The validity of discharge diagnosis of amyotrophic lateral sclerosis (ALS) and of the main procedures performed during hospitalization was assessed using as gold standard the data from the Piemonte and Valle d'Aosta Register for ALS (PARALS), a collaborative population-based registry aimed at determining prospectively the incidence and the factors related to ALS outcome. All patients discharged with ICD code 335.2 (primary and secondary diagnoses) in the period 1995-1996 in Piemonte, Italy, were considered. Out of the 1,049 cases identified, 433 remained after excluding patients not resident in Piemonte and repeated admissions. Of these, 258 had a correct diagnosis of ALS (168 incident and 90 prevalent cases) after a review of clinical records. The sensitivity of discharge diagnoses was 78.9%, and the positive predictive value was 38.8%. The sensitivity for main procedures (percutaneous endoscopic gastrostomy, noninvasive ventilation, and tracheostomy) ranged between 76 and 100%. ICD codes allowed to identify 22 cases that had not been ascertained with other sources. In conclusion, hospital discharge records appear to poorly reflect the incidence of ALS, and can be used only after clinical verification of the diagnosis.