Predictive Value of ICD-9-CM Codes Used in Vaccine Safety Research

摘要: Objectives: To assess how well selected ICD-9-CM diagnosis codes predict adverse events; to model bias and power loss when vaccine safety analyses rely on unverified codes. Methods: We extracted chart verification data for from six Vaccine Safety Datalink (VSD) publications modeled biases losses using positive predictive value (PPV) estimates ranges of code sensitivity. Results: Positive values were high type 1 diabetes (80%) in children, relative WHO criteria, intussusception (81%) young a standard published case definition. PPVs moderate (65%) inpatient emergency department childhood seizures low (21%) outpatient seizures, both physician investigator judgment. Codes incident central nervous system demyelinating disease adults had PPV (42%) physicians’ diagnoses. Modeled modest, but large increases frequencies events are required achieve adequate if used, especially associations weak. Conclusions: care settings, sufficiently Adverse event misclassification should be accounted statistical calculations.