Objective: To study changes in the epidemiology and management of meningococcal disease in one health district during a period of high local incidence of disease. Design: Prospective case ascertainment and data collection over 14 years, with retrospective analysis of cases. Setting: West Gloucestershire (population 320 000). Subjects: Residents developing meningococcal disease between 1 January 1982 and 31 December 1995. Results: 252 cases of invasive meningococcal disease were identified, of which 102 (40%) were officially notified and 191 (76%) were confirmed by culture from a deep site. The observed disease incidence of 5.6/100 000/year was about 2.7 times the national incidence (as measured by either statutory notifications or reference laboratory reports). The period 1983–90 was characterised by a prolonged localised outbreak due to serogroup B serotype 15 sulphonamide resistant (B15R) strains. General practitioners gave benzylpenicillin before hospital admission to 18% of patients who presented with meningococcal disease in the first half of the study period and to 40% who presented in the second half. The overall case fatality rate was 6.7% (17/252). Four deaths were directly or indirectly related to lumbar puncture. Of 120 patients whose lumbar puncture yielded meningococci, nine (8%) showed no abnormality on initial examination. Conclusions: Neither laboratory records nor formal notifications alone can give an accurate estimate of the incidence of meningococcal disease. Because of the dangers of lumbar puncture, the frequency of misleading negative initial findings, and the advent of new diagnostic techniques, the need for samples of cerebrospinal fluid should be critically questioned in each case of suspected meningococcal disease. The first doctor to suspect a diagnosis of meningococcal disease should start treatment immediately Education and encouragement can increase the proportion of patients with meningococcal disease receiving benzylpenicillin before admission to hospital Specimens for polymerase chain reaction testing should be collected from all patients suspected of having meningococcal disease in admission to hospital The need for lumbar puncture should be considered carefully in suspected meningococcal meningitis; it is contraindicated in meningococcal septicaemia Notifications of meningococcal disease remain incomplete; accurate estimation of numbers of cases requires reconciliation of notifications with laboratory diagnoses and local clinical registers