Article Text

Variation in completeness of coding external cause of injuries under ICD-10-CM
  1. Christine C Stewart1,
  2. Gregory Simon1,
  3. Brian K Ahmedani2,
  4. Arne Beck3,
  5. Yihe G Daida4,
  6. Frances L Lynch5,
  7. Ashli A Owen-Smith6,7,
  8. Sonya L Negriff8,
  9. Rebecca Rossom9,
  10. Stacy A Sterling10,
  11. Christine Y Lu11,
  12. Michael Schoenbaum12
  1. 1Kaiser Permanente Washington Health Research Institute, Seattle, Washington, USA
  2. 2Henry Ford Health Center for Health Services Research, Detroit, Michigan, USA
  3. 3Kaiser Permanente Colorado Institute for Health Research, Denver, Colorado, USA
  4. 4Kaiser Permanente Hawaii Center for Health Research, Honolulu, Hawaii, USA
  5. 5Kaiser Permanente Northwest Center for Health Research, Portland, Oregon, USA
  6. 6Kaiser Permanente Georgia Center for Health Research, Atlanta, Georgia, USA
  7. 7Georgia State University, Atlanta, Georgia, USA
  8. 8Department of Research and Evaluation, Kaiser Permanente Southern California, Pasadena, California, USA
  9. 9HealthPartners Institute, Bloomington, Minnesota, USA
  10. 10Kaiser Permanente Division of Research, Oakland, California, USA
  11. 11Department of Population Medicine, Harvard Medical School, Boston, Massachusetts, USA
  12. 12National Institute of Mental Health, Bethesda, Maryland, USA
  1. Correspondence to Dr Gregory Simon, Kaiser Permanente Washington Health Research Institute, Seattle, WA 98112, USA; gregory.e.simon{at}


Introduction Information about causes of injury is key for injury prevention efforts. Historically, cause-of-injury coding in clinical practice has been incomplete due to the need for extra diagnosis codes in the International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) coding. The transition to ICD-10-CM and increased use of clinical support software for diagnosis coding is expected to improve completeness of cause-of-injury coding. This paper assesses the recording of external cause-of-injury codes specifically for those diagnoses where an additional code is still required.

Methods We used electronic health record and claims data from 10 health systems from October 2015 to December 2021 to identify all inpatient and emergency encounters with a primary diagnosis of injury. The proportion of encounters that also included a valid external cause-of-injury code is presented.

Results Most health systems had high rates of cause-of-injury coding: over 85% in emergency departments and over 75% in inpatient encounters with primary injury diagnoses. However, several sites had lower rates in both settings. State mandates were associated with consistently high external cause recording.

Conclusions Completeness of cause-of-injury coding improved since the adoption of ICD-10-CM coding and increased slightly over the study period at most sites. However, significant variation remained, and completeness of cause-of-injury coding in any diagnosis data used for injury prevention planning should be empirically determined.

  • Epidemiology
  • Coding Systems
  • Suicide/Self-Harm

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Coding for external cause of injury under ICD-9-CM varied widely across health systems and over time.


  • Coding for external cause of injury under ICD-10-CM varied widely across health systems, and was generally higher in states with historical mandates to code external cause.


  • Researchers and policy makers relying on health records data to assess intent or cause of injury should carefully examine completeness of coding.


Injuries and poisonings figure prominently in top causes of death in the USA and are the reason for one in five emergency department (ED) visits1 and 6% of hospital stays.2 3 Information about cause and intent of injuries and poisonings guides prevention efforts, and the International Classification of Diseases (ICD) coding system includes specific indicators for the intent and mechanism of injuries. Use of those external cause codes is necessary to distinguish accidental injury from assault or self-harm and to distinguish firearm injuries from other mechanisms of injury. However, collection of external causes is not mandated federally, and relevant regulations vary by state.4 5 The nature of state mandates varies and examples of variation include both enacted laws and published guidelines for collection and submission of external cause-of-injury data.6 7

Prior to October 2015, the International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) was used for coding non-fatal injuries and poisonings. In addition to the primary injury or poisoning code (in the range 800–999), a separate ‘E’ code (E800–E999) was needed to specify external cause, as shown in figure 1A. In a study of 10 health systems in nine states, rates of recording an external cause varied considerably by state, health system and treatment setting.8

Figure 1

Change in scheme for coding of injuries and poisonings from the International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) to ICD-10-CM. (A) In ICD-9-CM, an additional external cause code was needed to completely specify (grey shaded areas are completely specified) both dimensions of injuries and poisoning: type and external cause. (B) In ICD-10-CM, both dimensions are specified in a single code for poisonings and toxic effects, but not most injuries, which still require an additional code. NSAIDs, non-steroidal anti-inflammatory drugs.

With the change to ICD-10-CM, external cause was embedded in the mechanism codes for most poisonings but not for most injuries. As shown in figure 1B, a single digit in the codes for poisoning and toxic effects (T36–T65) indicates the external cause. This is not the case for injuries (S00–S99), which still require a second code to indicate cause.

Because of the consolidation of information into one code in some cases, the increased use of electronic health record (EHR) systems to facilitate optimal diagnosis recording and the spread of state recording mandates,5 it is likely that completeness of external cause recording has improved. The Agency for Healthcare Research and Quality assessed the external cause recording in Healthcare Cost and Utilization Project (HCUP) data between 2016 and 2019 and found at least 90% completeness in three-fourths of participating states for inpatient stays and two-thirds of emergency encounters.8 9 In both settings, the vast majority of states reported >80% completeness for external causes. However, the definition of ‘injury’ in this analysis included drug and non-drug poisonings,10 which do not require an additional code. Here we measure completeness of external cause coding specifically for diagnosis codes (primarily injuries) that still require a separate code for external cause in 10 healthcare systems in nine states.


Diagnoses were extracted from federated data warehouses containing electronic medical record and insurance claims data11 at each of 10 participating health systems: HealthPartners (Minnesota), Harvard Pilgrim Health Care (Massachusetts), Henry Ford Health (Michigan) and seven Kaiser Permanente regions in California, Colorado, Georgia, Hawaii, Oregon and Washington.

These healthcare systems served an annual population of over 12 million members in 2019 which reflects the demographic diversity of the associated geographical areas.12 13 Responsible institutional review boards at each health system approved waivers of consent for this research use of deidentified records data. Patients and members of the general public were not involved in these analyses of records data.

All recorded diagnosis codes from emergency and inpatient encounters were extracted from October 2015 through December 2021 (code lists for included injuries and external causes will be included in the online supplemental materials). Briefly, in addition to the S chapter, the injury definition included all T07, T15–34, T66–70 and T79 codes, as well as all T14 codes (injury of unspecified body region), except for T14.91 (suicide attempt, which was considered an external cause). External cause codes included T14.91 and V00-Y69. For every encounter with a qualifying injury code as the principal or primary diagnosis, the presence or absence of an accompanying external cause code was noted. A secondary analysis removed the requirement that injury or poisoning be the principal or primary diagnosis. While guidelines or mandates only apply to encounters with those principal or primary diagnoses, accuracy of principal or primary coding could vary across care settings.

Supplemental material


The average number of ED encounters analysed was 159 520 per month, with numbers contributed by each site ranging from 992 to 64 401. The average number of inpatient encounters was 12 218 per month, with numbers contributed by each site ranging from 66 to 3534.

Figure 2 shows the proportion of encounters with a primary or principal injury diagnosis accompanied by an external cause code from October 2015 (the first use of ICD-10) to the end of 2021, stratified by care setting and if the state in which the health system operated had a mandate regarding use of external cause coding.5 For all sites combined, the proportion with external cause coding increased slightly from 2015 to 2021, from 82.9% to 88.30% among ED encounters (Mantel-Haenszel χ2=4442, df=1, p<0.0001) and from 73.3% to 79.5% among inpatient encounters (Mantel-Haenszel χ2=329, df=1, p<0.0001). For all years combined, the proportion of encounters with external cause coding varied significantly across sites (χ2=405 529, df=1, p<0.0001 for ED and χ2=18 375, df=9, p<0.0001 for inpatient). The proportion with external cause coding was higher for sites with state mandates for both ED (91% vs 76%, χ2=149 848, df=1, p<0.0001) and inpatient encounters (χ2=4287, df=1, p<0.0001).

Figure 2

Completeness of external cause coding with injuries in 10 US health systems from 2016 to 2021, stratfied by site of care (emergency department or inpatient) and presence/absence of historical state mandate for coding of external cause.


Completeness of external cause recording for injuries has increased, and the variability decreased, since 2010,5 even in this subset of codes where ICD-10-CM does not embed the external cause in the primary diagnosis code. The time trends did show some discontinuities and a slight upward trend in the early part of the study period. Interestingly, there was no disruption in 2020 during the COVID-19 pandemic, suggesting the mechanisms supporting external cause coding are resilient to changes in volume and workflow. Variability still existed between some health systems and between treatment settings.

State mandates were associated with consistently high rates of external cause recording in our sample, which suggests these regulations are effective for improving the quality of data available to inform public health policy.

The states in which our healthcare systems operate were represented in the HCUP analysis, with the exception of one site in which the state did not participate in the HCUP Emergency Department Report.9 10 These findings are broadly consistent with the HCUP findings in that most sites perform well, with a few exceptions. Our findings consistently indicate more complete recording in EDs than inpatient settings, whereas the opposite pattern is seen in the HCUP analysis, although the differences there are small. In general, the rates of external cause coding presented in the HCUP rates are higher than in this analysis. This difference is consistent with the inclusion of poisoning codes in the HCUP analysis which automatically qualify as complete.

Variation in completeness could also be attributed to health system policy and clinical support software. The differences we find within sites between treatment settings highlight the role of these health system-specific factors, such as the involvement of professional medical coders. EHR systems include customisable text-to-diagnosis code mapping, and this has been linked to differential rates of diagnoses observed between health systems using different implementations of the same EHR product.14

These specific findings should not be generalised beyond the participating health systems. In fact, the data presented here highlight the variation that may be hidden in state-level rates, arguing against the generalisability of findings from these or other health systems. The completeness of cause-of-injury coding in the underlying data should be assessed before using diagnosis data to guide or assess prevention efforts at all levels from the health system to nationwide.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by Kaiser Permanente Interregional Institutional Review Board. The Kaiser Permanente Interregional Institutional Review Board granted a waiver of consent to use deidentified records data for this research.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • X @GregSimonKPWHRI

  • Contributors Conception and design: CCS, GS, MS. Data analysis: CCS. Drafting of manuscript: CCS. Critical revision of manuscript: GS, BKA, AB, YGD, FLL, AAO-S, SLN, RR, SAS, CYL, MS. CCS and GS had full access to data and can take responsibility for all contents of the manuscript.

  • Funding Funded by NIMH Cooperative Agreement (U19 MH121738).

  • Disclaimer The views expressed here are those of the authors, and not necessarily of the National Institute of Mental Health, the National Institutes of Health or the federal government.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.