Background Recent estimates have shown a stagnation in maternal mortality reduction globally. Levels and trends in maternal mortality in urban settings in Africa are of interest due to health system challenges posed by rapid urbanisation. Given the high percentage of births occurring in city health facilities, tracking the institutional maternal mortality ratio (IMMR) can reveal key performance issues. We described the trends in IMMR between 2016 and 2021, and the causes and characteristics of maternal deaths in Kampala, Uganda. Methods This retrospective study utilised routine data from all facilities in Kampala that provided childbirth services from 2016 to 2021. Three key variables were used: number of deliveries, live births, and maternal deaths. Additionally, from medical records in three purposively selected hospitals, we extracted information on characteristics of women recorded as maternal deaths and their causes for the period 2016-2021 from death notification and review forms. We conducted descriptive analysis and assessed time trends in IMMR. Results For the six-year period, the Kampala-level IMMR was 189/100,000 deliveries (194/100,000 livebirths). The IMMR declined from 213 (2016) to 109/100,000 deliveries in 2019 (p=0.04) and then increased to 258 in 2021 (p=0.12). A review of 358 maternal death records showed that 65% of deaths were women referred from other health facilities, and 47% lived in the surrounding district (Wakiso). Obstetric haemorrhage (44%) and hypertensive disorders in pregnancy (28%) were the most commonly recorded causes of death. Health system factors such as lack of blood and other supplies contributed to 65% of the deaths reviewed. Conclusions IMMR in Kampala increased from 2019-2021. Health infrastructure changes and maternal health initiatives potentially accounted for the observed trend by affecting data quality. Referral and facility readiness were underlying reasons for maternal deaths. Addressing bottlenecks in emergency referral processes may contribute to mortality reduction. Further research should determine the relative contribution of key influences on mortality levels and examine the dynamics between women’s care-seeking pathways and service delivery arrangements for emergency obstetric care in urban settings.