A near-miss maternal mortality enquiry was performed at University College Obstetric Hospital, London, by reviewing retrospectively all 30 obstetric admissions to the intensive care unit (ITU) over a two-year period. The obstetric admission rate to ITU was 0.5 (95% CI 0.32-0.67%), or one per 200 women delivered. Haemorrhage and severe pre-eclampsia were the two commonest causes of admission. Sub-standard care was identified in 52% of cases. Blood loss was often massive ( 2000 ml), underestimated and required large volume transfusions (mean transfusion 6.4 units, range 1-24). Although there are problems with definitions, ascertainment and validity, 'near-miss' review is feasible. It is worthwhile for every hospital to carry out its own 'near-miss' enquiry using appropriate local criteria to identify potential areas for improvements. 'Near-misses' are more prevalent than deaths and are dominated by conditions that are amenable to treatment. They may be even more sensitive to improvement or deterioration in obstetric services than mortality data.