Accuracy of administrative data in ascertaining health conditions: a systematic review

Objectives: To conduct a systematic review of studies assessing the accuracy of International Classification of Diseases (ICD) and Diagnostic and Statistical Manual of Mental Disorders (DSM) codes in administrative data for ascertaining health conditions when compared to a reference standard. Materials and Methods: We searched MEDLINE, Embase, and PsycINFO, and reported study characteristics and accuracy measures, including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). We synthesized information about primary and validation sources of administrative data used in this literature and visually described accuracy measures by ICD chapter. Results: = 150). The IQRs of accuracy measures across studies and ICD chapters were 44.0-91.4, 90.0-99.6, 59.4-91.1, and 90.0-99.5 for sensitivity, specificity, PPV, and NPV, respectively. Discussion: The assessed literature is heavily skewed towards data from high-income countries. The variance we observed in accuracy measures, particularly for sensitivity and PPV, indicates that we need more evidence on and ongoing monitoring of the accuracy of ICD/DSM codes in administrative data, which are now widely used in population-based epidemiologic studies and therefore highly relevant for public health policymaking. Conclusion: Health conditions ascertained using ICD/DSM have moderate to high accuracy in identifying true positives and high to very high accuracy in identifying true negatives.