Taylor A. Burke

Background: Exposure and response prevention is a first-line intervention for obsessive-compulsive disorder (OCD), yet many individuals with contamination concerns do not access care. Virtual reality exposure–based therapy (VRET) may improve scalability and acceptability. Objective: This pilot study evaluated the feasibility and acceptability of a standardized single-session VRET protocol targeting contamination concerns and whether it elicited within-session anxiety and exploratory contamination symptom change at 1-month follow-up. Methods: We conducted a single-arm pilot study in adults with elevated contamination concerns and no formal OCD diagnosis, recruited via convenience sampling. Participants completed a baseline survey, an in-laboratory VRET session using a standardized virtual public toilet environment, and a follow-up survey. Outcomes included momentary anxiety (Subjective Units of Distress Scale) during exposure, affect (positive and negative affect schedule) across time points, and contamination symptoms (Obsessive-Compulsive Inventory–Revised contamination subscale) at baseline and follow-up. Usability (System Usability Scale) and VR sickness were also assessed. Within-session outcomes used repeated-measures ANOVA or Friedman tests; symptom change used paired t tests (α=.05); point estimates include 95% CIs. Missing data were addressed using multiple imputation (random forest; m=5); 37.5% of participants did not complete the follow-up survey (overall missingness: 5.47%). Results: Sixteen participants were included (aged 18‐32 years). Anxiety increased during exposure tasks and decreased after virtual hand washing in both trials (Exposure 1 Friedman Test: χ ² ₃ =28.56; P <.001; W=0.6); Exposure 2 repeated measures ANOVA: F _{1.85, 27.81} =5.35; P =.01; Greenhouse-Geisser corrected=0.058. Negative and positive affect both changed significantly across time points (negative affect: Friedman Test: χ ² ₃ =13.76; P =.003; W=0.29 and positive affect: repeated measure ANOVA: F _{3, 45} =4.60‐4.71; P =.006-.007; Greenhouse-Geisser corrected=0.07‐0.073). Contamination symptoms did not significantly change from baseline to follow-up (mean change 1.30, SD 3.39; 95% CI –0.36 to 2.96; P =.12). Usability was adequate (System Usability Scale mean 69.5, 95% CI 62.79-76.21). Conclusions: This study is among the first to systematically evaluate a standardized single-session contamination-focused VRET protocol in adults with elevated contamination concerns who did not meet diagnostic criteria for OCD, a subthreshold population underrepresented in prior VRET research, which has focused on clinically diagnosed samples and multisession protocols. The protocol proved feasible and acceptable, eliciting within-session anxiety and providing benchmarks in existing clinical literature. Contamination symptom change at 1-month follow-up was not statistically significant, and the effect size estimate was sensitive to the missingness assumption, underscoring the need for adequately powered multisession designs in future trials. The usability profile and VR sickness levels that did not prevent session completion suggest self-administered or minimally supervised delivery warrants evaluation, with implications for scalable early intervention in individuals with elevated contamination concerns who have not accessed formal treatment.

Background: Longitudinal measurement of depression severity in outpatient psychiatric care is limited by infrequent standardized assessments. Although psychiatric clinical notes capture illness burden and functional impairment, this information is rarely quantified for analysis. Objective: To evaluate whether large language models (LLMs) can infer clinically meaningful measures of depression severity from outpatient psychiatry notes. Methods: We sampled 91,651 outpatient psychiatry notes from 8,287 adult patients across 58 clinics within a large academic medical center between 2015 and 2021. A HIPAA-compliant LLM (OpenAI GPT-5.2) was prompted to independently estimate three depression severity scores (Patient Health Questionnaire-9 [PHQ-9], Hamilton Depression Rating Scale [HAM-D], and depression-specific Clinical Global Impression-Severity [CGI-S]) from notes, with patient-reported PHQ-9 content within notes redacted to prevent biasing. Convergent validity was assessed against patient-reported PHQ-9 (n=3,757), study-clinician chart review (n=125), and treating-clinician suicide risk assessments (SRA; n=2,985). Predictive validity was evaluated using survival models of antidepressant switching and psychiatric emergency visits. Discriminant validity across diagnoses and consistency across demographic groups and clinics were also evaluated. Results: 10.8% of eligible visits had a PHQ-9 recorded within 7 days before the encounter. LLM-inferred PHQ-9 scores showed moderate agreement with patient-reported PHQ-9 (Cohen’s κ=0.64, 95%CI:0.62-0.66; Pearson r=0.67, 95%CI: 0.65-0.68). Stronger agreement was found between LLM CGI-S and study-clinician chart review (κ _rater1 =0.79, 95%CI: 0.70-0.85; κ _rater2 =0.67, 95%CI: 0.58-0.77; r=0.86 with mean rating, 95%CI: 0.80-0.90). In prospective analyses, LLM CGI-S predicted antidepressant switching (C-index=0.60; CI95%: 0.58-0.62) and psychiatric emergency visits (C-index=0.63; 95%CI: 0.57-0.68), which was comparable to the predictive performance of patient-reported PHQ-9 and treating-clinician SRA. Correlations between LLM CGI-S and patient-reported PHQ-9 were consistent across clinics (I ² <0.1) but significantly lower among Black (r=0.48, 95%CI: 0.38-0.57) and Hispanic (r=0.43, 95%CI: 0.27-0.56) patients. Conclusions: LLM-inferred depression severity scores from psychiatric outpatient notes support longitudinal, standardized phenotyping of depression severity, such as for routine outcome monitoring. These results have implications for facilitating genetic, pharmacoepidemiologic, and antidepressant treatment effectiveness studies using real-world evidence.

Suicidal ideation (SI) fluctuates rapidly within individuals, with approximately 50% of its variance occurring within-person. Standard ecological momentary assessment (EMA) protocols use fixed schedules that fail to capture these rapid fluctuations, missing critical risk windows and opportunities for intervention. Here, we present ARES-SI (Adaptive Reinforcement-Enhanced Sampling for Suicidal Ideation), a reinforcement learning system that personalizes EMA timing based on individual risk patterns. ARES-SI combines random forest models to predict active SI probability and non-response likelihood with a Q-learning algorithm to determine optimal assessment intervals (30 minutes for high-risk states, 3 hours for low-risk states, and next day to reduce burden). We trained the system using EMA data from 98 participants across three studies with varied sampling frequencies (4–15 assessments per day; 2,359 positive active SI responses) over 28-day protocols. In validation testing (N = 36; 353 positive active SI responses), the random forest models achieved an AUC of 0.865 for active SI prediction and 0.750 for non-response prediction. ARES-SI demonstrated 65% higher action selection accuracy than random scheduling (0.781 vs. 0.474; macro F1 = 0.650 vs. 0.377), with superior detection of high-risk windows (recall = 0.810 for 30-minute assessments vs. 0.392 for random) and efficient low-risk scheduling (precision = 0.972 for 3-hour assessments vs. 0.876 for random). These findings demonstrate that reinforcement learning can identify when individuals are most vulnerable, providing the temporal precision needed for just-in-time adaptive interventions in suicide prevention.

Background: The distinction between passive and active suicidal ideation (SI) and their underlying etiologies remains poorly understood. The Interpersonal Theory of Suicide implicates guilt, loneliness, and hopelessness in these SI subtypes, but there is minimal work testing these relationships in real time, capturing clinically meaningful fluctuations in SI. We conducted the first ecological momentary assessment (EMA) study to distinguish between passive and active SI in adolescents, and the first study to evaluate moment-to-moment etiological factors and mediators of passive and active SI in this age group. Methods: Participants ( N = 104) were adolescent psychiatric inpatients ( M _age = 15.1; 72.12% female). They completed an EMA protocol including measures of guilt, loneliness, hopelessness, and passive and active SI for four weeks post-discharge. Multilevel modeling was used to evaluate guilt and loneliness, respectively, as predictors of prospective passive and active SI, respectively. We also evaluated whether hopelessness mediated the interaction between guilt and loneliness in predicting future SI. Hopelessness was also evaluated as a mediator between passive and active SI. Results: Guilt predicted prospective passive and active SI, respectively, whereas loneliness only predicted prospective passive SI. The interaction between guilt and loneliness did not predict active SI, and hopelessness did not mediate the association between guilt and active SI. Passive SI prospectively predicted active SI, but hopelessness did not mediate this association. Conclusions: Findings suggest that passive and active SI may share overlap but also differences in their etiologies. Their relationship with etiological factors and mediators may differ as a function of temporal scale.

Background: Youth suicide is a critical public health crisis. Subscription-based parental digital monitoring apps have emerged to monitor youths’ web-based activities and promptly alert parents in case of detected suicide risk. Parents’ responses to digital suicide risk alerts could significantly influence their children’s immediate and long-term well-being. However, parent experiences receiving these alerts and their impact on parent–child and co-parent relationships remain unclear. Objective: This study aimed to examine parental perceptions of digital suicide risk alerts, as well as characterize parents’ emotional, cognitive, and behavioral responses to alert receipts, and evaluate the impact on their relationships with their child and co-parent. Methods: Parents subscribed to the MMGuardian app (Pervasive Group Inc) who received a suicide risk alert were invited to complete a web-based survey. The survey assessed demographics, experiences with emotional, cognitive, and behavioral responses to alerts, and the impact on parent–child and co-parent relationships. Results: The final sample included 217 parents with an average age of 40.1 (SD 6.82) years and who were predominantly women (183/217, 84.3%), White (189/217, 87.1%), and non-Hispanic (183/217, 84.3%). Approximately 32% (69/217) of parents reported learning for the first time that their child may be at risk for suicide or self-harm through this alert. Parents reported a moderate level of concern about their child’s self-injurious thoughts and behaviors risk based on review of the alert (mean 2.12, SD 0.985). Parents reported experiencing a range of negative emotions in response to receiving a risk alert, with nervousness and sadness being the most common. Self-blame was the most strongly endorsed cognitive response, while thoughts of hopelessness, permanence (eg, “My child is never going to be okay”), and skepticism (eg, “My child is manipulating me”) were less common. Common behavioral actions taken after receiving the alert involved telling their child they loved them, talking with their child to learn more, validating their child’s thoughts and feelings, helping their child use coping skills, and discussing mental health with their child. Approximately 72% (156/217) of parents reported that the alert positively impacted their closeness with their child. Participants were between moderately and quite a bit in sync (mean 2.25, SD 1.26) with their co-guardian regarding alignment in perceptions of their child’s current risk as well as their corresponding thoughts, emotions, and plans related to child risk. Conclusions: Findings highlight several points for potential intervention to better support parents in navigating responding to their child’s possible suicide risk. As digital parental monitoring tools grow in popularity, research is essential to optimize these tools for the benefit of children and families. Improved understanding of the impact of digital suicide risk alerts will guide the development of effective digital support tools, empowering parents to help both themselves and their child.

Background: The months following psychiatric hospitalization are associated with heightened suicide risk among adolescents. Better characterizing predictors of trajectories of suicidal ideation (SI) post-discharge is critical. Method: We examined trajectories of SI over 18 months post-discharge and emotional processing variables (recognition, reactivity, and regulation) as predictors using a multi-method approach. Participants were 180 adolescents recruited from a pediatric psychiatric inpatient unit, assessed during hospitalization and 3, 6, 12, and 18-months post-discharge. At each time-point, participants reported on SI; at baseline, they completed measures of emotion dysregulation, reactivity, and a behavioral task measuring facial emotion recognition. Results: A three-group model best fits the data (Chronic SI, Declining SI, and Subthreshold SI groups). The Chronic SI group, compared to the Declining SI group, had greater difficulty identifying children’s sad facial expressions. The Declining SI group compared to the Subthreshold SI group reported greater overall emotion dysregulation and difficulties engaging in goal-directed behavior. No other emotional processing variable was significantly associated with specific SI trajectories. Conclusions: The findings suggest that difficulties in properly identifying peer emotions may be predictive of resolution of severe SI post-discharge. Furthermore, the results suggest that emotion regulation may be an important target for discharge planning.

Nonsuicidal self-injury (NSSI) is one of the strongest predictors of suicidal behavior. Despite this, the field still has a limited understanding of the mechanisms by which this relationship is conferred. As there has been a proliferation of research in this area over the past decade, an updated systematic review of the empirical research examining potential factors driving the relationship between NSSI and suicide-related outcomes is needed to move the field forward. We identified only 18 studies examining moderators or mediators of the relationship. Research to date does little to improve our understanding of the robust, prospective relationship between NSSI and suicide outcomes, highlighting a foundational gap in both the empirical and theoretical literature. Thus, we propose the Nonsuicidal to Suicidal Self-Injury Pathway Model, an expanded theoretical model of this relationship drawing on extant theory and empirical research, as well as discuss future directions for work in this area.

Objectives: Nonsuicidal self-injury (NSSI) is one of the strongest predictors of suicidal behavior. Despite this, the field still has a limited understanding of the mechanisms by which this relationship is conferred. Method: We conducted a systematic review of the empirical research examining potential factors driving (i.e., moderators, mediators) the relationship between NSSI and suicidal behavior to address this gap in the literature. Results: We identified only 15 studies examining moderators or mediators of this relationship, examining 40 unique mediators and 22 unique moderators. Three prominent weaknesses were identified in the reviewed literature: (1) limited intersection with existing theoretical models of the NSSI – suicidal behavior relationship, (2) little replication of findings across studies (i.e., only four mediators and four moderators assessed in multiple studies), and (3) only one of the included studies utilized a prospective design. Research to date does little to improve our understanding of the theoretical or prospective relationship between NSSI and suicidal behavior, highlighting a foundational gap in the literature. Discussion: We propose the Nonsuicidal to Suicidal Self-Injury Pathway Model, a new conceptual model of the relationship between NSSI and suicidal behavior, drawing on extant theory and empirical research; we discuss future directions for work in this area.

BACKGROUND Youth suicide is a critical public health crisis. Subscription-based parental digital monitoring apps have emerged to monitor youths’ online activities and promptly alert parents in the case of detected suicide risk. Parents’ responses to digital suicide risk alerts could significantly influence their children’s immediate and long-term well-being. Yet, parent experiences receiving these alerts, and their impact on parent-child and co-parent relationships, remain unclear. OBJECTIVE This study aimed to examine parental perceptions of digital suicide risk alerts, as well as characterize parents’ emotional, cognitive, and behavioral responses to alert receipt, and evaluate the impact on their relationships with their child and co-parent. METHODS Parents subscribed to the MMGuardian mobile app who received a suicide risk alert were invited to complete an online survey. The survey assessed demographics, experiences with and emotional, cognitive, and behavioral responses to alerts, and the impact on parent-child and co-parent relationships. RESULTS The final sample included 217 parents who were on average 40.1 years old (SD = 6.82) and who were predominantly women (84.3%), white (87.1%), and non-Hispanic (84.3%). Approximately 32% of parents reported learning for the first time that their child may be at risk for suicide or self-harm through this alert. Parents reported a moderate level of concern about their child’s SITB risk based on review of the alert (M = 2.12; SD = .985). Parents reported experiencing a range of negative emotions in response to receiving a risk alert, with nervousness and sadness being the most common. Self-blame was the most strongly endorsed cognitive response, while thoughts of hopelessness, permanence (e.g., “My child is never going to be okay”), and skepticism (e.g., “My child is manipulating me”) were less common. Common behavioral actions taken after receiving the alert involved telling their child they loved them, talking with their child to learn more, validating their child’s thoughts and feelings, helping their child use coping skills, and discussing mental health with their child. Approximately 72% of parents reported that the alert positively impacted their closeness with their child. Participants were between moderately and quite a bit in sync (M = 2.25; SD = 1.26) with their co-guardian regarding alignment in perceptions of their child’s current risk as well as their corresponding thoughts, emotions, and plans related to child risk. CONCLUSIONS Findings highlight several points for potential intervention to better support parents in navigating responding to their child’s possible SITB risk. As digital parental monitoring tools grow in popularity, research is essential to optimize these tools for the benefit of children and families. Improved understanding of the impact of digital suicide risk alerts will guide the development of effective digital support tools, empowering parents to help both themselves and their child.