Quality assessments of HMO diagnosis databases used to monitor childhood vaccine safety.

摘要: OBJECTIVE To assess the quality of automated diagnoses extracted from medical care databases by Vaccine Safety Datalink (VSD) study. METHODS Two methods are used to VSD diagnosis data. The first method compares common and abstracted diagnostic categories ("outcomes") in 1-2% simple random samples study populations. second estimates positive predictive values codes identify potential cases rare conditions (e.g., acute ataxia) for inclusion nested case-control record abstraction studies. RESULTS There was good agreement (64-68%) between outcomes at 3 4 sites poor (44%) 1 site. Overall sites, 56% children with cerebella ataxia (ICD-9 = 334) 22% "lack coordination" 781.3) met objective clinical criteria ataxia. CONCLUSIONS misclassification error rates screening substantially reduce power analyses limit usefulness moderate strong vaccine-outcome associations. Medical verification is needed definitive assessments.