The heart rate during atrial fibrillation (AF) is highly dependent on the conduction properties of the atrioventricular (AV) node, which can be affected using β-blockers or calcium channel blockers, often chosen empirically. Thus, characterization of the AV nodal conduction properties could contribute to personalized treatment of AF. We have created a mathematical network model of the AV node where continuous estimation of the refractory period and conduction delay from 24-hour ambulatory ECGs from patients with permanent AF (n=59) was achieved using a problem-specific genetic algorithm. Circadian variations in the resulting model parameter trends were quantified using cosinor analysis, and differences between treatment with β-blockers and calcium blockers were assessed using a linear-mixed effect approach. The mixed-effects analysis indicated increased refractoriness relative to baseline for all drugs. For the β-blockers, an additional decrease in circadian variation for parameters representing conduction delay was observed. This indicates that the two drug types have quantifiable differences in their effects on AV-nodal conduction properties. The proposed method enables analysis of circadian variation in AV node conduction delay and refractoriness from 24h ambulatory ECG, which can be used to monitor and possibly predict the effect of rate control drugs.