The matching rate was presented as the number of matched cases divided by the number of cases retrieved from the DNMC (the gold standard). The validity of using the ICD-9 410.xx code to identify matched cases of AMI was assessed by calculating the positive predictive value (PPV) using medical records (of confirmed cases after review by the cardiologists) as the gold standard. The agreement rate between the two reviewers was calculated using the agreement cases divided by the total cases. In addition, we estimated the PPV of principal diagnosis, antiplatelet therapy, and cardiac procedures of confirmed AMI cases. Further, different criteria were used to evaluate sensitivity and PPV of the diagnosis code of AMI in the NHIRD, such as “principal diagnosis with antiplatelet” or “principal diagnosis with percutaneous transluminal coronary angioplasty (PTCA)”.
To ensure validity of procedures and aspirin/clopidogrel exposure, we defined sensitivity as the probability that the procedure/antiplatelet agents recorded in the medical chart (denominator) by a doctor were also recorded in the NHIRD (numerator). PPV is the conditional probability that claims of procedures/antiplatelet agents in the NHIRD (denominator) were actually present in the DNMC records (numerator). For agreement among discharge diagnoses for each AMI hospitalization, percentage of consistency between the two databases was calculated for linkage cases.
All computations and 95% confidence intervals (CIs) for binominal proportions were performed with SAS version 9.2 (SAS Institute Inc, Cary, NC, USA). This study was reviewed and approved by the Institutional Review Board of the National Cheng Kung University Medical Center (ER-95-137).
To ensure validity of procedures and aspirin/clopidogrel exposure, we defined sensitivity as the probability that the procedure/antiplatelet agents recorded in the medical chart (denominator) by a doctor were also recorded in the NHIRD (numerator). PPV is the conditional probability that claims of procedures/antiplatelet agents in the NHIRD (denominator) were actually present in the DNMC records (numerator). For agreement among discharge diagnoses for each AMI hospitalization, percentage of consistency between the two databases was calculated for linkage cases.
All computations and 95% confidence intervals (CIs) for binominal proportions were performed with SAS version 9.2 (SAS Institute Inc, Cary, NC, USA). This study was reviewed and approved by the Institutional Review Board of the National Cheng Kung University Medical Center (ER-95-137).
Free full text: Click here
