Case identification of non-traumatic brain injury in youth using linked population data

We present a systematic approach to case identification of non-traumatic acquired brain injury in paediatric populations. We compiled ICD-9, ICD-10, and Read codes which were examined by a clinical consensus committee for inclusion in a case definition of non-traumatic brain injury in children. Using multiple datasets contained in the New Zealand Integrated Data Infrastructure, we identified over 8000 cases of NT-ABI among the youth population at a rate of 0.52% in the five year period examined. Our approach makes a significant contribution to NT-ABI research by using multiple classification systems to identify a wide range of potential codes and an expert review process to reach consensus regarding inclusion. In addition, results from an application of the method are presented for ages 0–24 years, including and excluding birth trauma, to account for international differences in classification of acquired brain injury.

Case identification is the foundational step for further research using population data to identify and monitor the healthcare service needs and utilization of clinical populations [28‐30]. Future studies could apply this definition to administrative data with the aim of improving the changing needs of people with NT-ABI. Service utilization is high in people living with a non-traumatic brain injury acquired in childhood [20]. Studies within distinct diagnostic groups indicate that education and quality of life sequelae for youth living with a NT-ABI play a significant role in longer-term outcomes across the lifespan [9‐11, 31] thereby presenting diverse support needs across health, education, employment, and social services. Monitoring trends in the data for the population of people with NT-ABI as a whole will contribute to the understanding of the unique service needs of this group of people.

A method for identification of NT-ABI allows for the investigation of the costs of ongoing service utilization following non-traumatic injury. Epidemiological measures of the costs associated with NT-ABI require a definition that can be used in population data. Calculation of the years of life lost (YLLs) and years lived with disability (YLDs) can be combined to an estimate of disability adjusted life years (DALYS). Across the age spectrum and a range of causes, neurological disorders carry a significant global burden and in 2016 had the highest cost of DALYs [32]. In children neurological injuries contribute to a high global burden in part due to years lived with disability from a young age [33]. Our results indicate the highest rate of NT-ABI occurs in those aged 0–4 years which is consistent with other reports of high rates of neurological insult and traumatic injury in the preschool years [33, 34]. The effects of early injury on developmental trajectory and on the higher number of years lived with disability contribute to the burden of NT-ABI in children [2, 35]. Burden of injury studies in TBI across all ages suggest high costs associated with injury in the first 12 months and ongoing costs associated with moderate to severe injury [36, 37]. Case identification of moderate to severe NT-ABI in young people could be used to estimate costs associated with persistent disability across the sub-groups of brain injury and identify the costs associated with specific service use.

We aimed to provide a method of case identification for NT-ABI in young people using national administrative data but the method was not designed with the intention of estimating prevalence. Several authors have discussed the limitations of administrative data for prevalence estimates and warned against using such estimates for funding decisions [13, 14]. Prevalence estimates from population-level administrative data are generally accepted to underestimate the brain injury rate substantially. In a validation study of the application of ICD-10 codes to a cohort of cases known to have TBI, only 18% of known TBI cases had appropriate ICD10 codes in administrative data [15]. Studies comparing TBI codes present in hospital level administrative data with medical records indicate that the rate of TBI is underestimated [38‐40]. Codes for skull fractures, intracranial lesions, and neurosurgical procedures may be more reliably present in hospital level data [40] while neurological conditions may be underestimated [39]. Internal neurological events leading to NT-ABI are therefore also likely to be underestimated in the hospital level data that feeds into population level datasets such as the IDI. A limitation of our method of case identification is that it is currently unvalidated and therefore the degree of undercount or false positives is unknown. Future studies using medical records to examine the application of appropriate NT-ABI codes in the youth population from medical records through to administrative data would assist in validating case identification. This would allow an iterative approach to refine this method of case identification.

The five-year rate of NT-ABI in our cohort will significantly underestimate the true rate. We used stringent criteria for identifying NT-ABI cases and included only codes where a brain injury was specified in the code or considered likely to have resulted in a brain injury according to the clinical consensus committee. There is no classification system for the severity of NT-ABI, and mild NT-ABI is likely undetected in hospital records that feed into administrative data. Therefore, the NT-ABI cases identified will represent only a proportion of severe and moderately severe NT-ABI cases. Clear severity classification guidelines exist for traumatic injuries [41]. Even with this system, it is well documented that mild traumatic injuries are more likely to be absent from administrative data [15]. To provide incidence estimates requires identification of the first time an NT-ABI code appears for a young person. This necessitates the availability of robust data with time coverage that extends beyond that currently present in the IDI. Our approach examined the rate of codes present in a five year period but does not provide information about the timing of injury. Calculation of incidence estimates is an important focus for future research as more data become available.

We used a clinical consensus approach similar to that used in the identification of youth mental health problems [21]. The clinical consensus committee that reviewed codes for inclusion in a definition of NT-ABI was comprised of clinicians with experience across a wide range of NT-ABI presentations. In New Zealand there is a single Pediatric Rehabilitation Service that provides specialist rehabilitation to children, therefore clinicians working in this service are familiar with the spectrum of causes of NT-ABI. However, a consensus committee comprised of experts across individual causes of neurological insult may have arrived at a different set of codes for inclusion. In depth knowledge of neurological conditions and their functional outcome over time is required to understand NT-ABI. Future studies could bring together clinicians with specific expertise in sub-groups of neurological conditions affecting children in an effort to determine codes for inclusion and to compare these with those identified here. Other approaches to case identification have also been reported. In their study of TBI, Chen and Colantonio (2011) used a systematic review of the literature to examine the use of ICD-10 codes for TBI. They identified 26 studies and noted a high degree of inconsistency between studies in use of ICD-10 codes [18]. The challenges in applying a systematic literature review approach to NT-ABI include the diverse range of neurological conditions that can cause NT-ABI. Decisions about which conditions to include in a literature search would still need to be informed by clinical knowledge of NT-ABI rehabilitation and ideally by consensus. Nevertheless, a systematic literature review would provide an alternative approach to case identification that could be compared to the clinical consensus method. An advantage of our method is the review of ICD-9, ICD-10, and Read codes to capture NT-ABI in administrative data that spans timeframes during which classification systems can change.

We found a clear socioeconomic gradient of case identification of non-traumatic ABI. Cases of NT-ABI were higher in young people living in the most economically deprived households. This gradient was present in birth trauma cases and NT-ABI cases with birth trauma excluded. Socioeconomic deprivation is a documented risk factor for traumatic brain injury [42‐44]. It is unclear why non-traumatic brain injuries occurring as the result of internal neurological processes are more common in high-deprivation households. Barriers to healthcare access and factors associated with deprivation may lead to circumstances where brain insults are more likely to occur for example increased childhood infections. Higher levels of socioeconomic deprivation in early childhood have been found to increase the risk for multimorbidity and chronic illness [45]. Socioeconomic factors represent a potentially modifiable risk factor that could reduce rates of NT-ABI in childhood. For traumatic brain injuries, prevention efforts have focused on reducing injury from road traffic accidents, falls, and sports injuries. This work demonstrates the possibilities for injury prevention that emerge from an understanding of modifiable risk factors. Our finding highlights the need to investigate the potential reasons for this deprivation gradient in future studies.

Most NT-ABI cases were identified from the NMDS dataset (hospital inpatient records), with birth trauma cases exclusively from this dataset. Given that children with medical events or diagnoses resulting in a non-traumatic brain injury are treated in a hospital setting, it is unsurprising that most cases were identified through the NMDS. No cases of NT-ABI came from the PRIMHD (community mental health) dataset. Although acquired brain injury is a risk factor for mental health problems [46], PRIMHD reflects tertiary-level mental health service use where mental health diagnoses are the focus. Physical health diagnoses underlying a NT-ABI would not necessarily appear in these services, and the lack of case identification through PRIMHD is not unexpected. Future studies using the IDI to explore NT-ABI could investigate mental health outcomes using PRIMHD rather than using this dataset to identify cases of NT-ABI. ACC (accident claims) and Socrates (disability support services) contributed a relatively small number of NT-ABI cases but did contribute to case identification. Previously, researchers have found that utilization of multiple databases in the IDI is preferable to capture cases more thoroughly [21]. Our findings suggest that using multiple datasets to identify cases of NT-ABI will result in better capture of cases, particularly for brain injuries occurring after the perinatal period.