Objectives — To validate an algorithm to identify cases of intussusception using the health administrative data of Ontario, Canada, and to apply the algorithm to estimate provincial incidence of intussusception, preceding the introduction of the universal rotavirus vaccination program.
Study Design — The researchers determined the accuracy of various combinations of diagnostic, procedural, and billing codes using the chart-abstracted diagnoses of patients of the Children's Hospital of Eastern Ontario as the reference standard. The researchers selected an algorithm that maximized positive predictive value while maintaining a high sensitivity and used it to ascertain annual incidence of intussusception for fiscal years 1995-2010. The researchers explored temporal trends in incidence using Poisson regression.
Results — The selected algorithm included only the International Classification of Diseases (ICD)-9 or ICD-10 code for intussusception in the hospitalization database and was sensitive (89.3%) and highly specific (>99.9%). The positive predictive value of the ICD code was 72.4%, and the negative predictive value was >99.9%. The researchers observed the highest mean incidence (34 per 100 000) in male children <1 year of age. Temporal trends in incidence varied by age group. There was a significant mean decrease in incidence of 4% per year in infants (<1 year) until 2004 and rates stabilized thereafter.
Conclusions — The researchers have demonstrated that intussusception can be accurately identified within health administrative data using validated algorithms. The researchers have described changes in temporal trends in intussusception incidence in Ontario and established a baseline to allow ongoing monitoring as part of vaccine safety surveillance.