With the number of patients with chronic diseases increasing, a hierarchical diagnosis and treatment system is a feature of long-term treatment. During the treatment, if patients with chronic diseases are relieved, they can be referred to community hospitals for rehabilitation treatment, and to promote the rational allocation of hospital medical resources. Therefore, hospitals need referral standards to encourage patients to refer down. Existing ranking research which is mainly based on referral criteria,has not considered the low reliability due to different opinions within each group when making referral standard. In view of this, we propose a probabilistic linguistic-multiple criteria decision-making (PL-MCDM) method based on the decision-making reliability of multi-group for patients with chronic diseases requiring downward referral. First, the decision information of different groups on referral criteria is expressed as probabilistic linguistic preferences, and then the reliability levels of groups are measured when setting standards respectively. After that, the decision information and referral standards are re-adjusted for the least reliable group to improve the whole reliability. Secondly, the patients requiring downward referral are ranked by the similarity measure based on statistics of probability preference to acquire the formal referral standards. Finally, compared with the methods without reliability improvement, the real case of coronary heart disease in First Hospital of Qinhuangdao is used to verify the effectiveness of the proposed method, which illustrates its superiority.

随着慢性病患者数量的增加，分级诊疗体系是长期治疗的一个特点。在治疗过程中，慢性病患者如病情缓解，可转诊至社区医院进行康复治疗，促进医院医疗资源的合理配置。因此，医院需要转诊标准来鼓励患者转诊。现有的排名研究主要基于推荐标准，在制定推荐标准时没有考虑到由于各组内部意见不同而导致的低信度。鉴于此，我们针对需要向下转诊的慢性病患者提出了一种基于多组决策可靠性的概率语言多标准决策（PL-MCDM）方法。第一的，将不同群体在推荐标准上的决策信息表达为概率语言偏好，然后分别在制定标准时衡量群体的可靠性水平。之后，对最不可靠的组重新调整决策信息和推荐标准，以提高整体可靠性。其次，通过基于概率偏好统计的相似性度量对需要向下转诊的患者进行排序，以获得正式的转诊标准。最后，与未提高可靠性的方法相比，以秦皇岛市第一医院冠心病的真实病例为例，验证了所提方法的有效性，说明了其优越性。