About

Arden Morris is the Robert L. and Mary Ellenburg Professor of Surgery, and also holds a courtesy appointment as a Professor of Health Policy at Stanford University. She is affiliated with the Center for Artificial Intelligence in Medicine & Imaging (AIMI) at Stanford. Her role involves leadership within the AIMI center, which focuses on advancing artificial intelligence applications in medicine and imaging. The biography emphasizes her academic and professional standing, highlighting her contributions to the integration of AI in healthcare through her leadership position and her association with Stanford University.

Research topics

• Medicine
• Internal medicine
• Surgery
• Oncology
• Emergency medicine
• Family medicine
• Intensive care medicine
• General surgery
• Environmental health
• Pediatrics
• Nursing
• Actuarial science
• Radiology
• Psychiatry
• Physical therapy
• Microbiology
• Pathology
• Nuclear medicine
• Medical emergency

Selected publications

• Outcomes Among Patients With Colon Cancer Living in Neighborhoods With Persistent Poverty

  JAMA Network Open · 2026-01-09

  articleOpen access

  Importance: Patients with cancer living in persistent poverty (PP) are at risk for worse oncologic outcomes. Existing welfare interventions typically focus on current poverty and may not benefit patients in PP if the underlying mechanisms are unique; thus, modifiable targets are needed to inform future policy efforts. Objectives: To compare clinical outcomes for patients diagnosed with colon cancer based on the share of census tracts per zip code that were in PP at the time of diagnosis and to explore 2 potential mechanisms connecting PP and disease-specific mortality. Design, Setting, and Participants: This retrospective cohort study using data from a statewide cancer registry included all patients diagnosed with colon cancer in California from 2017 to 2020. Patients with multiple malignant tumors and patients diagnosed based on only their death certificate were excluded. Data were analyzed from February 2024 to February 2025. Exposure: Proportion of census tracts per zip code designated as being in PP at the time of diagnosis (0, 0.01-0.25, 0.26-0.50, and >0.50). Main Outcomes and Measures: The primary outcome was disease-specific mortality. Secondary outcomes included overall mortality, stage at diagnosis, and receipt of guideline-concordant care. Fine-Gray competing risk survival models were used to calculate risk-adjusted mortality and to evaluate the relative contribution of access to care and quality of care as potential mediators of the association between PP and disease-specific mortality. Charlson comorbidity indices ranged from 0 to 14, with higher values indicating higher comorbidity. Results: In total, 20 015 patients (mean [SD] age at diagnosis, 65.9 [14.0] years; 51.3% male) met inclusion criteria, and the majority (66.3%) resided in zip codes with no PP. Patients living in areas with higher PP ratios were younger (eg, mean [SD] age at diagnosis, 64.3 [14.1] years for >50% PP vs 66.3 [14.1] years for no PP), more likely to identify as Hispanic (eg, 45.5% for >50% PP vs 19.2% for no PP) or non-Hispanic Black (eg, 15.7% for >50% PP vs 4.9% for no PP), and had higher Charlson comorbidity indices (eg, mean [SD] score, 1.3 [1.8] for >50% PP vs 1.2 [1.7] for no PP). After adjustment for demographic and clinical variables, higher shares of PP were associated with higher rates of disease-specific mortality: hazard ratios, 1.20 (95% CI, 1.07-1.36) and 1.19 (95% CI, 1.01-1.42) for PP ratios 0.26-0.50 and higher than 0.50, respectively. Health care practitioner density did not appear to mediate this association. However, adjusting for the receipt of guideline-concordant care affected both the magnitude and the statistical significance of the model, suggesting potential mediation. Conclusions and Relevance: In this cohort study, living in PP was associated with disease-specific mortality among patients diagnosed with colon cancer. The Persistent Poverty Initiative presents a unique opportunity to improve our understanding of PP and to support efforts to extend treatment to all US residents with cancer.

  PublisherOA PDFDOI

• Comparison of Surgery and Medicine on the Impact of Diverticulitis (COSMID) trial: a protocol for a pragmatic randomised study of diverticulitis treatment in the USA

  BMJ Open · 2026-01-01

  articleOpen access

  INTRODUCTION: Each year, millions of people experience recurrent diverticulitis episodes. Elective sigmoid colon resection reduces the risk of recurrence, but The American Society of Colon and Rectal Surgeons recommends individualising surgical decisions based on the impact of the condition on a patient's quality of life (QoL). However, no threshold for QoL impairment has been established to guide decision-making, and evidence comparing elective colectomy with medical management in terms of QoL limitation is limited. To address these gaps and to guide treatment decision-making, we designed the Comparison of Surgery and Medicine on the Impact of Diverticulitis (COSMID) trial.The COSMID trial is a large, pragmatic randomised trial including patients with QoL-limiting diverticulitis that aims to determine if partial colectomy is superior to medical management and explore subgroups that are more likely to respond to each treatment. METHODS AND ANALYSIS: COSMID will recruit 250 English-speaking and Spanish-speaking adults with imaging-confirmed and QoL-limiting diverticulitis (defined using a modified diverticulitis-related QoL survey). Participants are randomly assigned to undergo elective partial colectomy or receive comprehensive medical management (eg, selected from options including fibre, probiotics, mesalamine and rifaximin). A total of 100 patients who decline randomisation but consent to follow-up will be included in a parallel observational cohort. The primary outcome is the time-averaged score of the Gastrointestinal Quality of Life Index at 6, 9 and 12 months after randomisation. Secondary outcomes include clinical adverse events, healthcare utilisation, recurrent episodes of diverticulitis and additional patient-reported outcomes like the Diverticulitis Quality of Life instrument, decisional regret and work productivity. Exploratory analyses aim to identify differential treatment effects based on patients' characteristics. ETHICS AND DISSEMINATION: This trial was approved by the Vanderbilt Institutional Review Board (IRB) on 26 August 2019 (IRB #191217). Vanderbilt serves as the institutional review board of record for the following study sites: Albany Medical College, Allegheny Health, Atrium Health Carolinas Medical Center, Virginia Mason Medical Center, Boston University Medical Center, Cedars-Sinai Medical Center, UT Health Lyndon B. Johnson Hospital, Medical University of South Carolina, New York-Presbyterian Queens, Stanford University, University of Pennsylvania, University of California San Diego, University of California San Francisco, University of Colorado Denver, University of Florida, University of Iowa, University of Utah, University of Washington Medical Center, University of South Florida, University of Rochester Medical Center, University of Texas Southwestern Medical Center, Virginia Commonwealth University, Lahey Hospital & Medical Center, Weill Cornell Medical Center and Northwell Health. Rush University Medical Center (approved 8 January 2020), Columbia University Medical Center (approved 28 January 2020), Northwestern University (approved 19 March 2020), Mount Carmel Health System (approved 5 May 2020) and Memorial Health University Medical Center (approved 4 April 2022) are regulated and were approved by their respective IRBs. Results from this trial will be presented at international conferences and published in peer-reviewed journals. TRIAL REGISTRATION NUMBER: NCT04095663.

  PublisherOA PDFDOI

• Deep learning analysis of MRI to assess rectal cancer treatment

  Frontiers in Oncology · 2026-02-09

  articleOpen accessSenior author

  Introduction Traditional neoadjuvant therapy for locally advanced rectal cancer (LARC) results in pathologic complete response (pCR) in approximately 15% of patients, supporting non-operative strategies for those with clinical complete response (cCR). The subjectivity and variability in MRI-based cCR assessments highlight the need for objective, quantitative tools. Objective To develop deep learning models for automated rectal tumor segmentation on pre- and post-treatment MRIs, and to identify radiomic features differentiating cCR from non-cCR patients. Materials and methods We retrospectively analyzed pre- and post-treatment MRIs from 37 LARC patients enrolled in a Phase 2 TNT trial (NCT04380337). Rectal tumors were segmented on T2-weighted images by two data scientists, refined by a radiologist (reference standard), and independently segmented by a fellow. For pre-treatment segmentation, Model 1 (baseline; <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im1"> <mml:mrow> <mml:mi>n</mml:mi> <mml:mo>=</mml:mo> <mml:mn>37</mml:mn> </mml:mrow> </mml:math> ) was trained on reference cases, then used to generate pseudo-labels for 81 additional cases. Model 2 (semi-supervised; <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im2"> <mml:mrow> <mml:mi>n</mml:mi> <mml:mo>=</mml:mo> <mml:mn>118</mml:mn> </mml:mrow> </mml:math> ) was trained on the combined dataset. Model 3 (baseline; <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im3"> <mml:mrow> <mml:mi>n</mml:mi> <mml:mo>=</mml:mo> <mml:mn>37</mml:mn> </mml:mrow> </mml:math> ) was trained on post-treatment cases. Radiomic features were extracted from post-treatment ADC maps, filtered by reproducibility (ICC <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im4"> <mml:mrow> <mml:mo>≥</mml:mo> <mml:mn>0.8</mml:mn> </mml:mrow> </mml:math> ) and redundancy (Spearman <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im5"> <mml:mrow> <mml:mi>ρ</mml:mi> <mml:mo>≤</mml:mo> <mml:mn>0.95</mml:mn> </mml:mrow> </mml:math> ), then analyzed using unsupervised hierarchical clustering. Results For pre-treatment segmentation, radiologist-fellow inter-rater agreement was DSC <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im6"> <mml:mrow> <mml:mo>=</mml:mo> <mml:mn>0.748</mml:mn> <mml:mo>±</mml:mo> <mml:mn>0.092</mml:mn> </mml:mrow> </mml:math> . Model 1 achieved mean DSC <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im7"> <mml:mrow> <mml:mo>=</mml:mo> <mml:mn>0.682</mml:mn> <mml:mo>±</mml:mo> <mml:mn>0.254</mml:mn> </mml:mrow> </mml:math> versus the radiologist, significantly lower than inter-rater agreement. Model 2 improved performance to mean DSC <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im8"> <mml:mrow> <mml:mo>=</mml:mo> <mml:mn>0.769</mml:mn> <mml:mo>±</mml:mo> <mml:mn>0.214</mml:mn> </mml:mrow> </mml:math> (mean gain <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im9"> <mml:mrow> <mml:mo>=</mml:mo> <mml:mn>0.087</mml:mn> </mml:mrow> </mml:math> ; <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im10"> <mml:mrow> <mml:mn>12.8</mml:mn> <mml:mo>%</mml:mo> </mml:mrow> </mml:math> relative improvement; <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im11"> <mml:mrow> <mml:mi>p</mml:mi> <mml:mo>&lt;</mml:mo> <mml:mn>0.001</mml:mn> </mml:mrow> </mml:math> ), slightly outperforming inter-rater agreement. For post-treatment segmentation, inter-rater agreement declined to mean DSC <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im12"> <mml:mrow> <mml:mo>=</mml:mo> <mml:mn>0.362</mml:mn> <mml:mo>±</mml:mo> <mml:mn>0.256</mml:mn> </mml:mrow> </mml:math> , while Model 3 achieved mean DSC <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im13"> <mml:mrow> <mml:mo>=</mml:mo> <mml:mn>0.175</mml:mn> <mml:mo>±</mml:mo> <mml:mn>0.231</mml:mn> </mml:mrow> </mml:math> versus the radiologist, reflecting challenges from treatment-induced tissue changes affecting both automated models and human raters. Radiomic clustering revealed two distinct patient groups aligned with cCR and non-cCR status. Conclusion This study demonstrates the feasibility of deep learning-based automated segmentation and radiomic profiling for differentiating treatment response in rectal cancer. Semi-supervised learning with pseudo-labeled data significantly improved segmentation performance, offering a practical approach to overcome limited annotations. Radiomic features warrant validation in larger multi-center studies for clinical translation.

  PublisherOA PDFDOI

• Corrigendum to “Development of a universal child abuse screening tool (U-CAST): A qualitative analysis of medical expert perspectives” [Child Abuse &amp; Neglect 172 (2026) start page (2)–end page (3)/107838]

  Child Abuse & Neglect · 2026-02-12

  articleOpen access
  PublisherOA PDFDOI

• Transportation Insecurity, Social Support, and Adherence to Cancer Screening

  JAMA Network Open · 2025-01-30 · 22 citations

  articleOpen accessSenior authorCorresponding

  Importance: Transportation insecurity and lack of social support are 2 understudied social determinants of health that contribute to excess morbidity, mortality, and acute health care utilization. However, whether and how these social determinants of health are associated with cancer screening has not been determined and has implications for preventive care. Objective: To determine whether transportation insecurity or social support are associated with screening adherence for colorectal, breast, and cervical cancer. Design, Setting, and Participants: This cohort study used data from the publicly available 2018 in-person National Health Interview Survey (NHIS) comprising a noninstitutionalized, civilian adult population of the United States. Participants included adults eligible for colorectal, breast, or cervical cancer screening who participated in the in-depth NHIS interview (1 selected per household). Data were acquired in December 2023 and analyzed through July 31, 2024. Exposures: Transportation insecurity, represented dichotomously as adults who reported that they have or have not delayed medical care in the past year due to transportation difficulties, and neighborhood social support, represented as factor scores derived from 4 Likert-type questions. Main Outcomes and Measures: The primary outcome was adherence to the US Preventive Services Task Force screening recommendations in place during 2018 for colorectal, breast, and cervical cancer. Results: In 2018, of 25 417 NHIS respondents (55% female), 660 (3%) reported delaying medical care because they did not have transportation. In fully adjusted models, transportation insecurity was associated with adherence to breast cancer screening (odds ratio [OR], 0.59 [95% CI, 0.40-0.86]) but not to colorectal (OR, 0.87 [95% CI, 0.65-1.15]) or cervical (OR, 0.73 [95% CI, 0.46-1.13) cancer screening. Social support was associated with colorectal (OR, 1.12 [95% CI, 1.06-1.17]) and breast (OR, 1.13 [95% CI, 1.05-1.22]) cancer screening but not with cervical cancer screening (OR, 1.01 [95% CI, 0.93-1.10]). There were no significant interactions between transportation insecurity and social support for any cancer screening. Conclusions and Relevance: The presence of transportation insecurity was associated with a 41% reduction in the odds of breast cancer screening. Clinicians should consider screening for transportation needs at the time of mammography referral, as patients may be eligible for programs that can assist with medical transportation needs.

  PublisherDOI

• Predictive Value of Magnetic Resonance Complete Response After Neoadjuvant Therapy for Rectal Cancer

  Journal of Surgical Research · 2025-01-27 · 1 citations

  articleSenior author
  PublisherDOI

• Preimplementation expectations and perceptions of a preoperative frailty screening and optimization intervention: A qualitative analysis

  Surgery · 2025-09-22

  article
  PublisherDOI

• Phase II trial of organ preservation program using short-course radiation and FOLFOXIRI for rectal cancer (SHORT-FOX): Two-Year primary outcome analysis

  Radiotherapy and Oncology · 2025-04-08 · 2 citations

  articleOpen access

  BACKGROUND AND PURPOSE: As patients with rectal cancer with clinical complete response (cCR) after neoadjuvant therapy may be safely spared Total Mesorectal Excision (TME), strategies to maximize cCR are needed. MATERIALS AND METHODS: We conducted a single-arm phase II study to determine whether dose-escalated short-course radiotherapy (25 Gy/5 fractions + 5 Gy/1 fraction boost) followed by eight cycles of FOLFOXIRI increased cCR rates among adult patients with > T2N0M0 or low T2N0 rectal cancer. RESULTS: Between 2020 and 2023, we enrolled 37 patients, of whom 27 (73 %) had at least one high-risk feature (cT4, extramural vascular invasion [EMVI], N2, threatened circumferential resection margin, positive lateral node). At primary endpoint assessment, nine (24 %) patients had cCR on both endoscopy and MRI, and pursued organ preservation (OP). Fourteen (38 %) patients had cCR only on endoscopy, nine of whom pursued OP. Of the 18 patients who pursued OP, nine had local regrowth at two years from radiotherapy start, with two-year TME-free survival of 26 %. Baseline factors significantly associated with not achieving OP included age < 50 years and T4 disease. At mid-treatment restaging, patients who achieved OP were significantly less likely to have persistent node positivity, EMVI, and endoscopically visible tumor. Grade 3+ adverse events at least possibly attributed to chemotherapy and radiotherapy occured in 51% and 43% of patients, respectively. CONCLUSION: Short-course radiotherapy with a boost followed by FOLFIXIRI results in OP in one-quarter of patients with high-risk rectal cancer, with poorer response among younger patients and T4 disease. Mid-treatment response may help guide timely decision-making regarding treatment.

  PublisherDOI

• Perioperative Management of the Patient Receiving Maintenance Hemodialysis

  Anesthesiology · 2025-09-09 · 1 citations

  articleOpen access
  PublisherOA PDFDOI

• Development of a universal child abuse screening tool (U-CAST): A qualitative analysis of medical expert perspectives

  Child Abuse & Neglect · 2025-12-09

  articleOpen access
  PublisherOA PDFDOI

Frequent coauthors

• David R. Flum

  University of Washington

  91 shared

• Ardith Z. Doorenbos

  University of Washington

  76 shared

• Emily A. Haozous

  Pacific Institute For Research and Evaluation

  75 shared

• Heather M. Harris

  Cognizant (United States)

  64 shared

• Amber W. Trickey

  58 shared

• Cindy Kin

  48 shared

• Todd H. Wagner

  VA Palo Alto Health Care System

  44 shared

• Sarah T. Hawley

  Michigan Medicine

  37 shared

Education

• M.D.

  Stanford University

• B.S.

  Stanford University