The estimation of diagnostic sensitivity using stability data: An application to major depressive disorder