Evidence of second-order factor structure in a diagnostic problem space: Implications for medical education