Patients fall out of their beds frequently in hospitals, and such incidents are a critical issue because they often lead to serious accidents. Accidents of patient falling out of beds have increased with increasing numbers of the hospitalized elderly. Although bed sensors are commonly used as a fall-preventive measure in hospitals, these sensors cannot prevent accidents. We hypothesized that if bed sensors can monitor patients,behavioral changes continuously like vital sensors, nurses will be able to understand patient behaviors from the recorded data and provide appropriate care, thus decreasing patient falls. A three-dimensional measurement system was then developed. With this system, the position, movement, direction, and velocity of patients can be measured in a three-dimensional space in a non-contact, noninvasive manner. Using this system, we measured long-term behaviors of patients in bed. The analysis results allowed us to identify the behavior changes leading to a patient,s fall from bed. In addition, we quantified how behavioral changes in acute-phase patients change over several weeks after hospitalization. We believe that these quantified data will be useful in daily patient care and in evaluation of nursing-care plans.