In a distributed Clinical Decision Support environment, a decision model is usually defined and coupled with a group of decision rules that together drive the decision support process. These can then be applied to patient-specific data to provide customised clinical advices. Being useful and able to deliver improved clinical service, such systems are often closed and thus lead to the difficulties in keeping aligned with up-to-date best evidences, interoperating with local Electronic Health Records (EHRs) systems, and advising in multi-disciplinary decisions whereas disparate participants work towards the same decision goals. In this paper, a semantic-enabled agent-oriented distributed clinical decision model is proposed. Group decisions are modelled as goal trees, which can be managed altogether and later chosen, decomposed and distributed to appropriate decision participants. Individual decisions are modelled in the form of RDF-based argumentation rules, which can be universally interpreted by a Rule Engine (RE) against the local EHRs. An Interactive Decision Interface Engine (IDIE) is developed to guide local decisions step by step and a Choreographer Agent (CA) synthetizes decisions among a collaborative group in a distributed environment. Ultimately, specialists can easily maintain decision knowledge in open and shared triple stores, and decision makers can better understand and investigate decision recommendation rationale using knowledge graphs that are intimately integrated within their decision-making environment. We demonstrate the idea with a case study of triple assessment of breast cancer.