Morbidity registration by a network of sentinel general practitioners (SGPs) in Belgium raises a number of problems related to possible biases in the network procedure, such as unequal geographical distribution, non-participation of a segment of the target population of practitioners and difficulties in the estimation of the denominator population at risk for the health problems under study. Through the application of two hierarchical clustering procedures, the initial number of 43 districts in the country has been reduced to 15 homogeneous district clusters. These represent the new geographical framework from which the geographical spread of the network is checked. This network is subsequently corrected for such socio-demographic parameters as age, sex and occupation in order to match more closely the total population of Belgian general practitioners (GPs). The population covered by the network is estimated on the basis of the annual number of patient contacts. Application of the described procedures should result in a network allowing valid estimations for a number of health issues as seen by Belgian GPs.