ABSTRACT: Background: Upper limb functional exercise (ULFE) has a positive effect on promoting the rehabilitation of upper limb function. However, little is known, about what drives postoperative patients to engage in and even maintain the advised exercises. This study integrated the health action process approach (HAPA) and the theory of planned behavior theory (TPB) to investigate the psychosocial determinants on the initiation and maintenance of ULFE in breast cancer patients. In addition, this study also tests key hypotheses relating to reasoned and implicit pathways to ULFE and its maintenance among postoperative patients with breast cancer. Methods: Purposive sampling was used to recruit patients from two breast cancer wards in a provincial hospital in Jiangsu, China. Patients (N = 430) completed self-reported questionnaire about constructs from integrated theories concerning ULFE at an initial time point (T1): task self-efficacy, positive outcome expectations, negative outcome expectations, risk perception, attitude behavior, subjective norm, perceived behavioral control, behavioral intention, and ULFE-in hospital (ULFE-IH). Three months later (T2), patients self-reported: maintenance self-efficacy, action planning, coping planning, recovery self-efficacy, and ULFE-maintenance (ULFE-M). Results: The model has a good fit (GoF = 0.48). For behavioral intention of ULFE, subjective norm (β = 0.35) and perceived behavioral control (β = 0.61) were positively directly related to behavioral intention. Regarding the initiation of ULFE, perceived behavioral control (β = 0.47) and behavioral intention (β = 0.42) had a direct positive relation to ULFE-IH. In the maintenance of ULFE, action planning (β = 0.30), coping planning (β = 0.21), maintenance self-efficacy (β = 0.32), and recovery self-efficacy (β = 0.09) all had significant positive relation on ULFE-M. In addition, maintenance self-efficacy had a significant positive association on action planning (β = 0.80), coping planning (β = 0.74), and recovery self-efficacy (β = 0.67). Coping planning was significantly predicted by behavioral intention (β = 0.07). Additionally, behavioral intention is a mediator of subjective norm (β = 0.14) and perceived behavioral control (β = 0.25) to ULFE-IH. Action planning, coping planning, and recovery self-efficacy are mediators of maintaining self-efficacy to ULFE-M (β = 0.46). Conclusions: This study presents the first attempt to integrate the health behavior model in ULFE in postoperative patients with breast cancer. The study has shown that the HAPA-TPB integrated model has good applicability and effectiveness to explain and predict ULFE initiation and maintenance. Future work can be considered to develop appropriate intervention strategies based on this integrated behavioral theory.