Incidence of Sudden Infant Death Syndrome in Olmsted County, Minnesota: 1945 Through 1992

To apply uniform diagnostic criteria for sudden infant death syndrome (SIDS) for an extended period for comparison of incidence rates from 1945 to 1992 in Olmsted County, Minnesota, to investigate the influence of a person-time or live birth denominator on the estimate of incidence, to calculate the proportionate mortality rate for SIDS over time, and to evaluate the accuracy of death certificates for case ascertainment and the role of interobserver variation in case classification. We retrospectively reviewed the autopsy results and complete medical records for all infant deaths from 1945 through 1992 for residents of Olmsted County, Minnesota. Cases were identified from a computerized list of all Olmsted County deaths of infants occurring at ages 48 hours to 365 days. All resident cases were categorized as non-SIDS, possible SIDS, or incomplete, on the basis of findings from autopsy and clinical history. Incidence rates were calculated for two different SIDS groups and with use of denominators of person-time and live births. For the study period, 82 cases of SIDS were identified (97% white and 3% Asian). The mean age at death was 12.5 weeks; male infants constituted 59% of cases. No significant trend in seasonal distribution was noted; 73% of deaths occurred between midnight and noon. The incidence rate, defined as SIDS definite and possible deaths per 1,000 resident live births, increased from 0.55 in 1950 through 1953 to 1.28 in 1990 through 1992. The secular trend was best described by a linear model with a constant positive slope. Similar trends were observed with other definitions of incidence. During the study period, SIDS as a percentage of total infant deaths dramatically increased, ranging from 2.5 in 1950 through 1953 to 17.9 in 1990 through 1992. The death certificate diagnosis correctly predicted 72 % of SIDS cases before 1970 and 100% of cases after 1970. Since 1945, the incidence of SIDS apparently has increased, although diagnostic transfer from other causes of death probably contributes to the observed trend. The comparison of live births versus person-time as denominators showed no significant difference in incidence rates. Interobserver reliability is modest for SIDS cases diagnosed before 1970 and may contribute to the variability in reported SIDS incidence rates.