Applications of Latent Variable Models and Other Advanced Statistical Methods in Mental Health Settings

Empathic accuracy (EA) is the ability to understand the thoughts and emotions of another person, which plays an essential role in shaping social and psychological interactions. Conventional EA analyses often ignore the misalignment of the minds and feelings between people and yield significantly biased results. In this paper, we consider the empathy rating as a function and separate its temporal and vertical variability using the warping-invariant metric-based alignment, which can deal with arbitrary patterns of misalignment. In addition, we propose a penalized functional alignment approach that bounds the temporal alignment of the perceiver's response to the target's emotion to avoid over-alignment. To our knowledge, the proposed approach is among the first to adjust arbitrary patterns of misalignment in the EA study area. We demonstrate the effectiveness of the proposed method using simulated data and two EA data of video and music. 

Increased time on social media platforms (SMP) has contributed to mental health crisis, particularly among youth. This study aims to identify different types of SMP use engagement patterns (e.g., passive versus active) and investigate their relationship with anxiety symptoms among emerging young adults. Participants provided their SMP use data from Facebook, Instagram, Snapchat, Twitter, and YouTube. Generalized Linear Models (GLM) with Tweedie distribution were constructed to model outgoing engagement variabilities between passive and active use. Growth Mixture Models (GMMs) were also applied to identify latent SMP use patterns that maybe related to participants' anxiety symptoms. Daily outgoing variation between active and passive engagement was associated with anxiety score at follow-up, meaning more variation was associated with less anxiety symptoms. 3-Class latent patterns were identified by GMMs using overall data or split by daytime or nighttime use, and significant association with anxiety symptoms at follow-up were identified. SMP use and its impact on youth are important. Future applications will apply these methods to college students and other mental health domains. 

The Mental Illness Calibration Study (MICS) is a clinical follow-up study designed to assess mental health disorders and produce estimates of mental illness for the National Survey on Drug Use and Health (NSDUH). From 2008-2012, a similar study (the Mental Health Surveillance Study) was fielded and used to produce a NSDUH model for predicting mental illness, calibrated to the Diagnostic and Statistical Manual of Mental Disorders Fourth Edition (DSM-IV) standards. In 2023-2024, 4,000 adult NSDUH respondents aged 18 or older will participate in a clinical interview within 28 days of completing their initial NSDUH interview. Data collected from these MICS clinical interviews will be used to update the current NSDUH statistical model for mental illness based on DSM-5 criteria.

This presentation covers: (1) differences between the prior mental illness model and MICS model for estimating mental illness, (2) the modeling procedures planned for 2023-2024 MICS data, and (3) preliminary results of the updated model parameters and estimated rates of mental illness using the 2023 MICS data. 

Introduction
Depression, anxiety, and posttraumatic stress (PTS) symptoms are expected and common following a traumatic brain injury (TBI). These symptoms however have an often-overlapping presentation that can be challenging to address for rehabilitation researchers and clinicians. The current literature focuses on the relationship between these symptoms predominately after mild TBI. This study looks to evaluate several transdiagnostic, structural models to understand and better utilize reported psychological symptoms in the wake of moderate/severe TBI.
Methods
This study used 1,258 participants enrolled in both the TBI Model Systems and The Characterization and Treatment of Chronic Pain after TBI study. Confirmatory factor analysis (CFA) was initially conducted on the parent scales of the PHQ-9 (Patient Health Questionnaire-9), GAD-7 (General Anxiety Disorder-7), and PCL-5 (Posttraumatic Stress Disorder Checklist for DSM-5) to better understand the transdiagnostic relationship of these PTS measures. Exploratory factor analysis (EFA) was then performed to identify new structures of PTS and CFA to validate the model fit. EFA used a weighted least square mean and adjusted estimation with a Geomin or Bi-Geomin rotation. Global fit indices were calculated using the comparative fit index (CFI), Tucker-Lewis index (TLI), standardized root mean square residuals (SRMR), the root mean square error of approximation (RMSEA), and a chi-square test.
Results
The sample had an average age of 42 years and were mostly male and non-Hispanic white. The parent scales revealed evidence of a higher-order factor of psychological distress but these structures were found to be unstable. EFA then successfully identified a 3-factor, highly correlated single-order structure. The second-order structure of these factors revealed large factor loadings and evidence of the existence of a higher-order factor. The bifactor EFA identified a stable 4- and 5-factor structure. The CFA results of these structures found overall good model fit and internal consistency.
Conclusion
The transdiagnostic approach of this study revealed instability amongst our current PTS measures and also identified second-order and bifactor structural models of PTS symptoms for a moderate/severe TBI population. These findings may help improve statistical modeling, research, and treatment options for this population. 

Post-traumatic stress disorder (PTSD) and abnormal spirometry are highly prevalent mental and health conditions in World Trade Center (WTC) responders. We hypothesized that PTSD symptomatology and abnormal spirometry are synergistically associated with cognitive performance in WTC responders. PTSD symptomatology was assessed using the PCL-IV, and we calculated the FEV1/FVC ratio to measure pulmonary function and characterize abnormal spirometry. Cogstate assessment measured cognitive performance. We evaluated PTSD, pulmonary function and their interaction on cognitive performance by linear regressions adjusting for confounders. PTSD symptomatology and pulmonary function appeared to have a significant synergistic effect on cognitive performance in that higher severity of PTSD symptomatology in the presence of lower pulmonary function was associated with poorer cognitive performance. Results suggested chronic stress and lung damage might share underlying biological mechanisms, including inflammatory and oxidative stress pathways, which might also affect the brain. Early intervention efforts to mitigate preventable cognitive decline in high-risk populations should be studied. 

In this report, we attempt to identify subgroups of participants experiencing differential patterns of transitions over time. The data are from the Texas Youth Depression and Suicide Research Network (TX-YDSRN), a multi-site registry network of adolescents (8-17 years of age) who are under treatment for depression or suicidal ideation. We utilized Pediatric Patient Reported Outcomes Measurement Inventory (PROMIS -Pediatric 25): a self-reported measure of Physical Function and Mobility, Depression Severity, Anxiety, Fatigue, Peer Relationships, Pain Interference, and Pain Severity of these participants over a period of 6-months during which 4 data-points were collected. The model selection was based on fit statistics such as Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), Consistent AIC (CAIC), and Adjusted BIC (ABIC), etc.