The shortage of inpatient beds is a major cause of delays and cancellations in many hospitals. It may also lead to patients being admitted to inappropriate wards, resulting in a lower quality of care and a longer length of stay. Academic/practical relevance: Investment in additional beds is not always feasible. Instead, new and creative solutions for a more efficient use of existing resources must be sought. Methodology: We propose a new configuration of inpatient beds, which we call the clustered overflow configuration. In this configuration, patients who are denied admission to their primary wards as a result of beds being fully occupied are admitted to overflow wards, with each designated to serve overflows from a certain subset of specialties and providing the same quality of care as in primary wards. We propose two different formulations for partitioning and bed allocation in the proposed configuration: one minimizing the sum of average daily costs of turning patients away and nursing teams, and another minimizing the numbers turned away subject to nursing cost falling below a given threshold. We heuristically solve instances from both formulations.