Crop health management systems can be designed according to practices that help to reduce crop losses by restricting pathogen development and promoting host plant growth. A good understanding of pathogen and host dynamics, which are interdependent, is therefore needed. In this paper, we used a holistic approach to explain the behavior of coffee leaf rust (CLR), a major coffee disease. We monitored coffee plant and CLR dynamics simultaneously in plots under different disease management and agroforestry systems. Diseased leaves were also collected to characterize rust life stages (latent rust area, area with uredospores, necrosis due to rust) through picture analysis and inoculum stock. We used structural equation modeling to obtain an overview of CLR pathosystem functioning on a plant scale. This overview integrates processes such as disease dilution by host leaf renewal, direct and indirect effects of fruit load on CLR development, antagonistic effects of shading depending on rust life stages, the tonic effect of copper-based fungicides on leaf retention, and effects on rust life stages depending on fungicide types. From our results, we also deduced that the inoculum stock could be calculated in unsprayed plots from the rust area with uredospores, giving 58.10³ uredospores cm^-2, on average.