This report presents the findings of the May-October 1995 in-depth Sumve Survey on Adult and Childhood Mortality (SACM) in the Mwanza Region of northwestern Tanzania. The aim was to determine whether data collection by proxy from mothers' sisters could be used to estimate childhood mortality rates and to describe the demographic situation and use of maternity services in order to improve local programs. The sample was drawn from the Kwimba District of Mwanza Region and south of Lake Victoria. The area is largely subsistence-farmed, with low modernization and educational levels. The Sukuma group is the dominant ethnic group. The study population resided mostly within the primary health care program service area, which includes a large hospital for tertiary care needs. Phase I involved a representative sample of 1488 households and 2130 women aged 15-50 years. Data collection included full birth histories of own respondent and sibling histories. Phase II was conducted one month after Phase I and involved data collection from all living sisters aged 15-50 years in an expanded study area. The sample included 2123 sisters (96% of eligible sisters). Findings indicate that lifetime fertility was about 7.4 children/woman. Almost 60% had a first birth before the age of 20 years. Child mortality was an estimated 134 deaths/1000 live births. Infant mortality was an estimated 83 deaths/1000 live births. Infant mortality was twice as high among children born with short birth intervals compared to births with long intervals. Adult female mortality was 4/1000 person-years; male mortality was 5/1000 person-years. The maternal mortality ratio was around 500 maternal deaths/100,000 live births. Phase II birth history data revealed 3719 proxy-reported birth histories. The proxy data yielded fertility in the 5 years preceding the survey of 5.9 children/woman compared to 6.7 children/woman from own children data. Mortality estimates were similar in the 5 years before the survey for own and proxy data. However, proxy data yielded a 23% underestimate of child mortality during the 5-9 years before the survey and a 31% underestimate during the 10-14 years before the survey. It is argued that proxy data are not an appropriate measure of mortality. - See more at: http://www.popline.org/node/168564#sthash.ghUGwnHT.dpuf