Contact

  • Academic
    mchen20@stanford.edu
    University - Student Department: School of Medicine Position: Graduate, Medicine

All Publications

  • Surgery and Suicide Deaths Among Patients With Cancer. JAMA network open Chen, M. L., Gomez, S. L., O'Hara, R., John, E. M., Morris, A. M., Kurian, A. W., Linos, E. 2024; 7 (9): e2431414

    View details for DOI 10.1001/jamanetworkopen.2024.31414

    View details for PubMedID 39226059

  • Development of Melanoma and Other Nonkeratinocyte Skin Cancers After Thyroid Cancer Radiation. JAMA network open Rezaei, S. J., Chen, M. L., Kim, J., John, E. M., Sunwoo, J. B., Linos, E. 2024; 7 (9): e2434841

    View details for DOI 10.1001/jamanetworkopen.2024.34841

    View details for PubMedID 39298173

  • Performance and risk of harm of a large language model on dermatology continuing medical education questions Chen, M. L., Cai, Z., Kim, J., Novoa, R., Barnes, L. A., Beam, A., Linos, E. ELSEVIER SCIENCE INC. 2024: S25

    View details for Web of Science ID 001276879000142

  • Quality of images submitted by older patients to a teledermatology platform Onyeka, S., Kim, J., Eid, E., Cai, Z., Chen, M. L., Hinton, A., Linos, E. ELSEVIER SCIENCE INC. 2024: S54

    View details for Web of Science ID 001276879001081

  • Large language models outperform mental and medical health care professionals in identifying obsessive-compulsive disorder. NPJ digital medicine Kim, J., Leonte, K. G., Chen, M. L., Torous, J. B., Linos, E., Pinto, A., Rodriguez, C. I. 2024; 7 (1): 193

    Abstract

    Despite the promising capacity of large language model (LLM)-powered chatbots to diagnose diseases, they have not been tested for obsessive-compulsive disorder (OCD). We assessed the diagnostic accuracy of LLMs in OCD using vignettes and found that LLMs outperformed medical and mental health professionals. This highlights the potential benefit of LLMs in assisting in the timely and accurate diagnosis of OCD, which usually entails a long delay in diagnosis and treatment.

    View details for DOI 10.1038/s41746-024-01181-x

    View details for PubMedID 39030292

    View details for PubMedCentralID 10352922

  • Analysing common topics of secure patient messages in hidradenitis suppurativa: a text-embedding and natural language-processing approach. The British journal of dermatology Chen, M. L., Kim, J., Naik, H. B., Aleshin, M. A., Sarin, K. Y., Barnes, L. A., Linos, E. 2024

    View details for DOI 10.1093/bjd/ljae222

    View details for PubMedID 38975641

  • Incidence of Suicide Among Melanoma and Non-Keratinocyte Skin Cancer Patients in the US, 2000-2020. Journal of the American Academy of Dermatology Chen, M. L., Rezaei, S. J., Kim, J., Rodriguez, C., Swetter, S. M., O'Hara, R., Linos, E. 2024

    View details for DOI 10.1016/j.jaad.2024.05.028

    View details for PubMedID 38768863

  • Telehealth Utilization and Associations in the United States During the Third Year of the COVID-19 Pandemic: Population-Based Survey Study in 2022. JMIR public health and surveillance Kim, J., Cai, Z. R., Chen, M. L., Onyeka, S., Ko, J. M., Linos, E. 2024; 10: e51279

    Abstract

    BACKGROUND: The COVID-19 pandemic rapidly changed the landscape of clinical practice in the United States; telehealth became an essential mode of health care delivery, yet many components of telehealth use remain unknown years after the disease's emergence.OBJECTIVE: We aim to comprehensively assess telehealth use and its associated factors in the United States.METHODS: This cross-sectional study used a nationally representative survey (Health Information National Trends Survey) administered to US adults (≥18 years) from March 2022 through November 2022. To assess telehealth adoption, perceptions of telehealth, satisfaction with telehealth, and the telehealth care purpose, we conducted weighted descriptive analyses. To identify the subpopulations with low adoption of telehealth, we developed a weighted multivariable logistic regression model.RESULTS: Among a total of 6252 survey participants, 39.3% (2517/6252) reported telehealth use in the past 12 months (video: 1110/6252, 17.8%; audio: 876/6252, 11.6%). The most prominent reason for not using telehealth was due to telehealth providers failing to offer this option (2200/3529, 63%). The most common reason for respondents not using offered telehealth services was a preference for in-person care (527/578, 84.4%). Primary motivations to use telehealth were providers' recommendations (1716/2517, 72.7%) and convenience (1516/2517, 65.6%), mainly for acute minor illness (600/2397, 29.7%) and chronic condition management (583/2397, 21.4%), yet care purposes differed by age, race/ethnicity, and income. The satisfaction rate was predominately high, with no technical problems (1829/2517, 80.5%), comparable care quality to that of in-person care (1779/2517, 75%), and no privacy concerns (1958/2517, 83.7%). Younger individuals (odd ratios [ORs] 1.48-2.23; 18-64 years vs ≥75 years), women (OR 1.33, 95% CI 1.09-1.61), Hispanic individuals (OR 1.37, 95% CI 1.05-1.80; vs non-Hispanic White), those with more education (OR 1.72, 95% CI 1.03-2.87; at least a college graduate vs less than high school), unemployed individuals (OR 1.25, 95% CI 1.02-1.54), insured individuals (OR 1.83, 95% CI 1.25-2.69), or those with poor general health status (OR 1.66, 95% CI 1.30-2.13) had higher odds of using telehealth.CONCLUSIONS: To our best knowledge, this is among the first studies to examine patient factors around telehealth use, including motivations to use, perceptions of, satisfaction with, and care purpose of telehealth, as well as sociodemographic factors associated with telehealth adoption using a nationally representative survey. The wide array of descriptive findings and identified associations will help providers and health systems understand the factors that drive patients toward or away from telehealth visits as the technology becomes more routinely available across the United States, providing future directions for telehealth use and telehealth research.

    View details for DOI 10.2196/51279

    View details for PubMedID 38669075

  • Assessment of correctness, content omission, and risk of harm in large language model responses to dermatology continuing medical education questions. The Journal of investigative dermatology Cai*, Z. R., Chen*, M. L., Kim, J., Novoa, R. A., Barnes, L. A., Beam, A., Linos, E., (*co-first authors) 2024

    View details for DOI 10.1016/j.jid.2024.01.015

  • Prevalence and associations of poor mental health in the third year of COVID-19: U.S. population-based analysis from 2020 to 2022. Psychiatry research Kim, J., Linos, E., Rodriguez, C. I., Chen, M. L., Dove, M. S., Keegan, T. H. 2023; 330: 115622

    Abstract

    BACKGROUND: Poorer mental health was found early in the COVID-19 pandemic, yet mental health in the third year of COVID-19 has not been assessed on a general adult population level in the United States.METHODS: We used a nationally representative cross-sectional survey (Health Information National Trends Survey, HINTS 5 2020 n=3,865 and HINTS 6 2022 n=6,252). The prevalence of poor mental health was examined using a Patient Health Questionnaire-4 scale in 2020 and 2022. We also investigated the factors associated with poor mental health in 2022 using a weighted multivariable logistic regression adjusting for sociodemographic and health status characteristics to obtain the odds ratio (OR).OUTCOMES: The prevalence of poor mental health in adults increased from 2020 to 2022 (31.5% vs 36.3%, p=0.0005). U.S. adults in 2022 were 1.28 times as likely to have poor mental health than early in the pandemic. Moreover, individuals with food insecurity, housing instability, and low income had greater odds of poor mental health (ORs=1.78-2.55). Adults who were females, non-Hispanic Whites, or age 18-64 years were more likely to have poor mental health (ORs=1.46-4.15).INTERPRETATION: Mental health of U.S. adults worsened in the third year of COVID-19 compared to the beginning of the pandemic.

    View details for DOI 10.1016/j.psychres.2023.115622

    View details for PubMedID 38006717

  • Assessing Biases in Medical Decisions via Clinician and AI Chatbot Responses to Patient Vignettes. JAMA network open Kim, J., Cai, Z. R., Chen, M. L., Simard, J. F., Linos, E. 2023; 6 (10): e2338050

    View details for DOI 10.1001/jamanetworkopen.2023.38050

    View details for PubMedID 37847506

  • Evaluation of diagnosis diversity in artificial intelligence datasets: a scoping review. The British journal of dermatology Chen, M. L., Rotemberg, V., Lester, J. C., Novoa, R. A., Chiou, A. S., Daneshjou, R. 2023; 188 (2): 292-294

    View details for DOI 10.1093/bjd/ljac047

    View details for PubMedID 36763858

  • Incidence Trends of Primary Cutaneous T-Cell Lymphoma in the US From 2000 to 2018: A SEER Population Data Analysis. JAMA oncology Cai, Z. R., Chen, M. L., Weinstock, M. A., Kim, Y. H., Novoa, R. A., Linos, E. 2022

    View details for DOI 10.1001/jamaoncol.2022.3236

    View details for PubMedID 36048455

  • A Social Media‒Based Public Health Campaign Encouraging COVID-19 Vaccination Across the United States. American journal of public health Hunt, I. d., Dunn, T., Mahoney, M., Chen, M., Nava, V., Linos, E. 2022: e1-e4

    Abstract

    Tailored public health messaging encouraging COVID-19 vaccination may help increase vaccination rates and decrease the burden of COVID-19. We conducted a three-part COVID-19 vaccine uptake public health campaign disseminated on Facebook between April and June 2021. Our first campaign focused on reaching Black and Latinx communities; our second campaign focused on addressing vaccine access and scheduling in Latinx communities; and our third campaign focused on religious communities. Overall, we reached 25 million individuals with 171 million views across the United States. (Am J Public Health. Published online ahead of print July 7, 2022:e1-e4. https://doi.org/10.2105/AJPH.2022.306934).

    View details for DOI 10.2105/AJPH.2022.306934

    View details for PubMedID 35797502

  • A convolutional neural network highlights mutations relevant to antimicrobial resistance in Mycobacterium tuberculosis. Nature communications Green, A. G., Yoon, C. H., Chen, M. L., Ektefaie, Y., Fina, M., Freschi, L., Groschel, M. I., Kohane, I., Beam, A., Farhat, M. 2022; 13 (1): 3817

    Abstract

    Long diagnostic wait times hinder international efforts to address antibiotic resistance in M. tuberculosis. Pathogen whole genome sequencing, coupled with statistical and machine learning models, offers a promising solution. However, generalizability and clinical adoption have been limited by a lack of interpretability, especially in deep learning methods. Here, we present two deep convolutional neural networks that predict antibiotic resistance phenotypes of M. tuberculosis isolates: a multi-drug CNN (MD-CNN), that predicts resistance to 13 antibiotics based on 18 genomic loci, with AUCs 82.6-99.5% and higher sensitivity than state-of-the-art methods; and a set of 13 single-drug CNNs (SD-CNN) with AUCs 80.1-97.1% and higher specificity than the previous state-of-the-art. Using saliency methods to evaluate the contribution of input sequence features to the SD-CNN predictions, we identify 18 sites in the genome not previously associated with resistance. The CNN models permit functional variant discovery, biologically meaningful interpretation, and clinical applicability.

    View details for DOI 10.1038/s41467-022-31236-0

    View details for PubMedID 35780211

  • Differences in Thickness-Specific Incidence and Factors Associated With Cutaneous Melanoma in the US From 2010 to 2018. JAMA oncology Chen, M. L., de Vere Hunt, I. J., John, E. M., Weinstock, M. A., Swetter, S. M., Linos, E. 2022

    Abstract

    The recent incidence of cutaneous melanoma of different thicknesses in the US is not well described.To evaluate recent patterns in the incidence of melanoma by tumor thickness and examine associations of sex, race and ethnicity, and socioeconomic status with melanoma thickness-specific incidence.This population-based cohort study analyzed data for 187 487 patients with a new diagnosis of invasive cutaneous melanoma from the Surveillance, Epidemiology, and End Results Registry from January 1, 2010, to December 31, 2018. The study was conducted from May 27 to December 29, 2021. Data were analyzed from June 21 to October 24, 2021.Age-adjusted incidence rates of melanoma were calculated by tumor thickness (categorized by Breslow thickness) and annual percentage change (APC) in incidence rates. Analyses were stratified by sex and race and ethnicity. The associations with socioeconomic status were evaluated in 134 359 patients diagnosed with melanoma from 2010 to 2016.This study included 187 487 patients with a median (IQR) age of 62 (52-72) years and 58.4% men. Melanoma incidence was higher in men compared with women across all tumor thickness groups. Individuals in lower socioeconomic status quintiles and members of minority groups were more likely to be diagnosed with thicker (T4) tumors (20.7% [169 of 816] among non-Hispanic Black patients, 11.2% [674 of 6042] among Hispanic patients, and 6.3% [10 774 of 170 155] among non-Hispanic White patients). Between 2010 and 2018, there was no significant increase in incidence of cutaneous melanoma across the full population (APC, 0.39%; 95% CI, -0.40% to 1.18%). The incidence of the thickest melanomas (T4, >4.0 mm) increased between 2010 and 2018, with an APC of 3.32% (95% CI, 2.06%-4.60%) overall, 2.50% (95% CI, 1.27%-3.73%) in men, and 4.64% (95% CI, 2.56%-6.75%) in women.In this population-based cohort study, the incidence of the thickest cutaneous melanoma tumors increased from 2010 to 2018, in contrast with the incidence patterns for thinner melanomas. The findings suggest potential stabilization of overall melanoma incidence rates in the US after nearly a century of continuous increase in incidence. Patients with low socioeconomic status and Hispanic patients were more likely to be diagnosed with thick melanoma. The continued rise in incidence of thick melanoma is unlikely to be attributable to overdiagnosis given the stability of thin melanoma rates.

    View details for DOI 10.1001/jamaoncol.2022.0134

    View details for PubMedID 35323844

  • Refractive Outcomes for Cataract Surgery With Toric Intraocular Lenses at a Veterans Affairs Medical Center. Federal practitioner : for the health care professionals of the VA, DoD, and PHS Tran, E. M., Tang, K. S., Chen, A. J., Chen, M. L., Rivera, D. R., Rivera, J. J., Greenberg, P. B. 2020; 37 (3): 138-142

    Abstract

    Background: Refractive outcomes for cataract surgery with toric intraocular lenses (IOLs) are not well described in a teaching hospital setting. This study investigated the refractive outcomes of cataract surgery with toric IOLs at an academic-affiliated Veterans Affairs Medical Center (VAMC) and compared the accuracy of 2 biometric formulae for toric IOL power calculation.Methods: A retrospective chart review of patients who received cataract surgery with toric IOLs from November 2013 to May 2018 was conducted. The Holladay 2 and Barrett toric IOL formulae were used to predict the postoperative refraction for each cataract surgery. The main outcome measures were best-corrected visual acuity (BCVA) and the difference in cylinder between the preoperative and postoperative manifest refractions. The accuracy of each biometric formula was also assessed using 2-tailed t tests of the mean absolute error, and subgroup analyses were conducted for short, medium, and long eyes.Results: Of 325 charts reviewed, 283 patients met the inclusion criteria; 87% (248/283) of these surgeries were performed by resident surgeons. The median postoperative BCVA was 20/20, and 92% of patients had a postoperative BCVA of 20/25 or better. There was no statistically significant difference in mean absolute error between the 2 formulae for the entire axial length range (P = .21), as well as the short (P = .94), medium (P = .49), and long axial length (P = .08) groups.Conclusions: To our knowledge, this is the largest study that compared the performance of the Barrett toric and Holladay 2 formulae and the first that made the comparison in a teaching hospital setting. This study suggests that the 2 formulae have similar refractive outcomes across all axial lengths.

    View details for PubMedID 32317850

  • Beyond multidrug resistance: Leveraging rare variants with machine and statistical learning models in Mycobacterium tuberculosis resistance prediction EBIOMEDICINE Chen, M. L., Doddi, A., Royer, J., Freschi, L., Schito, M., Ezewudo, M., Kohane, I. S., Beam, A., Farhat, M. 2019; 43: 356-369

    Abstract

    The diagnosis of multidrug resistant and extensively drug resistant tuberculosis is a global health priority. Whole genome sequencing of clinical Mycobacterium tuberculosis isolates promises to circumvent the long wait times and limited scope of conventional phenotypic antimicrobial susceptibility, but gaps remain for predicting phenotype accurately from genotypic data especially for certain drugs. Our primary aim was to perform an exploration of statistical learning algorithms and genetic predictor sets using a rich dataset to build a high performing and fast predicting model to detect anti-tuberculosis drug resistance.We collected targeted or whole genome sequencing and conventional drug resistance phenotyping data from 3601 Mycobacterium tuberculosis strains enriched for resistance to first- and second-line drugs, with 1228 multidrug resistant strains. We investigated the utility of (1) rare variants and variants known to be determinants of resistance for at least one drug and (2) machine and statistical learning architectures in predicting phenotypic drug resistance to 10 anti-tuberculosis drugs. Specifically, we investigated multitask and single task wide and deep neural networks, a multilayer perceptron, regularized logistic regression, and random forest classifiers.The highest performing machine and statistical learning methods included both rare variants and those known to be causal of resistance for at least one drug. Both simpler L2 penalized regression and complex machine learning models had high predictive performance. The average AUCs for our highest performing model was 0.979 for first-line drugs and 0.936 for second-line drugs during repeated cross-validation. On an independent validation set, the highest performing model showed average AUCs, sensitivities, and specificities, respectively, of 0.937, 87.9%, and 92.7% for first-line drugs and 0.891, 82.0% and 90.1% for second-line drugs. Our method outperforms existing approaches based on direct association, with increased sum of sensitivity and specificity of 11.7% on first line drugs and 3.2% on second line drugs. Our method has higher predictive performance compared to previously reported machine learning models during cross-validation, with higher AUCs for 8 of 10 drugs.Statistical models, especially those that are trained using both frequent and less frequent variants, significantly improve the accuracy of resistance prediction and hold promise in bringing sequencing technologies closer to the bedside.

    View details for DOI 10.1016/j.ebiom.2019.04.016

    View details for Web of Science ID 000470091600048

    View details for PubMedID 31047860

    View details for PubMedCentralID PMC6557804

  • Association of Caffeine Intake and Caffeinated Coffee Consumption With Risk of Incident Rosacea in Women JAMA dermatology Li*, S., Chen*, M. L., Drucker, A. M., Cho, E., Geng, H., Qureshi, A. A., Li, W., (*co-first authors) 2018; 154 (12): 1394-1400

    View details for DOI 10.1001/jamadermatol.2018.3301

  • Complete resolution of erythema elevatum diutinum using oral sulfasalazine. Dermatology online journal Chen, M. L., Chlopik, A., Hoang, M. P., Smith, G. P. 2017; 23 (10)

    Abstract

    Erythema elevatum diutinum (EED) is a rare, chronic small-vessel vasculitis that presents as firm, red, violaceous, or brown papules and nodules on the extensor surfaces of the limbs. Oral dapsone is considered first-line therapy for EED; in the current case report, a patient presenting with EED began dapsone treatment and symptoms subsided within two weeks. Seven months later, the patient became pregnant and stopped dapsone owing to her concerns with dapsone use during pregnancy, resulting in recurrence of EED symptoms. We present a novel treatment approach with oral sulfasalazine, which was given to the patient in lieu of dapsone and resulted in complete resolution of EED symptoms.

    View details for PubMedID 29469798