Pathways to Prevention Workshop:

Advancing Research To Prevent Youth Suicide

March 29–30, 2016

Masur Auditorium

Clinical Center, Building 10

NIH Main Campus

Bethesda, Maryland



NIH Office of Disease Prevention website

Workshop Agenda

Day 1 – Tuesday, March 29, 2016
8:30 a.m. to 5:00 p.m. Eastern time

INTRODUCTORY SESSION
Time Agenda Item
8:30 a.m. Opening Remarks
Bruce Cuthbert, Ph.D.
Acting Director
National Institute of Mental Health
National Institutes of Health
8:40 a.m. Charge to the Panel
David M. Murray, Ph.D.
Associate Director of Disease Prevention
Director
Office of Disease Prevention
Division of Program Coordination, Planning, and Strategic Initiatives
Office of the Director
National Institutes of Health
8:50 a.m. Panel Activities
Todd D. Little, Ph.D.
Director
Institute for Measurement, Methodology, Analysis, and Policy
Professor
Educational Psychology and Leadership
Texas Tech University
9:00 a.m. Advancing Research To Prevent Youth Suicide: Overview of Topic
Amy B. Goldstein, Ph.D.
Associate Director for Prevention
Chief
Preventive Intervention Research Program
Treatment and Prevention Intervention Research Branch
Division of Services and Intervention Research
National Institute of Mental Health
National Institutes of Health

Peter A. Wyman, Ph.D.
Professor
Department of Psychiatry
School of Medicine and Dentistry
University of Rochester Medical Center

Suicide was the second leading cause of death for youth (10- to 24-year-olds) in 2014, resulting in 5,504 deaths in the United States. This mortality has not decreased compared with other external causes of death, and youth suicide attempts have remained at consistent rates for decades. According to the 2011 Youth Risk Behavior Surveillance System, 2.4% of high school students received medical treatment for attempted suicide, and 7.8% attempted suicide one or more times within the year. Some groups (e.g., American Indian youth, young adults with substance use problems, children of depressed parents, youth and young adults who identify as a sexual and gender minority) are at increased risk for suicidal behaviors. One of the challenges in suicide prevention research is that the primary outcome of interest is multidetermined, and depending on the target population, suicide can be a low base-rate occurrence. Many studies examining risk in important subgroups (e.g., racial, ethnic, sexual, and gender minorities) often lack sufficient power to accurately determine the effectiveness of the intervention. The goal of this session will be to provide the context and motivation for the workshop, “Advancing Research To Prevent Youth Suicide.” The overview will consist of two presentations. The first presentation will provide a summary of this topic from the federal perspective. Gaps in suicide prevention research previously identified by the National Action Alliance for Suicide Prevention Research Prioritization Task Force will be presented, and opportunities for ways forward to enhance and harmonize data efforts in the area of suicide prevention will be reviewed. The second presentation will present what is currently known about the benefits of early intervention on downstream suicidal behaviors. The presentation will also review existing research demonstrating the potential impact of family and community-based preventive interventions on future suicidal behavior, highlighting those interventions not designed to explicitly target suicidal behavior. Future directions and opportunities will be discussed.
9:30 a.m. Discussion
9:40 a.m. The Personal Perspective
Iden D. Campbell McCollum, CPRP
Leah Harris, M.A.
The Honorable Gordon H. Smith

Introduction by
Jane L. Pearson, Ph.D.
Program Chief
Suicide Treatment and Preventive Interventions Research Program
Treatment and Prevention Intervention Research Branch
Division of Services and Intervention Research
National Institute of Mental Health
National Institutes of Health

Suicide deaths and attempts clearly are events that signify significant suffering. These events also lead to distress for many family members, friends, and communities. While many of us probably know of someone who has died by suicide or has attempted suicide, the planners of this workshop felt it was critical to include in the meeting the voices of individuals who have experienced a suicide loss in their family, as well as the voices of “lived experience”—those who have attempted suicide and have found ways to survive, and even thrive. In this segment of the meeting, we ask you to listen to three individuals who have been willing to share their life experiences so that others might avoid the tragedy and suffering of suicide attempts and deaths. Each video clip is about 5 minutes long. Each of these individuals has devoted significant amounts of time to suicide prevention advocacy. You will hear, through their diverse backgrounds, some common themes that will serve to inform our discussions and deliberations for this meeting.
10:10 a.m. Break
10:20 a.m. Workshop Overview
Todd D. Little, Ph.D.
Director
Institute for Measurement, Methodology, Analysis, and Policy
Professor
Educational Psychology and Leadership
Texas Tech University
KEY QUESTION 1 IN THE EVIDENCE REPORT: How can national, state, and community data systems be linked to existing data from suicide prevention efforts in order to add possible value for stakeholders? What methods are available to link the data systems?
Time Agenda Item
10:30 a.m. Evidence-based Practice Center Presentation: Data Linkage Strategies and Methods To Advance Youth Suicide Prevention
Holly C. Wilcox, Ph.D.
Associate Professor of Psychiatry
Department of Psychiatry
Johns Hopkins School of Medicine
Department of Mental Health
Johns Hopkins Bloomberg School of Public Health

Objectives: Linking national, state, and community data systems, such as those used for medical service billing, to existing data from suicide prevention efforts could facilitate the assessment of longer term outcomes. Our objective was to identify and describe data systems that can be linked to data from studies of youth suicide prevention interventions and to identify analytic approaches to advance youth suicide prevention research. Data Sources: We conducted a systematic review to identify studies of suicide prevention interventions and three types of searches to identify data systems providing suicide-related outcomes: (1) a literature search; (2) an environmental scan of grey literature; and (3) a targeted search, through contact with relevant individuals, in six states, two cities, and one tribal community. Review Methods: Two independent reviewers screened all results. Studies and data systems had to be U.S. based, include individuals between age 0 and 25, and include suicide, suicide attempt, or suicide ideation as an outcome. Results: Of the 47 studies (described in 59 articles) of suicide prevention interventions identified in our systematic review, only 6 studied outcomes by linking to external data systems, and only 12 explored treatment heterogeneity through the effects of moderators such as gender or race/ethnicity. We identified 153 unique and potentially linkable external data systems, 66 of which we classified as “fairly accessible” with data dictionaries available. Conclusions: There is potential for linking existing data systems with suicide prevention efforts to assess the broader and extended impact of suicide prevention interventions. However, sparse availability of data dictionaries and lack of adherence to standard data elements limit the potential utility of linking prevention efforts with data systems.
10:50 a.m. Linking Data Systems To Support Suicide Prevention Efforts
Alex Crosby, M.D., M.P.H.
Branch Chief
Surveillance Branch of the Division of Violence Prevention
National Center for Injury Prevention and Control
Centers for Disease Control and Prevention

Surveillance systems are an important component to understanding health issues and to health planning. These systems involve the routine collection and timely dissemination of health data to guide appropriate public health action. Even though there are a number of existing data systems, many of them are isolated from each other. The utility of existing data can be enhanced significantly by developing linkages across jurisdictions or databases, which overcomes the limitations of separate databases and goes far in developing comprehensive information about an event and its circumstances. Data linkage is the process of matching records from separate files and trying to select the records that belong to the same entity. The linked data often afford a more complete description of the characteristics of an event or a topic, resulting in a better understanding of how to prevent or control a disease or adverse condition. There are significant barriers to successful linkage, such as high costs and limited resources for developing and maintaining databases and technical difficulties. This presentation will describe successes and challenges in surveillance data linkage.
11:10 a.m. Challenges and Opportunities To Link Administrative Data To Analyze Suicide Prevention
Robert M. Goerge, Ph.D.
Chapin Hall Senior Research Fellow
Senior Advisor for Master’s Program in Computational Analysis in Public Policy
Chicago Harris School of Public Policy
University of Chicago

This presentation will address how national, state, and local data systems can be better used to inform suicide prevention efforts. What is being collected in specific domains (health, crime, human services)? How can data across domains be combined or linked to add richness to existing data? In addition, given the variation in suicide across space and time, looking to see what geographic data are available and what data are available for youth and adults will also be explored. Challenges and potential solutions around access and linkage will also be presented.
11:30 a.m. Discussion
Noon Lunch
KEY QUESTION 3 IN THE EVIDENCE REPORT: Which statistical methods are reliable and valid for understanding possible mediators and moderators in suicide prevention programs to improve targeting interventions to populations?
Time Agenda Item
1:00 p.m. Evidence-based Practice Center Presentation: Data Linkage Strategies and Methods To Advance Youth Suicide Prevention: Methods for Understanding Moderators
Rashelle J. Musci, Ph.D.
Assistant Professor of Mental Health
Department of Mental Health
Johns Hopkins Bloomberg School of Public Health
1:10 p.m. A Methodologic Overview of Opportunities and Challenges in Moving Suicide Research Forward
C. Hendricks Brown, Ph.D.
Professor
Psychiatry and Behavioral Sciences and Medical Social Sciences
Northwestern University Feinberg School of Medicine

This presentation discusses a range of methodologic opportunities that could advance three types of research: epidemiologic, intervention, and large-scale implementation. Research on suicide prevention is truly challenging because of the following factors: limited or biased reporting, comparatively low base-rate outcomes, and trials with different interventions and different assessments. But there are methodologic approaches that can address many of these challenges. One fundamental approach involves combining data from multiple studies, either through a synthesis of similar trials to increase statistical power, linking individuals within different datasets, or combining administrative datasets. A second approach involves the routine use of standard measures, improved calibration of measures, or more efficient assessment procedures such as those involving computerized adaptive testing. A third approach is to apply methods that do not require individual-level linkage, including the use of small ecological levels and missing data approaches to handle uncertainty.

Examples are provided for epidemiologic, intervention, and implementation research. For epidemiologic data, we illustrate how the combined use of three datasets led to what we believe are large elevations in the rates of suicide among younger U.S. military personnel followed by more modest risk as these veterans age. In intervention analyses, we point out how synthesis of individual-level data from legacy randomized trials can provide greater overall power and discern important patterns not achievable by meta-analysis. We illustrate this with examples showing the mediational effects of antidepressants on suicide risk arising through the reduction of depressive symptoms. In addition, we explore how general preventive trials may impact suicide risk by targeting key suicide risk factors. We also point to the use of computerized adaptive testing that could allow much more efficient identification, prediction, and monitoring of suicide risk. Finally, we discuss approaches to link data longitudinally in an ongoing randomized trial of a school-based suicide prevention program. Throughout this talk, we present some lessons learned about the successes and failures in searching and assembling relevant data, bridging gaps in data from different studies, and tracking individuals.
1:30 p.m. Studying Mediation in Prevention Trials for Adolescent Depression: The State of the Field and Implications for Suicide Prevention Research
George W. Howe, Ph.D.
Professor of Clinical Psychology
Department of Psychology
The George Washington University

Randomized trials are the platinum standard for testing whether an intervention prevents or resolves emotional or behavioral disorders; randomized trials provide nuanced answers to the questions of who is most likely to benefit (questions of moderation) and why they do so (questions of mediation). Questions examining moderation and mediation are inherently intertwined and best considered together, as they are difficult or impossible to answer with a single randomized trial and are best approached through combining data from multiple trials, including existing legacy datasets and next-generation trials.

Prevention programs target a range of risk or protective mechanisms, based on the understanding of what contributes to the disorder. Demonstrating that trials suppress risk mechanisms or enhance protective mechanisms, and that these changes are associated with reductions in later rates of disorder, not only provides evidence for an intervention’s effectiveness, but also that these targets are causally implicated. It opens the possibility that proximal targets represent final common pathways that can be targeted in a variety of ways during next-generation program development. Identifying active targets may be one of the best ways to study questions of moderation. Although it is important to test moderation based on the usual suspects (age, gender, ethnic or racial variation, socioeconomic status), recent work suggests that baseline variation in intervention targets is often the most theoretically meaningful source of moderation. New designs for testing baseline target-moderated mediation can point directly to subgroups most likely to benefit from intervention, as well as provide guidance in the types of interventions that will work best for those groups. Combining individual-level data from multiple trials may also be essential for exploring mediation and moderation.

We present data from a recent review of randomized trials for the prevention of adolescent depression, indicating that only a few trials have systematically tested for mediation. We contrast this with information from a subset of these trials indicating that many potential mediators have been measured in those trials and suggesting that legacy datasets present opportunities for mining existing data. We end by noting some of the practical and methodological issues that need to be solved for this enterprise to move forward. These include developing new methods for harmonizing measurement of mediators and moderators, exploring the properties of quantitative methods for combining datasets to study moderation and mediation, and expanding work on causal inference to strengthen confidence in causal mediation and effect heterogeneity.
1:50 p.m. Promises and Challenges in the Psychometric Assessment of Low Base-Rate Behaviors Using Integrative Data Analysis
Patrick J. Curran, Ph.D.
Director
L.L.Thurstone Psychometric Laboratory
Professor
Department of Psychology and Neuroscience
University of North Carolina at Chapel Hill

Integrative data analysis (IDA) is a general methodological framework that allows for the fitting of statistical models to aggregated data that have been drawn from two or more independent samples. A central challenge in IDA is developing harmonized measures of key theoretical constructs based on multi-item scales that may differ in item stem or response across each contributing study. Fitting models to pooled data using IDA offers many strengths including increased statistical power, greater developmental coverage, enhanced construct validity of measures, greater generalizability findings, and direct tests of reproducibility that contribute to the ongoing quest of building a cumulative science. However, there are also several challenges that are in need of further study and development. An issue of key importance in the study of suicidality is the need for ongoing refinement of psychometric models for application to multi-item scales assessed over time, particularly when the behaviors under study are rare. Low base-rate behaviors are traditionally difficult to model because there are few empirical data points to analyze. However, using IDA to pool multiple datasets has the promise of increasing the number of participants who report rare behaviors, and this, in turn, can improve the reliability and validity of the resulting scales’ scores used in subsequent analyses. The goal of this talk is to review existing psychometric models most commonly used in IDA, describe why these models are difficult to estimate when measuring low base-rate behaviors, and propose future areas of research needed to make such applications possible.
2:10 p.m. Discussion
2:50 p.m. Break
KEY QUESTION 2 IN THE EVIDENCE REPORT: Which statistical methods are reliable and valid for analyzing linked national, state, and community data systems and suicide prevention data to avoid misleading conclusions?
  • A. What are potential sources of bias for these statistical methods?
  • B. What are the advantages and disadvantages of these different methods?
Time Agenda Item
3:10 p.m. Evidence-based Practice Center Presentation: Data Linkage Strategies and Methods To Advance Youth Suicide Prevention: Methods for Analyzing Linked Data
Rashelle J. Musci, Ph.D.
Assistant Professor of Mental Health
Department of Mental Health
Johns Hopkins Bloomberg School of Public Health
3:20 p.m. Propensity Score Methods in Suicide Prevention Research
Donna Coffman, Ph.D.
Research Associate Professor
College of Health and Human Development
Principal Investigator
The Methodology Center
The Pennsylvania State University

This presentation will discuss the motivation for and utility of propensity score methods in suicide prevention research. Because random assignment is frequently not possible in suicide prevention research, propensity scores are an alternative approach to inferring causality. Propensity scores will be defined, followed by information about how to estimate and apply them. Although there are advantages to using propensity scores, their use invokes assumptions. These assumptions will be discussed, as well as sensitivity analyses for assessing the implications of violations of assumptions. Issues of compliance and assessing mediation will also be discussed.
3:40 p.m. Using Multiple Studies To Assess Generalizability and Broaden Conclusions
Eloise E. Kaizar, Ph.D.
Associate Professor
Department of Statistics
The Ohio State University

There are many situations for which the joint analysis of multiple data sources is necessary to answer important scientific questions, including those related to suicide prevention. In particular, because different types of studies may provide evidence about different dimensions of a scientific inquiry, jointly analyzing data collected via different study designs may be the key to learning about difficult-to-study topics. The strong internal validity that results from randomization often places randomized trials atop “pyramids of evidence,” but restrictive or ad hoc recruitment may limit the applicability of their findings to more general populations. The strengths and weaknesses of observational studies mirror their randomized counterparts with treatment self-selection weakening causal conclusions, but broad participation lending confidence to generalizing the findings. Careful analyses of both types of data can exploit the strengths of both designs while mitigating their weaknesses. The question of a relationship between suicide and antidepressant use in children and adolescents will be used as a case study to demonstrate methods for exploring generalizability of randomized trials. Included is an overview of cross-design synthesis methodology as a more formal means to combine data and discussion of barriers to its widespread implementation.
4:00 p.m. The Measurement and Modeling of Population Heterogeneity in Suicide Risk Using Data From Multiple Sources at Multiple Levels
Katherine E. Masyn, Ph.D.
Associate Professor
Division of Epidemiology and Biostatistics
Georgia State University School of Public Health

There are several theoretical frameworks that inform youth suicide prevention research, including the social-ecological model and intersectionality theory. These guiding frameworks emphasize the complex, constitutive, multivariate, multilevel nature of the processes that shape heterogeneity in suicide risk for youth over time. These process conceptualizations present the persistent challenge of adequately representing key sources of (co)variation in the data analysis phase. For example, to analytically occupy a social-ecological approach, we must first integrate and harmonize ecological data sources with multiple individual-level data sources. In addition, we must enhance our capabilities related to operationalizing contextual risk factors—multilevel measurement models for ecological constructs, utilizing data from both individual-level and environment-level sources, should be extended to provide functional alternatives to the simple aggregation of individual-level measures. As another example, to analytically apply intersectionality theory in the investigation of social inequalities in suicide risk, we must enhance our person-centered measurement-modeling capabilities for effectively combining integral identity elements, leveraging the power of integrated data sources to avoid the empirical necessities of either singular examination of those elements in isolation or coarse categorizations of identity. Maximizing measurement reliability and validity for suicide risk and protective factors as well as potential intervention moderators at all levels is essential to avoiding misleading conclusions.
4:20 p.m. Discussion
5:00 p.m. Adjourn

Attendees will be responsible for meals and/or light refreshments on their own, at their own cost. The government and/or government contractors are not involved in facilitating the provision of food and/or light refreshments.