Two community-based retrospective studies conducted in the northern province of Zambia and a review of mortality data from Kasama General Hospital from 1991 to 1995 confirmed the existence of exceptionally high maternal mortality levels in this area. The sisterhood method was applied to a randomly selected sample of 3123 respondents from Kasama District and 2953 from Kaputa District. The life-time risk of dying from maternal causes was 5.4% in Kasama and 11.0% in Kaputa. The maternal mortality ratio was 764/100,000 live births in Kasama District, 1549/100,000 live births in Kaputa District, and 543/100,000 live births at Kasama District Hospital. 94% of women delivering at the hospital were within two hours' walking distance from the facility. In Kasama District, the population-attributable risk (PAR, "risk in the total population minus risk in the unexposed population") of maternal mortality from poor accessibility (more than 2 hours' walking distance) was 220 maternal deaths/100,000 live births, and the population etiologic fraction (PEF, "PAR/risk in total population") was 29%. In Kaputa District, where there is no hospital, the PAR from poor accessibility was 1006 maternal deaths/100,000 live births and the PEF was 65%. To reduce accessibility-related maternal mortality, both districts have established an ambulance service, set up strategic blood banks, and provided short wave radios to outlying health centers.