Rapid and Reliable Computational Markers for Predicting Daily Smoking Behavior and Smoking Cessation Treatment Outcome

• DOI: 10.31234/osf.io/bj7fd_v4
• Publication Date: 2025-03-31T19:35:38Z
• Authors: Jeung-Hyun Lee; Sang Ho Lee; Jaeyeong Yang; Hyeonjin Kim; Mark Pitt; Hyung Jun Park; Hee-Kyung Joh; Anna B. Konova; Woo-Young Ahn

Nicotine addiction is a complex disorder shaped by factors such as craving, mood, and neurocognitive processes. While ecological momentary assessment (EMA) provides real-time method for capturing dynamic changes in behavior, traditional tasks surveys are often too lengthy demanding repeated use clinical settings. Integrating EMA with computational approaches offers promising solution to predict smoking behavior dynamically while addressing the practical limitations of conventional assays, paving way more effective scalable interventions. To evaluate predictive value markers derived from decision-making data short-term (daily behavior) long-term (cessation success) outcomes, assess timing amount collection needed prediction. 79 daily smokers (mean age 25.64 years, 83% male) took part longitudinal experimental study involving psychological states tasks, delivered via smartphone app, undergoing 5–6 week cessation program. Using machine-learning methodology (adaptive design optimization, ADO) effectively generate task variables, we estimated just 20 30 trials per day, reducing length participants burden. A time-lagged model incorporating both self-reported provided most accurate prediction next-day behavior. Higher levels depression, ambiguity tolerance on previous day were significantly increased following day. Smoking status at end treatment was strongly predicted lower discounting rates, reduced craving stress, longer history, greater engagement (AUC = 0.76). Notably, models based collected during first follow-up, either 0.74) or variables 0.73), demonstrated comparable accuracy end-of-treatment cessation. Combining efficient approach predicting success holds promise applications.