Objective-To collect a valid, complete, con-, tinuous, and representative database of morbidity presenting to primary care and to use the data to help commission services on the basis of local need and effectiveness.Setting-Computerised general practices in Somerset.Methods-Participating general practices were selected to be representative of the district health authority population for general practice and population characteristics. All conditions presented at face to face consultations were assigned a Read code and episode type and the data were regularly validated. Data were sent by modem from the practices via a third party to the health authority each week.Main outcome measures-Proportion of consultations coded and accuracy of coding.Results-l1 practices agreed to participate. Validations for completeness during April 1994 to March 1995 revealed that 96.4% of the records were coded; 94 % ofthe 1090 records validated had appropriate episode types and 87% appropriate Read codes. The results have been used to help formulate the health authority's purchasing plans and have enabled a change in the local contracts for surgery for glue ear.Conclusions-The project has shown the feasibility of establishing a network of practices recording and reporting the morbidity seen in primary care. Early indications are that the data can be useful in evidence based purchasing.