Within-person associations between psychological and contextual factors and dietary lapse in adults: a machine learning–assisted systematic review and meta-analysis

Aims To systematically review and meta-analyze the evidence from ecological momentary assessment (EMA) studies on how psychological and contextual factors influence dietary lapses in adults striving for healthy eating habits. Background Dietary lapses, deviations from intended eating patterns, are influenced by both psychological and contextual factors. While psychological aspects like stress and negative emotional states have been widely studied, the role of contextual influences, such as food availability and social environments, remains underexplored. EMA offers a real-time, context-sensitive approach to better understanding these lapses. Methods We conducted a comprehensive search across Ovid MEDLINE, Embase, PsycINFO, and Web of Science, employing machine learning to enhance the efficiency and accuracy of literature screening. This search aimed to identify studies that utilized EMA to investigate the predictors of dietary lapses. Our analysis included a qualitative synthesis of the definitions of “lapse” and “relapse” used in the studies, as well as the theoretical frameworks underpinning EMA study designs. Results Eighty-seven articles were selected for the systematic review, and 53 articles were eligible for meta-analysis. Our analysis revealed a positive correlation between the occurrence of dietary lapses and negative emotional states (OR = 1.12, 95% CI = 1.07-1.18, P < .0001, I2 = 3.0%), environmental and social influences (OR = 1.07, 95% CI = 1.00-1.55, P = .048, I2 = 0%), and the craving (OR = 1.32, 95% CI = 1.11-1.55, P < .001, I2 = 21.5%). In addition, we found that the definition of “lapse” was relatively consistent across studies, while the definition of “relapse” varied significantly across studies. Meanwhile, although most studies mentioned at least 1 psychological theory to guide the assessment of psychological or situational variables, the theory was less used in the specific application of determining EMA frequency and study duration. Conclusions This study underscores the complexity of dietary behavior and the multifaceted influences that lead to lapses in healthy eating. By highlighting the significant predictors of dietary lapses, our findings provide valuable insights for developing targeted interventions that support individuals in achieving healthy eating habits. In addition, the application of machine learning in literature screening represents an innovative approach that could be utilized in systematic reviews and meta-analyses in the future to enhance the efficiency and outcomes of studies.