Value-based decision-making and non-suicidal self-injury in adolescent females: A dimensional and computational...

Non-suicidal self-injury (NSSI) commonly emerges during adolescence, conferring elevated psychiatric burden and risk for later suicide attempts. Contemporary models conceptualize NSSI as an affectively driven decision-making process. While measurement approaches to date have relied on categorical operationalizations of NSSI and broad indices of value-based decision-making, computational decomposition of behavioral tasks have promise for identifying latent cognitive processes underlying decision-making processes in relation to adolescent NSSI.

107 adolescents assigned female at birth (mean age = 15.6 years) completed a tangibly-incentivized Iowa Gambling Task (IGT) and the Self-Injurious Thoughts and Behaviors Interview. A series of regressions examined associations between lifetime and past-year NSSI frequency and aggregate IGT performance, in addition to five computationally derived parameters using hierarchical Bayesian Outcome-Representation Learning modeling - reward and punishment learning rates, memory decay, choice persistence, and win-frequency sensitivity.

Poorer aggregate IGT performance was associated with greater lifetime (IRR = 0.47, 95% CI [0.32, 0.71]) and past-year (IRR = 0.40, 95% CI [0.20, 0.80]) NSSI frequency. Parameter-level analyses revealed that reduced punishment learning rate (IRR = 0.13, 95% CI [0.06, 0.31]) and choice persistence (IRR = 0.56, 95% CI [0.36, 0.85]) were associated with greater past-year NSSI frequency. These associations largely survived adjustment for age, depression, and intelligence.

Impaired value-based decision-making is dimensionally associated with greater NSSI engagement, with reduced punishment-based learning and choice persistence emerging as preliminary cognitive markers. With further research, these computational targets may eventually inform early identification and intervention efforts.