Predictive Value of Pediatric Stroke Diagnoses in Administrative Data: A Systematic Literature Review

Abstract

<b><i>Background and Aim:</i></b> Administrative healthcare data are frequently used for studying incidence, prevalence, risk factors, and outcome of pediatric stroke. However, the accuracy of these data sources is uncertain. The aim of this study was to systematically analyze published data on the positive predictive value (PPV) and sensitivity of diagnoses used to identify pediatric stroke patients in administrative data. <b><i>Methods:</i></b> This systematic review was performed in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. We searched PubMed and Embase for studies, published in year 2000 or later, describing the PPV or sensitivity of diagnoses used to identify children with stroke in administrative data. The search was performed on June 9, 2022. Studies written in other languages than English, with less than 30 participants, and conference abstracts were excluded. <b><i>Results:</i></b> Eight studies were included after full-text review from 2,475 potentially eligible records. These included 3,137 children. All studies reported data from high-income countries. Reported PPVs varied considerably across studies and stroke subtypes: acute ischemic stroke, range 0.27–0.89; cerebral venous thrombosis, range 0.45–0.72; spontaneous subarachnoid hemorrhage, range 0.52–0.83; and spontaneous intracerebral hemorrhage, range 0.62–0.66. One study examined sensitivity of an ICD-9 search compared to a radiology report search and found that the ICD search had poor sensitivity (33%). <b><i>Conclusion:</i></b> Caution is recommended in the use and interpretation of nonvalidated administrative data for pediatric stroke. Data on the PPV and sensitivity of pediatric stroke diagnoses in administrative data remain limited and are only available from high-income countries.