The Quality Assurance Project (QAP), initiated in 1990, provides technical assistance to developing countries in designing and implementing effective strategies for monitoring the quality of health care. This fourth report in QAP's Quality Assurance Methodology Refinement Series compared the level of statistical agreement between four quality assessment methods used at Under Five clinics in 14 health facilities in Malawi's Lilongwe district. These methods included 436 observations of provider-patient encounters, 426 exit interviews with caretakers, reviews of 362 records, and interviews with 30 providers. The consistency of provider behavior varied according to the child's primary symptom (highest for diarrhea and lowest for cough). Agreement between observation and record review was poor for both general assessment of the child and management of specific conditions; provider interview data had only fair agreement with observation data on a limited number of items. Exit interview data had fair to good agreement with observation data for many items, especially tasks related to fever or diarrhea and concrete activities caretakers could see or hear. Record reviews involved the least time, while provider interviews were least expensive. Overall, these findings suggest that supervisors should first conduct provider interviews to ensure those supervised have the requisite knowledge for proper case management and record reviews to assess whether they use the correct drugs and dosages for the conditions diagnosed. Once supervisors are confident of their employees' knowledge, skills can be assessed through direct observation of performance. When skills have been confirmed, occasional unscheduled exit interviews can be conducted.