The association of workflow interruptions and hospital doctors' workload: a prospective observational study