Determination of patients’ requirements (PRs) importance is a critical issue of medical service design. Quality function deployment (QFD) is one of the very effective customer-driven quality system tools typically applied to fulfil PRs. Often the patients cannot easily express their judgements on PRs’ importance with exact numerical values, and they usually present their judgements in different scales. Therefore, this paper aims at providing a systematic method to simultaneously deal with PRs’ fuzziness and multi-granularity in the QFD. Compared to previous research, its contribution is threefold. First, it proposes use of the 2-tuple linguistic model which can effectively manage the imprecise and vague evaluation information in QFD because of its accuracy and no information loss. Second, it develops a new 2-tuple transformation function to solve the unification problem of multi-granular linguistic judgements in QFD. Third, it proposes a qualitative approach to deal with uncertainty existed extensively in establishing modified factors. Finally, a practical case of hospital PRs is provided to illustrate its feasibility and effectiveness of the proposed methodology.