Motivated by the effectiveness of condition-based maintenance (CBM) for single-unit systems, we develop in this paper an opportunistic predictive maintenance model for a k-out-of-n deteriorating system. This model reflects the joint effects of economic dependence and CBM decision on the maintenance cost. Unlike most existing CBM models whose maintenance decision-making relies directly on a condition index of degradation level, we base the decision on the reliability of each component computed conditionally on its degradation level. This makes the model robust with respect to (w.r.t.) measurement errors, more efficient and easier to optimize compared to a corresponding degradation-based maintenance model.