Accuracy of identification of patients with immune thrombocytopenic purpura through administrative records: a data...

Administrative data are commonly used to estimate the prevalence of a disease, but the validity of the coding system needs to be evaluated before its use. We assessed the validity of the International Classification of Disease, 9(th) version, Clinical Modification (ICD-9-CM) code of 287.3 for identifying patients with immune thrombocytopenic purpura (ITP). Administrative data from inpatients and outpatients seen were retrieved if the patient or insurer was billed with one of three ICD-9-CM codes for thrombocytopenic disorders, 287.3, 287.4, and 287.5, as a primary or secondary diagnosis; or was physician-identified as having ITP. The electronic medical records for these patients were systematically reviewed to identify patients with ITP and with non-ITP diagnoses. Sensitivity, specificity, positive and negative predictive values, and kappa scores were calculated separately for inpatients and outpatients. Four-hundred eighteen records were reviewed. Among inpatients, the sensitivity of code 287.3 for indicating a diagnosis of ITP was 100% [95% confidence interval 94-100%]. The specificity was 89% [95% confidence interval 84-94%]. The percent agreement was 92%, and the kappa statistic was 0.80. For outpatients, the sensitivity of the billing code 287.3 was 84% [95% confidence interval 76-91%], a conservative estimate because of how the patients with other diagnoses were selected. The specificity for outpatients was 66% [95% confidence interval 56-76%]. ICD-9-CM code 287.3 in administrative billing data is likely to be sufficiently sensitive and specific, particularly when inpatient data are used, for the estimation of the prevalence of ITP.