Episodes of adult bacterial meningitis (ABM) at a Danish hospital in 1991-2000 were identified from the databases of the Department of Clinical Microbiology, and compared with data from the Danish National Patient Register and the Danish National Notification System. Reduced penicillin susceptibility occurred in 21 (23%) of 92 cases of known aetiology, compared to an estimated 6% in nationally notified cases (p <0.001). Ceftriaxone plus penicillin as empirical treatment was appropriate in 97% of ABM cases in the study population, and in 99.6% of nationally notified cases. The notification rate was 75% for penicillin-susceptible episodes, and 24% for penicillin-non-susceptible episodes (p <0.001). Cases involving staphylococci, Pseudomonas spp. and Enterobacteriaceae were under-reported. Among 51 ABM cases with no identified risk factors, nine of 11 cases with penicillin-non-susceptible bacteria were community-acquired. Severe sequelae correlated independently with age, penicillin non-susceptibility, mechanical ventilation and non-transferral to a tertiary hospital (p <0.05; logistic regression). Other factors that correlated with severe sequelae by univariate analysis only were inappropriate clinical handling, abnormal consciousness, convulsions and nosocomial infection. Overall, the data indicated that neither age alone, community-acquired infection nor absence of identified risk factors can predict susceptibility to penicillin accurately. Recommendations for empirical antibiotic treatment for ABM should not be based exclusively on clinical notification systems with possible unbalanced under-reporting.