Sentinel practice networks, also referred to as practice research networks, have been in use for morbidity registration in Europe and North America for many years and have recently been recognised by the US Institutes of Medicine as a significant underpinning for studies in primary care. In such networks, office-based physicians are recruited on a (mostly) non-monetary basis to monitor defined events among their patients (e. g. influenza, mumps, asthma attacks). After adequate analysis, these data indicate spatial and temporal trends in event frequencies. The areas covered by such sentinel practice networks are communities, counties, or - at best - countries. Though surveillance systems can be very useful in small regions, national differences or global trends in the frequency and distribution of health problems can only be demonstrated by comparing results between sentinel networks. Comparability of national results is, however, limited. This is partly due to the different national health systems influencing the demand of medical care. One of the main reasons is, however, the lack of comparable data on the population from which the cases arise. Such "denominator data" is a vital premise for the comparison of data from primary care research between different regions or even different countries.The extent to which data are available about the population seeking medical care in the European Community countries depends mainly on the national health care system. In countries where most or all inhabitants are registered with certain GPs (Netherlands, United Kingdom), the population covered by physicians' offices is known in principle, though short-term migrations and double- ornon-registered persons may limit the accuracy of population-based measures of event frequencies. In countries where every resident may consult the physician of their own choice (e. g. Belgium, France, Germany), the population covered by medical practices is highly unstable and mostly unknown. A special problem in those countries is the estimation of the so-called "zero class", i.e., the portion of the population which does not see a certain physician at regular intervals but rather only when necessary. Depending on the type of health care system, the number of patients or consultations per unit time - e. g. as specified by health insurance companies - may be available as denominator data in those countries.In early 1993 an initiative was undertaken to gather a workgroup of scientists from five European countries with different health care systems in order to launch a joint grant application to the BIOMED programme of the European Community. The proposed methodological project on the denominator problem was recognised to have central relevance for the harmonisation of methodologies and protocols in health services research and for the further development of co-ordinated networks providing for adequate collection and interpretation of data on health status and health systems, and was therefore funded by the EC from 1 June 1994 to 30 May 1996. The fruits of this concerted research activity The European Denominator Project were finally condensed into an end report, which has now been made available to the public in book format. It provides an overview of the denominator problem itself, of approaches that are readily available at different levels of complexity, and of future ways of harmonising denominator data for primary health care research in order to facilitate the comparison of international data.
