1 - (IOP 23) Continuous-time Modeling of the Transition from Suicidal Ideation to Planning: Examining Mediating Risk Factors

Introduction
In 2023, 12.8 million Americans had serious suicidal thoughts, 3.7 million created suicide plans, and 1.1 million with both thoughts and plans attempted suicide. As nearly one-third of planners attempt suicide, understanding the suicidal ideation (SI) to suicidal planning (SP) relationship is critical for timely intervention. Our understanding of this transition and its timeline remains limited – likely due to methodological limitations such as overreliance on retrospective designs and discrete time models in ecological momentary assessment (EMA) studies. The latter limits findings to time lags of 3-5 hours while the transition from ideation to plan can occur within a shorter time frame. Thus, it is crucial to develop models that can detect risk at a finer temporal resolution. This study addresses this gap by modeling the SI-SP relationship in a continuous time framework and exploring potential mediating risk factors.

Methods
Participants (N = 74) with a past-month history of suicidality completed 28-days of EMA, receiving six daily surveys. Each EMA prompt assessed negative affect (NA), active SI, perceived burdensomeness (PB), and thwarted belongingness (TB). Continuous time residual dynamic structural equation modelling (CT-RDSEM) was used to investigate the relationship between SI and SP mediated by three potential risk factors: NA, PB, and TB. Effects were estimated via drift rates, which capture the influence of a variable on the instantaneous rate of change of another variable, where negative effects indicate resistance to shifts away from equilibrium (e.g., reducing rate of change). Drift rates were exponentiated to obtain autoregressive curves.

Results
CT-RDSEM revealed significant direct and indirect effects for all three models. Active SI had a significant direct effect on SP with NA, PB, and TB as mediators (c' = -0.029, -0.064, and -0.116) that lasted 0.5, 3.0, and 5.5 hours, respectively. The NA model revealed a significant indirect effect (ab = 0.009), where active SI had a negative effect on NA (a = -0.095) that persisted for 5.5 hours (-0.029 to -0.002), and NA had a negative effect on SP (b = -0.085) that persisted for 6.0 hours (-0.026 to -0.001). The PB model revealed a significant indirect effect (ab = 0.007), where active SI had a negative effect on PB (a = -0.118) that persisted for 6.5 hours (-0.036 to -0.001), and PB had a negative effect on SP (b = -0.061) that lasted 7.0 hours (-0.019 to -0.002). Lastly, the TB model revealed a significant indirect effect (ab = 0.006), where active SI had a negative effect on TB (a = -0.090) that lasted 5.0 hours (-0.028 to -0.003), and TB had a negative effect on SP (b = -0.072) that also persisted for 5.0 hours (-0.022 to -0.003).

Discussion
Negative drift rates across all models indicate that active SI and mediating risk factors (NA, TB, and PB) resist shifts away from equilibrium SP states, suggesting that these factors may play a more prominent role in making it difficult to transition out of SP, as opposed to causing SP. Direct effects of active SI on SP dissipated more rapidly (0.5, 3, and 5.5 hours) than indirect effects through mediators (6, 7, and 5.5 hours), suggesting that NA, TB, and PB may prolong the duration of the effects – a finding that discrete time modelling would not capture.