on CYP data be reduced and that multiple kinds of data be developed to monitor the flow of supplies through the commodity supply system.
ADMINISTRATIVE/PROGRAM RECORDKEEPING
James Palmore discussed specific problems inherent in using administrative records for evaluation purposes, the most glaring being that administrative records are usually collected for accounting and personnel purposes and are not set up in such a way as to be useful to evaluators. Palmore noted seven additional problems that make it difficult to incorporate administrative records in evaluation: (1) Multilevel data for corresponding units are not available to which administrative records can be linked; (2) data measuring the actual--as opposed to supposed--services rendered are not available; (3) the time lag between administratively recorded program inputs and consumer response in contraceptive prevalence is unclear; (4) service inputs may interact with each other or with the socioeconomic environment, thus necessitating models that contain complex interaction terms; (5) the relationships between administrative inputs and outputs may not be linear; (6) administrative records do not measure a population's use of clinics if clients leave the administrative area for services; and (7) reliable information concerning the demand for contraceptives--to link with input and output measures--is difficult to obtain. These problems are all compounded by the expansion of family planning services in the private sector (which increases the number of administrative recordkeeping systems), the need for more decentralized data, and
the integration of family planning services with other services such as maternal and child health.
Brant Fries discussed the newly developed administrative system used in nursing homes in the United States, characteristics of which are relevant to other sectors. Developed within the past several decades, the current nursing home system assesses different products of nursing home care, the case-mix of nursing homes, payment schedules, and characteristics of residents. The data collected can provide indicators of quality of care, help managers make staff decisions, and be used for policy research. The development of the system provides instructive guidelines: (1) well-designed data systems should include explicit examples, definitions, and delimiters on the data collection forms; (2) those collecting the data should be well trained; (3) collected data need to be detailed enough to have multiple uses, and the uses need to be made clear to service providers so that care can be improved; (4) data should be collected in such a way that they can be entered easily into a computer (even if that is not an initial goal); (5) standardized definitions enable a common language to be developed, which aids in communication among staff about patients and analysis of data by researchers. Fries emphasized that, if data are used for multiple purposes, they are more likely to be accurate, and he stressed that data will be more accurate if those collecting them have a vested interest.
During the discussion the importance of common language, data quality, political will, and information flow were stressed as key elements for administrative recordkeeping systems. Standardizing instruments to produce a common language would make data more useful to researchers and service providers. Data accuracy and usefulness could also be