Secondary Logo

Institutional members access full text with Ovid®

Accuracy of Administrative Health Data for Surveillance of Traumatic Brain Injury

A Bayesian Latent Class Analysis

Lasry, Olivera,b; Dendukuri, Nandinia; Marcoux, Judithb; Buckeridge, David L.a

doi: 10.1097/EDE.0000000000000888
Injury Epidemiology

Background: Traumatic brain injury surveillance provides information for allocating resources to prevention efforts. Administrative data are widely available and inexpensive but may underestimate traumatic brain injury burden by misclassifying cases. Moreover, previous studies evaluating the accuracy of administrative data surveillance case definitions were at risk of bias by using imperfect diagnostic definitions as reference standards. We assessed the accuracy (sensitivity/specificity) of traumatic brain injury surveillance case definitions in administrative data, without using a reference standard, to estimate incidence accurately.

Methods: We used administrative data from a 25% random sample of Montreal residents from 2000 to 2014. We used hierarchical Bayesian latent class models to estimate the accuracy of widely used traumatic brain injury case definitions based on the International Classification of Diseases, or on head radiologic examinations, covering the full injury spectrum in children, adults, and the elderly. We estimated measurement error-adjusted age- and severity-specific incidence.

Results: The adjusted traumatic brain injury incidence was 76 (95% CrI = 68, 85) per 10,000 person-years (underestimated as 54 [95% CrI = 54, 55] per 10,000 without adjustment). The most sensitive case definitions were radiologic examination claims in adults/elderly (0.48; 95% CrI = 0.43, 0.55 and 0.66; 95% CrI = 0.54, 0.79) and emergency department claims in children (0.45; 95% CrI = 0.39, 0.52). The most specific case definitions were inpatient claims and discharge abstracts (0.99; 95% CrI = 0.99, 1.00). We noted strong secular trends in case definition accuracy.

Conclusions: Administrative data remain a useful tool for conducting traumatic brain injury surveillance and epidemiologic research when measurement error is adjusted for.

From the aDepartment of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, Quebec, Canada

bDepartment of Neurology and Neurosurgery, McGill University-McGill University Health Centre, Montreal, Quebec, Canada.

Submitted January 17, 2018; accepted July 5, 2018.

Statistical analysis done by O.L. O.L. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Data access: The data and computing code can be accessed upon request from the authors.

O.L. reports having been supported by the Canadian Institutes of Health Research (CIHR) Vanier Canada Graduate Scholarship for his doctoral studies. N.D. reports having received a salary award from the Fonds de Recherche du Québec – Santé (FRQS). D.L.B. reports having received a salary award from the CIHR.

The authors report no conflicts of interest.

Supplemental digital content is available through direct URL citations in the HTML and PDF versions of this article (

Correspondence: Oliver Lasry, 1140 Pine Avenue West, Montreal, Quebec H3A 1A3. E-mail:

Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.