All baseline surveys in the Matlab study site were completed in 2000, including:

• A Household Demographic Survey in 88 647 households; a follow-up verbal and social autopsy survey was completed in early 2002;

• Health and Morbidity Survey on 2066 under-five children;

• Health Facility Survey in 19 first-level facilities.

Facility catchment areas were determined by the use of data from the demographic survey. Mortality rates were estimated for each facility and catchment area, and the 20 facilities were randomly selected for IMCI intervention or existing services (comparison areas). Randomization was designed to achieve a balance as regards number of facilities, facility type, geographical distribution, baseline mortality levels and size of catchment population.

The verbal autopsy survey was designed to estimate cause-specific mortality, and a social autopsy module investigated care-seeking practices prior to death.

IMCI clinical guidelines have been adapted to Bangladesh and translated and produced in Bengali. Training of health care workers from the selected facilities at Matlab was completed in November–December 2001. Health systems support was strengthened for IMCI implementation in the selected facilities and was in place by March 2001. Programme performance and community response and practices are continuously monitored. A package of interventions designed to improve selected family and community practices is under development and phased introduction. As part of this package, community-based workers have been trained in counselling skills to improve child care and feeding, and care-seeking practices.