Decline in maternal mortality in Matlab, Bangladesh: a cautionary tale

A study in Matlab, Bangladesh, has provided evidence favouring a community-based maternity-care delivery system. 3 years of this programme coincided with a significant reduction in direct obstetric mortality compared with the 3 years before the programme. We have examined whether the effects of the programme are sustained over time.

Using data from the continuing demographic survelliance system and from special investigations into the rates and causes of maternal mortality during 1976-93, we compared the trends in direct obstetric maternal mortality ratios in the Maternal and Child Health and Family Planning (MCH-FP) area (which has received extensive services in health and family planning since 1977) with those in the comparison area (with no such intensive health inputs). We divided the areas and time periods into discrete groups that best represented the effects of the introduction of the maternity-care programme.

Direct obstetric mortality declined by 3% per year (rate ratio 0.97 per year [95% CI 0.95-0.99]); there was no difference between the MCH-FP and comparison areas (1.00 [0.96-1.05]). Direct obstetric mortality halved between 1976-86 and 1987-89 in the northern MCH-FP area, where the maternity-care programme was initiated in 1987 (0.50 [0.22-0.99]), but showed no change in the southern MCH-FP area, which had no such intervention at that time (1.07 [0.64-1.72]). After 1990, when the programme was expanded throughout the MCH-FP area, the southern part showed a downward (non-significant) trend in direct obstetric mortality (0.68 [0.35-1.32]). However, direct obstetric mortality also declined between 1987 and 1989 in the southern comparison area (0.48 [0.26-0.83]) in the absence of an intense maternity-care programme, and remained stable thereafter. In the northern comparison area, there was no such decline in direct obstetric mortality (0.78 [0.40-1.40]).

Although the introduction of the maternity-care programme coincided with declining trends in direct obstetric mortality in the areas covered by the programme, a decline also occurred in one of the areas not receiving any such interventions. Caution is required in the interpretation of short-term trends in one indicator in studies designed without random allocation of interventions into treatment and control groups.