6 - (IOP 14) From Repetition to Relevance: Initial Development of a Computerized Adaptive Test for Intensive Longitudinal Assessment of Suicide Risk

Suicide is a major public health concern. Research strongly indicates that 1) risk for suicide is multidimensional (e.g., O’Connor & Kirtley, 2018) and 2) fluctuations in suicidal thoughts and risk factors can occur within short time intervals (e.g., hours; Coppersmith et al., 2023). However, comprehensive assessment of suicide risk factors in intensive longitudinal designs like ecological momentary assessment (EMA) is limited in existing research due to concerns of participant burden. To overcome this limitation, the present work proposes the computerized adaptive test for suicide risk pathways (CAT-SRP), which leverages multidimensional computerized adaptive testing (MCAT), to support simultaneous assessment of twelve empirically supported domains of suicide risk; passive and active suicidal ideation (SI) are also assessed. The CAT-SRP uses variable length MCAT algorithms to increase measurement efficiency and support measurement on a continuous (i.e., latent) scale that may alleviate difficulties associated with SI measurement in EMA (e.g., zero-inflation). CAT-SRP items were adapted for EMA from previously validated instruments based on theoretical models of suicide (Joiner, 2005; Klonsky & May, 2015; O’Connor & Kirtley, 2018) and consultation with area experts. Proposed CAT-SRP domains included hopelessness, thwarted belongingness, perceived burdensomeness, loneliness, defeat and entrapment, humiliation, psychological pain, anger and aggression, perceived stress, impulsivity (i.e., negative urgency), distress tolerance, passive SI, and active SI). A community sample (N = 1759, 54.41% Female, 65.32% White, 36.33% with lifetime SI history) was studied to evaluate the proposed CAT-SRP structure and calibrate CAT-SRP item parameters using a multidimensional graded response model. Empirically derived domains were generally consistent with the hypothesized CAT-SRP structure. Following calibration, the CAT-SRP was implemented in an EMA study using a sample reporting past month suicidal thoughts (N = 29, 62.1% Female, 69.0% White). Surveys were administered 6 times per day for 21 days resulting in 2,134 completed surveys. Findings suggested that the CAT-SRP administered surveys with low to moderate content overlap (mean Overlap Coefficient = 0.37) and high participant acceptability (89.4% would participate again). Intraclass correlation coefficients for domain scores ranged from .39 to .63 suggesting nontrivial within-person variability. Additionally, CAT-SRP risk domains accounted for 52.1% of variance in concurrent passive SI and 18.4% of variance in next timepoint passive SI (excluding lagged effect of SI); similar findings were observed for active SI (R2 = 0.392, R2 = 0.093). Excluding aggression (ps > 0.08), a fixed effect was found for each risk domain in at least one of the four models (ps < .05); lack of distress tolerance was present in all models (bs = 0.04 – 0.14, ps = .04 - <.001). Overall, these studies highlight the promise of the CAT-SRP for measuring multiple suicide risk factors simultaneously in EMA designs. As the first multidimensional CAT for intensive longitudinal data, the CAT-SRP represents a significant methodological advance in ambulatory assessment.