Due to the decline in physical and cognitive abilities, many frail elderly may have to lie in the bed most of their time. It is not feasible to monitor them continuously through manual observations alone. This issue can be resolved by embedding a set of multimodal sensors into the bed and providing automated activity recognition intelligence. But it is important to design and develop such multimodal sensing intelligence system desirable to the demands made by the clinicians. This paper presents the comparison and evaluation of different sensing bed configurations to observe different granularities of patient's contexts and activities in and around the bed. Based on the achievements and lessons learned from the experimental analysis, we propose improved sensing bed hardware and software systems to meet the real needs of in and around the bed patient monitoring.