Data collection involves decisions of what to count, how to count, and what to do with the count. The first of these will be determined by the third, since it is the goals which decide what data should be collected. Ways of categorizing and classifying patients, services, treatments, and personnel in relation to long-term medical care must be formulated, and development of comparable data requires definition of baselines or units of observation. Many decision makers with widely varying interests will be competing for information, and to ensure timeliness and accuracy, data collection should be restricted to a minimum of simple, easily obtained, and unambiguous items. Routine reporting of events, periodic censuses relating to persons, and sample surveys all have advantages and limitations. Probably the most useful tools are long-term care registers, as they enable longitudinal studies of patient cohorts. Reliable but confidential means for linking personal records must be found and all records in the system must flow into a central collecting point.