All cerebrospinal fluid (CSF) shunt operations performed at Oregon Health Sciences University from 1976 through 1989 were reviewed retrospectively. There were 727 insertions or revisions of CSF shunts. Outcomes from these operations were studied using statistical techniques for analysis of survival data. There were 671 simple, linear shunts and 56 complex shunt systems incorporating 3-way connectors or consisting of multiple independent linear shunts. The overall estimated median survival of simple CSF shunts was 73 months, whereas the median survival of complex shunts was only 11 months (p = 0.0069). Only simple shunts were subjected to further analysis. The most important determinant of risk of obstruction was the age of the patient: children less than 2 years old were at higher risk than older patients (p < 0.00005). The risk of obstruction of a revised shunt was not different overall from the risk of failure of a new shunt, but revision of a shunt after an interval shorter than 6 months encountered a higher risk (p < 0.00005) of failure than either new shunts or shunts revised after a longer interval. The following factors had no significant influence on risk of obstruction: attending surgeon, etiology of hydrocephalus, duration of operation, time of day of operation, and whether the patient had epilepsy. Totally revised shunts lasted no longer than partially revised shunts. In children less than 2 years old, trends were observed favoring new shunts over revised shunts (p = 0.0355) and flat bottom valves over Hakim valves (p = 0.0255). The risk of shunt infection by 1 year was 8.5%.(ABSTRACT TRUNCATED AT 250 WORDS)