We assessed the validity of hospital discharge diagnoses (HDDs) as a tracer of the Guillain-Barrè syndrome (GBS) in Lombardy Region, Northern Italy. The total number of HDDs with the ICD-9 code 357.0 as primary or any secondary diagnosis for the year 1996 was obtained from the Lombardy Region archives and matched to three sources of cases (an active GBS regional registry, a list of patients enrolled in an ongoing case-control study, regional hospital record files). Sensitivity and positive predictive value were calculated for the whole hospital population, by hospital department (neurology vs. other), and by diagnostic level (primary vs. other). Of a total of 1 443 071 HDDs, 361 had the ICD-9 code 357.0. Of these, 100 represented repeated admissions of the same patient. Of the 261 patients, 15 had been hospitalized in 1995, 15 were seropositive for human immunodeficiency virus, and three patients had unavailable records. Of the remaining 228 patients with complete medical records, the diagnosis of GBS was not confirmed in 103 cases. The sensitivity of the HDDs was 90.6% and the positive predictive value was 54.8%. The values were 81.9% and 76.4% for patients discharged from the neurology departments, and 79.7% and 61.8% when only the primary diagnostic level was considered. The crude annual incidence rate (calculated from the three sources combined) was 1.6 per 100 000 (95% CI, 0.8-2.8); the rate was 2.6 (95% CI, 1.7-3.7) when we used HDDs from all hospitals as tracers of disease, 1.7 (95% CI, 0.9-2.9) when only the HDDs from the neurology departments were considered, and 2.0 (95% CI, 1.1-3.2) when the analysis was limited to the primary diagnostic level. HDDs from neurology departments are a fairly valid surrogate of GBS incidence in Lombardy, Italy.