Hostname: page-component-586b7cd67f-t8hqh Total loading time: 0 Render date: 2024-12-01T20:36:13.649Z Has data issue: false hasContentIssue false

Guiding prevention initiatives by applying network analysis to systems maps of adverse childhood experiences and adolescent suicide

Published online by Cambridge University Press:  24 May 2024

Benjamin D. Maldonado
Affiliation:
Department of Computer Science and Software Engineering, Miami University, Oxford, OH, USA
Ryan Schuerkamp
Affiliation:
Department of Computer Science and Software Engineering, Miami University, Oxford, OH, USA
Cassidy M. Martin
Affiliation:
Department of Computer Science and Software Engineering, Miami University, Oxford, OH, USA
Ketra L. Rice
Affiliation:
National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, GA, USA
Nisha Nataraj
Affiliation:
National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, GA, USA
Margaret M. Brown
Affiliation:
Defense Suicide Prevention Office, Department of Defense, Washington, DC, USA
Christopher R. Harper
Affiliation:
National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, GA, USA
Curtis Florence
Affiliation:
National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, GA, USA
Philippe J. Giabbanelli*
Affiliation:
Department of Computer Science and Software Engineering, Miami University, Oxford, OH, USA
*
Corresponding author: Philippe J. Giabbanelli; Email: giabbapj@miamioh.edu
Rights & Permissions [Opens in a new window]

Abstract

Suicide is a leading cause of death in the United States, particularly among adolescents. In recent years, suicidal ideation, attempts, and fatalities have increased. Systems maps can effectively represent complex issues such as suicide, thus providing decision-support tools for policymakers to identify and evaluate interventions. While network science has served to examine systems maps in fields such as obesity, there is limited research at the intersection of suicidology and network science. In this paper, we apply network science to a large causal map of adverse childhood experiences (ACEs) and suicide to address this gap. The National Center for Injury Prevention and Control (NCIPC) within the Centers for Disease Control and Prevention recently created a causal map that encapsulates ACEs and adolescent suicide in 361 concept nodes and 946 directed relationships. In this study, we examine this map and three similar models through three related questions: (Q1) how do existing network-based models of suicide differ in terms of node- and network-level characteristics? (Q2) Using the NCIPC model as a unifying framework, how do current suicide intervention strategies align with prevailing theories of suicide? (Q3) How can the use of network science on the NCIPC model guide suicide interventions?

Type
Research Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2024. Published by Cambridge University Press

1. Introduction

Suicide is the second leading cause of death among adolescents ages 10–14 and third leading cause of death among adolescents ages 15–19 in the United States, accounting for 17.0% of all deaths among those aged 10–14 and 18.0% of all deaths among those aged 15–19 (Centers for Disease Control and Prevention's National Center for Injury Prevention and Control, 2020). Moreover, suicide rates among adolescents are on the rise: between 2009 and 2018, suicide rates among high school-aged children (i.e., ages ranging from 14 to 18) increased by 61.3% (Ivey-Stephenson et al., Reference Ivey-Stephenson, Demissie, Crosby, Stone, Gaylor, Wilkins, Lowry and Brown2020). Suicide attempts have likewise increased, doubling in a similar period between 2008 and 2015. Within that same high school age category, 18.8% of students reported seriously considering suicide, and 15.7% reported making a suicide plan. With suicide attempts and suicidal ideation (“a broad term used to describe a range of contemplations, wishes, and preoccupations with death and suicide” (Harmer et al., Reference Harmer, Lee, TvH and Saadabadi2020)) on the rise in recent decades, a public health solution is needed to effectively address the complex causes of suicidal behaviors.

Prior research has developed several models to capture the inherent complexity behind suicidal ideations (Brenas et al., Reference Brenas, Shin, Shaban-Nejad and Shabo2019; Chung, Reference Chung2016; Page et al., Reference Page, Atkinson, Campos, Heffernan, Ferdousi, Power and Hickie2018). Although these models use different approaches (e.g., ontologies, System Dynamics), they are all structured as networks consisting of concepts for nodes (e.g., homelessness, depression) and directed edges for the impact of one concept upon another. These models recognize suicide as a multifactorial problem, thus using networks to represent how suicide is shaped by (and contributes to) multiple causes instead of having a single “root” cause. As such networks are comprehensive, they can consist of a large number of factors and interrelationships. Visualizing all the content at once can be overwhelming for users and emphasizes the complexity of a problem, instead of focusing on specific solutions. Consequently, such network-based models are primarily used as decision-support systems either to understand the dynamics of a problem (e.g., by identifying key themes) or by examining the potential effects of interventions. Recently, researchers at the National Center for Injury Prevention and Control (NCIPC) within the CDC created a new causal map regarding the complexities linking Adverse Childhood Experiences (ACEs) to suicide ideation, attempt, and fatality among youth. The map was obtained by a participatory modeling project involving fifteen subject matter experts (SMEs) on various facets of ACEs and suicide (e.g., epidemiologists, psychiatrists, psychologists). We refer to this causal map as “the NCIPC model,” and its content is briefly summarized in the Appendix. It captures adolescent suicide through 361 nodes and 946 edges (Giabbanelli et al., Reference Giabbanelli, Galgoczy, Nguyen, Foy, Rice, Nataraj and Harper2021), which makes it a sizable systems map to both understand the problem and guide prevention initiatives.

The release of a map following a participatory modeling initiative is usually followed by analytical efforts rooted in network science. The Foresight Obesity Map (McPherson et al., Reference McPherson, Marsh and Brown2007) illustrates this approach: after its release in 2007, practitioners observed that the network was too large to be simply “looked at” (Siokou et al., Reference Siokou, Morgan and Shiell2014), hence network analyses were conducted to find useful patterns for obesity prevention. Such analyses have included the identification of communities, revealing themes that could centralize prevention efforts (Finegood et al., Reference Finegood, Merth and Rutter2010; Drasic and Giabbanelli, Reference Drasic and Giabbanelli2015) (e.g., body image, physical well-being). Node centrality has also been a common analysis, particularly with regard to finding leverage points to intervene in a system (McGlashan et al., Reference McGlashan, Johnstone, Creighton, de la Haye and Allender2016). The Foresight Obesity Map has also been compared with another map, using themes and node centrality to contrast the focal points of these two maps and their implications for obesity prevention initiatives (McGlashan et al., Reference McGlashan, Hayward, Brown, Owen, Millar, Johnstone, Creighton and Allender2018). The field of obesity is well-known for taking a “whole-system approach” resulting in an abundance of maps and comparative network studies, and other fields such as diabetes (Giles et al., Reference Giles, Findlay, Haas, LaFrance, Laughing and Pembleton2007) and socio-environmental management (Lavin et al., Reference Lavin, Giabbanelli, Stefanik, Gray and Arlinghaus2018) have also performed such analyses. Despite a recent growth of modeling studies applied to suicide (Schuerkamp et al., Reference Schuerkamp, Liang, Rice and Giabbanelli2023), fewer systems maps are devoted to suicide and no comparative network study has taken place into this context yet. In this paper, we perform the first comparative analyses through four such models for suicide and ACEs: two models with a focus on suicide (Chung, Reference Chung2016; Page et al., Reference Page, Atkinson, Campos, Heffernan, Ferdousi, Power and Hickie2018), one emphasizing ACEs (Brenas et al., Reference Brenas, Shin, Shaban-Nejad and Shabo2019), and the NCIPC model touching on both aspects.

Our study addresses three questions:

  1. Q1: What are the differences between network-based models for suicide prevention? We address this question by representing the four detailed models above-mentioned as a network. Following established practices in terms of contrasting themes and central nodes, we assess each network with respect to both overall structure (density, diameter) and key factors (via five measures of node centrality).

  2. Q2: Using a network model as a unifying framework, how are current suicide intervention packages aligned with prevailing theories for suicide? This question cannot be addressed solely based on the network, as we need to contextualize its content with respect to theories. We thus use a mixed-methods approach by examining network elements through a policy lens. Specifically, we categorize each node with respect to both the CDC suicide prevention technical package of policy, programs, and practices and the Social-Ecological Model (SEM) Framework (Stone et al., Reference Stone, Holland, Bartholow, Crosby, Davis and Wilkins2017; Centers for Disease Control and Prevention's National Center for Injury Prevention and Control, 2021).

  3. Q3: How can a network model guide suicide interventions? While Q1 emphasizes the comparison of structural elements among models, Q3 focuses on the NCIPC model since it was recently released with the aim of guiding the design of suicide prevention interventions. In Q3, our analysis complements a purely static perspective (“what is in the model”) with a dynamic approach, since interventions consist of what-if questions that require investigating potential pathways and rippling effects. We thus use algorithms to identify the major systems components (i.e., community detection) and demonstrate how to perform intervention-oriented tasks (e.g., identifying unintended consequences).

The remainder of this paper is organized as follows. In Section 2, we describe systems maps (including causal maps, stock and flow diagrams, and ontologies) together with the process for their creation and analysis. Section 3 details the mixed methods used to address the three questions above, with a subsection dedicated to each question. Similarly, our results in Section 4 are organized to provide quantitative and qualitative answers to each question. In Section 5, we discuss some of the limitations of this study inherent to the field of suicide prevention. Section 6 summarizes our main contributions at the interface of applied network science and suicide.

2. Background: systems maps

2.1 What is a systems map?

System mapping is a tool originating from systems thinking in which a given domain is represented as a set of components (nodes) and relationships (edges). While components always have a name (i.e., node labels), the type of edges depends on the specific representation of a system used in a given project. One such representation is a causal map, in which categorized edges labeled positive (+) or negative (−) link various quantifiable factors to create a directed network (Firmansyah et al., Reference Firmansyah, Supangkat, Arman and Giabbanelli2019). A positive edge from one factor to another implies positive causation (an increase in the first factor will cause an increase in the second), while a negative edge implies negative causation (an increase in the first factor will cause a decrease in the second). For example, a positive edge from the node representing suicide ideation to the node representing suicide attempts means that an increase in suicide ideation leads to an increase in suicide attempts. Other representations of a system include stock and flow diagrams used for System Dynamics, where an edge may have a numerical value that is encoded within a modeling software rather than displayed on a schema. Ontologies are a related representation, in which the meaning and characteristics of a system can be conveyed in more details by equipping concepts with properties or associating them with entries in a glossary.

Several features present in complex problems such as feedback loops Footnote 1 (when a change in one factor ultimately affects itself) and factors with a high number of outgoing relationships can be represented in systems maps (Giabbanelli et al., Reference Giabbanelli, Galgoczy, Nguyen, Foy, Rice, Nataraj and Harper2021). As a result, these maps are a powerful tool to analyze complex problems.

2.2 Why make a systems map?

Mapping a preexisting system is now one of the key stages in identifying and evaluating interventions for public health programs (Luna Pinzon et al., Reference Luna Pinzon, Stronks, Dijkstra, Renders, Altenburg, den Hertog and Waterlander2022). Creating and visualizing a systems map can exhibit the intricacies of the issue at hand and highlight the interdependence of factors through feedback loops (Finegood et al., Reference Finegood, Merth and Rutter2010). Depicting feedback loops is critical as it helps decision-makers to comprehend the relationships between variables and begin to weigh the ramifications of policies and interventions. Interventions are actions taken to change variables in a system to achieve a specific outcome—e.g., reducing access to lethal means among persons at risk of suicide (Stone et al., Reference Stone, Holland, Bartholow, Crosby, Davis and Wilkins2017). Identifying feedback loops also empowers decision-makers to utilize naturally occurring system dynamics in their interventions (Firmansyah et al., Reference Firmansyah, Supangkat, Arman and Giabbanelli2019). For example, implementing social-emotional learning programs in schools may increase coping and problem-solving skills, which may, in turn, reduce suicide attempts (Stone et al., Reference Stone, Holland, Bartholow, Crosby, Davis and Wilkins2017).

System maps also allow users to identify a group of factors, the variables affected by them (both directly and indirectly), and the nature of these effects (e.g., by monitoring their type in a causal map) (Firmansyah et al., Reference Firmansyah, Supangkat, Arman and Giabbanelli2019). Without a systems map, tracing the effects of variables throughout the system would likely be more difficult. The process of creating a systems map can have widespread benefits even if the map is not widely used. Bringing together a broad range of stakeholders to form the map can help “forge multisector, multidisciplinary relationships that support future decision-making based on evidence” (Finegood et al., Reference Finegood, Merth and Rutter2010). Moreover, stakeholder participation in creating a map can increase the legitimacy of decisions (Firmansyah et al., Reference Firmansyah, Supangkat, Arman and Giabbanelli2019). Combining both factors can help boost the quality and support of decision-making.

2.3 How to create a systems map?

The general process of constructing systems maps with participants has recently been covered in several articles, with respect to either knowledge representation (Giabbanelli and MacEwan, Reference Giabbanelli and MacEwan2024) or the interactions between facilitators and participants (Knox et al., Reference Knox, Furman, Jetter, Gray and Giabbanelli2024). We thus provide a more succinct overview here, based on the diagram in Figure 1. One approach to create a map is to engage in participatory modeling. In this case, modelers perform community-based participatory research (CBPR) by building a collaborative partnership with community members through participatory activities (McGlashan et al., Reference McGlashan, de la Haye, Wang and Allender2019). A model is created in CBPR by identifying and recruiting eligible participants, then obtaining informed consent. Participants are selected based on the problem boundaries; for instance, in the case of the model on adolescent suicide, individuals whose experience lies entirely on suicide in adulthood would not be engaged (Giabbanelli et al., Reference Giabbanelli, Galgoczy, Nguyen, Foy, Rice, Nataraj and Harper2021). Although research areas such as educational technology tend to use software so that each individual directly creates their map (e.g., cMap, jMap, Coggle), CBPR has fewer platforms (e.g., STICKE (Hayward et al., Reference Hayward, Morton, Johnstone, Creighton and Allender2020)) and tends to employ trained facilitators who will support participants in developing maps. Data elicitation techniques can either be performed with a group, which will automatically result in a group-level model or with individuals, whose models are later combined into a group-level model (Giabbanelli et al., Reference Giabbanelli, Galgoczy, Nguyen, Foy, Rice, Nataraj and Harper2021).

Figure 1. General process for creating a systems map.

When working with individuals, information is elicited using unstructured techniques, which rely on asking open-ended questions to externalize the participants" perspectives without imposing the bias of the interviewer on the interviewee. Transcribed interviews can be analyzed to identify variable names and the nature of their relationships, represented as nodes and edges, respectively. When building a causal map, relationships encode the direction and type of impact. For example, “traumatic experiences increase the risk for suicide ideation” and “having meaningful relationships with positive peers protects against suicide attempts.” Consequently, causal maps use directed, typed edges. In the two examples, the statements could be represented as traumatic experiences $\xrightarrow{\text{+}}$ suicide ideation and connectedness $\xrightarrow{-}$ suicide attempts, respectively. In the case of ontologies or stock and flow diagrams, relationships would have attributes or numerical values, respectively.

Even when they share the same interest, modeling projects may differ in some of the sub-steps (Figure 1) based on considerations such as experience and availability of the facilitator (e.g., do they prefer one-on-one sessions or group workshops?), access to participants (e.g., can they all come to the same room?), or time constraints (e.g., do we need to produce a map by the end of the workshop?). The NCIPC model employed one-on-one interviews with a trained facilitator and an observer, then a causal map was produced from each interview (including typed and directed edges), and maps were aggregated with the assistance of SMEs to identify semantically equivalent constructs (Giabbanelli et al., Reference Giabbanelli, Rice, Galgoczy, Nataraj, Brown, Harper, Nguyen and Foy2022). The NCIPC model thus only reflects the views of the participants involved. The stock and flow diagram produced by Page and colleagues involved workshops and a wide representation “from health and social policy agencies, local councils, non-government organizations, emergency services, primary care providers, program planners, research institutions, community groups and those with lived experience of suicide” (Page et al., Reference Page, Atkinson, Campos, Heffernan, Ferdousi, Power and Hickie2018). The model-building method thus differs from the NCIPC case, which relied solely on experts (e.g., epidemiologists, psychologists). As is commonly the case, the model by Page et al. was developed using the same protocols that the team applied for CBPR research in other domains, such as diabetes management (Freebairn et al., Reference Freebairn, Atkinson, Kelly, McDonnell and Rychetnik2016) and alcohol-related harms (Atkinson et al., Reference Atkinson, O’Donnell, Wiggers, McDonnell, Mitchell, Freebairn, Indig and Rychetnik2017). The model by Page et al. proceeds in a series of steps, from the general population to vulnerable individuals, who can become distressed and attempt suicide, leading either to fatality or post-attempt treatment.

CBPR is not the only way to create a map. For example, participants may be unavailable for a research team, or the modelers may consider that the evidence-base is sufficient to derive a map. The stock and flow model created by Chung thus indirectly builds on the work of experts, by expanding on the Interpersonal Theory of Suicide (Chu et al., Reference Chu, Buchman-Schmitt, Stanley, Hom, Tucker, Hagan, Rogers, Podlogar, Chiurliza, Ringer, Michaels, Patros and Joiner2017). This theory posits that suicide ideation happens when individuals feel that they are a burden and do not belong (i.e., the core constructs of belongingness and burdensomeness). In addition, the theory considers that attempts are enabled when individuals have a desire for suicide and no longer fear death (i.e., capability) (Vélez--Grau et al., Reference Vélez-Grau, Magan and Gwadz2023; Kelley and DeShong, Reference Kelley and DeShong2023). Chung created the model by building on theory and identified specific variables using data from the In-Home surveys administered as part of AddHealth, a widely studied dataset created by experts. As a result, the model includes constructs such as depression (characterized by e.g., average duration and depression level), perceived self-worth, desire to die by suicide, fear of death (which is reduced by attempts and exposure to painful experiences), hopelessness, burdensomeness, and cultural stigma associated with suicide. The ontology by Brenas et al. (Brenas et al., Reference Brenas, Shin, Shaban-Nejad and Shabo2019) also built on prior work, as it (i) reused the seminal study of Felitti and colleagues to identify relationships between Adverse Childhood Experiences (Felitti et al., Reference Felitti, Anda, Nordenberg, Williamson, Spitz, Edwards, Koss and Marks1998), (ii) reused medical ontologies to cover concepts such as abuse or mental illnesses, and (iii) connected the concepts as needed for the purpose of their study. The constructs thus cover mental health, physical harm (detailed as being hit, push, from thrown object, etc.), substance use and abuse, housing problems (e.g., low-quality housing or homelessness), and their effects on health (detailed via e.g., exposure to bug infestation, lead-based paint, mold, water leaks), triggering events (e.g., death, incarceration, parental separation or divorce), and access to care (e.g., transportation).

2.4 Analyzing systems maps

In a comparative network analysis, we need to assess the models based on shared features. Regardless of whether they originated as ontologies, causal maps, or stock and flow diagrams, all four models share a representation as directed, labeled networks. However, edges have different information, such as types (+ and − in causal maps), numerical weights or rates per time unit (stock and flow diagrams), or properties (ontologies). We can thus analyze all four models by accounting for the direction of their edges, but not in terms of edge weights or edge types.

A network can be analyzed at different levels of granularity. McGlashan and colleagues have summarized how several network analyses can be used on a systems map, defining each analysis (e.g., density, degree) along its interpretation in the context of a systems map and the takeaway message for interventions (c.f. Table 1 in McGlashan et al. (Reference McGlashan, Johnstone, Creighton, de la Haye and Allender2016)). The most granular analysis is performed at the level of individual nodes, using centrality measures to capture the “importance” of a node with respect to a desired property (e.g., degree, flow). Although node centrality is a well-known analytical method in network science, the interpretation of results is field-dependent. Table 1 lists the typical centrality measures used in systems maps together with their interpretation in the context of intervention planning for public health. Figure 2 exemplifies how centralities might be calculated on a fragment of the NCIPC model. These centrality measures have been used in numerous other studies on network science for public health (Power et al., Reference Power, Lambe and Murphy2022; Brauer et al., Reference Brauer, Wulff, Pawellek and Ziegeldorf2022; Smith et al., Reference Smith, Georgiou, King, Tieges and Chastin2022), as well as in other comparative studies on systems maps (Krabbe, Reference Krabbe2014; Kim and McCarthy, Reference Kim and McCarthy2021; Wang and Giabbanelli, Reference Wang and Giabbanelli2023).

Table 1. Node centrality measures and their interpretation in the context of systems maps on suicide

At an intermediate level of granularity, we can analyze groups of nodes. This can be achieved by examining very local groups, through the clustering coefficient. For a given node, this measure quantifies the density of connections among the nodes’ neighbors. For example, if a node has four neighbors and they have three edges (out of $4 \times 3 = 12$ possible edges among four nodes), then the node has a clustering coefficient of $\frac{3}{12} = 0.25$ . The clustering coefficient of a network is obtained by averaging the clustering coefficient of individual nodes. While the clustering coefficient is a familiar construct in the literature on suicide, it is more commonly employed in neurophysiology studies through brain networks (Kim et al., Reference Kim, Jang, Lee, Shim and Kim2024; Qin et al., Reference Qin, Li, Zhang, Yin, Wu and Pan2024) or in computational social science via online social networks (Masuda et al., Reference Masuda, Kurahashi and Onari2013; Colombo et al., Reference Colombo, Burnap, Hodorog and Scourfield2016). In this paper, we use the clustering coefficient to study a network model of suicide. Larger groups of nodes can be studied through communities. Communities exist in certain subgroups where connections within the group are denser than connections to other groups. Communities can be useful for revealing hidden relationships among nodes (Lee and Lee, Reference Lee and Lee2013), identifying themes (Chunaev, Reference Chunaev2020), and visualization purposes (Fan, Reference Fan2020). For example, communities can simplify a large system map into five or six themes, which become priority areas for prevention. Communities can serve to create “derivative maps” which are reduced to their groups, as illustrated by analyses of the Foresight Obesity Map which was reduced into a series of such derivative maps to guide policies (Finegood et al., Reference Finegood, Merth and Rutter2010).

Figure 2. Fragment of the National Center for Injury Prevention and Control (NCIPC) model with three centralities: node size represents betweenness centrality (i.e., a larger node has control over a larger share of shortest paths), colors show degree centrality (nodes with more ties are darker), and the number indicates closeness centrality (which is based on distances). Betweenness centrality tracks how many times a concept would be traversed, hence we see that “child getting help” and “visit to the ER” are commonly occurring steps (i.e., frequently traversed) through the journey of an individual experiencing suicide ideation or attempts. Degree centrality reveals the extent to which a construct directly relates to another construct (e.g., therapy success is driven by parental involvement and also contributes to handling different emotions).

The goal of community detection via algorithms is to divide nodes in such a way that there are many edges in subgroups and few edges between them (Newman, Reference Newman2018). Multiple community detection algorithms have been developed (Clauset et al., Reference Clauset, Newman and Moore2004; Lancichinetti and Fortunato, Reference Lancichinetti and Fortunato2009; Newman and Girvan, Reference Newman and Girvan2004), and we use two algorithms to support the triangulation of results (i.e., ensure that findings are not an artifact of one specific method). The Louvain Algorithm for community detection greedily maximizes modularity (Newman, Reference Newman2018, p. 511–512) by repeatedly applying two steps: joining individual nodes into groups to find the arrangement with the highest modularity (which measures the density of links within a group as compared to links going outside) and restructuring the network by aggregating communities into “super nodes”. This iterative process is performed until no increase of modularity is possible. Figure 3 exemplifies the application of this method onto the same section of the NCIPC model as shown in the previous figure for centrality. In the worst case, communities detected using the Louvain Algorithm can be badly connected or even disconnected. The Leiden Algorithm extends the Louvain Algorithm to guarantee well-connected communities by using partition refinement. This algorithm consists of three steps: 1) local moving of nodes, 2) refinement of the partition, and 3) aggregation of the network based on the refined partition (Traag et al., Reference Traag, Waltman and Van Eck2019). By refining communities, the algorithm has more room to identify high-quality partitions. The use of several community detection algorithms on a map helps to establish the stability of the communities, hence ensuring that they reflect fundamental properties of the empirical data instead of being an artifact of one specific method.

At the level of the whole network, four measures are commonly included (McGlashan et al., Reference McGlashan, Johnstone, Creighton, de la Haye and Allender2016; McGlashan, Reference McGlashan2018). The density refers to the portion of relationships that exist compared to how many relationships are possible. Table 2 shows examples of several policy application areas and their corresponding map densities to show some common densities for these types of maps. We note that the density in this application area is generally very low (i.e., we have sparse graphs). This property can be important for other aspects of the analysis, such as guiding the choice of centrality measures (Freund and Giabbanelli, Reference Freund and Giabbanelli2022). Second, distributions for the in- and out-degrees of all nodes provide some insight into the general structure of a model. Unlike other fields, the intent in systems maps for policy-making is not typically to fit a certain distribution (e.g., test for a power-law). Rather, the examination tends to focus on three categories of degrees (Eden et al., Reference Eden, Ackermann and Cropper1992; Alomia-Hinojosa et al., Reference Alomia-Hinojosa, Groot, Andersson, Speelman, McDonald and Tittonell2022; Konti and Damigos, Reference Konti and Damigos2018; Giles et al., Reference Giles, Findlay, Haas, LaFrance, Laughing and Pembleton2007): nodes without incoming edges (i.e., sources or “transmitters” with in-degree 0), which often constitute parameters of the model; nodes without outgoing edges (i.e., sinks or “receivers” with out-degree 0), which can serve as outputs; and hubs (large in- or out-degree), which can be prime targets for intervention planning due to their high connectivity. The Average Path Length is defined as the average of the shortest paths between each pair of nodes. It relates to the amount of effort that is needed to spread change, on average. Finally, the diameter measures the furthest distance between two nodes of the graph. Intuitively, a larger diameter indicates that a systems map has a wide “span” by covering far-away topics. The diameter has been extensively studied across various types of systems maps Strautmane (Reference Strautmane2012); Maksimenkova et al. (Reference Maksimenkova, Neznanov, Papushina and Parinov2018); Barbrook-Johnson and Carrick (Reference Barbrook-Johnson and Carrick2022).

Table 2. Number of nodes, edges, density, and number of participants in participatory modeling studies that produced maps. Note that most models have a low density (i.e., sparse graphs)

Figure 3. The Louvain method detected five communities in this fragment of the National Center for Injury Prevention and Control (NCIPC) model causal map, shown with distinct colors. Labels are only indicative, since the algorithm assigns concepts to communities but does not automatically name these communities.

3. Methodology

3.1 Comparing network models: node- and network-level metrics (Q1)

A network model can be built for different objectives. Small-scale models often serve to estimate the (direct and indirect) effects of a small set of variables of interest onto a suicide-related construct. For example, Kranzler et al. (Reference Kranzler, Fehling, Anestis and Selby2016) used an 11-node model to examine how self-injury and emotion dysregulation contribute to suicide attempt and Tang et al. (Reference Tang, Yen, Cheng, Yang, Chen, Yang, Huang, Jong and Yu2010) used a six-node model to estimate how mental health constructs (e.g., post-traumatic stress disorder) ultimately shape the risk of suicide. In contrast, detailed models intended for intervention planning and evaluation have a much larger number of constructs and associated edges. Owing to their level of details and the significant investment required to create them via participatory modeling efforts (Section 2.3), there are fewer such models. For each of the four models examined in this paper, we obtained a directed network either directly from a file provided by the authors in a repository or supplementary materials (NCIPC and Brenas et al. (Reference Brenas, Shin, Shaban-Nejad and Shabo2019)), or by manually re-creating all edges and nodes from the authors’ figures (Chung, Reference Chung2016; Page et al., Reference Page, Atkinson, Campos, Heffernan, Ferdousi, Power and Hickie2018). Using the Python library NetworkX (version 2.6.2), we extracted characteristics for each of these four models at the node- and network-level (Section 2.4). Network-level characteristics include the number of nodes, number of edges, density, average path length, and degree distribution. Node-level characteristics include five node centrality measures: degree, Katz, Betweenness, Load, and Closeness. For each network, we focused on the top centralities, hence nodes were ranked by centrality.

3.2 Comparing perspectives: model, prevention package, and prevailing theories (Q2)

The SEM is a common framework for ACEs and suicide prevention (Centers for Disease Control and Prevention's National Center for Injury Prevention and Control, 2021; Cramer and Kapusta, Reference Cramer and Kapusta2017). This approach considers that suicide-related factors belong to four different levels: individual, relationship, community, or societal. Taking the example of (Shagle and Barber, Reference Shagle and Barber1995), suicidal ideation is shaped by personal coping strategies (individual), family and peers (relationship), the school environment (community), and religion (societal). The CDC offers a complementary perspective through its suicide prevention technical package of policy, programs, and practices (Stone et al., Reference Stone, Holland, Bartholow, Crosby, Davis and Wilkins2017). In the technical package, concepts are classified into seven prevention strategies and approaches (Stone et al., Reference Stone, Holland, Bartholow, Crosby, Davis and Wilkins2017, p. 12–44), with an associated “shorthand notation” introduced here: strengthen economic supports (“economic”), strengthen access and delivery of suicide care (“care”), create protective environments (“environments”), promote connectedness (“connectedness”), teach coping and problem-solving skills (“coping”), identify and support people at risk (“support”), and lessen harms and prevent future risk (“prevent”).

In this study, we categorized the 361 nodes in the NCIPC Model per the SEM framework. This alignment is a qualitative endeavor: given the content of a model (i.e., the nodes’ names/labels), we seek to identify which aspect of a prevention strategy is covered and which level of the SEM framework is mobilized. Different individuals may interpret a model differently, which is a classic problem in reusing models (Freund and Giabbanelli, Reference Freund and Giabbanelli2021). This creates the risk that findings reflect the particular interpretation of a group rather than the inherent characteristics of a model. To mediate this risk, we employ multiple coders from different backgrounds. By analyzing where their classifications concur, we can identify reliable categories. Specifically, we assigned nodes to four coders comprising two SMEs (in either suicide and/or ACEs) and two lay persons. Note that the non-SMEs and SMEs were not included in the same categorization groups and were instead split because of the differences in experience and knowledge in the specific areas of suicide and ACEs. Each coder assigned categories to each of their assigned nodes. They could disagree on a category; for example, a person may have seen “culture of secrecy” as a matter of environments and prevention strategies, while another may have seen it as environments and coping strategies. Coding conflicts were identified and resolved with a discussion between the coders, who explained the reasons for their classification and arrived at a consensus. In this example, one possible outcome could be the final categories of environments and coping. It was also possible that a consensus led to adding a category that was used by neither coder initially. This process has been documented in our online supplementary materials, with a “Nodes Additions” table devoted to categories that emerged as a result of inter-coder discussions. For a given node, the inter-coder agreement is defined as the fraction of categories assigned to the node that were endorsed by both coders. In our guiding example, since coders only shared the “environments” category out of three categories used in total, the inter-coder agreement is $\frac{1}{3}$ .

3.3 Utilizing a model to guide policy: themes and intervention-oriented measures (Q3)

Complex problems such as adolescent suicide are addressed by myriad of actors, from the school (e.g., role models and after-school programs) to the community (e.g., psychologists and psychiatrists). Systems maps in other fields such as obesity have shown that the responsibilities of each actor may not fully align with the characteristics of the problem, thus leading to aspects for which nobody was clearly responsible, or aspects with shared responsibilities without clear communication lines. Cross-sector partnerships and whole-of-society models have thus emerged as multi-stakeholder approaches to policy-making in such problems (Dubé et al., Reference Dubé, Addy, Blouin and Drager2014; Bekker et al., Reference Bekker, Helderman, Jansen and Ruwaard2017), thus shifting from an actor- and problem-centric perspective (e.g., nutritionists and general practitioners for obesity) to a solution-centered approach (e.g., on themes such as body image and mental well-being). The decomposition of a map into themes is thus a key step for this paradigm shift, enabled by communication detection algorithms. Using the Python Library CDlib (version 0.2.5), we performed community detection on the NCIPC model using both the Louvain algorithm and the Leiden algorithm and compared their results to minimize the risk that communities are an artifact of one specific method. Each community was then named by manual inspection of its content and compared to key categories for suicide prevention from both the CDC suicide prevention technical package of policy, programs, and practices (Stone et al., Reference Stone, Holland, Bartholow, Crosby, Davis and Wilkins2017) and SEM framework (Centers for Disease Control and Prevention's National Center for Injury Prevention and Control, 2021) (Section 3.2). The identification of communities helps find themes and coordinate multiple stakeholders to take and monitor interventions. The next step in using a system map is to evaluate the potential effect of such interventions (Mkhitaryan et al., Reference Mkhitaryan, Giabbanelli, de Vries and Crutzen2020). Using the Actionable Systems tool (Giabbanelli and Baniukiewicz, Reference Giabbanelli and Baniukiewicz2018) to navigate networks, we show that the NCIPC model can support three types of interventions:

  1. 1. The multiple paths, as an intervention on one factor (e.g., access to mental health treatment), may have several direct and indirect effects onto an evaluation target (e.g., suicide ideation). We identify all disjoint paths between an intervention node and an outcome node.

  2. 2. The feedback loops, which can either work against an intervention (e.g., by creating inertia in the system and locking in certain dynamics) or help to amplify its effects (e.g., by reinforcing the initial impetus created by an intervention). We list all simple cycles in the map and categorize them as reinforcing or balancing based on standard terminology in systems science.Footnote 2

  3. 3. The rippling consequences, as an intervention in one part of the map seeks to affect another specific part but may inadvertently have effects elsewhere. Hence we perform a core-decomposition centered on an intervention node, to reveal the parts of the map that would be affected at different distances.

Identifying multiple paths, feedback loops, and rippling effects is essential to comprehensively assess the cost-effectiveness of an intervention and mitigate unintended consequences. These types of measurements are a systems science approach to network-level analysis, thus complementing classical network science metrics such an examination of the system’s dynamics.

4. Results

4.1 Comparing network models (Q1)

Network statistics for the four models are summarized in Table 3. We note that they are all sparse both globally (low density) and locally (low clustering), indicating that the SMEs were highly selective in identifying relationships between nodes. The relatively slow increase in the diameter and average path length compared with the increase in the number of nodes as the models get larger (from left to right in Table 3) suggests that experts increased granularity pertaining to certain facets of suicide (e.g., ACEs) instead of expanding onto new domains. For example, the NCIPC model has six times as many nodes as Chung (Reference Chung2016), but about the same average path length and only a small increase in diameter (from 11 to 13). For the two largest models, we see that the average path length is very short relative to the network sizeFootnote 3 so it is relatively easy for one factor of the model to affect another. From a participatory modeling viewpoint, nodes are thus generally related, and participants did not embark on long tangents with aspects that had little to do with other factors of the model. From a policy implication viewpoint, it also means that the existing system surrounding suicide is tightly interconnected, which likely necessitates the joint use of several intervention points to deliver a system-wide change. We also note that clustering is almost twice as large in the NCIPC model compared to the previous sizeable models. This higher clustering coefficient suggests that there are more tightly connected neighborhoods of nodes, which represent areas of related concepts. The clustering coefficient is thus further evidence that SMEs have delved further into suicide-related notions instead of expanding into previously uncovered themes. Distributions for in- and out-degree (Figure 4) are highly skewed in all models but Page et al. (Reference Page, Atkinson, Campos, Heffernan, Ferdousi, Power and Hickie2018), which is the smallest one with 18 nodes. Once models become sizeable, they clearly have a few “hub nodes” that attract a lot of attention and a large number of nodes with few connections. The identification of these hub nodes is thus of particular interest to examine the focal points of the models and their policy implications.

Table 3. Network measurements for all four network-based suicide models

Centrality measures help identify and compare important nodes across models. We identified important nodes for each model by observing the highly ranked nodes across centrality measures. For example, in the NCIPC Model, ACEs were ranked first according to degree, load, and betweenness centrality and third according to Katz and closeness centrality. This process led us to identify ACEs, mental health disorders, suicide ideation, suicide attempts, and substance abuse as driving nodes in the NCIPC Model. Table 4 summarizes the top five nodes for all models with respect to five centrality measures (degree, betweenness, load, closeness, and Katz centrality); our online supplementary material includes tables listing up to the top ten nodes. In other models, prominent factors are sleep and alcohol disorders, capability and desire for suicide as well as attempts, distress, and past attempts (Brenas et al., Reference Brenas, Shin, Shaban-Nejad and Shabo2019; Chung, Reference Chung2016; Page et al., Reference Page, Atkinson, Campos, Heffernan, Ferdousi, Power and Hickie2018).

Table 4. Some nodes have the same value for a centrality measure, indicated by multiple nodes with the same rank. Note that Brenas et al. (Reference Brenas, Shin, Shaban-Nejad and Shabo2019) have duplicated nodes which the above rankings reflect; Page et al. (Reference Page, Atkinson, Campos, Heffernan, Ferdousi, Power and Hickie2018) also have a special “Black Hole” node, which refers to an exit point where the relationships leave the scope of concepts in the model. Mental health is abbreviated as `mhs' by the authors.

Figure 4. In-degree and out-degree distributions for four models. We note a skewed distribution indicating a high prevalence of low-degree nodes in the first three and a more uniform distribution in the last (bottom) model. The NCIPC model refers to the causal map created by the National Center for Injury Prevention and Control (NCIPC). Note that scales are different, since some models have fewer edges than others.

At a high level, the top five nodes suggest that central factors across models are proximal to the individual, rather than community or societal constructs. This can be expected for network models, as their objective is to elucidate suicide of individuals, hence all contributing factors eventually point to the individual level. At a detailed level, three of the models have similar key constructs (NCIPC, Page et al. (Reference Page, Atkinson, Campos, Heffernan, Ferdousi, Power and Hickie2018); Chung (Reference Chung2016)): mental health (e.g., through disorders or distress) and suicide (re-attempts). Brenas et al. (Reference Brenas, Shin, Shaban-Nejad and Shabo2019) have more unique focal points due to emphasis on ACEs rather than suicide. Note that key concepts in Brenas et al. (Reference Brenas, Shin, Shaban-Nejad and Shabo2019) are not absent in larger maps such as the NCIPC map; rather, they are relatively less important in the wider scope of the model. For example, substance abuse is among the top-10 nodes for closeness centrality in the NCIPC model, hence reflecting the notion of substance use and misuse in Brenas et al. (Reference Brenas, Shin, Shaban-Nejad and Shabo2019).

When seeking to use highly central nodes for intervention design, it is important to note that such nodes may be the outcomes (i.e., cannot be directly changed) of the intervention or the leverage point (i.e., the action through which we change the system). For instance, suicide ideation or attempts are the outcomes that we wish to change, but we cannot directly change them. Leverage points that will ultimately affect these target outcomes would include firearm restriction, as it will lower access to lethal means and eventually prevent attempts. In the NCIPC model, the top five most central nodes per degree centrality are ACEs, suicide ideation, suicide attempt, mental health disorders, and poverty. The first three are target outcomes, while the latter two are potential intervention points. Since mental health disorders are also among the top five constructs per betweenness, we see that mental health is heavily involved throughout the journey of an individual from trauma to ideation and attempts. The complementary measures of centrality thus suggest that intervening on mental health disorders could have a high immediate effect (per the degree) and even more rippling effects (per betweenness). Ultimately, a network analysis alone cannot determine whether a highly central node is an outcome or a leverage point: this interpretation must be made by a SME. Other examples of leverage points suggested by the network analysis and domain expertise include addressing poverty (which has a high degree in the NCIPC model) and other social determinants of health (high degree in Brenas et al. (Reference Brenas, Shin, Shaban-Nejad and Shabo2019)) or improving housing and treatment for mental health disorders (which have high load centrality in Brenas et al. (Reference Brenas, Shin, Shaban-Nejad and Shabo2019) and the NCIPC model).

4.2 Comparing perspectives (Q2)

Figure 5 summarizes how the nodes of the NCIPC map were categorized by lay individuals (left) and experts (right) with respect to both the SEM framework and the CDC suicide prevention technical package of policy, programs, and practices. Each group had two coders. Inter-coder agreement among lay individuals was 73.6% on the social-ecological level and 48.3% on the prevention strategy. During resolution, the lay coders realized that the CDC suicide prevention efforts were not mutually exclusive, but highly overlapping approaches that examine the problem from all sides.

Both lay coders and SMEs agreed that a large proportion of nodes fell into two major prevention strategies (teaching coping skills and identifying and supporting people at risk) and two social-ecological levels (individual and relationship). The intersection of relationship-level issues being solved by learning coping skills makes up 23.55% of all concepts covered in the NCIPC model. In both the SME and lay heatmaps (Figure 5), there is a significant decrease in the number of nodes affected by societal-level frameworks, making up at most 7% of all concepts in the NCIPC model. This suggests that many of the interventions that would impact adolescent suicide come from family, friends, and the community. We find the inclusion of the lay coders to be important, as it helps display gaps in public knowledge and can suggest tailored communication initiatives to further educate community members with tools for suicide prevention.

Figure 5. Percent of nodes assigned to each combination of prevention strategy and socio-ecological level. For example, 18.84% of nodes belong at the individual level, and the support prevention strategy could help address it according to the expert plot. Nodes can belong to multiple socio-ecological levels, and several prevention strategies may address them hence the total exceeds 100%.

Table 5. Both community detection algorithms produced similar communities; entries are ordered such that these similar communities are placed next to each other for ease of visibility. The community numbers are also used in Figure 5

4.3 Utilizing a model to guide policy (Q3)

4.3.1 Leveraging community identification to identify suicide themes

Table 5 summarizes the communities found by the Louvain and Leiden algorithms on the NCIPC model, and how they compare to categories examined in Q2 (social-ecological theme, CDC prevention category). For example, when looking at the Personal and Community Acceptance of Identity community in the Louvain detection algorithm, we can see where many of the CDC’s current approaches, namely promoting connectedness of the community and teaching coping skills to individuals, can be applied to this area. Both algorithms selected their own communities without any prior bias being imposed on them by the researchers; they found many of the same communities, with some differences in the bottom half of the table. The Louvain and Leiden algorithms respectively extracted eleven and ten communities of nodes from the NCIPC Model. Five communities are nearly identical, and two are similar between the two algorithms. The Leiden algorithm detected one community where discrimination (i.e., discriminatory hiring practices, segregated housing) was prominent that Louvain did not. Moreover, the Louvain algorithm discovered a community encompassing the societal stigma around using mental health care that Leiden did not.

Figure 6. National Center for Injury Prevention and Control (NCIPC) model reduced based on the communities (comm.) detected by the Louvain (left) and Leiden algorithms (right). The adjacency matrices (top) provide a more compact view of the structures shown in node-and-link diagrams (bottom). Community numbers refer to Table 5.

Figure 6 displays the NCIPC Model reduced based on the communities detected by the Louvain and Leiden algorithms. The thickness of the arrow summarizes the strength of the connections between the communities based on the number of edges between their nodes. There are several points of agreement between the decompositions suggested by the two community detection algorithms. For example, ACEs (community #1) and access to health care and racial barriers (community #5) are strongly related to the rest of the system, which provides structural evidence that these communities form important themes for suicide prevention. However, the two algorithms offer widely different pictures on some themes, such as connectedness, suicide ideation, and attempt (community #6): it is weakly connected to anything with the Louvain algorithm (Figure 6; left) but one of the most highly connected themes per the Leiden algorithm (Figure 6; right). Suicide research would agree with the Leiden algorithm, as the vast majority of factors represented in the map are either protective or risk factors regarding suicide ideation and attempt, either directly or in relation with connectedness and ACEs. Furthermore, the Leiden algorithm correctly identifies the very strong relation coping skills and interpersonal relationships with peers (community #8) and ideation/attempt/connectedness, whereas the Louvain algorithm concludes that there is no such relation. This suggests that the Leiden algorithm is more apt to decompose our specific system into communities. Among the strong connections found by the Leiden Algorithm in Figure 6 (right), we note environmental factors on ACEs and development (community #7), socioeconomic factors, and discrimination (community #10). We also observe that household challenges and stress in the family (community #9) are involved in many other themes, such as emotional support from parents and crisis treatment (community #4) or socioeconomic factors and discrimination (community #10); these capture the impact of family stress onto youth and causes of family stress, respectively.

Figure 7. The ActionableSystems software allows users to automatically identify all pathways from one concept to another, as exemplified here by going from Adverse Childhood Experiences (ACEs) of parents (left) to suicide fatality (right).

4.3.2 Systems interventions

We asked a SME on suicide and another SME on ACEs to provide queries that would commonly be used in their work, in order to illustrate each of the types of policy-oriented questions listed in Section 3.3. We start by investigating the notion of multiple paths between a construct of interest and a suicide outcome. Specifically, in Figure 7, we examined the many ways in which ACEs of parents eventually contribute to suicide fatality of the child, thus highlighting intergenerational effects. These effects are well established in the literature on suicide, as a recent review showed that nearly all studies demonstrated the effect of parental ACEs onto their children, either directly or indirectly (Zhang et al., Reference Zhang, Mersky, Gruber and Kim2023).

Next, we exemplify the concept of feedback loops, whereby the effect of a factor ultimately impacts the same factor either in a reinforcing (i.e., amplifying) or balancing (i.e., dampening the dynamics) manner. In Figure 8, we selected a sample of six feedback loops covering different domains, from the benefits of therapy and changes in parental practices (Figure 8a/c) to the many loops involving parental frustration and coping via substance abuse or perpetration of ACEs (Figure 8b/d). We note that similar loops have been reported in other studies, such as loops between children’s well-being and parenting quality (Smith et al., Reference Smith, Alderman, Attell, Avila Rodriguez, DiGirolamo, Snyder, Covington, Manteuffel and Minyard2021). Thanks to its broad coverage of proximal and distal factors, the loops also include high-level societal constructs such as poverty, thus demonstrating the perpetuation of social issues (Figure 8e). Loops involving poverty in suicide have also been reported in recent mapping studies, for instance as poverty creates more unmet physical, emotional, and social needs, which ultimately impair learning and create a barrier to career attainment, hence reinforcing poverty (Vu et al., Reference Vu, Bishop, McGill, Valmadrid, Golden, Emmerling, Saeugling and Rubinelli2022).

Finally, we illustrate the idea of rippling effects, where an intervention targets a specific construct and its consequences are examined through the broader map. Economic policies (Figure 9) can directly improve financial stability and lower stress, which ultimately reduces ACEs. A systematic review of the evidence further supports that upstream socioeconomic interventions eventually lower ACEs, while noting that effect strengths are different depending on the ACE (e.g., modest for adverse parenting, strongest for childhood victimization) (Courtin et al., Reference Courtin, Allchin, Ding and Layte2019). Social support for parents (Figure 9) also reduces ACEs, in addition to preventing suicide ideation. The protection that social support for parents confer against ACEs in their children has been discussed in several studies (Narayan et al., Reference Narayan, Lieberman and Masten2021).

Figure 8. Sample loops from the map including (a) the benefits of therapy, (b) parental frustration and coping via substance use, (c) positive changes in practices that decrease capacity for suicide, parental perpetration of Adverse Childhood Experiences (ACEs), (e) mental health disorders, and (e) poverty.

Figure 9. Rippling effects of interventions on economic policies (a) and social support for parents (partial view; b). Effects are shown in concentric circles from most proximal to more distal. Both sub-figures only depict a part of the map, as the user controls the maximum distance up to which effects should be identified.

5. Discussion

Research on the associations between ACEs and suicide has typically been limited due to methodological constraints. The development of a large-scale system map of ACEs and suicide (i.e., the NCIPC model) served as a catalyst for expanding research between these two factors. In this paper, we applied network analysis to examine this system map to compare it against other maps and to find useful patterns between ACEs and suicide (ideation, attempt, and fatality) for informing suicide prevention efforts. While network science has been used to examine systems maps in fields such as obesity, there has been limited research at the intersection of suicide and network science. This paper helps to address this gap. In comparing the NCIPC model to other maps, the NCIPC model stands out as more comprehensive in concepts and relationships. It contains concepts captured by the previous models, such as the topics of suicide treatment, mental disorders, and other key concepts from Page et al. (Reference Page, Atkinson, Campos, Heffernan, Ferdousi, Power and Hickie2018), and mental health disorders from Chung (Reference Chung2016). The reason this model is noticeably more comprehensive is its inclusion of community-level factors (e.g., poverty, violence), policy-level factors (e.g., healthcare policies, responsible suicide awareness education), and the family unit (e.g., ACEs of parents, parent substance abuse). In short, the NCIPC model broadens the view of suicide by shifting from a predominantly individual-centric view into a social-ecological framework.

Trauma is not just an individual trait or behavior. Trauma is perpetuated through familial, social, and historical context. In the NCIPC model linking ACEs and suicide risk, families and parents appear to be an important contributor to the intergenerational transmission of ACEs, as evidenced in Figures 8 and 9 where both parental mental health and financial instability partially explain the relationship between exposure to ACEs and suicidal thoughts and behaviors. Thus, the map underscores the importance of trauma-informed interventions across all levels of the social ecology, and the importance of taking an intergenerational approach to addressing social determinants of health such as poverty, housing instability, and social disadvantage. Policies that increase access to mental health services (particularly for parents) and economic supports for families are critical for interrupting the intergenerational transmission of ACEs and suicide risk.

Our network analysis of the NCIPC model helps us understand the complex interactions that result in adolescent suicidal behavior. Given its large size, the NCIPC model is not meant to be visualized all at once. Rather, it provides a decision-support tool that can be utilized to address specific questions and to help identify direct and indirect risk and protective factors related to adolescent suicide.

There are two main technical limitations to this study. First, despite the recent increase in models for suicide prevention (Schuerkamp et al., Reference Schuerkamp, Liang, Rice and Giabbanelli2023), there are relatively few comprehensive systems maps. The comparative portion of our analysis was thus limited to the number of models available. We note that new models have emerged since our study was undertaken (Vu et al., Reference Vu, Bishop, McGill, Valmadrid, Golden, Emmerling, Saeugling and Rubinelli2022), and there are also several new systems maps on sub-systems involved in suicide and ACEs such as homelessness (OBrien et al., Reference OBrien, Lemieux, McGrew, Mohamud and Ford2024) and substance abuse (Smith et al., Reference Smith, Alderman, Attell, Avila Rodriguez, DiGirolamo, Snyder, Covington, Manteuffel and Minyard2021). Second, there is a seemingly endless number of measures that could be used to characterize networks at the level of individual elements, groups, or for the whole network. For instance, other studies have reported the “ruggedness” (i.e., number of components) (Strautmane, Reference Strautmane2012), the number of paths (Ifenthaler, Reference Ifenthaler2010), or custom measures (e.g., percentage of nodes with in- and out-degree larger than one in Plate (Reference Plate2010)). It was recently shown that many of these measures are related, hence using more measures does not necessarily yield more insight (Wang and Giabbanelli, Reference Wang and Giabbanelli2023). We thus focused on measures that are commonly used in systems science, so that our findings can be interpreted within the context of public health.

The goal of a model is not to re-create the complexity of a real-world problem. Rather, a model is necessarily a simplification, which provides guidance to identify and evaluate solutions. Continuously developing larger models or integrating existing ones into a form of “super-model” is thus likely to produce an overly complex model, as discussed in (Voinov and Shugart, Reference Voinov and Shugart2013). Since we showed that there are already several high-quality models for suicide and ACEs, future work could focus on the transition from comprehensive high-level qualitative models onto perhaps smaller but more data-driven and quantitative models. Such models would provide better support for systems intervention, as they would be able to measure the strength of an effect or when it would occur, in contrast to current qualitative models that can only list which factors would be impacted. Such a transition has recently started, as the content of the NCIPC model was contrasted with several national datasets, showing that some parts of the model could be measured (e.g., those closest to the individual) whereas other parts suffered from a lack of data (e.g., societal-level constructs) (Giabbanelli et al., Reference Giabbanelli, Rice, Nataraj, Brown and Harper2023). Other studies have built on systems maps to offer comprehensive simulation models which can report the cost-effectiveness of interventions (Crosland et al., Reference Crosland, Ho, Hosseini, Vacher, Skinner, Natsky, Rosenberg, Hasudungan, Huntley, Song, Lee, Marshall, Occhipinti and Hickie2024). A recent comment observed that “planners are not opposed to the use of systems models when the outputs help answer questions that align with the fundamental preferences of the decision-making structures within which they work” (Rock et al., Reference Rock, Cross and Rock2024). It is thus particularly promising to leverage the information structured in systems maps and use the available data to support smaller models that answer practical questions from policymakers, clinicians, or caregivers.

6. Conclusion

Our comparative study suggests that SMEs and modelers were highly selective in identifying relationships when creating systems maps for suicide and adverse childhood experiences. When models got larger, they tended to expand on existing themes (e.g., increased level of details on trauma) rather than add an entirely new aspect. The maps are usually constructed around suicide ideation and attempt (which are frequently among the most central factors), but other highly central factors include potential intervention points such as mental health disorders. The interventions most represented in the NCIPC model include teaching coping skills and identifying and supporting people at risk, with a focus on the individual and relationship levels of the social-ecological framework. Decomposing the NCIPC model into communities showed the strong relations between ACEs and access to health care and racial barriers.

Funding statement.

This publication was supported by the Centers for Disease Control and Prevention (CDC), National Center for Injury Prevention and Control (NCIPC), Intergovernmental Personnel Act Assignment Agreement 20IPA2009427.

Data availability.

Our study is a secondary analysis of network data. The NCIPC model was released by its authors on a third-party repository by the Open Science Framework (OSF), which can be publicly accessed without registration at https://osf.io/7nxp4/

Disclaimer.

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Supplementary material.

To view supplementary material for this article, please visit https://doi.org/10.1017/nws.2024.8

Competing interests.

None.

Appendix A: Brief summary of the NCIPC model

The NCIPC suicide map totals 361 suicide-related risk and protective factors related to adolescent suicide, stratified across four levels of the social ecology: individual, relationship, community, and societal. Each level provides risk and protective factors that were identified by at least one subject matter expert interviewed for constructing the map. Several factors may appear at more than one level given the interconnectedness of some risk and protective factors across the social ecology. At the individual level of the social ecology, our map identified variables that can be grouped into nine major categories. Personal identity includes beliefs that one’s identity is a personal choice, faith, identity as a sexual minority, and devaluation of one’s identity. Coping skills and resiliency cover constructs such as positive handling of stress, vulnerability, being equipped with psychological and interpersonal skills, unhealthy coping strategies, and tools for self-advocacy. Mental and emotional health is a prominent category, covering self-esteem, distress, psychological pain, self-destructive thoughts, feelings of self-worth, sense of fulfillment, feelings of insecurity, and interpersonal problems. Cognitive aptitudes are related to mental disability, brain development problems, and educational attainment. Physical health deals with chronic pain, co-morbidities, and physical disability. Impulsivity is an important trait, which includes acting in reaction to feelings, engaging in disruptive behaviors or risky behaviors. Factors in the map related to personal history include a history of ACEs or child maltreatment, prior exposure to violence, substance abuse, positive childhood experiences, historical trauma and its embedding in the family, previous suicide attempt, and experiences of failure. Experiences vary based on demographic and socioeconomic factors, including race and ethnicity, exposure to poverty, or homelessness. Finally, individuals have a different capacity for suicide, based on their access to lethal means, clarity of suicide planning (including the ability to identify suicide methods), and whether they feel capable to carry on suicide.

Moving one level away from the individual, the social ecology considers variables about relationships. For Adverse Childhood Experiences (ACEs), key relationships include parents (e.g., death/loss/separation from a parent, parents engage in illegal activities and are arrested or incarcerated, parent’s substance use or perpetration of ACEs) and the broader family unit (e.g., domestic violence, mental illness in the family). The family environment is characterized by many aspects, as supportive families can be beneficial (e.g., positive interactions with parents, nurturing environment, care at home, housing stability, cohesion of family) while dysfunctional households are detrimental (e.g., parental escalation of conflict and harsh discipline, family history of secrecy, marital relationship stress, fighting in the house, parental lack of education on therapy). Relationships beyond the family units fall under the umbrella of peer relationships. Peers can positively influence a person’s life, for example by informing friends of suicidal intentions and thoughts, having supportive peers to talk to and receive social support or healthy mentoring. However, there are issues when peers are abusive (e.g., bullying), cope in unhealthy ways (e.g., substance use, risky behaviors), or do not provide adequate role models (e.g., peers in the justice system, suicides among friends).

The two highest levels of the social ecology are the community and society. At the community level of the social ecology our map identified variables that can be grouped into six major categories. First, there is access and availability of health services, particularly for mental health. Access and provision of healthcare services depend on the availability of related psychiatrists and the physical access. In a similar way, there is access and availability of social services, which provide education to get jobs as well as health education, quality childcare, and a crisis center. Characteristics of the neighborhood can be examined with respect to the environment (e.g., local jobs and promotion of mental health, or a segregated living with exposure to suicide) or safety (e.g., unsafe to go outside as neighborhood violence is normalized and regularly witnessed). The school environment plays a particular role for youth, as schools can be places that provide services and nurturing starting with good education in early childhood, or problematic environments that tolerate injustice and victimization such as labeling kids. The sixth category deals with community inequities, including discriminatory hiring practices, a history of segregation with the medical system, and population-level inequities. Finally, the societal-level encompasses cultural norms, policies, and structural inequities such as adversity and racism. Cultural norms can promote ACEs and impair treatment (e.g., stigmatizing ACEs or not seeing them as a problem, engaging in practices such as spanking, considering that one is responsible for their own health) or help individuals by promoting help-seeking behaviors and seeing suicide as unacceptable. Policies are far-ranging, as they include economic policies (e.g., health insurance), social policies (e.g., to increase connectedness), or reduce access to lethal means; these policies eventually prevent suicide attempt and completion.

Footnotes

1 Systems Maps are similar to, yet distinct from, the directed acyclic graphs (DAGs) often found in the field of epidemiology. A system map can exhibit feedback loops, while DAGs cannot. Consequently, the assessment of cycles is central to the analysis of a systems map, but it cannot be done in DAGs.

2 In a causal map, each edge is categorized as “+” or “–” if its increase results in either an increase or decrease of a target, respectively. The category of a cycle depends on the category of its constituting edges. The effects of the edges multiply, hence two “-” cancel out and result in an increasing dynamic. Consequently, a cycle is “reinforcing” if it has none or an even number of “-” edges. Conversely, a cycle is “balancing” if it has an odd number of “-” edges. Being reinforcing or balancing does not suffice to know whether a cycle is “good” for an intervention, as it still depends on the content of a cycle. A cycle that reinforces suicidal ideation would be a problem, but a cycle that reinforces mental health treatment would be a benefit.

3 The average path length is less than the natural log of the number of nodes: $ln(361) = 5.88 \gt 5.668$ in the NCIPC model, $ln(296) = 5.69 \gt 3.94$ in Brenas et al. (Reference Brenas, Shin, Shaban-Nejad and Shabo2019). This qualifies as “short”.

References

Alomia-Hinojosa, V., Groot, J. C., Andersson, J. A., Speelman, E. N., McDonald, A. J., & Tittonell, P. (2022). Assessing farmer perceptions on livestock intensification and associated trade-offs using fuzzy cognitive maps; a study in mixed farming systems in the mid-hills of nepal. Systems Research and Behavioral Science, 40(1), 146–158.Google Scholar
Atkinson, J. A., O’Donnell, E., Wiggers, J., McDonnell, G., Mitchell, J., Freebairn, L., Indig, D., Rychetnik, L. (2017). Dynamic simulation modelling of policy responses to reduce alcohol-related harms: rationale and procedure for a participatory approach. Public Health Research and Practice, 27(1), e2711707.CrossRefGoogle ScholarPubMed
Barbrook-Johnson, P., & Carrick, J. (2022). Combining complexity-framed research methods for social research. International Journal of Social Research Methodology, 25(6), 835848.CrossRefGoogle Scholar
Bavelas, A. (1950). Communication patterns in task-oriented groups. The Journal of the Acoustical Society of America, 22(6), 725730.CrossRefGoogle Scholar
Bekker, M., Helderman, J. K., Jansen, M., & Ruwaard, D. (2017). The conditions and contributions of ‘whole of society’ governance in the dutch ‘all about health.’ programme. In: Greer, S. L., Wismar, M., Pastorino, G., & Kosinska, M., eds. Civil Society and Health (pp. 159–180). Copenhagen, Denmark: European Observatory on Health Systems and Policies.Google Scholar
Brauer, K., Wulff, H., Pawellek, S., & Ziegeldorf, A. (2022). Network analysis for a community-based school-and family-based obesity prevention program. In Healthcare (pp. 1501). MDPI.Google Scholar
Brenas, J. H., Shin, E. K., & Shaban-Nejad, A. (2019). An ontological framework to improve surveillance of adverse childhood experiences (aces). In: Shabo, A., et al., eds. ICT for Health Science Research (pp. 3135). The European Federation for Medical Informatics (EFMI) and IOS Press.Google Scholar
Centers for Disease Control and Prevention's National Center for Injury Prevention and Control (2020). Web-based injury statistics query and reporting system (wisqars) fatal injury and violence data. https://www.cdc.gov/injury/wisqars/index.html. accessed March 31, 2022.Google Scholar
Centers for Disease Control and Prevention's National Center for Injury Prevention and Control (2021). The social-ecological model: A framework for prevention. https://www.cdc.gov/violenceprevention/about/social-ecologicalmodel.html.Google Scholar
Christen, B., Kjeldsen, C., Dalgaard, T., & Martin-Ortega, J. (2015). Can fuzzy cognitive mapping help in agricultural policy design and communication? Land Use Policy, 45, 6475.CrossRefGoogle Scholar
Chu, C., Buchman-Schmitt, J. M., Stanley, I. H., Hom, M. A., Tucker, R. P., Hagan, C. R., Rogers, M. L., Podlogar, M. C., Chiurliza, B., Ringer, F. B., Michaels, M. S., Patros, C. H. G., Joiner, T. E. (2017). The interpersonal theory of suicide: a systematic review and meta-analysis of a decade of cross-national research. Psychological Bulletin, 143(12), 13131345.CrossRefGoogle ScholarPubMed
Chunaev, P. (2020). Community detection in node-attributed social networks: a survey. Computer Science Review, 37, 100286.CrossRefGoogle Scholar
Chung, S. Y. (2016). Suicide attempts from adolescence into young adulthood: a system dynamics perspective for intervention and prevention.Google Scholar
Clauset, A., Newman, M. E., & Moore, C. (2004). Finding community structure in very large networks. Physical Review E, 70(6), 066111.CrossRefGoogle ScholarPubMed
Colombo, G. B., Burnap, P., Hodorog, A., & Scourfield, J. (2016). Analysing the connectivity and communication of suicidal users on twitter. Computer Communications, 73, 291300.CrossRefGoogle ScholarPubMed
Courtin, E., Allchin, E., Ding, A. J., & Layte, R. (2019). The role of socioeconomic interventions in reducing exposure to adverse childhood experiences: a systematic review. Current Epidemiology Reports, 6(4), 423441.CrossRefGoogle Scholar
Cramer, R. J., & Kapusta, N. D. (2017). A social-ecological framework of theory, assessment, and prevention of suicide. Frontiers in Psychology, 1756.Google Scholar
Crosland, P., Ho, N., Hosseini, S. H., Vacher, C., Skinner, A., Natsky, A. N., Rosenberg, S., Hasudungan, R., Huntley, S., Song, Y. J. C., Lee, G. Y., Marshall, D. A., Occhipinti, J.-A., Hickie, I. B. (2024). Cost-effectiveness of system-level mental health strategies for young people in the australian capital territory: a dynamic simulation modelling study. The Lancet Psychiatry, 11(2), 123133.CrossRefGoogle ScholarPubMed
Drasic, L., & Giabbanelli, P. J. (2015). Exploring the interactions between physical well-being, and obesity. Canadian Journal of Diabetes, 39, S12S13.CrossRefGoogle Scholar
Dubé, L., Addy, N. A., Blouin, C., & Drager, N. (2014). From policy coherence to 21st century convergence: a whole-of-society paradigm of human and economic development. Annals of the New York Academy of Sciences, 1331, 201215.CrossRefGoogle ScholarPubMed
Eden, C., Ackermann, F., & Cropper, S. (1992). The analysis of cause maps. Journal of Management Studies, 29(3), 309324.CrossRefGoogle Scholar
Falcone, P. M., & De Rosa, S. P. (2020). Use of fuzzy cognitive maps to develop policy strategies for the optimization of municipal waste management: a case study of the land of fires (italy). Land Use Policy, 96, 104680.CrossRefGoogle Scholar
Fan, X. (2020). Visualization of large networks using recursive community detection.Google Scholar
Felitti, V. J., Anda, R. F., Nordenberg, D., Williamson, D. F., Spitz, A. M., Edwards, V., Koss, M. P., Marks, J. S. (1998). Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: the adverse childhood experiences (ace) study. American Journal of Preventive Medicine, 14(4), 245258.CrossRefGoogle ScholarPubMed
Finegood, D. T., Merth, T. D., & Rutter, H. (2010). Implications of the foresight obesity system map for solutions to childhood obesity. Obesity, 18(S1), S13.CrossRefGoogle ScholarPubMed
Firmansyah, H. S., Supangkat, S. H., Arman, A. A., & Giabbanelli, P. J. (2019). Identifying the components and interrelationships of smart cities in indonesia: supporting policymaking via fuzzy cognitive systems. IEEE Access, 7, 4613646151.CrossRefGoogle Scholar
Freebairn, L., Atkinson, J., Kelly, P., McDonnell, G., & Rychetnik, L. (2016). Simulation modelling as a tool for knowledge mobilisation in health policy settings: a case study protocol. Health Research Policy and Systems, 14(1), 112.CrossRefGoogle ScholarPubMed
Freeman, L. C. (1977). A set of measures of centrality based on betweenness. Sociometry, 35-41(1), 35.CrossRefGoogle Scholar
Freund, A. J., & Giabbanelli, P. J. (2021). Are we modeling the evidence or our own biases? A comparison of conceptual models created from reports. 2021 Annual Modeling and Simulation Conference (ANNSIM), IEEE (pp. 112).CrossRefGoogle Scholar
Freund, A. J., & Giabbanelli, P. J. (2022). An experimental study on the scalability of recent node centrality metrics in sparse complex networks. Frontiers in Big Data, 5.CrossRefGoogle Scholar
Giabbanelli, P. J., & Baniukiewicz, M. (2018). Navigating complex systems for policymaking using simple software tools. In Advanced data analytics in health (pp. 2140). Springer.CrossRefGoogle Scholar
Giabbanelli, P. J., Galgoczy, M. C., Nguyen, D. M., Foy, R., Rice, K. L., Nataraj, N.Harper, C. R. (2021). Mapping the complexity of suicide by combining participatory modeling and network science. Proceedings of the 2021 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (pp. 339342).CrossRefGoogle Scholar
Giabbanelli, P. J., & MacEwan, G. (2024). Leveraging artificial intelligence and participatory modeling to support paradigm shifts in public health: an application to obesity and evidence-based policymaking. Information, 15(2), 115.CrossRefGoogle Scholar
Giabbanelli, P. J., Rice, K. L., Galgoczy, M. C., Nataraj, N., Brown, M. M., Harper, C. R., Nguyen, M. D., & Foy, R. (2022). Pathways to suicide or collections of vicious cycles? understanding the complexity of suicide through causal mapping. Social Network Analysis and Mining, 12(1), 60.CrossRefGoogle ScholarPubMed
Giabbanelli, P. J., Rice, K. L., Nataraj, N., Brown, M. M., & Harper, C. R. (2023). A systems science approach to identifying data gaps in national data sources on adolescent suicidal ideation and suicide attempt in the united states. BMC Public Health, 23(1), 19.CrossRefGoogle ScholarPubMed
Giles, B. G., Findlay, C. S., Haas, G., LaFrance, B., Laughing, W., & Pembleton, S. (2007). Integrating conventional science and aboriginal perspectives on diabetes using fuzzy cognitive maps. Social Science & Medicine, 64(3), 562576.CrossRefGoogle ScholarPubMed
Goh, K. I., Kahng, B., & Kim, D. (2001). Universal behavior of load distribution in scale-free networks. Physical Review Letters, 87(27), 278701.CrossRefGoogle ScholarPubMed
Harmer, B., Lee, S., TvH, D., & Saadabadi, A. (2020). Suicidal ideation. http://www.ncbi.nlm.nih.gov/books/NBK565877/.Google Scholar
Hayward, J., Morton, S., Johnstone, M., Creighton, D., & Allender, S. (2020). Tools and analytic techniques to synthesise community knowledge in cbpr using computer-mediated participatory system modelling. NPJ Digital Medicine, 3(1), 16.CrossRefGoogle ScholarPubMed
Ifenthaler, D. (2010). Bridging the gap between expert-novice differences: the model-based feedback approach. Journal of Research on Technology in Education, 43(2), 103117.CrossRefGoogle Scholar
Ivey-Stephenson, A. Z., Demissie, Z., Crosby, A. E., Stone, D. M., Gaylor, E., Wilkins, N., Lowry, R., & Brown, M. (2020). Suicidal ideation and behaviors among high school students-youth risk behavior survey, United States, 2019. MMWR Supplements, 69(1), 4755.CrossRefGoogle Scholar
Katz, L. (1953). A new status index derived from sociometric analysis. Psychometrika, 18(1), 3943.CrossRefGoogle Scholar
Kelley, K., & DeShong, H. L. (2023). Examining facet-level distinctions within the trait-interpersonal model of suicide risk. Personality and Individual Differences, 202, 111973.CrossRefGoogle Scholar
Kim, M. K., & McCarthy, K. S. (2021). Using graph centrality as a global index to assess students’ mental model structure development during summary writing. Educational Technology Research and Development, 69(2), 9711002.CrossRefGoogle Scholar
Kim, S., Jang, K. I., Lee, H. S., Shim, S. H., & Kim, J. S. (2024). Differentiation between suicide attempt and suicidal ideation in patients with major depressive disorder using cortical functional network. Progress in Neuro-Psychopharmacology and Biological Psychiatry, 132, 110965.CrossRefGoogle ScholarPubMed
Knox, C. B., Furman, K., Jetter, A., Gray, S., & Giabbanelli, P. J. (2024). Creating an FCM with Participants in an Interview or Workshop Setting. Cham: Springer Nature Switzerland, pp. 1944.Google Scholar
Konti, A., & Damigos, D. (2018). Exploring strengths and weaknesses of bioethanol production from bio-waste in Greece using fuzzy cognitive maps. Energy Policy, 112, 411.CrossRefGoogle Scholar
Krabbe, H. (2014). Digital concept mapping for formative assessment, Digital knowledge maps in education: Technology-enhanced support for teachers and learners, pp. 275297.Google Scholar
Kranzler, A., Fehling, K. B., Anestis, M. D., & Selby, E. A. (2016). Emotional dysregulation, internalizing symptoms, and self-injurious and suicidal behavior: structural equation modeling analysis. Death Studies, 40(6), 358366.CrossRefGoogle ScholarPubMed
Lancichinetti, A., & Fortunato, S. (2009). Community detection algorithms: a comparative analysis. Physical Review E, 80(5), 056117.CrossRefGoogle ScholarPubMed
Lavin, E. A., Giabbanelli, P. J., Stefanik, A. T., Gray, S. A., & Arlinghaus, R. (2018). Should we simulate mental models to assess whether they agree? Proceedings of the Annual Simulation Symposium (pp. 112).Google Scholar
Lee, J., & Lee, J. (2013). Hidden information revealed by optimal community structure from a protein-complex bipartite network improves protein function prediction. PloS One, 8(4), e60372.CrossRefGoogle ScholarPubMed
Liche, C. A., Morar, L., & Campean, E. (2014). A fuzzy cognitive maps approach for analyzing the integration of two organizations of the same corporation. 2014 IEEE 16th Conference on Business Informatics, IEEE (pp. 149157).CrossRefGoogle Scholar
Luna Pinzon, A., Stronks, K., Dijkstra, C., Renders, C., Altenburg, T., den Hertog, K.Waterlander, W. (2022). The encompass framework: a practical guide for the evaluation of public health programmes in complex adaptive systems. International Journal of Behavioral Nutrition and Physical Activity, 19(1), 117.CrossRefGoogle ScholarPubMed
Maksimenkova, O., Neznanov, A., Papushina, I., & Parinov, A. (2018). On mind maps evaluation: a case of an automatic grader development. Teaching and Learning in a Digital World: Proceedings of the 20th International Conference on Interactive Collaborative Learning–Volume 2, Springer (pp. 210221).Google Scholar
Masuda, N., Kurahashi, I., & Onari, H. (2013). Suicide ideation of individuals in online social networks. PloS One, 8(4), e62262.CrossRefGoogle ScholarPubMed
McGlashan, J., de la Haye, K., Wang, P., & Allender, S. (2019). Collaboration in complex systems: multilevel network analysis for community-based obesity prevention interventions. Scientific Reports, 9(1), 110.CrossRefGoogle ScholarPubMed
McGlashan, J., Hayward, J., Brown, A., Owen, B., Millar, L., Johnstone, M., Creighton, D., & Allender, S. (2018). Comparing complex perspectives on obesity drivers: action-driven communities and evidence-oriented experts. Obesity Science & Practice, 4(6), 575581.CrossRefGoogle ScholarPubMed
McGlashan, J., Johnstone, M., Creighton, D., de la Haye, K., & Allender, S. (2016). Quantifying a systems map: network analysis of a childhood obesity causal loop diagram. PloS One, 11(10), e0165459.CrossRefGoogle ScholarPubMed
McGlashan, J. A. (2018). The use of network analysis to evaluate community-based obesity prevention interventions.CrossRefGoogle Scholar
McPherson, K., Marsh, T., & Brown, M. (2007). Foresight report on obesity. The Lancet, 370(9601), 1755.CrossRefGoogle ScholarPubMed
Mkhitaryan, S., Giabbanelli, P. J., de Vries, N. K., & Crutzen, R. (2020). Dealing with complexity: how to use a hybrid approach to incorporate complexity in health behavior interventions. Intelligence-Based Medicine, 3, 100008.CrossRefGoogle Scholar
Narayan, A. J., Lieberman, A. F., & Masten, A. S. (2021). Intergenerational transmission and prevention of adverse childhood experiences (aces). Clinical Psychology Review, 85, 101997.CrossRefGoogle ScholarPubMed
Newman, M. E. (2018). Networks, 1st Ed. Oxford, UK: Oxford University Press.CrossRefGoogle Scholar
Newman, M. E., & Girvan, M. (2004). Finding and evaluating community structure in networks. Physical Review E, 69(2), 026113.CrossRefGoogle ScholarPubMed
OBrien, J. R. G., Lemieux, A., McGrew, E., Mohamud, Z., & Ford, M. (2024). Youth-driven systems alignment strategies to optimize health among youth experiencing homelessness. Journal of Adolescent Health, 74(3), S15S16.CrossRefGoogle Scholar
Page, A., Atkinson, J. A., Campos, W., Heffernan, M., Ferdousi, S., Power, A.Hickie, I. (2018). A decision support tool to inform local suicide prevention activity in greater Western Sydney (Australia). Australian & New Zealand Journal of Psychiatry, 52(10), 983993.CrossRefGoogle ScholarPubMed
Plate, R. (2010). Assessing individuals’ understanding of nonlinear causal structures in complex systems. System Dynamics Review, 26(1), 1933.CrossRefGoogle Scholar
Power, D. D., Lambe, B. M., & Murphy, N. M. (2022). Using systems science methods to enhance the work of national and local walking partnerships: practical insights from ireland. European Journal of Public Health, 32(Supplement_1), i8i13.CrossRefGoogle ScholarPubMed
Qin, K., Li, H., Zhang, H., Yin, L., Wu, B., Pan, N.et al. (2024). Transcriptional patterns of brain structural covariance network abnormalities associated with suicidal thoughts and behaviors in major depressive disorder. Biological Psychiatry. In press.CrossRefGoogle ScholarPubMed
Rock, D., Cross, S., & Rock, L. (2024). Incorporating systems modelling into mental health system planning. The Lancet Psychiatry, 11(2), 8586.CrossRefGoogle ScholarPubMed
Sabidussi, G. (1966). The centrality index of a graph. Psychometrika, 31(4), 581603.CrossRefGoogle Scholar
Schuerkamp, R., Liang, L., Rice, K. L., & Giabbanelli, P. J. (2023). Simulation models for suicide prevention: a survey of the state-of-the-art. Computers, 12(7), 132.CrossRefGoogle ScholarPubMed
Shagle, S. C., & Barber, B. K. (1995). A social-ecological analysis of adolescent suicidal ideation. American Journal of Orthopsychiatry, 65(1), 114124.CrossRefGoogle ScholarPubMed
Shahvi, S., Mellander, P. E., Jordan, P., & Fenton, O. (2021). A fuzzy cognitive map method for integrated and participatory water governance and indicators affecting drinking water supplies. Science of the Total Environment, 750, 142193.CrossRefGoogle ScholarPubMed
Siokou, C., Morgan, R., & Shiell, A. (2014). Group model building: a participatory approach to understanding and acting on systems. Public Health Research & Practice, 25, e2511404.CrossRefGoogle ScholarPubMed
Smith, J. C., Alderman, L., Attell, B. K., Avila Rodriguez, W., DiGirolamo, A.M., Snyder, S.M., Covington, J., Manteuffel, B., & Minyard, K. (2021). Dynamics of parental opioid use and children’s health and well-being: Aan integrative systems mapping approach. Frontiers in Psychology, 2402.Google Scholar
Smith, N., Georgiou, M., King, A. C., Tieges, Z., & Chastin, S. (2022). Factors influencing usage of urban blue spaces: a systems-based approach to identify leverage points. Health & Place, 73, 102735.CrossRefGoogle ScholarPubMed
Stone, D. M., Holland, K. M., Bartholow, B. N., Crosby, A. E., Davis, S. P., & Wilkins, N. (2017). Preventing suicide: a technical package of policies, programs, and practice. Technical Report.Google Scholar
Strautmane, M. (2012). Concept map-based knowledge assessment tasks and their scoring criteria: An overview. Concept maps: Theory, methodology, technology. Proceedings of the fifth international conference on concept mapping (pp. 8088).Google Scholar
Tang, T.-C., Yen, C.-F., Cheng, C.-P., Yang, P., Chen, C.-S., Yang, R.-C., Huang, M.-S., Jong, Y.-J., Yu, H.-S. (2010). Suicide risk and its correlate in adolescents who experienced typhoon-induced mudslides: a structural equation model. Depression and Anxiety, 27(12), 11431148.CrossRefGoogle ScholarPubMed
Traag, V. A., Waltman, L., & Van Eck, N. J. (2019). From louvain to leiden: guaranteeing well-connected communities. Scientific Reports, 9(1), 112.CrossRefGoogle ScholarPubMed
Vélez-Grau, C., Magan, I. M., & Gwadz, M. (2023). The burden of not belonging: A qualitative study of the applicability of the interpersonal theory of suicide constructs of belongingness and burdensomeness to ethnocultural minoritized youth. Behavior Therapy, 54(5), 777–793.CrossRefGoogle Scholar
Voinov, A., & Shugart, H. H. (2013). Integronsters’, integral and integrated modeling. Environmental Modelling & Software, 39, 149158.CrossRefGoogle Scholar
Vu, T. H., Bishop, J., McGill, L., Valmadrid, L., Golden, S., Emmerling, D., Saeugling, S., Rubinelli, S. (2022). Using systems-mapping to address adverse childhood experiences (aces) and trauma: a qualitative study of stakeholder experiences. Plos One, 17(8), e0273361.CrossRefGoogle ScholarPubMed
Wang, B., & Giabbanelli, P. J. (2023). Identifying informative features to evaluate student knowledge as causal maps. International Journal of Artificial Intelligence in Education, 131. In press.Google Scholar
Zhang, L., Mersky, J. P., Gruber, A. M. H., & Kim, J. Y. (2023). Intergenerational transmission of parental adverse childhood experiences and children’s outcomes: a scoping review. Trauma, Violence, & Abuse, 24(5), 32513264.CrossRefGoogle ScholarPubMed
Figure 0

Figure 1. General process for creating a systems map.

Figure 1

Table 1. Node centrality measures and their interpretation in the context of systems maps on suicide

Figure 2

Figure 2. Fragment of the National Center for Injury Prevention and Control (NCIPC) model with three centralities: node size represents betweenness centrality (i.e., a larger node has control over a larger share of shortest paths), colors show degree centrality (nodes with more ties are darker), and the number indicates closeness centrality (which is based on distances). Betweenness centrality tracks how many times a concept would be traversed, hence we see that “child getting help” and “visit to the ER” are commonly occurring steps (i.e., frequently traversed) through the journey of an individual experiencing suicide ideation or attempts. Degree centrality reveals the extent to which a construct directly relates to another construct (e.g., therapy success is driven by parental involvement and also contributes to handling different emotions).

Figure 3

Table 2. Number of nodes, edges, density, and number of participants in participatory modeling studies that produced maps. Note that most models have a low density (i.e., sparse graphs)

Figure 4

Figure 3. The Louvain method detected five communities in this fragment of the National Center for Injury Prevention and Control (NCIPC) model causal map, shown with distinct colors. Labels are only indicative, since the algorithm assigns concepts to communities but does not automatically name these communities.

Figure 5

Table 3. Network measurements for all four network-based suicide models

Figure 6

Table 4. Some nodes have the same value for a centrality measure, indicated by multiple nodes with the same rank. Note that Brenas et al. (2019) have duplicated nodes which the above rankings reflect; Page et al. (2018) also have a special “Black Hole” node, which refers to an exit point where the relationships leave the scope of concepts in the model. Mental health is abbreviated as `mhs' by the authors.

Figure 7

Figure 4. In-degree and out-degree distributions for four models. We note a skewed distribution indicating a high prevalence of low-degree nodes in the first three and a more uniform distribution in the last (bottom) model. The NCIPC model refers to the causal map created by the National Center for Injury Prevention and Control (NCIPC). Note that scales are different, since some models have fewer edges than others.

Figure 8

Figure 5. Percent of nodes assigned to each combination of prevention strategy and socio-ecological level. For example, 18.84% of nodes belong at the individual level, and the support prevention strategy could help address it according to the expert plot. Nodes can belong to multiple socio-ecological levels, and several prevention strategies may address them hence the total exceeds 100%.

Figure 9

Table 5. Both community detection algorithms produced similar communities; entries are ordered such that these similar communities are placed next to each other for ease of visibility. The community numbers are also used in Figure 5

Figure 10

Figure 6. National Center for Injury Prevention and Control (NCIPC) model reduced based on the communities (comm.) detected by the Louvain (left) and Leiden algorithms (right). The adjacency matrices (top) provide a more compact view of the structures shown in node-and-link diagrams (bottom). Community numbers refer to Table 5.

Figure 11

Figure 7. The ActionableSystems software allows users to automatically identify all pathways from one concept to another, as exemplified here by going from Adverse Childhood Experiences (ACEs) of parents (left) to suicide fatality (right).

Figure 12

Figure 8. Sample loops from the map including (a) the benefits of therapy, (b) parental frustration and coping via substance use, (c) positive changes in practices that decrease capacity for suicide, parental perpetration of Adverse Childhood Experiences (ACEs), (e) mental health disorders, and (e) poverty.

Figure 13

Figure 9. Rippling effects of interventions on economic policies (a) and social support for parents (partial view; b). Effects are shown in concentric circles from most proximal to more distal. Both sub-figures only depict a part of the map, as the user controls the maximum distance up to which effects should be identified.

Supplementary material: File

Maldonado et al. supplementary material 1

Maldonado et al. supplementary material
Download Maldonado et al. supplementary material 1(File)
File 55.6 KB
Supplementary material: File

Maldonado et al. supplementary material 2

Maldonado et al. supplementary material
Download Maldonado et al. supplementary material 2(File)
File 51.4 KB