Publications

BACKGROUND: Suicide is an important public health concern in the United States and around the world. There has been significant work examining machine learning approaches to identify and predict intentional self-harm and suicide using existing data sets. With recent advances in computing, deep learning applications in health care are gaining momentum.

OBJECTIVE: This study aimed to leverage the information in clinical notes using deep neural networks (DNNs) to (1) improve the identification of patients treated for intentional self-harm and (2) predict future self-harm events.

METHODS: We extracted clinical text notes from electronic health records (EHRs) of 835 patients with International Classification of Diseases (ICD) codes for intentional self-harm and 1670 matched controls who never had any intentional self-harm ICD codes. The data were divided into training and holdout test sets. We tested a number of algorithms on clinical notes associated with the intentional self-harm codes using the training set, including several traditional bag-of-words-based models and 2 DNN models: a convolutional neural network (CNN) and a long short-term memory model. We also evaluated the predictive performance of the DNNs on a subset of patients who had clinical notes 1 to 6 months before the first intentional self-harm event. Finally, we evaluated the impact of a pretrained model using Word2vec (W2V) on performance.

RESULTS: The area under the receiver operating characteristic curve (AUC) for the CNN on the phenotyping task, that is, the detection of intentional self-harm in clinical notes concurrent with the events was 0.999, with an F1 score of 0.985. In the predictive task, the CNN achieved the highest performance with an AUC of 0.882 and an F1 score of 0.769. Although pretraining with W2V shortened the DNN training time, it did not improve performance.

CONCLUSIONS: The strong performance on the first task, namely, phenotyping based on clinical notes, suggests that such models could be used effectively for surveillance of intentional self-harm in clinical text in an EHR. The modest performance on the predictive task notwithstanding, the results using DNN models on clinical text alone are competitive with other reports in the literature using risk factors from structured EHR data.

Background: Clinical trials are key to ensuring high-quality, effective, and safe health care interventions, but there are many barriers to their successful and timely implementation. Difficulties with participant recruitment and enrollment are largely affected by difficulties with obtaining informed consent. Teleconsent is a telemedicine- based approach to obtaining informed consent and offers a unique solution to limitations of traditional consent approaches. Methods: We conducted a survey among 134 clinical trial researchers in academic/university-, industry-, and clinically based settings. The survey addressed important aspects of teleconsent, potential teleconsent enhancements, and other telehealth capabilities to support clinical research. Results: The majority of respondents viewed teleconsent as an important approach for obtaining informed consent and indicated that they would likely use teleconsent if available. Consenting participants at remote sites, increasing access to clinical trials, and consenting participants in their homes were viewed as the greatest opportunities for teleconsent. Features for building, validating, and assessing understanding of teleconsent forms, mobile capabilities, three-way teleconsent calls, and direct links to forms via recruitment websites were viewed as important teleconsent enhancements. Other telehealth capabilities to support clinical research, including surveys, file transfer, three-way video, screenshare, and photo capture during telemedicine visits, and proposed telemedicine capabilities such as video call recording, ID information capture, and integration of medical devices, were also viewed as important. Conclusions: Teleconsent and telemedicine are promising solutions to some common challenges to clinical trials. Many barriers to study recruitment and enrollment might be overcome by investing time and resources and further evaluating this technology.

Background and Introduction: Comprehensive monitoring and follow-up after traumatic injury is important for psychological recovery. However, scalable services to facilitate this are limited. Automated text message-based symptom self-monitoring (SSM) may be a feasible approach. This study examined its implementation and utility in identifying patients at risk for mental health difficulties after traumatic injury.Materials and Methods: Five hundred two patients admitted to a Level I trauma center between June 20, 2016 and July 31, 2017 were offered enrollment in a text message-based SSM service. Patients who enrolled received daily text message prompts over 30 days and most participated in a mental health screening 30 days postbaseline.Results: Approximately 67% of patients enrolled in the service; of these, 58% responded to the text messages, with an average response rate of 53%. Younger patients and those with elevated peritraumatic distress were more likely to enroll. Patients with higher levels of mental health stigma, who were White, or had been in a motor vehicle collision were more likely to enroll and respond to text messages once enrolled. Patients' daily ratings of distress detected clinically elevated 30-day mental health screens with high sensitivity (83%) and specificity (70%).Discussion and Conclusions: Text message-based SSM can be implemented as a clinical service in Level I trauma centers, and patient participation may increase engagement in mental health follow-up. Further, it can inform the use of risk assessments in practice, which can be used to identify patients with poor psychological recovery who require additional screening.

Following publication of the original article [1], the authors reported an error in one of the authors' names.

BACKGROUND: Social isolation is an important social determinant that impacts health outcomes and mortality among patients. The National Academy of Medicine recently recommended that social isolation be documented in electronic health records (EHR). However, social isolation usually is not recorded or obtained as coded data but rather collected from patient self-report or documented in clinical narratives. This study explores the feasibility and effectiveness of natural language processing (NLP) strategy for identifying patients who are socially isolated from clinical narratives.

METHOD: We used data from the Medical University of South Carolina (MUSC) Research Data Warehouse. Patients 18 years-of-age or older who were diagnosed with prostate cancer between January 1, 2014 and May 31, 2017 were eligible for this study. NLP pipelines identifying social isolation were developed via extraction of notes on progress, history and physical, consult, emergency department provider, telephone encounter, discharge summary, plan of care, and radiation oncology. Of 4195 eligible prostate cancer patients, we randomly sampled 3138 patients (75%) as a training dataset. The remaining 1057 patients (25%) were used as a test dataset to evaluate NLP algorithm performance. Standard performance measures for the NLP algorithm, including precision, recall, and F-measure, were assessed by expert manual review using the test dataset.

RESULTS: A total of 55,516 clinical notes from 3138 patients were included to develop the lexicon and NLP pipelines for social isolation. Of those, 35 unique patients (1.2%) had social isolation mention(s) in 217 notes. Among 24 terms relevant to social isolation, the most prevalent were "lack of social support," "lonely," "social isolation," "no friends," and "loneliness". Among 1057 patients in the test dataset, 17 patients (1.6%) were identified as having social isolation mention(s) in 40 clinical notes. Manual review identified four false positive mentions of social isolation and one false negatives in 154 notes from randomly selected 52 controls. The NLP pipeline demonstrated 90% precision, 97% recall, and 93% F-measure. The major reasons for a false positive included the ambiguities of the experiencer of social isolation, negation, and alternate meaning of words.

CONCLUSIONS: Our NLP algorithms demonstrate a highly accurate approach to identify social isolation.

BACKGROUND: Although at least 1 in 10 veterans meet criteria for Posttraumatic Stress Disorder (PTSD) related to their military service, treatment seeking is strikingly low due to perceived stigma and other barriers. The National Center for PTSD produced AboutFace, * a web-based video gallery of veterans with PTSD who share their personal stories about PTSD and how treatment has turned their lives around.

INTRODUCTION: We conducted a two-stage evaluation of AboutFace, which included (1) a usability testing phase and (2) a randomized, controlled trial phase to explore the feasibility of incorporating AboutFace into a specialized outpatient clinic for PTSD.

MATERIALS AND METHODS: Twenty veterans participated in the usability testing phase in which they answered moderator posed questions regarding AboutFace, while actively exploring the website. Sixty veterans participated in the study after completing a PTSD clinic evaluation and were randomized to receive an educational booklet about PTSD treatment or AboutFace before starting treatment. Stigma and attitudes about treatment seeking were assessed at baseline and 2 weeks later.

RESULTS: Veterans had positive attitudes about AboutFace and gave suggestions for improvement. Veterans in both conditions reported improved attitudes toward mental illness and treatment seeking from baseline to the 2-week follow-up.

DISCUSSION: AboutFace is a promising peer-to-peer approach that can be used to challenge stigma and promote help seeking.

CONCLUSIONS: This use of an online peer approach is innovative, relevant to a wide range of healthcare conditions, and has the potential to increase access to care through trusted narratives that promote hope in recovery.

Mental health systems need scalable solutions that can reduce the efficacy-effectiveness gap and improve mental health outcomes in community mental health service settings. Two major challenges to delivery of high-quality care are providers' fidelity to evidence-based treatment models and children's and caregivers' engagement in the treatment process. We developed a novel, tablet-based application designed to enhance via technology the quality of delivery of trauma-focused cognitive-behavioral therapy (TF-CBT). We piloted its use in four community mental health service organizations using a blocked randomized controlled trial to examine the feasibility of implementing tablet-facilitated TF-CBT versus standard TF-CBT with 13 providers and 27 families. Provider fidelity and child engagement in treatment were observationally measured via session audio recording. Parent and child perceptions of the tablet application were assessed using structured interviews and mixed-method analyses. Providers actively and appropriately used tablet TF-CBT to facilitate treatment activities. Providers and families expressed high satisfaction with its use, demonstrating acceptability of this approach. Youth and caregivers in both conditions reported high alliance with their providers. Overall, we found that tablet-facilitated treatment is accepted by providers and families and may be integrated into mental health treatment with minimal training. Further study is needed to examine the extent to which technology-based applications may enhance the reach, quality, and clinical outcomes of mental health treatment delivered to children and families.

Selective mutism (SM) is an anxiety disorder marked by withdrawal of speech in particular social situations. Treatment is often difficult, requiring attention to several characteristics particular to the disorder. Therapeutic tools and activities such as games and mobile applications (apps) may be particularly advantageous to behavioral therapy for SM. A 2-session hierarchy for shaping successive approximations of speech in SM was piloted with 15 children, 5 to 17 years old, who were randomly assigned to shaping while using mobile apps, other therapeutic tools/activities, and reinforcement alone. Very strong treatment gains were observed: 13 of 15 (88.7%) children completed the hierarchy during the first session and 14 (93.3%) did so during the second session, with the final child completing all but the final step (i.e., to ask and respond to at least 5 open-ended questions). Moreover, all 15 children spoke to the clinician within 59 minutes of treatment (M = 17 minutes), and 14 (93.3%) children held five, 5-minute conversations with additional unknown adults during the second session. This occurred regardless of the inclusion of therapeutic tools/activities, although preliminary patterns of responding were observed such that children shaped while using mobile apps tended to show less self-reported and physiologically measured anxious distress. The utility of therapeutic activities and mobile apps when treating SM is discussed as well as areas for future research.

The present review examines the relations between sleep disturbance and anxiety in children and adolescents. The review begins with a detailed discussion of normative developmental trends in sleep, and the relation between sleep quality and emotion dysregulation in children. The extant literature on sleep disturbance in clinically anxious children with a focus on subjective versus objective measures of sleep is then summarized in detail. Finally, a review of the reciprocal relationship between sleep and emotion regulation is provided. The available research suggests that sleep disturbance is quite prevalent in children with anxiety disorders, although the directionality of the association between sleep disturbance and anxiety in children remains unclear. Despite this limitation, a reciprocal relationship between sleep quality and anxiety appears to be well established. Research using objective measures of sleep quality (e.g. polysomnography, sleep actigraphy, sleep bruxism) is warranted to better understand this relation. Further, complicating factors such as the environment in which sleep quality is measured, the developmental stage of participants, varying severity of anxiety and the timeframe during which assessment takes place should all be considered when examining sleep disturbance in this population.