The effects of external-cause-of-injury coding system transition on injury hospitalization trends – Winner of the Jess Kraus Award

The Jess Kraus Award is given each year to the authors of the best paper published in Injury Epidemiology. This Year’s winning paper exemplifies several criteria for the award, such as simplicity and clarity. In this guest blog, Editor-In-Chief Professor Guohua Li, chats with the lead author of the study, Dr Svetla Slavova on the importance of the paper.

Editor in Chief, Guohua Li: Could you please tell us a bit about your study and what motivated you to do this research?  

Svetla Slavova – Winner of the Jess Kraus Award

Svetla Slavova: Injury morbidity surveillance and epidemiology are based heavily on administrative billing data. External-cause-of-injury codes listed on claims records are used to describe an injury by mechanism and intent but are not required codes for billing reimbursement. On October 1, 2015, the billing coding in the U.S. switched from one coding system, ICD-9-CM, to a considerably different coding system, ICD-10-CM, which includes about 5 times as many codes.

Preliminary studies on the transition to ICD-10-CM reported that the coding time for an inpatient hospitalization increased on average by 60% and there was a concern that the decreased coding productivity would lead to omission of external-cause-of-injury codes. The initial motivation for our study was to evaluate if there was a drop in the percentage of Kentucky injury hospitalization records (e.g. records with a principal diagnosis for fracture, cut, strain, etc.) that were supplemented with external-cause-of-injury codes describing the mechanism and the intent immediately after the switch to ICD-10-CM coding. Another motivation for this study was to provide a methodology that researchers can replicate to assess trends in injuries in ICD-9-CM- and ICD-10-CM-coded hospital claims data.

GL:  Injury coding is tedious and meticulous work. What were the major challenges to make the transition from ICD-9 to ICD-10?

SS: The transition to ICD-10-CM coding was expected to be challenging for medical coders; however, their professional organizations have developed and provided a variety of educational materials, training and continuing education.

The coding transition was also expected to be challenging for epidemiologists because of the volume of new injury codes, the introduction of injury codes without analogy in the ICD-9-CM coding system, and the inclusion on few concepts (e.g. “encounter” of care, “suspected” abuse, inclusion of “intent” information within the poisoning diagnoses codes).

The National Center for Health Statistics (NCHS) is the agency responsible for the use and development of the ICD-10-CM and in anticipation of the coding transition, the NCHS developed the “Proposed Framework for Presenting Injury Data using ICD-10-CM External Cause of Injury Codes”, followed by the “Proposed ICD–10–CM Surveillance Case Definitions for Injury Hospitalizations and Emergency Department Visits”.The Safe States also formed the Injury Surveillance Workgroup 9 on the Transition from ICD-9-CM to ICD-10-CM. This group of experts developed guidance for epidemiologists titled “Guidance for Analysis and Reporting of Injuries by Mechanism and Intent”. All of this work was done prior to anyone having real ICD-10-CM-coded claims data.

This work is still ongoing, but it has been instrumental for building capacity for ICD-10-CM-based injury epidemiology and surveillance.

The CDC’s National Center for Injury Prevention and Control (NCIPC) collaborated with the Council of State and Territorial Epidemiologists (CSTE) to form the CSTE Injury ICD-10-CM Transition Workgroup.  The workgroup aimed at engaging state and local injury epidemiologists in assessing the quality of their jurisdictional ICD-10-CM-coded hospital and ED data in the first months after the coding transition.

The workgroup members also started testing the proposed external cause matrix for reporting of ICD-10-CM-coded injury data in order to develop standardized definitions for injury morbidity surveillance. This work is still ongoing, but it has been instrumental for building capacity for ICD-10-CM-based injury epidemiology and surveillance. I participated in this group for 2 years and provided Kentucky data for the workgroup reports and it has given me an understanding and appreciation for the challenges of accurate and informed analysis of injury data, another motivating factor for the research paper.

GL:  Would you please describe the data you used in the study?  On average, how many injury hospital admissions per month are there in KY?

SS: We used Kentucky statewide inpatient hospital discharge administrative billing data for Kentucky residents treated in Kentucky acute care hospitals from 2012-2017. On average, there are about 2,000 Kentucky resident inpatient hospital discharges every month with a principal diagnosis for injury.

GL: What were the most notable findings from your study?

SS: Our study suggested that the CDC/NCHS’s proposed ICD-10-CM injury reporting framework provides a reasonable transition from the ICD-9-CM framework. This means that most of the statistically significant immediate disruptions in the Kentucky injury hospitalization trends could be explained by the structural and conceptual changes in the ICD-10-CM system.

For example, colleagues should be aware that the ICD-10-CM has a new rule on injury intent coding stating that “undetermined intent” is only for use if the documentation in the record specifies that the intent cannot be determined. Therefore, when the intent is unknown or unspecified, it should be coded as accidental (unintentional). This rule reflected in an immediate, significant increase in the percentage of injury hospitalization records in Kentucky that were coded as unintentional and a reciprocal drop in the undetermined intent injuries.

Colleagues from other jurisdictions may observe significant jumps in the rates of poisoning and suffocation injuries due to changes in how these injuries are coded in the ICD-10-CM. It was important for the Kentucky stakeholders to learn that the drop in the completeness in the external-cause-of-injury coding in the statewide hospital data was temporary, specifically within the first month after the transition.

GL:  What are the implications of your findings for researchers and practitioners in the field?

SS: Our findings are intended to raise awareness that changes in the ICD-10-CM coding system must be understood to assure accurate interpretation of injury trends. The methodology (segmented regression analysis for interrupted time series) presented in our paper is applicable for evaluation of the effect of the ICD-10-CM transition on any health condition trend (not just injuries).

GL: You have six coauthors from three different institutions. How did you form and manage this interdisciplinary team?

We are hoping that colleagues from other states will do analysis of their jurisdictional data and will contribute to the body of literature on the topic.

SS: We had an exceptional multidisciplinary team. Barbara Gabella chaired the work of the Safe States Injury Surveillance Workgroup 9 on the Transition from ICD-9-CM to ICD-10-CM. Ms. Gabella assisted with data interpretation and her comments on the manuscript helped translate this work to the broader audience of applied injury epidemiologists working in public health. Dr. Huong Luu assisted with the data analysis and injury trend visualizations. It was informative to discuss with Dr. Sergey Tarima different issues related to the statistical approach and model fit. Prior to deciding on the final data analysis, we jointly investigated multiple statistical methods potentially applicable to our data. Drs. Costich and Bunn helped with the manuscript development and study result discussion. Judith Fields is a medical coder. She provided interpretation and guidance for us on specific codes and changes between the two coding systems that would drive some of the trend interruptions that we observed.

GL: Do you intend to extend the study to a nationwide evaluation?

SS: We have some promising ideas but we are also hoping that colleagues from other states will do analysis of their jurisdictional data and will contribute to the body of literature on the topic. In an attempt to facilitate such research, in Appendix A of the manuscript, we provided a sample data set, the SAS code for the model used in our paper, and the SAS output.

GL:  Any advice for graduate students and junior faculty working in the field of injury epidemiology and prevention?

SS: It is an exciting time for graduate students, junior faculty, and even experienced injury epidemiologists and public health researchers. There is so much methodological work to be done in the field as a result to the transition to the ICD-10-CM coding. We need to validate new injury codes, new injury code groupings, and develop consensus on injury definitions. Studies utilizing state and local data are very important to test proposed definitions before they are utilized as standard definitions for jurisdictional comparison.

GL: Thank you for your time. I look forward to seeing you in the award ceremony at Columbia University on May 23rd.

Svetla Slavova is an Associate Professor in the Department of Biostatistics at the University of Kentucky and a faculty associate with the Kentucky Injury Prevention and Research Center. Her main research interests are in drug overdose surveillance methodology, injury surveillance quality improvement, analytical enhancement of prescription drug monitoring programs, and motor vehicle crash modeling. She is currently the principal investigator on grants from the U.S. Department of Justice, the Centers for Disease Control and Prevention, and the Food and Drug Administration.

View the latest posts on the On Health homepage