Can risk factors for prolonged mechanical ventilation and weaning failure be gleaned from studies using differing definitions of these outcomes?