factors correctly classified 52.83% (95% CI, 52.07%-53.59%) of individuals in an independent testing sample. Conclusions and Relevance: Results of this study suggest that the candidate genetic variants included in the approved algorithm do not meet reasonable standards of efficacy in identifying OUD risk. Given the algorithm's limited predictive accuracy, its use in clinical care would lead to high rates of both false-positive and false-negative findings. More clinically useful models are needed to identify individuals at risk of developing OUD.

  PublisherOA PDFDOI

• Increasing Value in the Veterans Affairs Healthcare System (VA) with Precision Health: A Continuing Landmark Collaboration with the Department of Energy

  Journal of the American Medical Informatics Association · 2025-08-24

  articleOpen access

  OBJECTIVE: Phase II of MVP-CHAMPION, a federal collaboration between the Veterans Affairs Healthcare System (VA) and the Department of Energy (DoE), leveraged large-scale clinical, geo-spatial, and genetic data with state-of-the-art artificial intelligence (AI), and high-performance computing (HPC) to improve value in healthcare. MATERIALS AND METHODS: Eight clinical priority projects for which AI was a critical missing capability were initiated to address: lung cancer screening (MVP 061), suicide risk screening (MVP 062), cardiovascular risk in obstructive sleep apnea (MVP 063), checkpoint inhibitor toxicity (MVP 064), heart failure (MVP 065), renal complications in diabetes (MVP 066), post COVID-19 sequelae (MVP 067), and antipsychotic medication toxicity (MVP 068). RESULTS: Building on a strong regulatory and administrative foundation, we developed multimorbidity-aware analytic frameworks, reusable computational tools, and analytic pipelines. These greatly facilitated identification of novel risk factors including genetic variants and specification of more discriminating prediction models. Novel genetic risk factors are informing development and repurposing of medications and discriminating prediction models promise to improve healthcare value. DISCUSSION: The research foundation developed in Phase I and extended in Phase II of MVP CHAMPION has supported an unprecedented federal collaboration and yielded significant scientific advances. Our clinical findings are poised for near-term application, while advances in machine learning and high-performance computing may accelerate the broader adoption of artificial intelligence in healthcare. CONCLUSION: This maturing VA-DoE federal collaboration is poised to transform the future of Veterans' healthcare and the broader national landscape of precision health.

  PublisherOA PDFDOI

• Examining the timing of trauma, PTSD onset, and DNA methylation: A multicohort investigation of candidate CpG sites

  Journal of Psychiatric Research · 2025-07-31

  articleOpen access
  PublisherOA PDFDOI

• Polygenic and developmental profiles of autism differ by age at diagnosis

  Nature · 2025-10-01 · 24 citations

  articleOpen access

  Abstract Although autism has historically been conceptualized as a condition that emerges in early childhood 1,2 , many autistic people are diagnosed later in life 3–5 . It is unknown whether earlier- and later-diagnosed autism have different developmental trajectories and genetic profiles. Using longitudinal data from four independent birth cohorts, we demonstrate that two different socioemotional and behavioural trajectories are associated with age at diagnosis. In independent cohorts of autistic individuals, common genetic variants account for approximately 11% of the variance in age at autism diagnosis, similar to the contribution of individual sociodemographic and clinical factors, which typically explain less than 15% of this variance. We further demonstrate that the polygenic architecture of autism can be broken down into two modestly genetically correlated ( r g = 0.38, s.e. = 0.07) autism polygenic factors. One of these factors is associated with earlier autism diagnosis and lower social and communication abilities in early childhood, but is only moderately genetically correlated with attention deficit–hyperactivity disorder (ADHD) and mental-health conditions. Conversely, the second factor is associated with later autism diagnosis and increased socioemotional and behavioural difficulties in adolescence, and has moderate to high positive genetic correlations with ADHD and mental-health conditions. These findings indicate that earlier- and later-diagnosed autism have different developmental trajectories and genetic profiles. Our findings have important implications for how we conceptualize autism and provide a model to explain some of the diversity found in autism.

  PublisherOA PDFDOI

• Trauma, posttraumatic stress disorder, and incident chronic disease.

  PubMed · 2025-01-04

  articleOpen access

  BACKGROUND: Posttraumatic stress disorder (PTSD) is associated with chronic disease risk, particularly cardiovascular disease (CVD). However, few studies have combined detailed measurements of trauma exposure and PTSD with incident chronic disease outcomes assessed using electronic health records (EHRs). PURPOSE: Our study examined associations between traumatic stress (combat exposure, lifetime trauma exposure, PTSD symptoms, and PTSD diagnosis) and chronic disease outcomes, including 7 clinical risk factors and 11 major chronic disease diagnoses assessed using EHRs. METHODS: Participants included 3696 post-9/11 US veterans enrolled in the VISN 6 (Veterans Integrated Service Networks 6) MIRECC (Mental Illness Research, Education, and Clinical Center)'s Post-Deployment Mental Health Study cohort who averaged 38.1 years old at baseline with 13.3 years of follow-up. RESULTS: At baseline, greater PTSD symptoms were associated with higher body mass, more alcohol use, higher rates of smoking, hypertension, and hyperlipidemia. Over follow-up, veterans with more combat exposure (HR, 1.11; 95% CI, 1.04-1.19; P = .002), trauma exposure (HR, 1.15; 95% CI, 1.08-1.23; P < .001), PTSD symptoms (HR, 1.22; 95% CI, 1.14-1.30; P < .001), or a diagnosis of PTSD (HR, 1.39; 95% CI, 1.21-1.59; P < .001) developed more chronic disease. PTSD symptoms and diagnostic status showed consistent associations with incident onset of CVD, diabetes, and pulmonary disease, and associations remained when accounting for non-PTSD psychiatric diagnoses. Compared to veterans with current PTSD, veterans with past PTSD had reduced risk of developing chronic diseases. CONCLUSIONS: Future research should examine if treating PTSD and the sequelae of trauma has the potential to reduce risk for chronic disease, particularly CVD, diabetes, and pulmonary disease.

  PublisherOA PDFDOI

• Anger Mediates the Relationship between Changes in Alcohol Consumption and violence: A Longitudinal Examination Among Iraq/Afghanistan-Era Veterans

  Drug and Alcohol Dependence · 2025-02-01

  article
  PublisherDOI

• Defining suicidality phenotypes for genetic studies: perspectives of the Psychiatric Genomics Consortium Suicide Working Group

  Molecular Psychiatry · 2025-09-25 · 5 citations

  reviewOpen access

  Suicidality phenotypes, consisting of suicidal ideation (SI), suicide attempt (SA), and suicide death (SD), are all heritable but present unique challenges in genome-wide association studies (GWAS) due to their individual complexity, overlap with each other and with related self-harm phenotypes, and varying associations with psychiatric disorders. GWAS have uncovered several loci associated with suicidality phenotypes by meta-analyzing data from multiple cohorts. However, combining datasets from many research groups, where each group may use different study designs, phenotyping instruments, and definitions of suicidality phenotypes, presents challenges. Heterogeneity resulting from these differences can limit genetic discovery; harmonizing phenotype definitions to ensure consistency will greatly improve results. Here, we describe a standardized phenotyping protocol that draws on the expertise of a subgroup of clinicians, researchers, and experts from the Psychiatric Genomics Consortium Suicide Working Group to propose consensus definitions for SI, SA, and SD for genetic studies.

  PublisherOA PDFDOI

• The Externalizing Spectrum and Suicide Risk: Insights from Genomics and Electronic Health Records in over 500,000 Veterans

  medRxiv · 2025-08-24 · 1 citations

  preprintOpen access

  Suicidal thoughts and behaviors originate from heterogeneous mechanisms, including behavioral disinhibition characteristic of "externalizing" disorders (e.g., substance use disorders, antisocial personality disorder, etc.). Prior work has demonstrated strong genetic overlap between externalizing and suicide attempts. In the current analysis, we investigate the co-occurrence between a broader array of suicide phenotypes (i.e., suicide deaths, non-fatal attempts, suicidal ideation) and the externalizing spectrum using data from the Million Veteran Program (MVP) Cohort. We leverage the large-scale MVP database to (1) estimate a latent genomic factor for externalizing comprised of MVP data (MVP-EXT) using genomic structural equation modeling (GenomicSEM), (2) validate these results against prior externalizing models and other traits, (3) examine the genetic overlap between externalizing and suicide outcomes using multiple approaches (e.g., genetic correlations, polygenic scores, and post mortem brain tissue of suicide deaths), and (4) explore whether phenotypic externalizing is prospectively associated with death by suicide. We identify 155 loci in our meta-analysis of European-like (EUR-like, N = 310,498) and African-like (AFR-like, N = 99,949) MVP participants. MVP-EXT showed a strong genetic correlation with a prior, non-MVP externalizing factor (rG = 0.87, 95% CI = 0.83, 0.91) and suicide attempt in both EUR-like (rG = 0.67, 95% CI = 0.60, 0.74) and AFR-like (rG = 0.62, 95% CI = 0.42, 0.81) veterans. MVP-EXT polygenic scores were associated with suicidal ideation (OR = 1.09, 95% CI = 1.05, 1.13) and suicide attempts (OR = 1.20, 95% CI = 1.13, 1.27) in independent cohorts. MVP-EXT associated genes showed significant enrichment particularly within inhibitory neurons in suicide deaths compared to deaths from other causes. A phenotypic score for externalizing was prospectively associated with death by suicide in MVP (HR = 1.39, 95% CI = 1.33, 1.45). In total, our results reiterate that, while the relation between suicide with internalizing disorders has generally received more attention, externalizing is an important risk factor for suicide related behaviors. Greater attention should be paid to these problems as potential antecedents of suicide-related behaviors.

  PublisherOA PDFDOI

Recent grants

• NIH Grant R29MH051752

  NIH · $462k · 2000

• NIH Grant I01RX001301

  NIH · 2019

• NIH Grant K24DA016388

  NIH · $1.6M · 2014

• CSR&D Research Career Scientist Award

  NIH · 2016–2023

• NIH Grant R01CA081595

  NIH · $1.7M · 2012

Frequent coauthors

• Patrick S. Calhoun

  Duke University

  1678 shared

• Nathan A. Kimbrel

  Duke University

  906 shared

• Michelle F. Dennis

  669 shared

• Eric A. Dedert

  Health Services Research & Development

  522 shared

• Paul A. Dennis

  Durham VA Health Care System

  413 shared

• Eric B. Elbogen

  Duke University

  388 shared

• Elizabeth E. Van Voorhees

  Duke University

  343 shared

• Henry R. Wagner

  Duke Medical Center

  289 shared