Aims To test prospective pathways of a Comprehensive Reminder System based on the Health Belief Model (CRS-HBM), stroke knowledge, health belief in health behaviour, blood pressure (BP) control, and disability in hypertensive ischaemic stroke patients at 6-month postdischarge. Design A nested cohort study design. Methods Data were derived from a randomized controlled trial evaluating the effects of the intervention (N = 174, performed during February 2015 - March 2016). Data were collected by questionnaires and analysed in structural equation modelling in Mplus software. Results The proposed model provided a good fit to the data. This model accounted for 51.5% of the variance in health behaviour, 34.1% in BP control, and 5.7% in modified Rankin Scale score at 6-month postdischarge. The CRS-HBM had: (a) direct positive effect (beta = .391, p < .001) and indirect positive effects (beta = .186, p = .002) on health behaviour; (b) direct positive effect (beta = .356, p < .001) and indirect positive effects (beta = .183, p = .009) on BP control; and (c) indirect negative effect (beta = -.146, p = .008) on disability. Being female was linked to better health behaviour. Higher education predicted higher level of stroke knowledge and health belief. Conclusions The CRS-HBM can not only directly but also indirectly improve patients' health behaviours by improving their health knowledge or health belief. Better health behaviour can improve patients' BP control and reduce disability. Therefore, nurses need to pay more attention to not only patients' health knowledge but also their health belief when providing education. Impact The CRS-HBM intervention accounted for 51.5% of variance in health behaviour, 34.1% in BP control, and 5.7% in modified Rankin Scale score at 6-month postdischarge. This research can help nurses improve health education strategies in postdischarge and community contexts to achieve better health results.