Predicting risk of suicidal behavior from insurance claims data vs. linked data from insurance claims and electronic health records.

PURPOSE: Observational studies assessing effects of medical products on suicidal behavior often rely on health record data to account for pre-existing risk. We assess whether high-dimensional models predicting suicide risk using data derived from insurance claims and electronic health records (EHRs) are superior to models using data from insurance claims alone. METHODS: Data were from seven large health systems identified outpatient mental health visits by patients aged 11 or older between 1/1/2009 and 9/30/2017. Data for the 5 years prior to each visit identified potential predictors of suicidal behavior typically available from insurance claims (e.g., mental health diagnoses, procedure codes, medication dispensings) and additional potential predictors available from EHRs (self-reported race and ethnicity, responses to Patient Health Questionnaire or PHQ-9 depression questionnaires). Nonfatal self-harm events following each visit were identified from insurance claims data and fatal self-harm events were identified by linkage to state mortality records. Random forest models predicting nonfatal or fatal self-harm over 90 days following each visit were developed in a 70% random sample of visits and validated in a held-out sample of 30%. Performance of models using linked claims and EHR data was compared to models using claims data only. RESULTS: Among 15 845 047 encounters by 1 574 612 patients, 99 098 (0.6%) were followed by a self-harm event within 90 days. Overall classification performance did not differ between the best-fitting model using all data (area under the receiver operating curve or AUC = 0.846, 95% CI 0.839-0.854) and the best-fitting model limited to data available from insurance claims (AUC = 0.846, 95% CI 0.838-0.853). Competing models showed similar classification performance across a range of cut-points and similar calibration performance across a range of risk strata. Results were similar when the sample was limited to health systems and time periods where PHQ-9 depression questionnaires were recorded more frequently. CONCLUSION: Investigators using health record data to account for pre-existing risk in observational studies of suicidal behavior need not limit that research to databases including linked EHR data.

Associations between clinicians' emotional responses, therapeutic alliance, and patient suicidal ideation.

BACKGROUND: Mental health clinicians frequently experience intense negative emotional responses to suicidal patients, which have been related to treatment outcome. This study examines the therapeutic alliance as a mediator of the relationship between clinicians' negative emotional responses at the initial encounter and patients' suicidal ideation (SI) concurrently and 1 month later. METHODS: We assessed 378 adult psychiatric outpatients (62.7% female; mean age = 39.1 ± 14.6 years) and their 61 treating clinicians. Following the initial encounter, self-report questionnaires assessed clinicians' emotional responses to their patients, patients' and clinicians' perception of the therapeutic alliance, and patients' SI. The SI was reassessed 1 month after the initial visit. Multilevel mediation analyses were performed. RESULTS: Patients' (but not clinicians') perception of the therapeutic alliance mediated the relationship between clinicians' negative emotional responses to patients and patients' SI 1 month following the initial visit (indirect effect estimate = 0.015; p < .001). CONCLUSIONS: The association between clinicians' negative emotional response and patients' prospective SI appears to be transmitted, at least partly, through the patients' perception of the poorer early quality of the therapeutic alliance. Thus, clinicians' awareness and management of their emotional states appear essential both for the identification of suicidal risk and to enhance therapeutic alliance and treatment outcomes.

Correlates of Chronic Suicidal Ideation Among Community-Based Minor-Attracted Persons.

Elevated suicidal risk has been documented in adults who are sexually attracted to minors but the topic has not been adequately investigated, particularly outside the context of the criminal justice system. In this study, risk factors for chronic suicidal ideation were assessed in 333 community-based minor-attracted persons (95% male) via an online survey. Chronic suicidal ideation was endorsed by 38.1% of the participants but was associated neither to history of sexually engaging with a child nor to prior contact with the criminal justice system. In bivariate logistic regression analyses, significant unadjusted correlates included young age, less education, prior mental health treatment, weaker attraction to adult women, history of sexual abuse in the participants' own childhood, and the psychosocial effect of perceived stigma against pedophilia. In multivariable analysis, all these factors except education were uniquely associated with suicidal ideation. These results identify meaningful clinical risk factors and treatment targets in this population.

Anhedonia and suicidal thoughts and behaviors in psychiatric outpatients: The role of acuity.

BACKGROUND: Anhedonia-impairment related to the experience of pleasure-has been identified as a potential risk factor for suicide, with some mixed findings. The current study sought to clarify the role of acuity of anhedonia in the relationship between anhedonia and suicidal thoughts and behaviors by comparing acutely and chronically anhedonic subjects on severity of suicidal ideation (SI) and suicide attempt (SA) history. METHODS: Psychiatric outpatients (N = 395) were administered the Columbia Suicide Severity Rating Scale, the Beck Scale for Suicidal Ideation and a modified version of the Snaith-Hamilton Pleasure Scale (SHPS); SI measures were readministered at a 1-month follow-up (N = 289, 73%). Participants were classified as acutely anhedonic, chronically anhedonic and nonanhedonic based on their responses to the SHPS at initial assessment. RESULTS: Controlling for symptoms of anxiety and depression, acute anhedonia was cross-sectionally and prospectively associated with greater severity of SI compared to the nonanhedonic group; no differences in severity of SI were found between the chronically anhedonic and nonanhedonic group at either time point. Anhedonia grouping was not associated with SA history. CONCLUSION: Changes in capacity to experience pleasure may be more informative of near-term SI than typically low pleasure levels. Future investigation should focus on the relationship between acute anhedonia and imminent suicidal behavior.

Comparison of Self-Identified Minor-Attracted Persons Who Have and Have Not Successfully Refrained From Sexual Activity With Children.

The present study compares community-based minor-attracted persons (MAPs) with and without histories of sexual activity with children. MAP actors (N = 342) were significantly older than non-actors (N = 223), with longer duration of pedophilic attraction, more antisocial traits, greater attraction to boys, greater difficulty controlling their attraction, and more positive attitudes toward adult-child sexual activity. Additionally, more MAP actors reported prior mental health treatment, nonsexual offenses, and childhood sexual and nonsexual abuse. Over a third of the whole sample reported chronic suicidal ideation. These findings support the existence of MAPs who successfully refrain from sexually engaging with children, identifying multiple protective and risk factors.

Qualitative Analysis of Attitudes Toward Adult-Child Sexual Activity Among Minor-Attracted Persons.

This study uses qualitative methodology to explore narrative responses to a question regarding the harmful versus beneficial effects of adult-child sex on children. Data were gathered from a survey of self-identified minor-attracted persons (MAPs). Two hundred and sixty-seven survey participants provided narrative responses. Results indicated a significant amount of variability in perceptions of harm and of mitigating or aggravating factors. Understanding the subjective perspectives of MAPs, the range of their attitudes, and the issues that they identify as pertinent is critically important for clinical efficacy in the treatment of pedophilia.

Entrapment as a mediator of suicide crises.

BACKGROUND: Prior research has validated the construct of a suicide crisis syndrome (SCS), a specific psychological state that precedes and may precipitate suicidal behavior. The feeling of entrapment is a central concept of the SCS as well as of several other recent models of suicide. However, its exact relationship with suicidality is not fully understood. In efforts to clarify the exact role of entrapment in the suicidal process, we have examined if entrapment mediates the relationship of other components of the SCS, including ruminative flooding, panic-dissociation, fear of dying and emotional pain, with suicidal ideation (SI) in recently hospitalized psychiatric inpatients. METHODS: The Suicide Crisis Inventory (SCI) and Beck Scale for Suicidal Ideation (BSS) were administered to 200 high-risk adult psychiatric inpatients hospitalized following SI or suicide attempt, assessing SCS and SI levels at admission, respectively. The possible mediation effects of entrapment on the relationship between the other components of the SCS and SI at admission were evaluated. RESULTS: Entrapment significantly and fully mediated the relationship of ruminative flooding, panic-dissociation, and fear of dying with SI, with no direct relationships between these variables and SI reaching statistical significance. Further, no reverse mediation relationships between these variables and SI were found, indicating that the mediation effects of entrapment were unidirectional. While entrapment did mediate the association between emotional pain and SI, the direct relationship between emotional pain and SI was also significant. Moreover, in reverse mediational analysis, emotional pain was a partial mediator of the relationship between entrapment and SI. CONCLUSION: Entrapment and emotional pain may have a more direct association with SI than the other components of the SCS, including ruminative flooding, panic-dissociation, and fear of dying, the effects of which are mediated by the former. This suggests entrapment and emotional pain may represent key symptomatic targets for intervention in acutely suicidal individuals. Further research is needed to determine the relationship of these constructs to suicidal behavior.

Prediction of suicidal behavior in high risk psychiatric patients using an assessment of acute suicidal state: The suicide crisis inventory.

BACKGROUND: We have developed the Suicide Crisis Inventory (SCI) to evaluate the intensity of the Suicidal Crisis Syndrome, an acute state hypothesized to precede suicide attempt. The psychometric properties of the SCI, including predictive validity for suicidal behavior (SB), were assessed. METHODS: Adult psychiatric patients (n = 201) hospitalized for high suicide risk were assessed. Logistic regression models assessed the SCI's predictive validity for SB in the 4-8 weeks following hospital discharge and its incremental predictive validity over traditional risk factors (n = 137, 64% f/u rate). Internal structure, reliability, convergent and discriminant validity, and state versus trait properties were also assessed. RESULTS: The SCI had excellent internal consistency (Cronbach's α 0.970). The SCI total score at discharge predicted short-term SB with 64% sensitivity 88% specificity (OR = 13, P = .003) at its optimal cut score. In a test of its incremental predictive validity, SCI total score at discharge improved prediction of SB over traditional risk factors (Chi-squared 5.597, P = .024, model P = .001), with AOR 2.02 (P = .030). The SCI admission versus discharge test-retest reliability and score distributions showed it to be an acute state measure. CONCLUSION: The SCI was predictive of future SB in high-risk psychiatric inpatients during the crucial weeks following their hospital discharge. Further validation in diverse patient populations is needed.

Clinicians' conflicting emotional responses to high suicide-risk patients-Association with short-term suicide behaviors: A prospective pilot study.

BACKGROUND: Clinician's emotional responses to patients have been recognized as potentially relating to treatment outcome, however they have received little attention in the literature on suicide risk. We examine the relationship between a novel targeted measure of clinicians' emotional responses to high-risk psychiatric inpatients and their short-term post-discharge suicide behavior. METHODS: First-year psychiatry residents' emotional responses to their patients were assessed anonymously with the novel self-report 'Therapist Response Questionnaire-Suicide Form' (TRQ-SF). Patient outcomes were assessed at 1-2months post-discharge, and post-discharge suicide outcomes were assessed with the Columbia Suicide Severity Rating Scale. Following exploratory factor analysis of the TRQ-SF, scores on the resultant factors were examined for relationships with clinical and demographic measures and post-discharge suicide behavior. RESULTS: A two-factor model fit the data, with factors reflecting dimensions of affiliation/rejection and distress/non-distress. Two items that did not load robustly on either factor had face validity for hopefulness and hopelessness and were combined as a measure along a hopefulness/hopelessness dimension. The interaction Distress×Hopefulness, reflecting a conflicting emotional response pattern, significantly predicted post-discharge suicide outcomes even after covarying for depression, entrapment, and suicidal ideation severity. CONCLUSION: Clinicians' conflicting emotional responses to high-risk patients predicted subsequent suicidal behavior, independent of traditional risk factors. Our findings demonstrate the potential clinical value of assessing such responses.

The Modular Assessment of Risk for Imminent Suicide (MARIS): A proof of concept for a multi-informant tool for evaluation of short-term suicide risk.

BACKGROUND: To date, no diagnostic tool has demonstrated clinical value for the assessment of short-term suicide risk among high-risk individuals. To this end we have developed the Modular Assessment of Risk for Imminent Suicide (MARIS), a modular patient and clinician informed risk evaluation instrument. Here we assess its predictive validity for suicidal behavior (SB) in psychiatric patients following discharge from an inpatient unit. METHODS: The MARIS and a psychological test battery were administered to 136 adult psychiatric patients hospitalized for high risk of suicide. Eighty-six participants had complete data at discharge and 59 (69.8%) were reached for follow-up. Logistic regressions were used to assess the predictive validity of the MARIS for SB over the 4-8weeks following hospital discharge and its incremental predictive validity over standard risk factors. RESULTS: Patients who exhibited SB in the period between initial assessment and follow-up had significantly higher MARIS scores. ROC analysis indicated good sensitivity and specificity of MARIS in identifying risk (OR=19). Further, MARIS total score significantly improved predictive validity by the standard risk factors when added to a model comprising global depression severity, hopelessness, and lifetime number of suicide attempts, and was the only factor that approached independent significance. CONCLUSION: The MARIS appears to be a practical and effective tool for detecting short-term suicide risk following hospital discharge, supporting the validity of modular multi-informant approach to suicide risk evaluation. Replication of these findings and further investigation of the applications of such an approach are warranted.

Assessment of Sexual Fantasies in Psychiatric Inpatients With Mood and Psychotic Disorders and Comorbid Personality Disorder Traits.

INTRODUCTION: Sexuality is an important aspect of quality of life and sexual fantasies comprise a normal part of human sexuality. However, the nature of sexuality and sexual fantasies of patients with mental illness remains an understudied area. AIM: To investigate the nature and frequency of sexual fantasies in psychiatric patients, the present study compared the frequency of four types of sexual fantasies across four different mood and psychotic diagnoses and three personality disorder clusters. METHODS: Study participants included 133 psychiatric inpatients recruited from an urban hospital. Sexual fantasies were compared across patients with schizophrenia, bipolar disorder, schizoaffective disorder, major depressive disorder and three nonclinical samples from the literature and then correlated with personality cluster scores. MAIN OUTCOME MEASURES: Subjects were administered the Structured Clinical Interview for DSM-IV for Axis I and for Axis II Disorders. Sexual fantasies were assessed by the Wilson Sexual Fantasies Questionnaire, which measures four types of sexual fantasies (exploratory, intimate, impersonal, and sadomasochistic). RESULTS: Within the entire sample, there were significant differences across sexual fantasy types, with subjects scoring highest on intimate sexual fantasies and then exploratory, impersonal, and sadomasochistic. There were no significant differences across mood and psychotic diagnostic groups for any of the sexual fantasy scales and the scores were within the normative range of nonclinical samples. Patients with high cluster B scores scored significantly higher on all four fantasy scales than those without. Patients with high cluster A scores scored lower on intimate fantasies, but there was no association between cluster C scores and sexual fantasies. The association between cluster B and sexual fantasies remained consistent across Structured Clinical Interview for DSM-IV for Axis I diagnoses (no interaction effect). CONCLUSION: Patients with severe mental illness report sexual fantasies that are largely affiliative in nature and consistent with normative patterns. This suggests that assessment and treatment of sexual issues in the mentally ill should be part of the clinical routine as it is in healthy individuals.

Increased sleep duration precedes the improvement of other symptom domains during the treatment of acute mania: a retrospective chart review.

BACKGROUND: Understanding trajectories of symptom changes may help gauge treatment response and better identify therapeutic targets in treatment of acute mania. We examined how symptoms of sleep disturbance, mania, and psychosis resolved in a naturalistic treatment setting, hypothesizing that improvement in sleep would precede improvement in manic and psychotic symptoms. METHODS: Charts of 100 patients with admitting diagnoses of bipolar mixed or manic episode were retrospectively reviewed. Medications and demographic variables were recorded, and the Clinician-Administered Rating Scale for Mania (CARS-M) mania and psychosis ratings and sleep hours were determined for 8 observation points. Times to minimum symptom level in each domain were compared via Wilcoxon signed-rank tests. Symptom correlations and trajectories and medication effects were explored using repeated measures ANOVA and regression models. RESULTS: Manic and psychotic symptom resolution was linear over the time of hospitalization. In contrast, sleep showed a slow initial response, followed by rapid increase to peak, preceding peak improvement in mania and psychosis (p < 0.001). Rate of sleep restoration was a predictor of rate but not of magnitude of treatment response for symptoms mania and psychosis. Patterns of medication use did not affect symptom trajectories. CONCLUSIONS: In acute mania, improvement in sleep with treatment is dissociable from resolution in symptoms of mania and psychosis, but there appears to be no therapeutic advantage to patient oversedation. Sleep improves first and may be both a predictor of the rate of treatment response and a useful therapeutic target.

Attitudes of acceptability and lack of condemnation toward suicide may be predictive of post-discharge suicide attempts.

BACKGROUND: Suicide attempts (SA) after psychiatric hospitalization continue to be a major cause of morbidity. Implicit measures may enhance our ability to assess suicide risk. In this context, we describe the first use of the Suicide Opinion Questionnaire (SOQ) to identify post-discharge suicide attempters. METHODS: Adult psychiatric inpatients admitted for suicidality (N = 91) were administered a battery of measures including the SOQ, and forty were reached and reassessed for SA at two months post-discharge. Exploratory factor analysis (EFA) on items associated with suicidality was performed to identify latent constructs. Linear discriminant analysis (LDA) was used to optimize factor combination for suicide identification. Results were compared with explicit measures of suicidality, and logistic regression was used to control for other risk factors. Finally, a simplified 9-item scale was derived from the results and its performance compared to that of the linear discriminant function. RESULTS: Twenty items differed between patients with and without SA at intake or follow-up. EFA on these identified two factors: suicide attempters indicated greater acceptability and less moral condemnation of suicide. The LDA-derived discriminant function and 9-item scale was significantly sensitive and specific for post-discharge SA. CONCLUSIONS: Attitudes of acceptability and lack of condemnation toward suicide may constitute an implicit measure of suicidality that could contribute to risk assessment in a high-risk population.

A Pilot Study of Safety and Efficacy of Cranial Electrotherapy Stimulation in Treatment of Bipolar II Depression.

This double-blind, sham-controlled study sought to investigate the effectiveness of cranial electrotherapy stimulation (CES) for the treatment of bipolar II depression (BD II). After randomization, the active group participants (n = 7) received 2 mA CES treatment for 20 minutes five days a week for 2 weeks, whereas the sham group (n = 9) had the CES device turned on and off. Symptom non-remitters from both groups received an additional 2 weeks of open-label active treatment. Active CES treatment but not sham treatment was associated with a significant decrease in the Beck Depression Inventory (BDI) scores from baseline to the second week (p = 0.003) maintaining significance until week 4 (p = 0.002). There was no difference between the groups in side effects frequency. The results of this small study indicate that CES may be a safe and effective treatment for BD II suggesting that further studies on safety and efficacy of CES may be warranted.

Panic symptom clusters differentially predict suicide ideation and attempt.

Increasingly strong evidence links anxiety disorders in general and panic attacks in particular to suicidality. The underlying causes and specifics of this relation, however, remain unclear. The present article sought to begin addressing this question by clarifying the association between panic symptoms and suicidality. Data were sampled from the NESARC epidemiological data set from the US and analyzed as four independently, randomly selected subsets of 1000 individuals using structural equation modeling analyses and replicating results across samples. Evidence is presented for four symptom clusters (cognitive symptoms, respiratory distress, symptoms of alpha and beta adrenergic activation) and the differential association of each with suicidal ideation and attempts. Symptoms of alpha adrenergic activation predicted prior suicide attempt whereas cognitive symptoms predicted prior suicidal ideation. These findings were independent of comorbid major depressive disorder. It is suggested that assessment of suicide risk in the community includes the presentation of cognitive symptoms and symptoms related to alpha adrenergic activation.

Stability of Suicide Risk Prediction Models During Changes in Health Care Delivery.

OBJECTIVE: The authors aimed to use health records data to examine how the accuracy of statistical models predicting self-harm or suicide changed between 2015 and 2019, as health systems implemented suicide prevention programs. METHODS: Data from four large health systems were used to identify specialty mental health visits by patients ages ≥11 years, assess 311 potential predictors of self-harm (including demographic characteristics, historical risk factors, and index visit characteristics), and ascertain fatal or nonfatal self-harm events over 90 days after each visit. New prediction models were developed with logistic regression with LASSO (least absolute shrinkage and selection operator) in random samples of visits (65%) from each calendar year and were validated in the remaining portion of the sample (35%). RESULTS: A model developed for visits from 2009 to mid-2015 showed similar classification performance and calibration accuracy in a new sample of about 13.1 million visits from late 2015 to 2019. Area under the receiver operating characteristic curve (AUC) ranged from 0.840 to 0.849 in the new sample, compared with 0.851 in the original sample. New models developed for each year for 2015-2019 had classification performance (AUC range 0.790-0.853), sensitivity, and positive predictive value similar to those of the previously developed model. Models selected similar predictors from 2015 to 2019, except for more frequent selection of depression questionnaire data in later years, when questionnaires were more frequently recorded. CONCLUSIONS: A self-harm prediction model developed with 2009-2015 visit data performed similarly when applied to 2015-2019 visits. New models did not yield superior performance or identify different predictors.

Fear of dying in panic attacks predicts suicide attempt in comorbid depressive illness: prospective evidence from the National Epidemiological Survey on Alcohol and Related Conditions.

CONTEXT: As a group, comorbid anxiety disorders among depressed persons have consistently been found to increase risk of suicide attempt (SA). Growing evidence supports the link between panic attacks (PAs) and suicidality, but prospective evidence is limited and the nature of the linkage remains unclear. The positive-feedback model of suicide suggests that PAs eliciting catastrophic cognitions may drive SA. OBJECTIVE: To prospectively examine the relationship between PAs, panic symptoms and suicidality in individuals meeting DSM-IV criteria for past-year major depressive episodes in a large epidemiological study. METHODS: In data on 2,864 participants of the National Epidemiological Survey on Alcohol and Related Conditions (NESARC) wave 1 and 2 surveys with depressive disorders, the associations of PAs and panic symptoms with subsequent suicidal ideation (SI) and SAs were assessed using logistic regression. RESULTS: Past-year PAs in wave 1 significantly increased odds for subsequent SI and attempt in the 3-year follow-up interval; however, in multivariate analyses, PAs were not a significant predictor, but PAs featuring fear of dying were. Further, among subjects with PAs, fear of dying during a PA increased the odds of subsequent SA sevenfold, even after controlling for comorbid disorders, demographic factors, and other PA symptoms. CONCLUSION: PAs characterized by prominent morbid catastrophic cognitions may mediate the transition to SIs and SAs in subjects with depressive episodes. Presence of these symptoms in clinical settings may serve as a warning sign for future suicidality.

Distinctive emotional responses of clinicians to suicide-attempting patients--a comparative study.

BACKGROUND: Clinician responses to patients have been recognized as an important factor in treatment outcome. Clinician responses to suicidal patients have received little attention in the literature however, and no quantitative studies have been published. Further, although patients with high versus low lethality suicidal behaviors have been speculated to represent two distinct populations, clinicians' emotional responses to them have not been examined. METHODS: Clinicians' responses to their patients when last seeing them prior to patients' suicide attempt or death were assessed retrospectively with the Therapist Response/Countertransference Questionnaire, administered anonymously via an Internet survey service. Scores on individual items and subscale scores were compared between groups, and linear discriminant analysis was applied to determine the combination of items that best discriminated between groups. RESULTS: Clinicians reported on patients who completed suicide, made high-lethality attempts, low-lethality attempts, or died unexpected non-suicidal deaths in a total of 82 cases. We found that clinicians treating imminently suicidal patients had less positive feelings towards these patients than for non-suicidal patients, but had higher hopes for their treatment, while finding themselves notably more overwhelmed, distressed by, and to some degree avoidant of them. Further, we found that the specific paradoxical combination of hopefulness and distress/avoidance was a significant discriminator between suicidal patients and those who died unexpected non-suicidal deaths with 90% sensitivity and 56% specificity. In addition, we identified one questionnaire item that discriminated significantly between high- and low-lethality suicide patients. CONCLUSIONS: Clinicians' emotional responses to patients at risk versus not at risk for imminent suicide attempt may be distinct in ways consistent with responses theorized by Maltsberger and Buie in 1974. Prospective replication is needed to confirm these results, however. Our findings demonstrate the feasibility of using quantitative self-report methodologies for investigation of the relationship between clinicians' emotional responses to suicidal patients and suicide risk.

Teaching evidence-based approaches to suicide risk assessment and prevention that enhance psychiatric training.

This report describes one in a series of National Institute of Health (NIH) supported conferences aimed at enhancing the ability of leaders of psychiatry residency training to teach research literacy and produce both clinician-scholars and physician-scientists in their home programs. Most psychiatry training directors would not consider themselves research scholars or even well-schooled in evidence based practice. Yet they are the front line educators to prepare tomorrow's psychiatrists to keep up with, critically evaluate, and in some cases actually participate in the discovery of new and emerging psychiatric knowledge. This annual conference is meant to help psychiatry training directors become more enthusiastic, knowledgeable and pedagogically prepared to create research-friendly environments at their home institutions, so that more trainees will, in turn, become research literate, practice evidence-based psychiatry, and enter research fellowships and careers. The overall design of each year's meeting is a series of plenary sessions introducing participants to new information pertaining to the core theme of that year's meeting, integrated with highly interactive small group teaching sessions designed to consolidate knowledge and provide pragmatic teaching tools appropriate for residents at various levels of training. The theme of each meeting, selected to be a compelling and contemporary clinical problem, serves as a vehicle to capture training directors' attention while teaching relevant brain science, research literacy and effective pedagogy. This report describes the content and assessment of the 2011 annual pre-meeting, "Evidence-based Approaches to Suicide Risk Assessment and Prevention: Insights from the Neurosciences and Behavioral Sciences for use in Psychiatry Residency Training."

Complex modeling with detailed temporal predictors does not improve health records-based suicide risk prediction.

Suicide risk prediction models can identify individuals for targeted intervention. Discussions of transparency, explainability, and transportability in machine learning presume complex prediction models with many variables outperform simpler models. We compared random forest, artificial neural network, and ensemble models with 1500 temporally defined predictors to logistic regression models. Data from 25,800,888 mental health visits made by 3,081,420 individuals in 7 health systems were used to train and evaluate suicidal behavior prediction models. Model performance was compared across several measures. All models performed well (area under the receiver operating curve [AUC]: 0.794-0.858). Ensemble models performed best, but improvements over a regression model with 100 predictors were minimal (AUC improvements: 0.006-0.020). Results are consistent across performance metrics and subgroups defined by race, ethnicity, and sex. Our results suggest simpler parametric models, which are easier to implement as part of routine clinical practice, perform comparably to more complex machine learning methods.