Simulation of a ‘suicidal mind’: Using the Integrated Motivational Volitional model of suicide to demonstrate dynamic suicidal states

Simulation of a ‘suicidal mind’: Using the Integrated Motivational Volitional model of suicide to demonstrate dynamic suicidal states | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Simulation of a ‘suicidal mind’: Using the Integrated Motivational Volitional model of suicide to demonstrate dynamic suicidal states Gabriel McDonnell Maayan, Andrew Page, Elizabeth Conroy, Rory O'Connor, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6605926/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Suicidal behaviors are characterised by complex, multi-factorial aetiology. Dynamic simulation models (DSMs) are computational approaches that can explicitly capture salient dimensions of complex suicidal thinking and behavior. This study describes the development of a DSM of a ‘suicidal mind’ to capture nonlinear, dynamic changes in suicidal states based on an integration of the (i) Integrated Motivational-Volitional (IMV) model of suicide, (ii) Fluid Vulnerability Theory of suicide, and (iii) Cusp-Catastrophe model from dynamical systems theory. A system-dynamics model was developed to estimate the level of ‘suicidality’ in an individual over time, capturing cognitions and behaviors with transitions between the ‘pre-motivational’, ‘motivational,’ and ‘volitional’ phases in the IMV model. Validation of the DSM and the underlying theoretical synthesis consisted of testing whether parameter changes - hypothesised to increase or decrease the level of suicidality - resulted in corresponding effects in the model, and whether the DSM displayed the expected nonlinear pathways to suicidal states. The model’s behavior in response to varying parameters of interest was consistent with expectations. The model could recreate the ‘stable,’ ‘dysregulated,’ and ‘discontinuous’ nonlinear pathways proposed in prior research supporting the validity of the DSM and its underlying theoretical synthesis. Mental health service access resulted in stabilization of suicidal ideation, but the effect varied by frequency of contact. This model demonstrates that DSMs can quantify and refine theories of suicidal behavior, which suggests the potential for using DSMs in virtual case studies to assist clinical decision making and training, or to investigate population-level interventions. Biological sciences/Psychology/Human behaviour Health sciences/Health care/Public health/Epidemiology Health sciences/Health care/Health policy Health sciences/Risk factors Health sciences/Diseases/Psychiatric disorders Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Significance Statement This is the first study to develop a computational simulation of an individual ‘suicidal mind’ based on a conceptual synthesis of three current psychological theories. The simulation model is able to quantify the proposed transitions from ‘pre-motivational’, ‘motivational’ and ‘volitional’ states in the Integrated Motivational-Volitional model, and also capture the non-linear and dynamic nature of suicidal ideation and behavior as proposed by prior research. This model of a suicidal mind demonstrates the coherence of a synthesis between several dissimilar theories of suicidality. It can also lead to further refinement of current theoretical understandings of suicidal behavior, the use of virtual case studies to guide clinical decision-making, training, and the evaluation of population-level interventions on individuals with specific characteristics. Introduction Suicide remains an important global public health priority, with over 720,000 deaths by suicide globally per year, two-thirds of deaths occurring in low- and middle-income countries. [1] Suicide is the third leading causes of death among young people (15-29 years), and intentional self-harm is of a similar magnitude to global rates in traffic injury. [2] Suicidal behaviors are characterised by a complex, multi-factorial aetiology, with interacting social, cultural, biological, psychological, and environmental determinants that can affect individuals across the life-course. [1] Dynamic simulation models (DSMs) are computational representations of the real world that aim to capture the dynamics and behavior of a whole system or population, [3] and are methods that can explicitly capture the complex behavior of suicide. Computational simulation has a long history in disciplines such as ecology, physics, and engineering. In population health, DSMs have mainly been used for infectious diseases - and became even more prominent in guiding responses to the recent COVID-19 pandemic. DSM’s are increasingly recognized for their broader value in helping to solve complex problems for other population health outcomes [4] Previous uses of computational simulation in suicide prevention have predominantly addressed policy related questions for whole populations, including identifying combinations of interventions likely to avert future suicidal behavior and to inform mental health service capacity requirements. [5-14] DSMs of suicidal behavior have been developed mainly in the US, Canadian, and Australian contexts, [5] and have been developed for national [6-7] and local populations [8-14] to investigate the potential impact of interventions on subsequent trends in suicide and attempted suicide. These approaches provided insights that have informed service planning, [11, 13-14] the commissioning of local mental health services, [13-14] potential impacts of firearms restrictions, [9-10] and identifying priority areas where the greatest population impact in reducing suicidal behavior might be achieved. [5-14] These previous approaches have largely considered aggregate patterns of suicidal behavior in populations. However, there is also scope to use DSMs to describe and understand individual behaviors, to operationalise and potentially validate specific psychological theories that attempt to incorporate the range of complex determinants of suicidal behavior and articulate why some individuals might engage in suicidal behavior, whereas other do not. Previous psychological theories of an individual’s suicidal behavior include the Interpersonal Theory of Suicide, [15] the Three-Step Theory (3ST) of suicide, [16] and the Fluid Vulnerability Theory of suicide. [17] These have culminated in the most recent iteration of the Integrated Motivational-Volitional model (IMV) [18] that integrates previous theories with current literature, and describes the emergence of suicidal ideation and behavior. These theories aim to be consistent with observational studies of risk factors for suicidal behavior, but they have not been without criticism. This includes the simplification of complex multi-factorial behavior to a few key constructs, the selected use of suicide prevention literature, and the challenges associated with empirical validation. [19] Psychological theories can clarify the drivers and emergence of suicidal behavior, but they are qualitative, conceptual and static, and with potentially limited utility in quantifying specific individual-level behavior. While suicidal acts are uniquely individual, being able to empirically quantify current psychological theories and demonstrate that patterns of individual behavior are logically consistent with these theories and research evidence, can help to generate hypotheses regarding clinical responses to those presenting in suicidal crisis (or with a past history of self-harm), and broader prevention approaches among those who have never previously shown a propensity for suicidal behavior. Additionally, dynamic modelling of psychological theories can help refine the conceptualisation of suicidal behavior and provide insights into how theories of suicidal behavior might be revised or modified to better capture individual behavior. Accordingly, this study describes the development of a DSM of a suicidal mind that synthesizes three theories as the basis for capturing the dynamic changes in different suicidal states within a simulated individual. Methods Theories to inform the model The IMV model is the first of the three synthesized theories. It integrates several current psychological theories, and its development was guided by the need to incorporate a synthesis of current literature into a single tenable theoretical framework. [18] This resulted in a ‘tri-partite’ model that describes the biopsychosocial context in which suicidal ideation and behavior may emerge (a ‘pre-motivational phase’), factors that determine the emergence of suicidal ideation (a ‘motivational phase’), and factors that determine the transition from ideation to behavior (a ‘volitional phase’) (Figure 1A). Transitions between the ‘motivational’ and ‘volitional’ phases (and the emergence of ideation and behavior, respectively) are determined by ‘threat-to-self moderators’ (such as social problem solving, coping, memory biases, ruminative processes), ‘motivational moderators’ (such as thwarted belongingness, burdensomeness, resilience, social support), and ‘volitional moderators’ (such as access to means, planning, exposure to suicide, impulsivity, physical pain sensitivity, fearlessness about death, past behavior). [18] However, the IMV model as proposed is monodirectional in its characterisation of how suicidal ideation, intent, and behavior develop (although dynamic bidirectional relationships between ideation and behavior are included in a more recent update to the IMV model) [18]. Our study captures the dynamic nature of suicidal behavior, the waxing and waning of ideation, of psychological distress, and environmental factors that can affect these outcomes, the temporal dynamics of these factors, and the behavioural (and psychological) feedback loops that can be hypothesised between them by integrating two other theories into the IMV model. We draw on the Fluid Vulnerability Model of suicidal behavior [17] which conceptualises suicidal behavior as inherently dynamic and non-linear, and which has been extended by Bryan et al., by incorporating the Cusp-Catastrophe model from dynamical systems theory (Figure 1B; hereafter called the FV+CC model). [20] The Cusp-Catastrophe model is useful in that it can capture nonlinear change processes, emergent behavior, and bifurcating patterns of behavior in systems. It can also provide a foundation for explaining phenomena such as the sudden emergence of suicidal behavior without prior suicidal planning. [20] The ‘suicidal mind’ model structure The ‘suicidal mind’ is a system-dynamics model of ‘suicidality’ in an individual (Figure 2A). Systems-dynamics modelling was first described by Jay Forrester [21] and is a type of DSM composed of stocks (quantities that accumulate) and flows (rates of change). The units of this model are largely arbitrary and can be conceptualised as levels of the relevant cognitions or behaviors. In the simulation model, the individual has stocks representing a quantity of states that correspond to each of the components of the IMV model (e.g. one stock representing an amount of defeat/humiliation, one representing an amount of entrapment, etc.). Firstly, the assumption is that ‘suicidal ideation and intention’ begins via ‘negative cognitions’ that are triggered by ‘life events’, ‘diathesis,’ and ‘environmental’ factors. Life events are perturbations in the individual’s life, such as losing a job or relationship breakdown. There are two types of life-events, categorised as ‘normal’ events or ‘severe’ events. Drawing from the fluid vulnerability theory, normal events and severe events are those to which the individual is less or more sensitive, respectively. Additionally, the individual has different dispositional tolerance (which can be modified in the model) that acknowledges a response to a given life event may be ‘more’ or ‘less’ stressful. ‘Diathesis’ relates to factors such as chronic illness, genetic vulnerabilities, or other medical conditions. ‘Environment’ includes factors such as long-term financial distress, material circumstances including social disadvantage, isolation, or other sustained external stressors. ‘Diathesis’ and ‘environment’ factors have their own effects on ‘negative cognitions,’ but can also compound the effects from life events. ‘Negative cognitions’ lead to feelings of ‘defeat/humiliation’ (i.e. ‘negative cognitions’ increase the ‘defeat/humiliation’ stock). In turn, ‘defeat/humiliation’ can lead to feelings of ‘entrapment’ through the presence or absence of ‘threat-to-self moderators’, such as social problem solving, coping skills, or memory biases. ‘Entrapment’ leads to increased levels of ‘suicidal ideation’, based on the influence of ‘motivational moderators’ including feelings of thwarted belongingness, burdensomeness, or social support. From ‘suicidal ideation’, the individual can begin to exhibit active ‘suicidal behavior’, based on the influence of ‘volitional moderators’ including access to means, suicidal planning, and fearlessness about death. The ‘suicidal mind’ model includes feedback loops to capture the dynamic and cyclical nature of moving through each of these states. For example, levels of ‘defeat/humiliation’ affect ‘negative conditions’ in a reinforcing loop, as do levels of ‘entrapment’, ‘suicidal ideation and intent’, and ‘suicidal behavior’. These feedback loops acknowledge the importance of past levels of suicidality in potential responses to the triggering of life events. From the fluid vulnerability theory, we also include a ‘suicidal mode’ in the form of ‘low’ and ‘high’ suicidal risk states. The individual may enter a high-risk state based on the level of suicidal ideation, referred to as ‘suicidal desire’, and on ‘capability’. The high-risk state loosely corresponds to the upper plateau of the FV+CC [20] model’s synthesis of the Fluid Vulnerability Theory and the Cusp Catastrophe model (Figure 1B). Accordingly, it can be more difficult for the individual to leave a high-risk state than it was to enter. While in a high-risk state, the individual’s capability increases, perhaps under the impact of past suicidal behavior; they might make suicidal plans, which increases capability further, and become more likely to make a suicide attempt. To make a suicide attempt, the individual must have very high levels of overall suicidal risk, defined by desire, capability and ‘risk-state’. At these levels, the individual may attempt suicide with probability proportional to their desire and capability (see Supplementary Materials for a detailed definition). If the individual attempts suicide, they may also die by suicide. Following an non-fatal suicide attempt, individuals may have a wide range of responses including disappointment, anger at a thwarted attempt, emotional relief or further feelings of entrapment. [22] One study found 90% of attempters felt relief following a non-fatal attempt, but around 50% intended to make a repeat attempt. [23] Following a non-fatal attempt in the ‘suicidal mind’ model, the individual’s ‘entrapment,’ ‘suicidal ideation,’ and ‘suicidal behavior’ stocks return to zero. This represents some amount of emotional release following a suicide attempt. The ‘defeat/humiliation’ stock is not reset because the attempt itself does nothing to resolve underlying and external drivers to suicidality unless there are other interventions, such as from the health services sector. Repeat attempts are likely without interventions. The ‘suicidal mind’ model simulates an individual’s internal cognitive-motivational mechanisms, so it also includes interactions with health services to investigate the effect of service use on modifying trajectories of suicidal ideation and behavior (Figure 2B). Health services are implemented as a simplified and idealized stepped-care model [24] with arrows indicating referral pathways between different health service states (Figure 2B). ‘Universal care’ refers to population-level interventions, for example, a government mental health promotion campaign, and these services reduce incoming negative cognitions. ‘Primary care’ services relate to presentation to a general practitioner setting for mental health assessment, but the individual must choose to interact with these services. Without mental health assessments, the individual is not referred to other services until they make a suicide attempt. ‘Specialist community services’ more directly target suicidal ideation and behavior, and may include regular sessions with a qualified mental health practitioner. These services provide more frequent risk assessments and improve the individual’s cognitions and coping skills over time. ‘Acute care’ represents short-term high-intensity services, such as the hospital emergency room. In the simplified health service system, the individual attends acute care services directly following a suicide attempt, which provides both stabilization and triage for the individual’s next treatment step. In a real health service system, acute care presentations are commonly for ideation or planning and not just for attempts. Finally, ‘psychiatric hospital’ services offer medium-term high-intensity services that temporarily suppress triggering life events and provide psychiatric and pharmaceutical interventions that directly target symptoms and underlying conditions. An individual enters the ‘psychiatric hospital’ state if they are at high suicidal risk and have been so for a sufficient length of time. An individual starts in the ‘universal care’ state but immediately moves to the ‘primary care’ state if they choose to use the primary care system. Otherwise, they move to the ‘acute care’ state following a suicide attempt, or into ‘specialist community’ and ‘psychiatric hospital’ services if they have had previous risk assessments and are assessed to need further care. From the other health service states, the individual moves to ‘acute care’ following a suicide attempt. ‘Assessed risk’ is a moving average of several previous mental health assessments. Various aspects of the health service system, such as assessment frequency and efficacy, are included as model parameters and may be adjusted to investigate alternative scenarios (e.g. service availability or efficacy). Model parameterisation A detailed list of parameters is in the ‘Overview, Design concepts and Details (ODD)’ protocol [25] (Supplementary Materials). In general, parameters fall into three categories. Parameters in the first set determine the strength and frequency of triggering factors, such as environmental or diathesis-related risks and life events. Parameters in the second set determine how the individual responds to the triggering factors, such as coping abilities or the moderators from the IMV model. The final set of parameters define external intervention characteristics, such as the frequency of primary care evaluations or the effectiveness of specialist/psychiatric hospital care. By modifying the three sets of parameters, we can describe an individual’s internal and external state, thereby investigating virtual case studies and “what-if” scenarios. The units of each time step in the model are days. The units of parameters relating to time-based events (e.g. frequency of primary care visits) are therefore days, while parameters relating to the size of effects (e.g. coping ability) have more abstract units. Scenario testing The main analytic questions for the ‘suicidal mind’ model are (i) whether changes in parameters that the underlying theory assumes increase or decrease the level of suicidality result in corresponding increases or decreases in suicidality as estimated by the model, and (ii) whether the model displays suicidal states as suggested by the FV+CC model. [20] For example, for (i), does increasing exposure to protective factors that might prevent an individual moving from feelings of defeat to entrapment lead to decreases in suicidality? Or do improvements in coping strategies, or access to health services, lead to decreases in suicidality? Conversely, does an increase in exposure to risk factors lead to increases in suicidality? To answer this first question, we ran 4.2 million simulations, varying key parameters, to build a full description of the model’s behavior. For each varied parameter, we summarize the behavior with the following outputs: presence of at least one suicide attempt, health service access, date of first suicide attempt, suicide capability at the end of the simulation, average attempt-recurrence rate, and the suicidal-desire inter-quartile distance to measure the amplitude of suicidal desire fluctuations. For (ii), the question is can the ‘suicidal mind’ model can replicate the types of temporal processes for suicide risk as proposed in the FV+CC model [20] (Figure 1B): slow, smooth, and mild fluctuations around a homeostatic equilibrium point (‘stable’, scenario A); rapid, large fluctuations around a homeostatic equilibrium point (‘dysregulated’, scenario B); and a sudden, dramatic departure from a homeostatic equilibrium point (‘discontinuous’, scenario C). Scenario A reflects the slow and smooth fluctuations in suicidal ideation among lower risk suicidal individuals. [26-29] Scenario B reflects the large fluctuations in suicidal ideation associated with a dysregulated change process among individuals who engage in repeated intentional self-harm and associated with a gradual increase in suicidal ‘capability’. [26, 28-31] Scenario C reflects the sudden, dramatic shift in suicide risk states associated with a discontinuous change process, where an individual begins with high suicidal ‘capability’ and with low suicidal ideation, reflecting an ‘impulsive’ response to a stressor or life-event. [20] We also investigated the impact of access to health services for an individual on a ‘dysregulated’ pathway (Scenario B) to determine the extent to which different patterns of mental health assessments and subsequent health service referrals were associated with changing trajectories of suicidal ideation and suicide attempts. Two scenarios were investigated: a higher frequency of mental health assessments (where the individual has a service contact and was assessed every 7 days), and a lower frequency of mental health assessments (where the individual has a service contact and was assessed every 30 days). To investigate whether the model displays the suicidal pathways suggested in FV+CC [20], we first set the simulation parameters to appropriate values. In each case, only small changes from the default were needed: scenario A uses high coping, scenario B uses low coping and low volitional characteristics, and scenario C uses low coping and high volitional characteristics. The health service access scenarios used the same parameters as the ‘dysregulated’ scenario B, but with health services enabled. The simulation was then run 100 times with stochastic life events. At each time step, we recorded the individual’s suicidal capability, desire, and risk. Statistical Analyses This data forms a set of three-dimensional time series. We conducted k-means clustering on these time series to determine if model behavior was consistent within each scenario. K-means clustering is an unsupervised machine learning algorithm that divides data into groups such that the data within a group is similar to each other and different from data in other groups. [32] To ensure consistent time series lengths, we extended time series shorter than 1,000 days with the capability, desire, and risk values of 100, 400, and 150 respectively, until they reached 1,000 days. Time series may be shorter than 1,000 days when the individual dies by suicide early in the simulation, so these values allow the clustering algorithm to group simulation runs that result in death. We resampled any series longer than 1,000 days to shorten them. We then computed k-mean clusters using the Soft-DTW [33] distance metric with smoothing parameter γ=0.1. The simulation model was developed using AnyLogic PLE Version 8.9.3, and post-hoc statistical analyses on model output were conducted use R Studio Version 24.12.0, R Version 4.3.1, and Python Version 3.11. Results Following a large number of simulations, the model’s behavior, in response to varying key parameters, was consistent with expectation. For example, Figure 3 displays the validation results for the ‘coping’ parameter, which affects the individual’s response to life events and moderates the effect of feelings of defeat and humiliation on feelings of entrapment. The validation results show that, as coping abilities increase, there is a resulting decrease in the percentage of simulation runs containing a suicide attempt, as well as lower likelihood of the need for health services beyond primary care (Figure 3). Additionally, with higher coping, the first suicide attempt (if present) takes longer to occur, and the individual ends the simulation with lower suicidal capability, less frequent recurrence of suicide attempts, and suicidal desire fluctuations of a smaller amplitude. The full analysis of all key parameters can be found in Supplementary Materials. The simulation model also displayed the hypothesised behavior of the three proposed pathways on the Cusp-Catastrophe model (Figure 4). Pathway A (‘stable’) is characterised by waxing and waning of ‘desire’ in relation to events and event frequency. Pathway B (‘dysregulated’) is characterised by increased desire and capability in response to moving into a suicide risk state. The increasing capability leads to increasing risk and, eventually, a suicide attempt at day 53. Pathway C (‘discontinuous’) is characterised by pre-existing high capability leading to an immediate attempt in response to moving into a high suicide risk state, reflecting a situation of an impulsive attempt, or behavior without extended ideation. For an individual on a ‘dysregulated’ pathway (pathway B), the impact of access to health services on subsequent suicidal ideation or behavior is in Figure 5, with lower (Figure 5A) and higher (Figure 5B) frequency of mental health assessment in two separate simulations. With access to health services enabled at a high assessment frequency of every 7 days (Figure 5A), the individual receives services as soon as they are needed. The individual starts the simulation in primary care. As their suicidal risk increases, they are referred to specialist services. The individual’s suicide risk and capability continue to increase, and so they are referred to a psychiatric hospital for stabilization and more intensive interventions. Following a hospital admission, they are referred back to specialist services for maintenance of care, before referral back to primary care following longer-term stabilization of suicidal ideation. Through these health service referrals, the individual’s risk becomes steady and low for the remainder of the simulation. In contrast, when access to health services for mental health assessment is less frequent (check-ins every 30 days, instead of 7 days) (Figure 5B), health services do not respond to the rising suicide risk quickly enough, and the individual makes a suicide attempt instead of being referred to more intensive care. K-means clustering showed that model behavior was consistent within each scenario, and also substantially different between scenarios (see Supplementary Materials for individual scenario clustering visualizations). Figure 6 contains the results of clustering combining data from 100 simulations of each scenario. Cluster 1 is composed of scenario C (‘discontinuous’) and the scenario B (‘dysregulated’) simulations that resulted in death by suicide (Figure 6, Table 1). Cluster 2 is also composed of scenarios B and C, but captures those simulations that did not result in death by suicide. Cluster 3 represents the stabilized behavior of scenarios A (‘stable’) and the impact of health service access. Finally, Cluster 4 is entirely composed of scenario B, and captures simulations with a death by suicide that occurred later in the time series. Discussion This study has described the development of a simulation model of a ‘suicidal mind’ using an integration of the IMV model, [18] the Fluid Vulnerability Model of suicidal behavior, [17] and the Cusp-Catastrophe model [20] in order to quantify the dynamic and nonlinear nature of suicidal behavior in an individual. The simulation model displayed the hypothesised behavior of the three proposed pathways to suicidal ideation and suicide attempt in the Cusp-Catastrophe model: the ‘stable’ pathway characterised by waxing and waning of ‘desire’ in relation to events and event frequency; the ‘dysregulated’ pathway characterised by increased desire and capability in response to moving into a suicide risk state; and the ‘discontinuous’ pathway characterised by high desire in the context of already existing high capability leading to an immediate suicide attempt in response to increased suicide risk. Additionally, parameters relating to motivational and volitional moderating factors that decreased suicide risk were associated with lower suicidal desire; conversely those motivational and volitional moderating factors that increased suicide risk were associated with higher suicidal desire. Comparative scenarios also showed the stabilization of suicidal ideation following access to mental health services, compared to sustained, or deteriorating, suicidal ideation and subsequent suicide attempts where mental health services were not accessed, or were accessed less frequently. There are a number of methodological considerations for the current study. Firstly, this study aimed to quantify the hypothesised outcomes of current psychological theories of suicidal behavior in individuals, for which no data sources exist to derive parameter values and outcomes. Thus, parameter values in the model were defined arbitrarily. That is, a ‘score’ on the scale of ‘suicidal desire’ is not a meaningful quantity that relates to an empirical biological value (such as output from a sphygmomanometer, or a thermometer). However, patterns over time and relative differences between measures of suicidal ideation, suicidal behavior, capability, and suicide risk behaved in a logically coherent way and were consistent with both the IMV model and the scenarios presented in Bryan et al.’s FCT+CC model. For example, increases in the values of parameters relating to protective factors that might prevent an individual from moving from feelings of defeat to entrapment (such as coping strategies, social connection, or access to health services) led to decreases in suicidality. Conversely, decreases in the values of these parameters led to increased frequency of suicidal ideation and suicide attempts. Validation of parameter values was based on 4.2 million simulations, where a range of different values for each parameter in the model, in combination, were tested (see ODD protocol in Supplementary Materials). It is also worth noting that we did not operationalise all pre-motivational, motivational, and volitional moderators from the IMV model. Rather, we selected major indicators from each phase and used combined parameters for the rest (e.g. personal threat-to-self characteristics represented all threat-to-self moderators besides coping ability). Although not used in our analyses, there are other theories of suicidal behavior that could have formed the basis of the model structure, potentially resulting in different insights than those presented in the current study. The IMV model was selected as the basis of the model structure as it explicitly aims to integrate existing models of suicidal behavior consistent with current research evidence. [18] The Bryan et al.’s model using the Fluid Vulnerability Theory and Cusp Catastrophe model [20] was used to capture the non-linear dynamics of suicidal behavior explicitly. The defined transitions in the health services sector of the model were also based on a set of referral pathways that are likely to be context specific, and may differ from the idealized ‘stepped care’ model [24] of mental health service provision used in the current study. Alternative referral pathways and assumptions relating to the effectiveness of services provided in the prevention of suicidal behavior may also result in different insights relating to how health service interactions may change the trajectory of suicidal ideation and behavior. Future work will include investigating the effects of different types of health services and less idealized care on simulated outcomes. Our model provides a beginning step for other future research to investigate ‘virtual case studies’ developed in partnership with clinicians and those with lived experience. For example, information from a specific clinical case (or ‘type’ of clinical case) can be used to convert information about a clinical presentation into a set of parameter values with which to initialize the model. The model can then be used to investigate the potential outcomes of proposed clinical responses, specific types of health service use, different sequences of triggering life events, or other changes to the patient’s environment. There has been previous use of computational simulation to investigate mechanisms of the human mind, in the areas of cognitive science, neuroscience, [34-35] and the burgeoning area of computational psychiatry. [34] These approaches often aim to identify particular mental health phenotypes using machine learning and artificial intelligence approaches to identify patterns in observational data for risk prediction. [34] To date there have been no applications of simulation modelling to an individual’s suicidal behavior. The suicidal mind presented in the current paper does not employ curve-fitting algorithms to model existing observational data; rather, the simulation was guided by causal reasoning and a synthesis of current theory derived from contemporary literature. In summary, this study has described the development of a simulation model of a ‘suicidal mind’ that captures the non-linear dynamics of suicidal behavior based on integrated prevailing psychological theories. Simulation model output was consistent with the different pathways to suicidal behavior that are proposed by current theoretical models of suicidal behavior, and also showed the plausible impact of different patterns of health service access on subsequent suicidal ideation and behavior within an individual. Further validation of the model incorporating wider clinical and lived-experience perspectives, and application to a wider array of different scenarios, may lead to the use of such simulation models to help refine theories of suicidal behavior. There is also potential to use such models in virtual case studies based on the specific characteristics of individuals to potentially assist in clinical decision making, or to estimate how clinical, psychosocial, or population-level policy interventions might affect individual suicidal behavior. References World Health Organization, “Suicide.” Accessed: Apr. 02, 2025. [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/suicide Global Burden of Disease Collaborative Network, “Global Burden of Disease Study 2021 (GBD 2021),” Institute for Health Metrics and Evaluation (IHME), Seattle, United States, 2024. Accessed: Apr. 02, 2025. [Online]. Available: http://vizhub.healthdata.org/gbd-compare A. M. El-Sayed and S. Galea, Systems Science and Population Health . Oxford University Press, 2017. S. Galea, M. Riddle, and G. A. Kaplan, “Causal thinking and complex system approaches in epidemiology,” Int. J. Epidemiol. , vol. 39, no. 1, pp. 97–106, Feb. 2010, doi: 10.1093/ije/dyp296. R. Schuerkamp, L. Liang, K. L. Rice, and P. J. Giabbanelli, “Simulation Models for Suicide Prevention: A Survey of the State-of-the-Art,” Computers , vol. 12, no. 7, Art. no. 7, Jul. 2023, doi: 10.3390/computers12070132. A. Page, J.-A. Atkinson, M. Heffernan, G. McDonnell, and I. Hickie, “A decision-support tool to inform Australian strategies for preventing suicide and suicidal behaviour,” Public Health Res. Pract. , vol. 27, no. 2, 2017, doi: 10.17061/phrp2721717. R. Zahan, J. Mikuliak, and N. D. Osgood, “Developing a Tool to Assess the Impact of Simulated Intervention Strategies on Suicide and Suicidal Behaviours in Canada: A Dynamic Modelling & Machine Learning Approach,” May 13, 2024. doi: 10.20944/preprints202405.0813.v1. J.-A. Atkinson et al. , “Systems modelling and simulation to inform strategic decision making for suicide prevention in rural New South Wales (Australia),” Aust. N. Z. J. Psychiatry , vol. 54, no. 9, pp. 892–901, Sep. 2020, doi: 10.1177/0004867420932639. M. Cerdá, A. D. Hamilton, M. Tracy, C. Branas, D. Fink, and K. M. Keyes, “Would restricting firearm purchases due to alcohol- and drug-related misdemeanor offenses reduce firearm homicide and suicide? An agent-based simulation,” Inj. Epidemiol. , vol. 9, no. 1, p. 17, Jun. 2022, doi: 10.1186/s40621-022-00381-x. K. M. Keyes, A. Hamilton, J. Swanson, M. Tracy, and M. Cerdá, “Simulating the Suicide Prevention Effects of Firearms Restrictions Based on Psychiatric Hospitalization and Treatment Records: Social Benefits and Unintended Adverse Consequences,” Am. J. Public Health , vol. 109, no. S3, pp. S236–S243, Jun. 2019, doi: 10.2105/AJPH.2019.305041. K. W. McKinley et al. , “Impact of Universal Suicide Risk Screening in a Pediatric Emergency Department: A Discrete Event Simulation Approach,” Healthc. Inform. Res. , vol. 28, no. 1, pp. 25–34, Jan. 2022, doi: 10.4258/hir.2022.28.1.25. J.-A. Occhipinti et al. , “Sound Decision Making in Uncertain Times: Can Systems Modelling Be Useful for Informing Policy and Planning for Suicide Prevention?,” Int. J. Environ. Res. Public. Health , vol. 19, no. 3, Art. no. 3, Jan. 2022, doi: 10.3390/ijerph19031468. A. Page et al. , “A decision support tool to inform local suicide prevention activity in Greater Western Sydney (Australia),” Aust. N. Z. J. Psychiatry , vol. 52, no. 10, pp. 983–993, Oct. 2018, doi: 10.1177/0004867418767315. A. Skinner, J.-A. Occhipinti, Y. J. C. Song, and I. B. Hickie, “Regional suicide prevention planning: a dynamic simulation modelling analysis,” BJPsych Open , vol. 7, no. 5, p. e157, Sep. 2021, doi: 10.1192/bjo.2021.989. K. A. Van Orden, T. K. Witte, K. C. Cukrowicz, S. R. Braithwaite, E. A. Selby, and T. E. Joiner, “The interpersonal theory of suicide.,” Psychol. Rev. , vol. 117, no. 2, pp. 575–600, 2010, doi: 10.1037/a0018697. E. D. Klonsky and A. M. May, “The Three-Step Theory (3ST): A New Theory of Suicide Rooted in the ‘Ideation-to-Action’ Framework,” Int. J. Cogn. Ther. , vol. 8, no. 2, pp. 114–129, Jun. 2015, doi: 10.1521/ijct.2015.8.2.114. M. D. Rudd, “Fluid Vulnerability Theory: A Cognitive Approach to Understanding the Process of Acute and Chronic Suicide Risk.,” in Cognition and suicide: Theory, research, and therapy. , T. E. Ellis, Ed., Washington: American Psychological Association, 2006, pp. 355–368. doi: 10.1037/11377-016. R. C. O’Connor and O. J. Kirtley, “The integrated motivational–volitional model of suicidal behaviour,” Philos. Trans. R. Soc. B Biol. Sci. , vol. 373, no. 1754, p. 20170268, Jul. 2018, doi: 10.1098/rstb.2017.0268. H. Hjelmeland and B. and Loa Knizek, “The emperor’s new clothes? A critical look at the interpersonal theory of suicide,” Death Stud. , vol. 44, no. 3, pp. 168–178, Mar. 2020, doi: 10.1080/07481187.2018.1527796. C. J. Bryan et al. , “Nonlinear change processes and the emergence of suicidal behavior: A conceptual model based on the fluid vulnerability theory of suicide,” New Ideas Psychol. , vol. 57, p. 100758, Apr. 2020, doi: 10.1016/j.newideapsych.2019.100758. J. W. Forrester, “Industrial Dynamics,” J. Oper. Res. Soc. , vol. 48, no. 10, pp. 1037–1041, Oct. 1997, doi: 10.1057/palgrave.jors.2600946. K. Michel and L. Valach, “Suicide as goal-directed action,” Arch. Suicide Res. , vol. 3, no. 3, pp. 213–221, Jul. 1997, doi: 10.1080/13811119708258273. G. Dedic, S. Djurdjevic, and B. Golubovic, “Psychological assessment of persons following suicide attempt by self-poisoning,” Vojnosanit. Pregl. , vol. 67, no. 2, pp. 151–158, 2010, doi: 10.2298/VSP1002151D. S. Cross and I. Hickie, “Transdiagnostic stepped care in mental health,” Public Health Res. Pract. , vol. 27, no. 2, 2017, doi: 10.17061/phrp2721712. V. Grimm et al. , “The ODD Protocol for Describing Agent-Based and Other Simulation Models: A Second Update to Improve Clarity, Replication, and Structural Realism,” J. Artif. Soc. Soc. Simul. , vol. 23, no. 2, p. 7, 2020. C. J. Bryan, J. E. Butner, S. Sinclair, A. B. O. Bryan, C. M. Hesse, and A. E. Rose, “Predictors of Emerging Suicide Death Among Military Personnel on Social Media Networks,” Suicide Life. Threat. Behav. , vol. 48, no. 4, pp. 413–430, 2018, doi: 10.1111/sltb.12370. C. J. Bryan, M. D. Rudd, A. L. Peterson, S. Young-McCaughan, and E. G. Wertenberger, “The ebb and flow of the wish to live and the wish to die among suicidal military personnel,” J. Affect. Disord. , vol. 202, pp. 58–66, Sep. 2016, doi: 10.1016/j.jad.2016.05.049. E. M. Kleiman, B. J. Turner, S. Fedor, E. E. Beale, J. C. Huffman, and M. K. Nock, “Examination of real-time fluctuations in suicidal ideation and its risk factors: Results from two ecological momentary assessment studies,” J. Abnorm. Psychol. , vol. 126, no. 6, pp. 726–738, 2017, doi: 10.1037/abn0000273. T. K. Witte, K. K. Fitzpatrick, T. E. Joiner, and N. B. Schmidt, “Variability in suicidal ideation: A better predictor of suicide attempts than intensity or duration of ideation?,” J. Affect. Disord. , vol. 88, no. 2, pp. 131–136, Oct. 2005, doi: 10.1016/j.jad.2005.05.019. C. J. Bryan, D. C. Rozek, J. Butner, and M. D. Rudd, “Patterns of change in suicide ideation signal the recurrence of suicide attempts among high-risk psychiatric outpatients,” Behav. Res. Ther. , vol. 120, p. 103392, Sep. 2019, doi: 10.1016/j.brat.2019.04.001. C. J. Bryan and M. D. Rudd, “Nonlinear Change Processes During Psychotherapy Characterize Patients Who Have Made Multiple Suicide Attempts,” Suicide Life. Threat. Behav. , vol. 48, no. 4, pp. 386–400, Aug. 2018, doi: 10.1111/sltb.12361. J. Wu, “Cluster Analysis and K-means Clustering: An Introduction,” in Advances in K-means Clustering: A Data Mining Thinking , J. Wu, Ed., Berlin, Heidelberg: Springer, 2012, pp. 1–16. doi: 10.1007/978-3-642-29807-3_1. M. Cuturi and M. Blondel, “Soft-DTW: a Differentiable Loss Function for Time-Series,” in Proceedings of the 34th International Conference on Machine Learning , D. Precup and Y. W. Teh, Eds., in Proceedings of Machine Learning Research, vol. 70. PMLR, Aug. 2017, pp. 894–903. [Online]. Available: https://proceedings.mlr.press/v70/cuturi17a.html P. F. Hitchcock, E. I. Fried, and M. J. Frank, “Computational Psychiatry Needs Time and Context,” Annu. Rev. Psychol. , vol. 73, no. Volume 73, 2022, pp. 243–270, Jan. 2022, doi: 10.1146/annurev-psych-021621-124910. C. Langley, B. I. Cirstea, F. Cuzzolin, and B. J. Sahakian, “Theory of Mind and Preference Learning at the Interface of Cognitive Science, Neuroscience, and AI: A Review,” Front. Artif. Intell. , vol. 5, Apr. 2022, doi: 10.3389/frai.2022.778852. Table 1 Table 1: Cluster scenario-composition for clusters shown in Figure 7 Cluster Scenario A Scenario B Scenario C Health service 1 0.7% 29.6% 67.4% 2.2% 2 0.0% 68.8% 28.1% 3.1% 3 50.8% 0.0% 0.0% 49.2% 4 0.0% 100.0% 0.0% 0.0% Additional Declarations There is NO Competing Interest. Supplementary Files SuicidalMindnhbsuppanon.docx Supplemental Material Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6605926","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":453711623,"identity":"e3922931-cedc-4d5b-a1e7-862b182ea8d9","order_by":0,"name":"Gabriel McDonnell Maayan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAtElEQVRIiWNgGAWjYBADOQMQyUOKFmPStSRuIFqLfNjhh595au6kb5dIYHzwto0ILYa304yleY49y905I4HZcC5RWmYnGEjzsB3O3XAjgU2alzgt6Z9/8/w7nG5wI4H9N1Fa5KVzzICGH04AamFjJkqLgXROmeXcvsOGO3seNkvOOUeMLbPTN9948+2wvDl78sEPb8qIseUAAwMTJDoYG4hQD7IFqI7xB3FqR8EoGAWjYKQCAL8NN3KXBB0qAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0001-5725-7622","institution":"Boston University","correspondingAuthor":true,"prefix":"","firstName":"Gabriel","middleName":"McDonnell","lastName":"Maayan","suffix":""},{"id":453711624,"identity":"0762c06c-a5ec-4917-b7cc-7122e9bf73b9","order_by":1,"name":"Andrew Page","email":"","orcid":"","institution":"Western Sydney University","correspondingAuthor":false,"prefix":"","firstName":"Andrew","middleName":"","lastName":"Page","suffix":""},{"id":453711625,"identity":"ee5eb497-d688-48bb-9f52-3ed4d5d73c5c","order_by":2,"name":"Elizabeth Conroy","email":"","orcid":"","institution":"Western Sydney University","correspondingAuthor":false,"prefix":"","firstName":"Elizabeth","middleName":"","lastName":"Conroy","suffix":""},{"id":453711626,"identity":"cd826874-b17b-4e46-beed-76e00d648bf2","order_by":3,"name":"Rory O'Connor","email":"","orcid":"","institution":"University of Glasgow","correspondingAuthor":false,"prefix":"","firstName":"Rory","middleName":"","lastName":"O'Connor","suffix":""},{"id":453711627,"identity":"5021e818-6f9b-4b3e-a0a4-b85c95c86de5","order_by":4,"name":"Greg Carter","email":"","orcid":"","institution":"The University of Newcastle Australia","correspondingAuthor":false,"prefix":"","firstName":"Greg","middleName":"","lastName":"Carter","suffix":""},{"id":453711628,"identity":"233d25fd-6a9f-4d6e-89cc-0c1f32671227","order_by":5,"name":"Wesley Wildman","email":"","orcid":"https://orcid.org/0000-0002-7571-1259","institution":"Center for Mind and Culture","correspondingAuthor":false,"prefix":"","firstName":"Wesley","middleName":"","lastName":"Wildman","suffix":""}],"badges":[],"createdAt":"2025-05-06 19:16:02","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6605926/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6605926/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":85618226,"identity":"669768a6-aa99-48d5-bc69-9cd57d7328bf","added_by":"auto","created_at":"2025-06-29 14:49:24","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1054425,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe Integrated Motivational Volitional model of suicide [18] (A), and the Cusp-Catastrophe model of suicide (adapted from [20])\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e*Note: Pathways ‘A’, ‘B’, and ‘C’ depict three different potential responses to suicidal triggers. Path ‘A’ reflects a ‘stable’ response wherein an individual’s suicidal desire fluctuates but always returns to a nonsuicidal state. ‘B’ is a dysregulating pathway in which the individual gradually increases suicidal capability until they may experience active suicidal behavior. Finally, path ‘C’ is the catastrophic response where an individual with initially high suicidal capability can immediately jump to suicidal behavior as desire increases.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6605926/v1/d2189df68934ed795e3afc67.png"},{"id":85619960,"identity":"71627744-8c76-4b1c-8ffd-2ab781163b75","added_by":"auto","created_at":"2025-06-29 15:05:24","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1182753,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe structure of the ‘suicidal mind’ model.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6605926/v1/c4777588ef48962e6312a996.png"},{"id":85619961,"identity":"43d144d7-b3c1-49fe-8bee-2602085a34e3","added_by":"auto","created_at":"2025-06-29 15:05:24","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":349177,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eParameter sweep results for the 'coping' parameter.*\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e*Note: Y axis in first row of the panel refers to Percent who Accessed: specialist community services (SCS), acute care, and psychiatric hospital. Service usage is calculated as true/false if the individual used the service at least once in a simulation. Error bars in the second row of the panel are the 95% CI.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6605926/v1/54dcde771f12ada6094e1c6f.png"},{"id":85618235,"identity":"0559fe86-6098-4427-a1f8-109ff4b55a03","added_by":"auto","created_at":"2025-06-29 14:49:24","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":494450,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSimulated behavior for ‘stable’ (Scenario A), ‘dysregulated’ (Scenario B), and ‘discontinuous’ (Scenario C) pathways to suicidal ideation and behavior (as per Bryan et al. 2020)* Parameters for Scenario A: \u003c/strong\u003ecoping = 0.6, disable health services = 1; \u003cstrong\u003eScenario B: \u003c/strong\u003ecoping = 0.2, disable health services = 1; \u003cstrong\u003eand Scenario C: \u003c/strong\u003ecoping = 0.2, personal volitional characteristics = 1, access to means = 1, disable health services = 1\u003c/p\u003e\n\u003cp\u003e*Note: Each of simulation was run with the random seed set to ‘123’ and no access to health services.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6605926/v1/d571b0ddbacd40995e1ed632.png"},{"id":85618230,"identity":"8287f969-ba6b-4c2e-9521-0979f3645ecb","added_by":"auto","created_at":"2025-06-29 14:49:24","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":489237,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSimulated behavior following access to health services, with annotation pertaining to each service type for higher frequency mental health assessment, every 7 days (A) and for lower frequency mental health assessment, every 30 days (B). Parameters used: \u003c/strong\u003ecoping = 0.2, primary care assessment frequency = 7 days or 30 days.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6605926/v1/319b42514d7df5a2694e554d.png"},{"id":85619320,"identity":"a6d86f94-8b7b-4de4-a547-98321cf72bfe","added_by":"auto","created_at":"2025-06-29 14:57:24","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":922471,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCluster centres (red line) for combined time-series data based on 100 model runs (black lines) of each scenario*\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e*Note: Death by suicide is represented by the line moving to the top of the window. Cluster 1 is characterized by early deaths by suicide. Cluster 2 is characterized by large fluctuations in suicidal desire and risk. Cluster 3 is characterized by low, steady values for each variable. Cluster 4 is characterized by gradual increases in capability, large fluctuations in desire and risk, and eventual death.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6605926/v1/dfb751becffa59be0bf3eb85.png"},{"id":85620324,"identity":"a8ed2749-ff32-4f78-a67a-b8a3155787be","added_by":"auto","created_at":"2025-06-29 15:13:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5892952,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6605926/v1/36c74e8b-d787-421b-b419-81cd7347433e.pdf"},{"id":85619318,"identity":"51804a8c-8d63-424f-a37b-deb7b81a19d6","added_by":"auto","created_at":"2025-06-29 14:57:24","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2208701,"visible":true,"origin":"","legend":"Supplemental Material","description":"","filename":"SuicidalMindnhbsuppanon.docx","url":"https://assets-eu.researchsquare.com/files/rs-6605926/v1/a007efc2da02e3abe5c132f5.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"\u003cp\u003eSimulation of a ‘suicidal mind’: Using the Integrated Motivational Volitional model of suicide to demonstrate dynamic suicidal states\u003c/p\u003e","fulltext":[{"header":"Significance Statement","content":"\u003cp\u003eThis is the first study to develop a computational simulation of an individual \u0026lsquo;suicidal mind\u0026rsquo; based on a conceptual synthesis of three current psychological theories. The simulation model is able to quantify the proposed transitions from \u0026lsquo;pre-motivational\u0026rsquo;, \u0026lsquo;motivational\u0026rsquo; and \u0026lsquo;volitional\u0026rsquo; states in the Integrated Motivational-Volitional model, and also capture the non-linear and dynamic nature of suicidal ideation and behavior as proposed by prior research. This model of a suicidal mind demonstrates the coherence of a synthesis between several dissimilar theories of suicidality. It can also lead to further refinement of current theoretical understandings of suicidal behavior, the use of virtual case studies to guide clinical decision-making, training, and the evaluation of population-level interventions on individuals with specific characteristics.\u0026nbsp;\u003c/p\u003e\n"},{"header":"Introduction","content":"\u003cp\u003eSuicide remains an important global public health priority, with over 720,000 deaths by suicide globally per year, two-thirds of deaths occurring in low- and middle-income countries. [1] Suicide is the third leading causes of death among young people (15-29 years), and intentional self-harm is of a similar magnitude to global rates in traffic injury. [2] Suicidal behaviors are characterised by a complex, multi-factorial aetiology, with interacting social, cultural, biological, psychological, and environmental determinants that can affect individuals across the life-course. [1] \u003c/p\u003e\n\n\u003cp\u003eDynamic simulation models (DSMs) are computational representations of the real world that aim to capture the dynamics and behavior of a whole system or population, [3] and are methods that can explicitly capture the complex behavior of suicide. Computational simulation has a long history in disciplines such as ecology, physics, and engineering. In population health, DSMs have mainly been used for infectious diseases - and became even more prominent in guiding responses to the recent COVID-19 pandemic. DSM\u0026rsquo;s are increasingly recognized for their broader value in helping to solve complex problems for other population health outcomes [4] \u003c/p\u003e\n\n\u003cp\u003ePrevious uses of computational simulation in suicide prevention have predominantly addressed policy related questions for whole populations, including identifying combinations of interventions likely to avert future suicidal behavior and to inform mental health service capacity requirements. [5-14] DSMs of suicidal behavior have been developed mainly in the US, Canadian, and Australian contexts, [5] and have been developed for national [6-7] and local populations [8-14] to investigate the potential impact of interventions on subsequent trends in suicide and attempted suicide. These approaches provided insights that have informed service planning, [11, 13-14] the commissioning of local mental health services, [13-14] potential impacts of firearms restrictions, [9-10] and identifying priority areas where the greatest population impact in reducing suicidal behavior might be achieved. [5-14]\u003c/p\u003e\n\n\u003cp\u003eThese previous approaches have largely considered aggregate patterns of suicidal behavior in populations. However, there is also scope to use DSMs to describe and understand individual behaviors, to operationalise and potentially validate specific psychological theories that attempt to incorporate the range of complex determinants of suicidal behavior and articulate why some individuals might engage in suicidal behavior, whereas other do not. \u003c/p\u003e\n\n\u003cp\u003ePrevious psychological theories of an individual\u0026rsquo;s suicidal behavior include the Interpersonal Theory of Suicide, [15] the Three-Step Theory (3ST) of suicide, [16] and the Fluid Vulnerability Theory of suicide. [17] These have culminated in the most recent iteration of the Integrated Motivational-Volitional model (IMV) [18] that integrates previous theories with current literature, and describes the emergence of suicidal ideation and behavior. These theories aim to be consistent with observational studies of risk factors for suicidal behavior, but they have not been without criticism. This includes the simplification of complex multi-factorial behavior to a few key constructs, the selected use of suicide prevention literature, and the challenges associated with empirical validation. [19] Psychological theories can clarify the drivers and emergence of suicidal behavior, but they are qualitative, conceptual and static, and with potentially limited utility in quantifying specific individual-level behavior. \u003c/p\u003e\n\n\u003cp\u003eWhile suicidal acts are uniquely individual, being able to empirically quantify current psychological theories and demonstrate that patterns of individual behavior are logically consistent with these theories and research evidence, can help to generate hypotheses regarding clinical responses to those presenting in suicidal crisis (or with a past history of self-harm), and broader prevention approaches among those who have never previously shown a propensity for suicidal behavior. Additionally, dynamic modelling of psychological theories can help refine the conceptualisation of suicidal behavior and provide insights into how theories of suicidal behavior might be revised or modified to better capture individual behavior. Accordingly, this study describes the development of a DSM of a suicidal mind that synthesizes three theories as the basis for capturing the dynamic changes in different suicidal states within a simulated individual. \u003c/p\u003e\n\n"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eTheories to inform the model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe IMV model is the first of the three synthesized theories. It integrates several current psychological theories, and its development was guided by the need to incorporate a synthesis of current literature into a single tenable theoretical framework. [18] This resulted in a \u0026lsquo;tri-partite\u0026rsquo; model that describes the biopsychosocial context in which suicidal ideation and behavior may emerge (a \u0026lsquo;pre-motivational phase\u0026rsquo;), factors that determine the emergence of suicidal ideation (a \u0026lsquo;motivational phase\u0026rsquo;), and factors that determine the transition from ideation to behavior (a \u0026lsquo;volitional phase\u0026rsquo;) (Figure 1A). Transitions between the \u0026lsquo;motivational\u0026rsquo; and \u0026lsquo;volitional\u0026rsquo; phases (and the emergence of ideation and behavior, respectively) are determined by \u0026lsquo;threat-to-self moderators\u0026rsquo; (such as social problem solving, coping, memory biases, ruminative processes), \u0026lsquo;motivational moderators\u0026rsquo; (such as thwarted belongingness, burdensomeness, resilience, social support), and \u0026lsquo;volitional moderators\u0026rsquo; (such as access to means, planning, exposure to suicide, impulsivity, physical pain sensitivity, fearlessness about death, past behavior). [18] \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHowever, the IMV model as proposed is monodirectional in its characterisation of how suicidal ideation, intent, and behavior develop (although dynamic bidirectional relationships between ideation and behavior are included in a more recent update to the IMV model) [18]. Our study captures the dynamic nature of suicidal behavior, the waxing and waning of ideation, of psychological distress, and environmental factors that can affect these outcomes, the temporal dynamics of these factors, and the behavioural (and psychological) feedback loops that can be hypothesised between them by integrating two other theories into the IMV model. We draw on the Fluid Vulnerability Model of suicidal behavior [17] which conceptualises suicidal behavior as inherently dynamic and non-linear, and which has been extended by Bryan et al., by incorporating the Cusp-Catastrophe model from dynamical systems theory (Figure 1B; hereafter called the FV+CC model). [20] The Cusp-Catastrophe model is useful in that it can capture nonlinear change processes, emergent behavior, and bifurcating patterns of behavior in systems. It can also provide a foundation for explaining phenomena such as the sudden emergence of suicidal behavior without prior suicidal planning. [20] \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe \u0026lsquo;suicidal mind\u0026rsquo; model structure\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe \u0026lsquo;suicidal mind\u0026rsquo; is a system-dynamics model of \u0026lsquo;suicidality\u0026rsquo; in an individual (Figure 2A). Systems-dynamics modelling was first described by Jay Forrester [21] and is a type of DSM composed of stocks (quantities that accumulate) and flows (rates of change). The units of this model are largely arbitrary and can be conceptualised as levels of the relevant cognitions or behaviors. In the simulation model, the individual has stocks representing a quantity of states that correspond to each of the components of the IMV model (e.g. one stock representing an amount of defeat/humiliation, one representing an amount of entrapment, etc.). Firstly, the assumption is that \u0026lsquo;suicidal ideation and intention\u0026rsquo; begins via \u0026lsquo;negative cognitions\u0026rsquo; that are triggered by \u0026lsquo;life events\u0026rsquo;, \u0026lsquo;diathesis,\u0026rsquo; and \u0026lsquo;environmental\u0026rsquo; factors. Life events are perturbations in the individual\u0026rsquo;s life, such as losing a job or relationship breakdown. There are two types of life-events, categorised as \u0026lsquo;normal\u0026rsquo; events or \u0026lsquo;severe\u0026rsquo; events. Drawing from the fluid vulnerability theory, normal events and severe events are those to which the individual is less or more sensitive, respectively. Additionally, the individual has different dispositional tolerance (which can be modified in the model) that acknowledges a response to a given life event may be \u0026lsquo;more\u0026rsquo; or \u0026lsquo;less\u0026rsquo; stressful. \u0026lsquo;Diathesis\u0026rsquo; relates to factors such as chronic illness, genetic vulnerabilities, or other medical conditions. \u0026lsquo;Environment\u0026rsquo; includes factors such as long-term financial distress, material circumstances including social disadvantage, isolation, or other sustained external stressors. \u0026lsquo;Diathesis\u0026rsquo; and \u0026lsquo;environment\u0026rsquo; factors have their own effects on \u0026lsquo;negative cognitions,\u0026rsquo; but can also compound the effects from life events.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026lsquo;Negative cognitions\u0026rsquo; lead to feelings of \u0026lsquo;defeat/humiliation\u0026rsquo; (i.e. \u0026lsquo;negative cognitions\u0026rsquo; increase the \u0026lsquo;defeat/humiliation\u0026rsquo; stock). In turn, \u0026lsquo;defeat/humiliation\u0026rsquo; can lead to feelings of \u0026lsquo;entrapment\u0026rsquo; through the presence or absence of \u0026lsquo;threat-to-self moderators\u0026rsquo;, such as social problem solving, coping skills, or memory biases. \u0026lsquo;Entrapment\u0026rsquo; leads to increased levels of \u0026lsquo;suicidal ideation\u0026rsquo;, based on the influence of \u0026lsquo;motivational moderators\u0026rsquo; including feelings of thwarted belongingness, burdensomeness, or social support. From \u0026lsquo;suicidal ideation\u0026rsquo;, the individual can begin to exhibit active \u0026lsquo;suicidal behavior\u0026rsquo;, based on the influence of \u0026lsquo;volitional moderators\u0026rsquo; including access to means, suicidal planning, and fearlessness about death.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;The \u0026lsquo;suicidal mind\u0026rsquo; model includes feedback loops to capture the dynamic and cyclical nature of moving through each of these states. For example, levels of \u0026lsquo;defeat/humiliation\u0026rsquo; affect \u0026lsquo;negative conditions\u0026rsquo; in a reinforcing loop, as do levels of \u0026lsquo;entrapment\u0026rsquo;, \u0026lsquo;suicidal ideation and intent\u0026rsquo;, and \u0026lsquo;suicidal behavior\u0026rsquo;. These feedback loops acknowledge the importance of past levels of suicidality in potential responses to the triggering of life events.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;From the fluid vulnerability theory, we also include a \u0026lsquo;suicidal mode\u0026rsquo; in the form of \u0026lsquo;low\u0026rsquo; and \u0026lsquo;high\u0026rsquo; suicidal risk states. The individual may enter a high-risk state based on the level of suicidal ideation, referred to as \u0026lsquo;suicidal desire\u0026rsquo;, and on \u0026lsquo;capability\u0026rsquo;. The high-risk state loosely corresponds to the upper plateau of the FV+CC [20] model\u0026rsquo;s synthesis of the Fluid Vulnerability Theory and the Cusp Catastrophe model (Figure 1B). Accordingly, it can be more difficult for the individual to leave a high-risk state than it was to enter. While in a high-risk state, the individual\u0026rsquo;s capability increases, perhaps under the impact of past suicidal behavior; they might make suicidal plans, which increases capability further, and become more likely to make a suicide attempt.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;To make a suicide attempt, the individual must have very high levels of overall suicidal risk, defined by desire, capability and \u0026lsquo;risk-state\u0026rsquo;. At these levels, the individual may attempt suicide with probability proportional to their desire and capability (see Supplementary Materials for a detailed definition). If the individual attempts suicide, they may also die by suicide. Following an non-fatal suicide attempt, individuals may have a wide range of responses including disappointment, anger at a thwarted attempt, emotional relief or further feelings of entrapment. [22] One study found 90% of attempters felt relief following a non-fatal attempt, but around 50% intended to make a repeat attempt. [23] Following a non-fatal attempt in the \u0026lsquo;suicidal mind\u0026rsquo; model, the individual\u0026rsquo;s \u0026lsquo;entrapment,\u0026rsquo; \u0026lsquo;suicidal ideation,\u0026rsquo; and \u0026lsquo;suicidal behavior\u0026rsquo; stocks return to zero. This represents some amount of emotional release following a suicide attempt. The \u0026lsquo;defeat/humiliation\u0026rsquo; stock is not reset because the attempt itself does nothing to resolve underlying and external drivers to suicidality unless there are other interventions, such as from the health services sector. Repeat attempts are likely without interventions.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;The \u0026lsquo;suicidal mind\u0026rsquo; model simulates an individual\u0026rsquo;s internal cognitive-motivational mechanisms, so it also includes interactions with health services to investigate the effect of service use on modifying trajectories of suicidal ideation and behavior (Figure 2B). Health services are implemented as a simplified and idealized stepped-care model [24] with arrows indicating referral pathways between different health service states (Figure 2B). \u0026lsquo;Universal care\u0026rsquo; refers to population-level interventions, for example, a government mental health promotion campaign, and these services reduce incoming negative cognitions. \u0026lsquo;Primary care\u0026rsquo; services relate to presentation to a general practitioner setting for mental health assessment, but the individual must choose to interact with these services. Without mental health assessments, the individual is not referred to other services until they make a suicide attempt. \u0026lsquo;Specialist community services\u0026rsquo; more directly target suicidal ideation and behavior, and may include regular sessions with a qualified mental health practitioner. These services provide more frequent risk assessments and improve the individual\u0026rsquo;s cognitions and coping skills over time. \u0026lsquo;Acute care\u0026rsquo; represents short-term high-intensity services, such as the hospital emergency room. In the simplified health service system, the individual attends acute care services directly following a suicide attempt, which provides both stabilization and triage for the individual\u0026rsquo;s next treatment step. In a real health service system, acute care presentations are commonly for ideation or planning and not just for attempts. Finally, \u0026lsquo;psychiatric hospital\u0026rsquo; services offer medium-term high-intensity services that temporarily suppress triggering life events and provide psychiatric and pharmaceutical interventions that directly target symptoms and underlying conditions. An individual enters the \u0026lsquo;psychiatric hospital\u0026rsquo; state if they are at high suicidal risk and have been so for a sufficient length of time. An individual starts in the \u0026lsquo;universal care\u0026rsquo; state but immediately moves to the \u0026lsquo;primary care\u0026rsquo; state if they choose to use the primary care system. Otherwise, they move to the \u0026lsquo;acute care\u0026rsquo; state following a suicide attempt, or into \u0026lsquo;specialist community\u0026rsquo; and \u0026lsquo;psychiatric hospital\u0026rsquo; services if they have had previous risk assessments and are assessed to need further care. From the other health service states, the individual moves to \u0026lsquo;acute care\u0026rsquo; following a suicide attempt. \u0026lsquo;Assessed risk\u0026rsquo; is a moving average of several previous mental health assessments. Various aspects of the health service system, such as assessment frequency and efficacy, are included as model parameters and may be adjusted to investigate alternative scenarios (e.g. service availability or efficacy).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eModel parameterisation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA detailed list of parameters is in the \u0026lsquo;Overview, Design concepts and Details (ODD)\u0026rsquo; protocol [25] (Supplementary Materials). In general, parameters fall into three categories. Parameters in the first set determine the strength and frequency of triggering factors, such as environmental or diathesis-related risks and life events. Parameters in the second set determine how the individual responds to the triggering factors, such as coping abilities or the moderators from the IMV model. The final set of parameters define external intervention characteristics, such as the frequency of primary care evaluations or the effectiveness of specialist/psychiatric hospital care. By modifying the three sets of parameters, we can describe an individual\u0026rsquo;s internal and external state, thereby investigating virtual case studies and \u0026ldquo;what-if\u0026rdquo; scenarios. The units of each time step in the model are days. The units of parameters relating to time-based events (e.g. frequency of primary care visits) are therefore days, while parameters relating to the size of effects (e.g. coping ability) have more abstract units.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eScenario testing\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe main analytic questions for the \u0026lsquo;suicidal mind\u0026rsquo; model are (i) whether changes in parameters that the underlying theory assumes increase or decrease the level of suicidality result in corresponding increases or decreases in suicidality as estimated by the model, and (ii) whether the model displays suicidal states as suggested by the FV+CC model. [20] For example, for (i), does increasing exposure to protective factors that might prevent an individual moving from feelings of defeat to entrapment lead to decreases in suicidality? Or do improvements in coping strategies, or access to health services, lead to decreases in suicidality? Conversely, does an increase in exposure to risk factors lead to increases in suicidality?\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo answer this first question, we ran 4.2 million simulations, varying key parameters, to build a full description of the model\u0026rsquo;s behavior. For each varied parameter, we summarize the behavior with the following outputs: presence of at least one suicide attempt, health service access, date of first suicide attempt, suicide capability at the end of the simulation, average attempt-recurrence rate, and the suicidal-desire inter-quartile distance to measure the amplitude of suicidal desire fluctuations.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; For (ii), the question is can the \u0026lsquo;suicidal mind\u0026rsquo; model can replicate the types of temporal processes for suicide risk as proposed in the FV+CC model [20] (Figure 1B): slow, smooth, and mild fluctuations around a homeostatic equilibrium point (\u0026lsquo;stable\u0026rsquo;, scenario A); rapid, large fluctuations around a homeostatic equilibrium point (\u0026lsquo;dysregulated\u0026rsquo;, scenario B); and a sudden, dramatic departure from a homeostatic equilibrium point (\u0026lsquo;discontinuous\u0026rsquo;, scenario C). Scenario A reflects the slow and smooth fluctuations in suicidal ideation among lower risk suicidal individuals. [26-29] \u0026nbsp;Scenario B reflects the large fluctuations in suicidal ideation associated with a dysregulated change process among individuals who engage in repeated intentional self-harm and associated with a gradual increase in suicidal \u0026lsquo;capability\u0026rsquo;. [26, 28-31] Scenario C reflects the sudden, dramatic shift in suicide risk states associated with a discontinuous change process, where an individual begins with high suicidal \u0026lsquo;capability\u0026rsquo; and with low suicidal ideation, reflecting an \u0026lsquo;impulsive\u0026rsquo; response to a stressor or life-event. [20]\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;We also investigated the impact of access to health services for an individual on a \u0026lsquo;dysregulated\u0026rsquo; pathway (Scenario B) to determine the extent to which different patterns of mental health assessments and subsequent health service referrals were associated with changing trajectories of suicidal ideation and suicide attempts. Two scenarios were investigated: a higher frequency of mental health assessments (where the individual has a service contact and was assessed every 7 days), and a lower frequency of mental health assessments (where the individual has a service contact and was assessed every 30 days).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;To investigate whether the model displays the suicidal pathways suggested in FV+CC [20], we first set the simulation parameters to appropriate values. In each case, only small changes from the default were needed: scenario A uses high coping, scenario B uses low coping and low volitional characteristics, and scenario C uses low coping and high volitional characteristics. The health service access scenarios used the same parameters as the \u0026lsquo;dysregulated\u0026rsquo; scenario B, but with health services enabled. The simulation was then run 100 times with stochastic life events. At each time step, we recorded the individual\u0026rsquo;s suicidal capability, desire, and risk.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eStatistical Analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis data forms a set of three-dimensional time series. We conducted k-means clustering on these time series to determine if model behavior was consistent within each scenario. K-means clustering is an unsupervised machine learning algorithm that divides data into groups such that the data within a group is similar to each other and different from data in other groups. [32] To ensure consistent time series lengths, we extended time series shorter than 1,000 days with the capability, desire, and risk values of 100, 400, and 150 respectively, until they reached 1,000 days. Time series may be shorter than 1,000 days when the individual dies by suicide early in the simulation, so these values allow the clustering algorithm to group simulation runs that result in death. We resampled any series longer than 1,000 days to shorten them. We then computed k-mean clusters using the Soft-DTW [33] distance metric with smoothing parameter \u0026gamma;=0.1.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe simulation model was developed using AnyLogic PLE Version 8.9.3, and post-hoc statistical analyses on model output were conducted use R Studio Version 24.12.0, R Version 4.3.1, and Python Version 3.11.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eFollowing a large number of simulations, the model\u0026rsquo;s behavior, in response to varying key parameters, was consistent with expectation. For example, Figure 3 displays the validation results for the \u0026lsquo;coping\u0026rsquo; parameter, which affects the individual\u0026rsquo;s response to life events and moderates the effect of feelings of defeat and humiliation on feelings of entrapment. The validation results show that, as coping abilities increase, there is a resulting decrease in the percentage of simulation runs containing a suicide attempt, as well as lower likelihood of the need for health services beyond primary care (Figure 3). Additionally, with higher coping, the first suicide attempt (if present) takes longer to occur, and the individual ends the simulation with lower suicidal capability, less frequent recurrence of suicide attempts, and suicidal desire fluctuations of a smaller amplitude. The full analysis of all key parameters can be found in Supplementary Materials.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;The simulation model also displayed the hypothesised behavior of the three proposed pathways on the Cusp-Catastrophe model (Figure 4). Pathway A (\u0026lsquo;stable\u0026rsquo;) is characterised by waxing and waning of \u0026lsquo;desire\u0026rsquo; in relation to events and event frequency. Pathway B (\u0026lsquo;dysregulated\u0026rsquo;) is characterised by increased desire and capability in response to moving into a suicide risk state. The increasing capability leads to increasing risk and, eventually, a suicide attempt at day 53. Pathway C (\u0026lsquo;discontinuous\u0026rsquo;) is characterised by pre-existing high capability leading to an immediate attempt in response to moving into a high suicide risk state, reflecting a situation of an impulsive attempt, or behavior without extended ideation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;For an individual on a \u0026lsquo;dysregulated\u0026rsquo; pathway (pathway B), the impact of access to health services on subsequent suicidal ideation or behavior is in Figure 5, with lower (Figure 5A) and higher (Figure 5B) frequency of mental health assessment in two separate simulations. With access to health services enabled at a high assessment frequency of every 7 days (Figure 5A), the individual receives services as soon as they are needed. The individual starts the simulation in primary care. As their suicidal risk increases, they are referred to specialist services. The individual\u0026rsquo;s suicide risk and capability continue to increase, and so they are referred to a psychiatric hospital for stabilization and more intensive interventions. Following a hospital admission, they are referred back to specialist services for maintenance of care, before referral back to primary care following longer-term stabilization of suicidal ideation. Through these health service referrals, the individual\u0026rsquo;s risk becomes steady and low for the remainder of the simulation. In contrast, when access to health services for mental health assessment is less frequent (check-ins every 30 days, instead of 7 days) (Figure 5B), health services do not respond to the rising suicide risk quickly enough, and the individual makes a suicide attempt instead of being referred to more intensive care.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;K-means clustering showed that model behavior was consistent within each scenario, and also substantially different between scenarios (see Supplementary Materials for individual scenario clustering visualizations). Figure 6 contains the results of clustering combining data from 100 simulations of each scenario. Cluster 1 is composed of scenario C (\u0026lsquo;discontinuous\u0026rsquo;) and the scenario B (\u0026lsquo;dysregulated\u0026rsquo;) simulations that resulted in death by suicide (Figure 6, Table 1). Cluster 2 is also composed of scenarios B and C, but captures those simulations that did not result in death by suicide. Cluster 3 represents the stabilized behavior of scenarios A (\u0026lsquo;stable\u0026rsquo;) and the impact of health service access. Finally, Cluster 4 is entirely composed of scenario B, and captures simulations with a death by suicide that occurred later in the time series.\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study has described the development of a simulation model of a \u0026lsquo;suicidal mind\u0026rsquo; using an integration of the IMV model, [18] the Fluid Vulnerability Model of suicidal behavior, [17] and the Cusp-Catastrophe model [20] in order to quantify the dynamic and nonlinear nature of suicidal behavior in an individual. The simulation model displayed the hypothesised behavior of the three proposed pathways to suicidal ideation and suicide attempt in the Cusp-Catastrophe model: the \u0026lsquo;stable\u0026rsquo; pathway characterised by waxing and waning of \u0026lsquo;desire\u0026rsquo; in relation to events and event frequency; the \u0026lsquo;dysregulated\u0026rsquo; pathway characterised by increased desire and capability in response to moving into a suicide risk state; and the \u0026lsquo;discontinuous\u0026rsquo; pathway characterised by high desire in the context of already existing high capability leading to an immediate suicide attempt in response to increased suicide risk. Additionally, parameters relating to motivational and volitional moderating factors that decreased suicide risk were associated with lower suicidal desire; conversely those motivational and volitional moderating factors that increased suicide risk were associated with higher suicidal desire. Comparative scenarios also showed the stabilization of suicidal ideation following access to mental health services, compared to sustained, or deteriorating, suicidal ideation and subsequent suicide attempts where mental health services were not accessed, or were accessed less frequently. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThere are a number of methodological considerations for the current study. Firstly, this study aimed to quantify the hypothesised outcomes of current psychological theories of suicidal behavior in individuals, for which no data sources exist to derive parameter values and outcomes. Thus, parameter values in the model were defined arbitrarily. That is, a \u0026lsquo;score\u0026rsquo; on the scale of \u0026lsquo;suicidal desire\u0026rsquo; is not a meaningful quantity that relates to an empirical biological value (such as output from a sphygmomanometer, or a thermometer). However, patterns over time and relative differences between measures of suicidal ideation, suicidal behavior, capability, and suicide risk behaved in a logically coherent way and were consistent with both the IMV model and the scenarios presented in Bryan et al.\u0026rsquo;s FCT+CC model. For example, increases in the values of parameters relating to protective factors that might prevent an individual from moving from feelings of defeat to entrapment (such as coping strategies, social connection, or access to health services) led to decreases in suicidality. Conversely, decreases in the values of these parameters led to increased frequency of suicidal ideation and suicide attempts. Validation of parameter values was based on 4.2 million simulations, where a range of different values for each parameter in the model, in combination, were tested (see ODD protocol in Supplementary Materials). It is also worth noting that we did not operationalise all pre-motivational, motivational, and volitional moderators from the IMV model. Rather, we selected major indicators from each phase and used combined parameters for the rest (e.g. personal threat-to-self characteristics represented all threat-to-self moderators besides coping ability).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Although not used in our analyses, there are other theories of suicidal behavior that could have formed the basis of the model structure, potentially resulting in different insights than those presented in the current study. The IMV model was selected as the basis of the model structure as it explicitly aims to integrate existing models of suicidal behavior consistent with current research evidence. [18] The Bryan et al.\u0026rsquo;s model using the Fluid Vulnerability Theory and Cusp Catastrophe model [20] was used to capture the non-linear dynamics of suicidal behavior explicitly.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;The defined transitions in the health services sector of the model were also based on a set of referral pathways that are likely to be context specific, and may differ from the idealized \u0026lsquo;stepped care\u0026rsquo; model [24] of mental health service provision used in the current study. Alternative referral pathways and assumptions relating to the effectiveness of services provided in the prevention of suicidal behavior may also result in different insights relating to how health service interactions may change the trajectory of suicidal ideation and behavior. Future work will include investigating the effects of different types of health services and less idealized care on simulated outcomes.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Our model provides a beginning step for other future research to investigate \u0026lsquo;virtual case studies\u0026rsquo; developed in partnership with clinicians and those with lived experience. For example, information from a specific clinical case (or \u0026lsquo;type\u0026rsquo; of clinical case) can be used to convert information about a clinical presentation into a set of parameter values with which to initialize the model. The model can then be used to investigate the potential outcomes of proposed clinical responses, specific types of health service use, different sequences of triggering life events, or other changes to the patient\u0026rsquo;s environment.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;There has been previous use of computational simulation to investigate mechanisms of the human mind, in the areas of cognitive science, neuroscience, [34-35] and the burgeoning area of computational psychiatry. [34] These approaches often aim to identify particular mental health phenotypes using machine learning and artificial intelligence approaches to identify patterns in observational data for risk prediction. [34] To date there have been no applications of simulation modelling to an individual\u0026rsquo;s suicidal behavior. The suicidal mind presented in the current paper does not employ curve-fitting algorithms to model existing observational data; rather, the simulation was guided by causal reasoning and a synthesis of current theory derived from contemporary literature.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;In summary, this study has described the development of a simulation model of a \u0026lsquo;suicidal mind\u0026rsquo; that captures the non-linear dynamics of suicidal behavior based on integrated prevailing psychological theories. Simulation model output was consistent with the different pathways to suicidal behavior that are proposed by current theoretical models of suicidal behavior, and also showed the plausible impact of different patterns of health service access on subsequent suicidal ideation and behavior within an individual. Further validation of the model incorporating wider clinical and lived-experience perspectives, and application to a wider array of different scenarios, may lead to the use of such simulation models to help refine theories of suicidal behavior. There is also potential to use such models in virtual case studies based on the specific characteristics of individuals to potentially assist in clinical decision making, or to estimate how clinical, psychosocial, or population-level policy interventions might affect individual suicidal behavior. \u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eWorld Health Organization, \u0026ldquo;Suicide.\u0026rdquo; Accessed: Apr. 02, 2025. [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/suicide\u003c/li\u003e\n\u003cli\u003eGlobal Burden of Disease Collaborative Network, \u0026ldquo;Global Burden of Disease Study 2021 (GBD 2021),\u0026rdquo; Institute for Health Metrics and Evaluation (IHME), Seattle, United States, 2024. Accessed: Apr. 02, 2025. [Online]. Available: http://vizhub.healthdata.org/gbd-compare\u003c/li\u003e\n\u003cli\u003eA. M. El-Sayed and S. Galea, \u003cem\u003eSystems Science and Population Health\u003c/em\u003e. Oxford University Press, 2017.\u003c/li\u003e\n\u003cli\u003eS. Galea, M. Riddle, and G. A. Kaplan, \u0026ldquo;Causal thinking and complex system approaches in epidemiology,\u0026rdquo; \u003cem\u003eInt. J. Epidemiol.\u003c/em\u003e, vol. 39, no. 1, pp. 97\u0026ndash;106, Feb. 2010, doi: 10.1093/ije/dyp296.\u003c/li\u003e\n\u003cli\u003eR. Schuerkamp, L. Liang, K. L. Rice, and P. J. Giabbanelli, \u0026ldquo;Simulation Models for Suicide Prevention: A Survey of the State-of-the-Art,\u0026rdquo; \u003cem\u003eComputers\u003c/em\u003e, vol. 12, no. 7, Art. no. 7, Jul. 2023, doi: 10.3390/computers12070132.\u003c/li\u003e\n\u003cli\u003eA. Page, J.-A. Atkinson, M. Heffernan, G. McDonnell, and I. Hickie, \u0026ldquo;A decision-support tool to inform Australian strategies for preventing suicide and suicidal behaviour,\u0026rdquo; \u003cem\u003ePublic Health Res. Pract.\u003c/em\u003e, vol. 27, no. 2, 2017, doi: 10.17061/phrp2721717.\u003c/li\u003e\n\u003cli\u003eR. Zahan, J. Mikuliak, and N. D. Osgood, \u0026ldquo;Developing a Tool to Assess the Impact of Simulated Intervention Strategies on Suicide and Suicidal Behaviours in Canada: A Dynamic Modelling \u0026amp; Machine Learning Approach,\u0026rdquo; May 13, 2024. doi: 10.20944/preprints202405.0813.v1.\u003c/li\u003e\n\u003cli\u003eJ.-A. Atkinson \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Systems modelling and simulation to inform strategic decision making for suicide prevention in rural New South Wales (Australia),\u0026rdquo; \u003cem\u003eAust. N. Z. J. Psychiatry\u003c/em\u003e, vol. 54, no. 9, pp. 892\u0026ndash;901, Sep. 2020, doi: 10.1177/0004867420932639.\u003c/li\u003e\n\u003cli\u003eM. Cerd\u0026aacute;, A. D. Hamilton, M. Tracy, C. Branas, D. Fink, and K. M. Keyes, \u0026ldquo;Would restricting firearm purchases due to alcohol- and drug-related misdemeanor offenses reduce firearm homicide and suicide? An agent-based simulation,\u0026rdquo; \u003cem\u003eInj. Epidemiol.\u003c/em\u003e, vol. 9, no. 1, p. 17, Jun. 2022, doi: 10.1186/s40621-022-00381-x.\u003c/li\u003e\n\u003cli\u003eK. M. Keyes, A. Hamilton, J. Swanson, M. Tracy, and M. Cerd\u0026aacute;, \u0026ldquo;Simulating the Suicide Prevention Effects of Firearms Restrictions Based on Psychiatric Hospitalization and Treatment Records: Social Benefits and Unintended Adverse Consequences,\u0026rdquo; \u003cem\u003eAm. J. Public Health\u003c/em\u003e, vol. 109, no. S3, pp. S236\u0026ndash;S243, Jun. 2019, doi: 10.2105/AJPH.2019.305041.\u003c/li\u003e\n\u003cli\u003eK. W. McKinley \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Impact of Universal Suicide Risk Screening in a Pediatric Emergency Department: A Discrete Event Simulation Approach,\u0026rdquo; \u003cem\u003eHealthc. Inform. Res.\u003c/em\u003e, vol. 28, no. 1, pp. 25\u0026ndash;34, Jan. 2022, doi: 10.4258/hir.2022.28.1.25.\u003c/li\u003e\n\u003cli\u003eJ.-A. Occhipinti \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Sound Decision Making in Uncertain Times: Can Systems Modelling Be Useful for Informing Policy and Planning for Suicide Prevention?,\u0026rdquo; \u003cem\u003eInt. J. Environ. Res. Public. Health\u003c/em\u003e, vol. 19, no. 3, Art. no. 3, Jan. 2022, doi: 10.3390/ijerph19031468.\u003c/li\u003e\n\u003cli\u003eA. Page \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;A decision support tool to inform local suicide prevention activity in Greater Western Sydney (Australia),\u0026rdquo; \u003cem\u003eAust. N. Z. J. Psychiatry\u003c/em\u003e, vol. 52, no. 10, pp. 983\u0026ndash;993, Oct. 2018, doi: 10.1177/0004867418767315.\u003c/li\u003e\n\u003cli\u003eA. Skinner, J.-A. Occhipinti, Y. J. C. Song, and I. B. Hickie, \u0026ldquo;Regional suicide prevention planning: a dynamic simulation modelling analysis,\u0026rdquo; \u003cem\u003eBJPsych Open\u003c/em\u003e, vol. 7, no. 5, p. e157, Sep. 2021, doi: 10.1192/bjo.2021.989.\u003c/li\u003e\n\u003cli\u003eK. A. Van Orden, T. K. Witte, K. C. Cukrowicz, S. R. Braithwaite, E. A. Selby, and T. E. Joiner, \u0026ldquo;The interpersonal theory of suicide.,\u0026rdquo; \u003cem\u003ePsychol. Rev.\u003c/em\u003e, vol. 117, no. 2, pp. 575\u0026ndash;600, 2010, doi: 10.1037/a0018697.\u003c/li\u003e\n\u003cli\u003eE. D. Klonsky and A. M. May, \u0026ldquo;The Three-Step Theory (3ST): A New Theory of Suicide Rooted in the \u0026lsquo;Ideation-to-Action\u0026rsquo; Framework,\u0026rdquo; \u003cem\u003eInt. J. Cogn. Ther.\u003c/em\u003e, vol. 8, no. 2, pp. 114\u0026ndash;129, Jun. 2015, doi: 10.1521/ijct.2015.8.2.114.\u003c/li\u003e\n\u003cli\u003eM. D. Rudd, \u0026ldquo;Fluid Vulnerability Theory: A Cognitive Approach to Understanding the Process of Acute and Chronic Suicide Risk.,\u0026rdquo; in \u003cem\u003eCognition and suicide: Theory, research, and therapy.\u003c/em\u003e, T. E. Ellis, Ed., Washington: American Psychological Association, 2006, pp. 355\u0026ndash;368. doi: 10.1037/11377-016.\u003c/li\u003e\n\u003cli\u003eR. C. O\u0026rsquo;Connor and O. J. Kirtley, \u0026ldquo;The integrated motivational\u0026ndash;volitional model of suicidal behaviour,\u0026rdquo; \u003cem\u003ePhilos. Trans. R. Soc. B Biol. Sci.\u003c/em\u003e, vol. 373, no. 1754, p. 20170268, Jul. 2018, doi: 10.1098/rstb.2017.0268.\u003c/li\u003e\n\u003cli\u003eH. Hjelmeland and B. and Loa Knizek, \u0026ldquo;The emperor\u0026rsquo;s new clothes? A critical look at the interpersonal theory of suicide,\u0026rdquo; \u003cem\u003eDeath Stud.\u003c/em\u003e, vol. 44, no. 3, pp. 168\u0026ndash;178, Mar. 2020, doi: 10.1080/07481187.2018.1527796.\u003c/li\u003e\n\u003cli\u003eC. J. Bryan \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Nonlinear change processes and the emergence of suicidal behavior: A conceptual model based on the fluid vulnerability theory of suicide,\u0026rdquo; \u003cem\u003eNew Ideas Psychol.\u003c/em\u003e, vol. 57, p. 100758, Apr. 2020, doi: 10.1016/j.newideapsych.2019.100758.\u003c/li\u003e\n\u003cli\u003eJ. W. Forrester, \u0026ldquo;Industrial Dynamics,\u0026rdquo; \u003cem\u003eJ. Oper. Res. Soc.\u003c/em\u003e, vol. 48, no. 10, pp. 1037\u0026ndash;1041, Oct. 1997, doi: 10.1057/palgrave.jors.2600946.\u003c/li\u003e\n\u003cli\u003eK. Michel and L. Valach, \u0026ldquo;Suicide as goal-directed action,\u0026rdquo; \u003cem\u003eArch. Suicide Res.\u003c/em\u003e, vol. 3, no. 3, pp. 213\u0026ndash;221, Jul. 1997, doi: 10.1080/13811119708258273.\u003c/li\u003e\n\u003cli\u003eG. Dedic, S. Djurdjevic, and B. Golubovic, \u0026ldquo;Psychological assessment of persons following suicide attempt by self-poisoning,\u0026rdquo; \u003cem\u003eVojnosanit. Pregl.\u003c/em\u003e, vol. 67, no. 2, pp. 151\u0026ndash;158, 2010, doi: 10.2298/VSP1002151D.\u003c/li\u003e\n\u003cli\u003eS. Cross and I. Hickie, \u0026ldquo;Transdiagnostic stepped care in mental health,\u0026rdquo; \u003cem\u003ePublic Health Res. Pract.\u003c/em\u003e, vol. 27, no. 2, 2017, doi: 10.17061/phrp2721712.\u003c/li\u003e\n\u003cli\u003eV. Grimm \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;The ODD Protocol for Describing Agent-Based and Other Simulation Models: A Second Update to Improve Clarity, Replication, and Structural Realism,\u0026rdquo; \u003cem\u003eJ. Artif. Soc. Soc. Simul.\u003c/em\u003e, vol. 23, no. 2, p. 7, 2020.\u003c/li\u003e\n\u003cli\u003eC. J. Bryan, J. E. Butner, S. Sinclair, A. B. O. Bryan, C. M. Hesse, and A. E. Rose, \u0026ldquo;Predictors of Emerging Suicide Death Among Military Personnel on Social Media Networks,\u0026rdquo; \u003cem\u003eSuicide Life. Threat. Behav.\u003c/em\u003e, vol. 48, no. 4, pp. 413\u0026ndash;430, 2018, doi: 10.1111/sltb.12370.\u003c/li\u003e\n\u003cli\u003eC. J. Bryan, M. D. Rudd, A. L. Peterson, S. Young-McCaughan, and E. G. Wertenberger, \u0026ldquo;The ebb and flow of the wish to live and the wish to die among suicidal military personnel,\u0026rdquo; \u003cem\u003eJ. Affect. Disord.\u003c/em\u003e, vol. 202, pp. 58\u0026ndash;66, Sep. 2016, doi: 10.1016/j.jad.2016.05.049.\u003c/li\u003e\n\u003cli\u003eE. M. Kleiman, B. J. Turner, S. Fedor, E. E. Beale, J. C. Huffman, and M. K. Nock, \u0026ldquo;Examination of real-time fluctuations in suicidal ideation and its risk factors: Results from two ecological momentary assessment studies,\u0026rdquo; \u003cem\u003eJ. Abnorm. Psychol.\u003c/em\u003e, vol. 126, no. 6, pp. 726\u0026ndash;738, 2017, doi: 10.1037/abn0000273.\u003c/li\u003e\n\u003cli\u003eT. K. Witte, K. K. Fitzpatrick, T. E. Joiner, and N. B. Schmidt, \u0026ldquo;Variability in suicidal ideation: A better predictor of suicide attempts than intensity or duration of ideation?,\u0026rdquo; \u003cem\u003eJ. Affect. Disord.\u003c/em\u003e, vol. 88, no. 2, pp. 131\u0026ndash;136, Oct. 2005, doi: 10.1016/j.jad.2005.05.019.\u003c/li\u003e\n\u003cli\u003eC. J. Bryan, D. C. Rozek, J. Butner, and M. D. Rudd, \u0026ldquo;Patterns of change in suicide ideation signal the recurrence of suicide attempts among high-risk psychiatric outpatients,\u0026rdquo; \u003cem\u003eBehav. Res. Ther.\u003c/em\u003e, vol. 120, p. 103392, Sep. 2019, doi: 10.1016/j.brat.2019.04.001.\u003c/li\u003e\n\u003cli\u003eC. J. Bryan and M. D. Rudd, \u0026ldquo;Nonlinear Change Processes During Psychotherapy Characterize Patients Who Have Made Multiple Suicide Attempts,\u0026rdquo; \u003cem\u003eSuicide Life. Threat. Behav.\u003c/em\u003e, vol. 48, no. 4, pp. 386\u0026ndash;400, Aug. 2018, doi: 10.1111/sltb.12361.\u003c/li\u003e\n\u003cli\u003eJ. Wu, \u0026ldquo;Cluster Analysis and K-means Clustering: An Introduction,\u0026rdquo; in \u003cem\u003eAdvances in K-means Clustering: A Data Mining Thinking\u003c/em\u003e, J. Wu, Ed., Berlin, Heidelberg: Springer, 2012, pp. 1\u0026ndash;16. doi: 10.1007/978-3-642-29807-3_1.\u003c/li\u003e\n\u003cli\u003eM. Cuturi and M. Blondel, \u0026ldquo;Soft-DTW: a Differentiable Loss Function for Time-Series,\u0026rdquo; in \u003cem\u003eProceedings of the 34th International Conference on Machine Learning\u003c/em\u003e, D. Precup and Y. W. Teh, Eds., in Proceedings of Machine Learning Research, vol. 70. PMLR, Aug. 2017, pp. 894\u0026ndash;903. [Online]. Available: https://proceedings.mlr.press/v70/cuturi17a.html\u003c/li\u003e\n\u003cli\u003eP. F. Hitchcock, E. I. Fried, and M. J. Frank, \u0026ldquo;Computational Psychiatry Needs Time and Context,\u0026rdquo; \u003cem\u003eAnnu. Rev. Psychol.\u003c/em\u003e, vol. 73, no. Volume 73, 2022, pp. 243\u0026ndash;270, Jan. 2022, doi: 10.1146/annurev-psych-021621-124910.\u003c/li\u003e\n\u003cli\u003eC. Langley, B. I. Cirstea, F. Cuzzolin, and B. J. Sahakian, \u0026ldquo;Theory of Mind and Preference Learning at the Interface of Cognitive Science, Neuroscience, and AI: A Review,\u0026rdquo; \u003cem\u003eFront. Artif. Intell.\u003c/em\u003e, vol. 5, Apr. 2022, doi: 10.3389/frai.2022.778852.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table 1","content":"\u003cp\u003e\u003cstrong\u003eTable 1: Cluster scenario-composition for clusters shown in Figure 7\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCluster\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eScenario A\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eScenario B\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eScenario C\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHealth service\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e0.7%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e29.6%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e67.4%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e2.2%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e0.0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e68.8%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e28.1%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e3.1%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e50.8%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e0.0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e0.0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e49.2%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e\u003cstrong\u003e4\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e0.0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e100.0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e0.0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 20%;\"\u003e\n \u003cp\u003e0.0%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6605926/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6605926/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Suicidal behaviors are characterised by complex, multi-factorial aetiology. Dynamic simulation models (DSMs) are computational approaches that can explicitly capture salient dimensions of complex suicidal thinking and behavior. This study describes the development of a DSM of a ‘suicidal mind’ to capture nonlinear, dynamic changes in suicidal states based on an integration of the (i) Integrated Motivational-Volitional (IMV) model of suicide, (ii) Fluid Vulnerability Theory of suicide, and (iii) Cusp-Catastrophe model from dynamical systems theory. A system-dynamics model was developed to estimate the level of ‘suicidality’ in an individual over time, capturing cognitions and behaviors with transitions between the ‘pre-motivational’, ‘motivational,’ and ‘volitional’ phases in the IMV model. Validation of the DSM and the underlying theoretical synthesis consisted of testing whether parameter changes - hypothesised to increase or decrease the level of suicidality - resulted in corresponding effects in the model, and whether the DSM displayed the expected nonlinear pathways to suicidal states. The model’s behavior in response to varying parameters of interest was consistent with expectations. The model could recreate the ‘stable,’ ‘dysregulated,’ and ‘discontinuous’ nonlinear pathways proposed in prior research supporting the validity of the DSM and its underlying theoretical synthesis. Mental health service access resulted in stabilization of suicidal ideation, but the effect varied by frequency of contact. This model demonstrates that DSMs can quantify and refine theories of suicidal behavior, which suggests the potential for using DSMs in virtual case studies to assist clinical decision making and training, or to investigate population-level interventions.","manuscriptTitle":"Simulation of a ‘suicidal mind’: Using the Integrated Motivational Volitional model of suicide to demonstrate dynamic suicidal states","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-29 14:49:19","doi":"10.21203/rs.3.rs-6605926/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-mental-health","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"natmentalhealth","sideBox":"Learn more about [Nature Mental Health](https://www.nature.com/natmentalhealth/)","snPcode":"44220","submissionUrl":"https://mts-natmentalhealth.nature.com/cgi-bin/main.plex","title":"Nature Mental Health","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Research","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"23d702c3-cf61-45f7-ac7c-fae1d543f870","owner":[],"postedDate":"June 29th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":50564622,"name":"Biological sciences/Psychology/Human behaviour"},{"id":50564623,"name":"Health sciences/Health care/Public health/Epidemiology"},{"id":50564624,"name":"Health sciences/Health care/Health policy"},{"id":50564625,"name":"Health sciences/Risk factors"},{"id":50564626,"name":"Health sciences/Diseases/Psychiatric disorders"}],"tags":[],"updatedAt":"2025-06-29T14:49:19+00:00","versionOfRecord":[],"versionCreatedAt":"2025-06-29 14:49:19","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6605926","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6605926","identity":"rs-6605926","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}