34 - Accuracy of ICD Codes in Identifying Pediatric Firearm Injury, Injury Intentionality and Firearm Type

Poster Number: 34
Publication Number: 34.103

Katie Donnelly, Children's National Health System, Washington, DC, United States; Marci Fornari, Children's National Health System, Washington, DC, United States; Megan M. Attridge, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, United States; Monika Goyal, Children's National Health System, Washington, DC, United States; Megan L. Ranney, Brown University, Providence, RI, United States; Elizabeth R. Alpern, Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, IL, United States

Background: International Classification of Disease (ICD) codes are used to understand the epidemiology of firearm injuries in population-based studies. Yet ICD codes show significant variability in the accuracy and completeness of identifying diseases, including firearm injuries.

Objective: Assess the accuracy of ICD codes to identify pediatric firearm injury, injury intentionality and firearm type in charts identified as potential firearm injuries.

Design/Methods: This is a retrospective cohort study of visits by patients (aged 0-17 years) presenting to one of seven Pediatric Emergency Care Applied Research Network (PECARN) emergency departments (ED) with a potential firearm injury between January 1, 2011 and December 31, 2019. Visits were identified by ICD9 (E965, E955, E922, E985, E970) or ICD10 (X93-X95, X72-X74, W32-W34, Y22-24, Y35) codes for firearm injury. To identify most potential firearm visits, ICD codes for “other and unspecified firearms” and subsequent visit were included. ED visit narratives for these potential firearm injuries underwent medical narrative review by authors using a standardized template to determine if an acute firearm injury (AFI) occurred and the intentionality and type of firearm used, if documented. Sensitivity and specificity of ICD codes was calculated for identification of firearm injury, injury intent and type of firearm in this potential firearm injury population.

Results: 2,418 ED visits for potential firearm injury were identified in the PECARN database by the ICD codes listed above. Of these potential firearm injuries, AFI was confirmed in 1228 charts by medical narrative review. ICD codes identifying AFI were associated with 1395 of the potential firearm injury visits (Table 1). Medical narrative review and ICD codes agreed on identification of an AFI for 1201 charts. For this potential firearm injury population, ICD codes had a sensitivity of 97.8% (96-8-98.3%) and specificity of 83.7% (81.5-85.8%) for identifying AFI. Of confirmed AFI, medical narrative review was able to determine intent in 946 (77%) and type of weapon used in 131 (10.6%) (Table 2). Test characteristics for ICD code identification of intent and type of firearm are documented in Table 3. Conclusion(s): ICD codes have excellent sensitivity and reasonable specificity in identification of pediatric firearm injury in this pool of potential firearm injuries. They have variable sensitivity and specificity in identifying intent and poor sensitivity in identifying type of firearm. Future directions include investigating other methods of identification, such as natural language processing.
Table 1:Potential Firearm Injuries as Identified by ICD Codes
Table 2: Intent and Type of Firearm for Acute Firearm Injuries (Nf1228)