The problem of assessment bias when measuring the hospice effect on nursing home residents' pain