Objective. To define the rate of severe maternal morbidity in different modes of delivery and to find out if the rate of severe morbidity has changed over a 5-year time span. Design. Retrospective register-based study. Setting. Finnish Medical Birth Registry and Hospital Discharge Registry. Population. All singleton deliveries in Finland in 1997 and 2002 (n = 110,717). Methods. Diagnoses and operative interventions recorded in the Hospital Discharge Registry indicating a severe maternal complication were linked with Birth Register data and compared by mode of delivery: spontaneous vaginal delivery (VD), instrumental VD, elective cesarean section and non-elective cesarean section. Main outcome measures were severe maternal morbidity: deep venous thromboembolism and amniotic fluid embolism, major puerperal infection, severe hemorrhage, events requiring operative intervention after delivery, uterine rupture and inversion, and intestinal obstruction. Results. Severe maternal morbidity was more frequent in cesarean than vaginal deliveries (p<0.001), and more frequent in non-elective than in elective operations (p<0.001). The rate of severe maternal morbidity increased considerably from 1997 to 2002; from 5.9 to 7.6 per 1,000 in all deliveries (p<0.001), from 4.0 per 1,000 to 5.2 per 1,000 in spontaneous vaginal deliveries (p = 0.005), from 9.9 per 1,000 to 12.1 per 1,000 in elective cesarean sections (CSs) (p = 0.164), and from 19.6 per 1,000 to 27.2 per 1,000 in non-elective CSs (p = 0.090), respectively. Conclusions. Severe maternal morbidity has increased both in cesarean and vaginal deliveries from 1997 to 2002. Cesarean delivery, even an elective one, carries a significantly higher risk of life-threatening maternal complications than VD.