Data on specialist pain management is scarce. We evaluated PainDB, a database which aggregates this information from UK pain clinics. PainDB entries for 1120 patients (2648 consultations) were compared to records at 30 pain clinics. Staff were surveyed about normal practice at 28 sites. First consultations (17 135) on the aggregated PainDB were analysed for 2003 for omissions. Those consultations included on PainDB (54.6%) showed good concurrence with written notes (88.1%), with no pattern for the missing visits. Questionnaire responses were often absent from notes (56%) and diagnosis was most frequently omitted from PainDB (12.4–18.4%). Clinic staff overestimated completeness. Despite commitment, PainDB is currently unsuitable for research or audit. As routine hospital data should provide information on activity, specific questions on severity and outcome could be answered by short-term recording of predefined variables. [ABSTRACT FROM AUTHOR]