OBJECTIVES: To evaluate and compare the three most commonly used perinatal death classification systems: (1) the Nordic-Baltic; (2) the Aberdeen; and (3) the Wigglesworth, and assess their applicability in a developing country (Sudan) with a high perinatal mortality rate, and their justification for practical use in quality assurance and audit activities. METHODS: At Omdurman Maternity Hospital (OMH), Khartoum, Sudan, 166 perinatal deaths were prospectively assessed during a 3-month period (May-August 2000) with a total of 2260 births. Narratives of 166 perinatal deaths were prepared for the purpose of audit. A panel of two Danish and one Sudanese obstetrician categorized the cases according to: (1) the Nordic-Baltic; (2) the Aberdeen; and (3) the Wigglesworth classification. RESULTS: By all three classifications a similar fraction of cases (approx. 85%) were allocated to one category only, and in 15% of cases the assessors were in doubt into which of two categories the cases should be allocated. The necessary information is often not available, giving at least 40% classified as 'unknown' in the Aberdeen classification, whereas the Wigglesworth classification results in an even larger group of unspecified asphyxia. CONCLUSION: Classification of perinatal deaths in developing countries is associated with problems regarding application, validity and usefulness. The Nordic-Baltic classification seems to be most suitable for appropriate stratification using routinely recorded variables and providing categories associated with specific levels of care.