Defining neurotrauma in administrative data using the International Classification of Diseases Tenth Revision

The World Health Organization (WHO) has recognized the importance of surveillance systems for neurotrauma to inform prevention programs, as well as to inform the population of the significance of brain and spinal cord injuries [1]. When creating a surveillance system the definition of the condition of interest will affect the estimate and comparability of the estimate both within a country and between countries. The use of administrative medical data is attractive for neurotrauma surveillance because it is collected systematically and it is less expensive than independently collecting data. It is particularly attractive in countries with universal health insurance because data collection will also be comprehensive. However, the use of administrative data requires the translation of the clinical definition of neurotrauma to the International Classification of Diseases 10^th revision (ICD-10) codes, which are used in administrative data in many countries including Canada, Australia, European countries, and in the US for mortality data. It is necessary to provide information for the newest system of coding given that the ICD-10 is substantially different than ICD-9. Using ICD codes can be problematic as lack of specificity, human error, lack of time when processing medical records, and ranges of codes may affect the quality of data collection. In addition, the sensitivity of the definition depends on the codes chosen. Using a broad set of codes may incorporate more people who have suffered a traumatic brain injury (TBI) or spinal cord injury (SCI), but may also result in false positives. As a result, it is important to determine the best set of codes that will be comprehensive but also discerning of diagnosis [2]. Neurotrauma surveillance is highly important given the considerable rates of death and disability resulting from TBI and SCI worldwide, and the high costs these injuries place on health care systems and individuals alike.

Traumatic brain injury is a leading cause of death and disability among children and young adults in North America. It is estimated that 1.5 million Americans sustain a TBI annually, with an estimated 5.3 million living with permanent residual disability [3]. A study in Canada analyzed data between 1992 and 2002 and found that the age-sex standardized incidence rate of TBI hospitalizations decreased from 83.1 per 100,000 population per year to 50.4 per 100,000 population per year during this time frame, still contributing significantly to disabilities in Canada [4]. A European review of mild to severe TBI found a range of incidence from 91-105 per 100,000 population per year to a high of 546 per 100,000 population per year; the higher estimate included hospital admission, emergency department visits and death, while the lower estimate only included hospitalized patients [5].

The incidence rate of SCI in Ontario has been estimated at 24.2 per million population in 2003 and 23.1 per million population in 2006 [6]. A worldwide review of incidence studies found a range of 10.4 to 83 per million population per year [7] Although SCI does not have a large incidence rate, these injuries are typically very serious, leading to long-term disability, poor quality of life, and mortality, with high social and economic costs to the individual and the community [6]. It is estimated that in the United States, 190,000 persons live with paralysis due to a spinal cord injury, with an estimated cost of $4.5 billion annually for their care, equipment, supplies and services [8]. Although incidence is known to be significant, it is not known how accurate these numbers are and whether comparisons between the U.S. and Canada are fair due to differences in the definition used, quality of data, and methods used to collect data.

Reviews of the incidence and prevalence of neurotrauma found large ranges in estimates worldwide with broad and narrow clinical, ICD-9 and ICD-10 definitions [5, 7, 9]. As a result, it is important to standardize definitions and methodology to accurately estimate incidence and prevalence. In Ontario, an initiative funded by the Ontario Neurotrauma Foundation (ONF) and advised by a steering committee of stakeholders from the Ministry of Health and Long Term Care (MOHLTC), The Ontario Agency for Health Protection and Promotion (OAHPP), SMARTRISK, and the Ministry of Transportation aimed to create a neurotrauma surveillance system in order to better inform prevention in the province. The steering committee was arranged by ONF as stakeholders from the results of the surveillance system. Additional conversations were held with the Centers for Disease Control and Prevention, those involved with the Victoria Neurotrauma Initiative in Australia, as well as coding experts of the Canadian Institute for Health Information. The goals of the surveillance system are to: 1. Contribute to the reduction of injuries and related deaths in Ontario by identifying, describing and quantifying neurotrauma; 2. Collect, process and analyze summary data on neurotrauma in Ontario regarding cause of injuries, risk factors, and estimates of incidence; 3. Increase awareness of neurotrauma as a public health problem in Ontario; 4. Assist neurotrauma prevention and treatment programs by providing the data to identify trends over time, to identify high risk groups and to evaluate programs; and 5. Support neurotrauma-related research by providing data and identifying research priorities. One of the first steps towards reach these goals was to determine a definition to identify those affected by neurotrauma in mandated collection of administrative data in the emergency rooms (National Ambulatory Care Registry System, NACRS) and hospitalizations (Discharge Abstract Database).

The purposes of this study are to 1) examine the ICD-10 codes used worldwide to define traumatic brain and spinal cord injuries, 2) examine code specificity and sensitivity for both types of injury 3) use international standards to inform ICD-10 codes to use for neurotrauma surveillance in Ontario.

To our knowledge, this is the first paper that systematically reviews ICD-10 codes worldwide for the purpose of neurotrauma surveillance. In light of the wider range of codes used, we make recommendations for surveillance in Ontario that may be applicable to other jurisdictions utilizing mandatory, linkable administrative data for surveillance.

Despite a recommended definition by the WHO, there is no international consensus on the ICD-10 codes to use for the definition of traumatic brain injuries. In designing a neurotrauma surveillance system in Ontario, the definition used would have important implications for data usage and interpretation, as a broad definition may result in inflated numbers and a narrow definition may exclude some who may have experienced a brain injury. A balance of sensitivity and positive predictive value is needed for a reasonable estimate of true incidence. The diverse definitions used worldwide may also reflect the range and quality of data (e.g. the way data are reported and used).

To decide which codes should be included for an Ontario neurotrauma surveillance system, we undertook a comparison of codes used internationally and gathered evidence on the reasoning for the inclusion of codes, and on the correlation between codes and brain injury (See Tables 5 and 6).

This review makes a recommendation for a standard Canadian definition of neurotrauma based on experience and studies of injury correlations. Since the international literature is inconsistent in defining brain and spinal cord injuries using ICD-10 codes, the definition used for neurotrauma (TBI and SCI) surveillance in Ontario was chosen based on international consensus between studies/reports, a review of the correlation between codes and brain injuries, a review of the reason for inclusion of codes in definitions used in other countries, and ultimately a discussion with the steering committee. Reviews of incidence and prevalence of TBI and SCI recognize the importance of an international consensus on definitions of these injuries based on evidence to allow comparability of data across countries.

Future research should explore the quality of data in Canada using a reabstraction study similar to the one conducted by Hagen in Norway [28]. Estimates of sensitivity, specificity, positive predictive value and likelihood ratio will quantify how incidence rates would be affected by the inclusion or exclusion of specific codes. Replication of coding quality studies in other countries would inform definitions to be used internationally. This study provides a preliminary definition based on codes used internationally and available quality indicators and known correlations. Moving forward, there are certain codes that still require discussion on the merit of inclusion. In particular the inclusion of unspecified codes is contentious due to inflated rates. From the data gathered by the CDC and previous reports, the proposed definition in Ontario will result in an underestimate in the rate of brain injuries by not including specific S09 codes in the definition. However, the decision not to include S09 codes will likely increase the specificity of the definition. From our experience and from recommendations from the CDC and journal articles, we decided not to include the codes in S09 in our definition. A study on the specificity and sensitivity of codes, in particular S09 would be helpful in making a more accurate estimate of rate. If the majority of any of the S09 codes, with a particular interest in S09.9, are in fact brain injuries from abstraction of medical records in accordance to clinical definitions, the inclusion of the S09 code or an adjustment in our rate estimates would be warranted. Our current definition is a conservative definition that would be applicable in other countries that similarly use ICD-10 administrative codes for morbidity and mortality. Continued research, discussion and monitoring of trends over time are recommended for both changes in coding practices and injury patterns. Ongoing review of coding practices, injury patterns and evolving evidence of related injuries should be conducted to review these recommendations over time.