About

Kinh Luan Phan, MD, is a Professor of Psychiatry at Ohio State College of Medicine. His research aims to better understand mental disorders and their causes, as well as to identify new, more effective treatments, including both medicine and behavioral therapy. His goal is to promote recovery and resilience by studying mental illness from childhood through adulthood to better understand how it impacts individuals differently. His research primarily focuses on post-traumatic stress disorder, depression, and anxiety disorders, but also includes studies on drug and alcohol addiction, and the connection that stress has with depression and anxiety.

Research topics

• Psychiatry
• Psychology
• Clinical psychology
• Computer Science
• Neuroscience
• Social psychology
• Medicine
• Data science
• Cognitive psychology

Selected publications

• Rural, Regional, Remote Residence, Anxiolytic Use, Multiple Site Surgery, Smoking Status Associated With Patient‐Initiated Communication After Mohs Micrographic Surgery

  Australasian Journal of Dermatology · 2026-03-26

  article1st author

  Mohs micrographic surgery is increasingly used for the surgical excision of skin cancers, particularly those located on anatomically complex areas [1-3]. Patients may initiate communication following surgery if they have unaddressed concerns relating to their surgery [4]. With the rise of telemedicine, phone calls and emails are becoming a more common form of communication, especially for those who live in regional and remote areas with reduced access to face-to-face specialist medical services [5]. The aim of this study was to determine reasons why patients initiate contact with the Mohs surgeon's office after surgery and to provide insight into the timing of these contacts, as well as factors that may predispose patients to initiate communication. Data was prospectively collected from 548 consecutive patients who underwent Mohs surgery and reconstruction with a single dermatologist (M.J.L) at The Skin Hospital, Darlinghurst from August 2020 to July 2023. Data on patient self-initiated contact with the hospital within 90 days of their Mohs surgery with a question or concern related to their surgery, as well as the form, timing and outcome of communications and nature of concern were collected. Statistical analysis was performed in IBM SPSS version 23 (Chicago, IBM Corp). Sixty-five patients (11.9%) initiated communication following Mohs surgery. Demographic, medical, tumour and surgical characteristics are summarised in Table 1. In the final multivariate analysis, 4 factors were found to be independently associated with patient-initiated communication (Table 2): rural/regional/remote residence, perioperative anxiolytic use, multiple-site surgery and smoking. Most of the communication was via phone (Table 3). The mean number of encounters was 1.4, and the mean time of communication postoperative was 5.3 days. The main concerns were related to dressings, wound care and surgical site appearance, unrelated to bleeding/infection. There is a significant shortage and geographical maldistribution of dermatologists in Australia. Only 4% of Australian dermatologists work in rural areas (Mohs surgery or non-Mohs clinical practice), and 2% in regional areas [6], meaning patients in these areas face significant challenges in accessing dermatologic care. The Australasian College of Dermatologists approved register of Mohs practitioners in New South Wales (NSW) lists 30 clinicians, of whom only 2 (6.6%) are primarily based in a regional or rural setting (Modified Monash Model MM 2–7) [6, 7]. Furthermore, 2 Mohs practitioners also provide regular services to an MM3 location in addition to their primary MM1 site of practice. Most patients in rural/regional/remote Australia who require access to specialist Mohs surgery will need to travel to consult and undertake surgery in a metropolitan setting. In the worst-case scenario, patients may be discharged after day surgery to a region where they may not have access to a local dermatologist, general practitioners or an emergency department. The use of perioperative anxiolytics in this study was used as a surrogate for patient anxiety. Patients who are anxious may require more frequent communication with their Mohs surgical team to allay their concerns during the perioperative period. Furthermore, anxiolytic use may reduce patient comprehension of postoperative instructions. With regards to multiple-site surgery, increased operative sites mean there are more wound sites required to be nursed by the patient postoperatively and an increased risk of wound concerns and complications. Smoking increases the risk of surgical complications, including wound dehiscence, flap or graft failure, prolonged healing times and infections [8] and may also be a correlate with increased baseline patient anxiety, which could increase the likelihood of the patient contacting the medical team after surgery. Most patient-initiated communication in our study was in direct relation to postoperative wound care. All patients received a combination of verbal and written wound care instructions. However, it is possible that these instructions may not have been understood or may have required elaboration or clarification. Paradoxically, information overload may result in decreased retention of information. This study has multiple limitations. The series was based on Mohs surgery cases from a single Mohs surgeon in a single institution, and as such, results may not be applicable across different groups. The included case indications were BCCs and SCCs, whereas other tumour types, including melanoma, were not represented. The size of the cancer/defect, as well as the nature of pre-operative consultation (in-person, teledermatology, on the day of procedure vs. prior separate appointment) are likely to impact rates of patient-initiated communication, which could be further evaluated in future studies. Rural/regional/remote residence cases were grouped together; however, some patients returned home immediately post-operatively, whilst some stayed in Sydney locally for day(s) before and after surgery. Other parameters such as socioeconomic status, health and language literacy and preoperative mental health status were not specifically collected and analysed. Furthermore, the scope of the study was to investigate patient-initiated communication after surgery, and we did not collect data on nurse-initiated communication after surgery. This study demonstrates risk factors associated with patient-initiated communication after Mohs surgery, including rural/regional/remote residence, anxiety, multiple-site surgery and smoking. Understanding these factors may help guide perioperative counselling, enhance the patient experience and improve surgical recovery. Study conception: K.P., M.J.L. Ethics application: K.P., M.J.L. Data collection: M.J.L., Data analysis and interpretation: K.P., M.J.L. Drafting of manuscript: K.P., M.J.L., Approval of final manuscript: K.P., M.J.L. The authors have nothing to report. The authors declare no conflicts of interest. The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

  PublisherDOI

• EZ-QC: A User-Friendly Tool for Multi-site, Multimodal Neuroimaging Data Triage and Visual Quality Control

  Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2025-09-16

  article

  Motivation: Multisite MRI studies face QC delays and inconsistencies due to data volume and variations in rater expertise. Prioritizing scans for review and supporting consistent QC could enhance data quality. Goal(s): Develop EZ-QC, a tool that triages data for efficient QC review while promoting consistency across raters. Approach: EZ-QC integrates with MRIQC to prioritize presenting data that could indicate systematic data-collection issues. It also includes tools to ensure consistent QC across raters, supporting reliable assessments. Results: EZ-QC improves QC efficiency by prioritizing the most informative data, providing structured training, and tracking interrater reliability, ensuring consistent QC for large, multisite datasets. Impact: EZ-QC improves QC efficiency by prioritizing data that are most likely to indicate systemic errors and standardizing rater training. This approach supports consistent and efficient QC, helping to maximize usable data and ensure reliable data quality in large, multisite studies.

  PublisherDOI

• Attenuated cognitive control network connectivity as a mechanism in mother–daughter intergenerational transmission of depression: a preliminary study

  Scientific Reports · 2025-11-20

  articleOpen access

  Identifying brain mechanisms implicated in the intergenerational transmission of major depressive disorder (MDD) is crucial for early detection and developing novel interventions. One promising mechanism involves altered intrinsic connectivity patterns in brain networks supporting emotion processing, including within the cognitive control network (CCN). The current preliminary study used resting state functional magnetic resonance imaging (fMRI) to examine whether altered CCN connectivity patterns are a brain-based mechanism of intergenerational risk for depression. We tested whether CCN connectivity patterns (1) differentiated mothers with and without recurrent MDD, (2) differentiated their high-risk (HR) and low-risk (LR) daughters, and (3) served as prospective predictors of daughters' depressive symptoms over a multi-wave follow-up. Participants were 56 mother-daughter pairs who completed a resting state fMRI scan. Mothers with, versus without, a history of MDD exhibited reduced connectivity between the CCN and other regions within the CCN, such as the middle frontal gyrus and dorsal anterior cingulate cortex (ACC). Reduced connectivity between the CCN and dorsal ACC was also observed in HR, relative to LR, daughters, correlated significantly among mothers and daughters, and associated with higher depression symptoms in daughters across 18 months. Reduced connectivity within the CCN may constitute one brain-based marker to further investigate as a target for prevention to attenuate the intergenerational transmission of depression.

  PublisherOA PDFDOI

• Pretreatment Dorsal Anterior Cingulate Cortex and Dorsolateral Prefrontal Cortex Activation Moderate Outcomes of Exposure-Focused Cognitive Behavioral Therapy in Pediatric Anxiety

  Biological Psychiatry Cognitive Neuroscience and Neuroimaging · 2025-12-01

  articleOpen access
  PublisherOA PDFDOI

• Neural alcohol cue reactivity as a risk factor for future drinking in youth with limited alcohol exposure

  Alcohol Clinical and Experimental Research · 2025-09-30

  article

  BACKGROUND: Heightened alcohol cue reactivity is associated with alcohol problems and poor alcohol use disorder outcomes. Theory suggests that this reflects a conditioned response, whereby cues repeatedly paired with chronic alcohol use become more salient. However, few studies have investigated the relative emergence of heightened alcohol cue reactivity. It is possible that this response occurs very early in individual drinking trajectories and may play a role in shaping future alcohol use behavior. METHODS: We tested this hypothesis in a sample of youth (n = 159; ages 16-19) with limited lifetime alcohol exposure (<100 lifetime drinks). Participants completed a baseline cue reactivity task in which they viewed images of alcoholic beverages, high-calorie foods (reward control), and neutral objects. The late positive potential (LPP), measured using electroencephalography, is a positive-going event-related potential measured 400 ms after a visual cue. The LPP was used to index cue reactivity and scored as the average amplitude from parietal site Pz. At baseline and 12 months, participants completed a retrospective calendar of alcohol use. Participants were classified into groups based on lifetime alcohol exposure: (1) ≤ 10 drinks (n = 50), (2) ≤50 drinks (n = 74), (3) >50 drinks (n = 35). RESULTS: We ran a repeated measures analysis of variance to compare the effects of task condition (alcohol cues/food cues > neutral) and drink groups on LPP amplitude. Our results revealed a significant condition × drink group interaction. Follow-up analyses revealed that, for alcohol cues only, there was a significant group effect. The highest drink exposure group exhibited greater LPP relative only to the low drink exposure group. Next, we examined whether baseline LPP to alcohol cues predicted total drinks consumed 12 months later, while controlling for baseline drinking behavior. Greater LPP to alcohol cues was associated with an increase in drinks consumed at one year. CONCLUSIONS: Heightened alcohol cue reactivity emerges with limited alcohol use and can be predictive of future drinking behaviors.

  PublisherDOI

• Comparative Study of Machine Learning Models for Textual Medical Notes Classification

  Preprints.org · 2025-11-18

  preprintOpen accessSenior author

  The expansion of electronic health records (EHRs) has generated a large amount of unstructured textual data, such as clinical notes and medical reports, which contain diagnostic and prognostic information. Effective classification of these textual medical notes is critical for improving clinical decision support and healthcare data management. This study presents a comparative analysis of four traditional machine learning algorithms, Random Forest, Logistic Regression, Multinomial Naive Bayes, and Support Vector Machine, for multiclass classification of medical notes into four disease categories: Neoplasms, Digestive System Diseases, Nervous System Diseases, and Cardiovascular Diseases. A dataset containing 9,633 labeled medical notes was preprocessed through text cleaning, lemmatization, stop-word removal, and vectorization using term frequency–inverse document frequency (TF-IDF) representation. Each model was tuned using grid search and cross validation to optimize classification performance. Evaluation metrics, including accuracy, precision, recall, and F1-score, were used to assess model performance. The results indicate that Logistic Regression achieved the highest overall accuracy (0.83), followed closely by Random Forest, Support Vector Machine and Naive Bayes (0.80 each). These findings confirm that traditional machine learning models remain robust, interpretable, and computationally efficient tools for textual medical note classification.

  PublisherOA PDFDOI

• Δ9-Tetrahydrocannabinol Alters Limbic and Frontal Functional Brain Connectomes Among Young Adult Cannabis Users

  Biological Psychiatry Cognitive Neuroscience and Neuroimaging · 2025-09-14

  articleOpen access

  BACKGROUND: -tetrahydrocannabinol (THC) disrupts brain connectivity. Few studies have examined this on a whole-brain level. We examined the effects of a single moderate dose of THC on resting-state functional brain networks among young adult cannabis users. METHODS: In a within-subject, double-blind, randomized study, 33 healthy occasional cannabis users received THC (7.5 mg, oral) and placebo before completing resting-state functional magnetic resonance imaging (rs-fMRI) during peak intoxication. Group-information-guided independent component analysis was performed on resting-state brain data to identify whole-brain networks associated with each scan. Within-samples t tests assessed for differences in intrinsic network functional connectivity and between-network functional connectivity after THC versus placebo. Additional linear models examined relationships between brain connectivity, subjective drug effects, and past-month cannabis use. RESULTS: THC reduced within-network intrinsic connectivity in corticostriatal circuits and other networks associated with sensory systems, interoceptive experiences, and spatial reasoning. THC reduced connectivity between 2 networks characterized by the anterior cingulate cortex and dorsal insula regions as well as the ventral insula and lingual gyrus, respectively. Network connectivity during THC (vs. placebo) was not related to subjective measures of drug effect or recent cannabis use. CONCLUSIONS: Our findings add to a growing literature showing that THC decreases rs-fMRI throughout the brain, impacting networks linked to the many behavioral and perceptual changes associated with THC. Future work is needed to extend these findings to clinical samples and to assess the extent to which these networks are associated with negative outcomes of chronic THC use.

  PublisherDOI

• Risk and Resilience Factors of Post-Traumatic Stress Disorder Symptom Severity Among Veterans

  Military Medicine · 2025-11-11

  article

  INTRODUCTION: Veterans experience high rates of post-traumatic stress disorder (PTSD) which impacts their daily functioning and long-term health outcomes; however, the deployment and post-deployment factors that contribute to the development and severity of symptoms are not fully understood. This study investigated risk and resilience factors during and after deployment that may impact the development of PTSD. MATERIALS AND METHODS: This Institutional Review Board-approved study examined risk and resilience factors of PTSD symptom severity using the Deployment Risk and Resilience Inventory-2 and the PTSD checklist for Diagnostic and Statistical Manual of Mental Disorders - Fifth Edition (DSM-5) among 54 Operation Enduring Freedom/Operation Iraqi Freedom/Operation New Dawn Veterans. Multiple regressions and moderated regressions examined the influence of the risk and resilience factors on PTSD symptoms. RESULTS: Adverse deployment environmental conditions were associated with worse PTSD outcomes, while post-deployment social support predicted better outcomes. Contrary to expectation, support from family and friends during deployment was associated with worse PTSD outcomes. Deployment environmental conditions significantly moderated the relationships between post-deployment social support and PTSD and between post-deployment stressors and PTSD such that the impact of these post-deployment factors on PTSD was weakened when deployment environmental conditions were worse. CONCLUSION: Consistent with previous research, social support post-deployment was an important predictor of PTSD outcomes and should be a focus of societal reintegration for Veterans. In addition, adverse conditions during deployment (e.g., cleanliness, access to showers) was a significant factor predicting post-traumatic stress disorder (PTSD) outcomes and should be prioritized to optimize Veteran health. These deployment conditions were associated with poor PTSD outcomes even for Veterans with high social support and with low life stressors. Findings suggest that enhancing environmental conditions-such as improving safety and cleanliness-may significantly reduce PTSD symptom severity among Veterans.

  PublisherDOI

• Functional Connectivity Predicting Transdiagnostic Treatment Outcomes in Internalizing Psychopathologies

  JAMA Network Open · 2025-09-03 · 3 citations

  articleOpen access

  Importance: Predicting treatment outcomes for internalizing psychopathologies (IPs), such as depression and anxiety, holds promise for advancing precision medicine. The extent to which whole-brain functional connectivity (FC) can predict treatment responses for patients with IPs across different therapeutic modalities remains unclear. Objective: To examine whether pretreatment FC patterns predict multidimensional treatment outcomes in patients with IPs and whether predictive performance generalizes across diagnoses and treatment modalities. Design, Setting, and Participants: This prognostic study analyzed baseline neuroimaging and clinical data from patients with IPs enrolled in 1 of 2 randomized clinical trials (conducted from December 2013 to February 2018 and September 2017 to December 2020). Data analysis for predictive modeling was conducted from September 2024 through March 2025. Exposures: Participants were randomized to receive 12 weeks of cognitive-behavioral therapy (CBT), selective-serotonin reuptake inhibitor (SSRI) treatment, or supportive therapy (ST). Main Outcomes and Measures: A regularized canonical correlation analysis model was trained with pretreatment FC patterns. The ability of the model to predict multidimensional treatment outcomes spanning depression, anxiety, worry, rumination, and emotion regulation was tested. The predictive model was evaluated across diagnostic categories and treatment modalities. Results: In 181 patients with IPs (mean [SD] age, 27.7 [9.2] years; 127 women [71%] and 52 men [29%]) randomized to receive CBT (n = 89), SSRI treatment (n = 46), or ST (n = 46), baseline whole-brain connectivity robustly predicted multidimensional symptom changes. Predictions were significant at the individual level (r = 0.37, P = .009, permutation test), across diagnoses (r = 0.24, P = .02) and across treatment modalities (ST: r = 0.28, P = .02; SSRI treatment: r = 0.39, P = .006; and CBT: r = 0.32, P = .003). Connections significantly contributing to the FC variate were distributed across the brain, but especially within the default mode network and the dorsal and ventral attention networks. Predictive performance decreased in models incorporating fewer neural systems or clinical outcome dimensions. Conclusions and Relevance: In this prognostic study assessing predictive models of 181 patients with IPs, whole-brain FC reliably predicted multidimensional treatment outcomes across diagnoses and treatment modalities. These results suggest an association between neural connectivity patterns within specific neural networks and clinical improvements induced by varying treatment modalities, thereby advancing efforts toward personalized treatment approaches in psychiatry.

  PublisherOA PDFDOI

• 17. Rostral Anterior Cingulate Cortex Activity During Attentional Control: A Transdiagnostic Predictor of Treatment Outcome

  Biological Psychiatry · 2025-04-09

  article
  PublisherDOI

Recent grants

• Dimensional Brain Behavior Predictors of CBT Outcomes in Pediatric Anxiety

  NIH · $2.4M · 2016–2021

• Clinical and Translational Science Award

  NIH · $33.7M · 2016–2025

• NIH Grant R01MH086517

  NIH · $2.8M · 2016

• NIH Grant I01CX000913

  NIH · 2017

• NIH Grant K23MH076198

  NIH · $754k · 2011

Frequent coauthors

• Stephanie M. Gorka

  182 shared

• Amy E. Kennedy

  152 shared

• Israel Liberzon

  Mitchell Institute

  136 shared

• Scott A. Langenecker

  The Ohio State University

  106 shared

• Mike Angstadt

  University of Michigan–Ann Arbor

  105 shared

• Daniel A. Fitzgerald

  University of Illinois Chicago

  105 shared

• Heide Klumpp

  University of Illinois Chicago

  101 shared

• Darrin M. Aase

  90 shared

Education

• Master degree, International Business

  Can Tho University

  2015