The Changing Picture of Physical Child Abuse in Trauma Centers: How ICD-10 Alters Prior Understanding of Pediatric Non-accidental Trauma

Background: Identification of pediatric non-accidental trauma (NAT) in large database research has traditionally relied on International Classification of Disease (ICD) codes. Much of the literature has utilized ICD-9-encoded data and yet under-diagnosis was a known issue of this revision. ICD-10, implemented in 2015, introduced “suspected” abuse codes, which have the potential to address that ICD-9 shortcoming by enabling abuse coding in clinical scenarios where abuse is highly suspected but not confirmed. This study sought to understand how ICD-10 implementation altered identification of NAT and changes understanding of NAT from prior ICD-9-encoded data.

Methods: Patients under age 19 were identified in the National Trauma Data Bank (NTDB) using ICD-9 codes from 2007-2015 and ICD-10 codes from 2015-2018. NTDB is a national collection of patient data from participating trauma centers. Demographics, resource utilization, injury patterns, and outcomes were analyzed using descriptive statistics.

Results: Compared to ICD-9 NAT cases, ICD-10 cases were less likely to be white or uninsured and more likely to be Black/African American, Hispanic, or publicly insured. ICD-10 cases were also less severely injured (injury severity score of 11.25 versus 13.03) and required shorter average length of stays in the hospital and ICU (6.26 and 2.14 days, respectively, versus 7.08 and 2.73 days). Mortality rate was 6.05% for ICD-10 cases, versus 8.45% for ICD-9 cases. ICD-10 cases were significantly less likely to suffer rib fracture, intrathoracic or intraabdominal organ injury, and upper or lower extremity fracture and more likely to suffer skull fractures and contusions/abrasions.

Conclusion: As expected, ICD-10 and its more flexible abuse codes appear to have increased NAT identification, with a 28.9% increase in cases in NTDB between the first three years of ICD-10 and last three years of ICD-9. However, while 98.8% of ICD-9 codes followed CMS coding guidelines and therefore also have an identified perpetrator of abuse in the database, only 63.4% of ICD-10 cases follow updated CMS coding guidelines. Thus, while every ICD-10 NAT case should be classified as suspected or abused, one-third of cases lack that categorization. Furthermore, only 24.4% of ICD-10 NAT cases now have an identified perpetrator.
From this analysis of NTDB, NAT cases encoded by ICD-10 significantly differ from those encoded by ICD-9 in terms of demographics, resource utilization, injury patterns, and outcomes. However, with current coding inconsistences, analysis of the differences in suspected versus confirmed cases, as well as study of perpetrators of abuse, are plagued by missing data. Without uniformity in coding, a major tool for large database analysis of NAT that was previously useful will be lost. Overall, the profile of pediatric NAT appears to have changed, but the confidence in these findings and future database studies of NAT relies upon increasing uniform usage of ICD-10 for NAT.