%0 Journal Article %A Maryam Oskoui %A Pamela Ng %A Marc Dorais %A Nicole Pigeon %A Louise Koclas %A Céline Lamarre %A Francine Malouin %A Carol L. Richards %A Michael Shevell %A Lawrence Joseph %T Accuracy of administrative claims data for cerebral palsy diagnosis: a retrospective cohort study %D 2017 %R 10.9778/cmajo.20170013 %J CMAJ Open %P E570-E575 %V 5 %N 3 %X Background: Cerebral palsy is the most common cause of childhood physical disability, with multiple associated comorbidities. Administrative claims data provide population-level prevalence estimates for cerebral palsy surveillance; however, their diagnostic accuracy has never been validated in Quebec. This study aimed to assess the accuracy of administrative claims data for the diagnosis of cerebral palsy.Methods: We conducted a retrospective cohort study of children with cerebral palsy born between 1999 and 2002 within 6 health administrative regions of Quebec. Provincial cerebral palsy registry data (reference standard) and administrative physician claims were linked. We explored differences between true-positive and false-negative cases using subgroup sensitivity analysis.Results: A total of 301 children were identified with confirmed cerebral palsy from the provincial registry, for an estimated prevalence of 1.8 (95% confidence interval [CI] 1.6-2.1) per 1000 children 5 years of age. The sensitivity and specificity of administrative claims data for cerebral palsy were 65.5% (95% CI 59.8%-70.8%) and 99.9% (95% CI 99.9%-99.9%), respectively, yielding a prevalence of 2.0 (95% CI 1.9-2.3) per 1000 children 5 years of age. The positive and negative predictive values were 58.8% (95% CI 53.3%-64.1%) and 99.9% (95% CI 99.9%-99.9%), respectively. The κ value was 0.62 (95% CI 0.57-0.67). Administrative claims data were more sensitive for children from rural regions, born preterm, with spastic quadriparesis and with higher levels of motor impairment.Interpretation: Administrative claims data do not capture the full spectrum of children with cerebral palsy. This suggests the need for a more sensitive case definition and caution when using such data without validation. %U https://www.cmajopen.ca/content/cmajo/5/3/E570.full.pdf