A server computer system screens for a health state in a human patient by applying a composite model that combines language, acoustic, and visual models, to a captured audiovisual signal of the patient's speech. Accuracy of the screening is improved by measuring the quality of a number of candidate questions and presenting those with the highest quality to the patient. Repeated screenings are kept both fresh and consistent by substituting equivalent questions. Multi-dimensional vectors representing the patient and screening score are used to automatically adapt health care of the patient to achieve goals.
展开▼