Research Implications of ICD-10 Coding

Rx for Researchers Adopting and Adapting to ICD-10 Coding: Caution and Collaboration

Alex J. Mainor, Nancy E. Morden, Jeremy Smith, Jonathan Skinner, Stephanie Tomlin

The 2015 U.S. health care billing codes update, from ICD-9 to ICD-10, challenges researchers and could bias health measures. Here we illustrate the risks and suggest a collaborative approach to the problem.

A problem faces researchers now analyzing health care data from 2015 forward. On October 1, 2015, the U.S. abruptly switched health care billing codes, from the International Classification of Disease (ICD) Version 9 to Version 10 (ICD-9 to ICD-10). Claims researchers use billing codes to identify and measure the health and health care of populations. The switch introduces the risk of artificial changes in these measurements over time. To help researchers address the challenges, ICD crosswalks (i.e. translation software creating code "equivalents") are available on the web. Unfortunately, ICD-9 and ICD-10 definitions emerging from these resources do not always align. Poor alignment of code diagnosis "equivalents" produces discontinuity in measures of disease prevalence over time; it may also compromise risk-adjustment methods that rely on ICD disease classifications.

On this website, using analysis of Medicare data before and after the switch, we illustrate potential pitfalls researchers may encounter using crosswalks and new definitions. We test some available translations by measuring month-to-month coding frequencies of common conditions during the transition and reveal the potential discontinuity of measures. We will repeat and make public similar analyses of outpatient (Part B) Medicare claims, and later, we will tackle procedures codes, which are subject to the same challenges and risks as diagnosis codes.

We begin by exploring and testing discontinuity in the following definitions using the Dartmouth Atlas Project Medicare inpatient (MedPAR) claims data. We examine month-to-month rates of appearance of diseases before and after October 2015 using diverse resources for our disease definitions:

    1. CMS Chronic Conditions Data Warehouse (CCW) ICD-9 and ICD-10 definitions are used by CMS to create 60 chronic conditions flags in the CMS research data. The flags and definitions are also commonly used by researchers. These current, published definitions provide an ideal opportunity to test the impact of the code update on measures. The figure below illustrates the discontinuity discovered for three conditions with distinct patterns of change following ICD-10 adoption.

    2. General Equivalency Mappings Software (GEMS) is an ICD-9 to ICD-10 translation product developed by CMS and the CDC to support ICD-10 adoption. We test this tool on ICD-9 Charlson-Deyo disease definitions. Since these conditions are commonly used for risk adjustment, understanding direction and magnitude of discontinuities is essential to effective biostatistical modeling.

    3. Researcher published ICD-9 and ICD-10 translations: Canada began adopting ICD-10 in 2001; their researchers have since been tackling the ICD update challenge. We explored one Canadian team's rigorous approach to ICD translation of Charlson-Deyo definitions from ICD-9 to ICD-10. This team created an algorithm comparable to GEMs and used output combined with clinician judgement to inform both ICD-10 definitions and revision of previously-used ICD-9 definitions to optimize alignment over time. Exploration of discontinuities in these definitions reveals the outcome (and residual discontinuity) of extensive effort aimed at harmonizing past and present measurements.

A crowd-sourced ICD-9 to ICD-10 crosswalk test and solution website. Given the public-good nature of research methods, we propose a common web-page for carefully derived and, ideally, tested, definitions. We do not presume to judge which are best, but would simply post the computer code in files that are specific to both date and research group, allowing researchers to either reproduce previous studies, or to compare existing conversion approaches. We have begun by using the Dartmouth Dataverse website to post our code, and would provide the access for others to post valuable code, but we are open to alternative posting locations.

For collaboration opportunities or questions, please contact Stephanie Tomlin.

Figure 1.

Figure 1


© 2019 The Trustees of Dartmouth College | 35 Centerra Parkway, Lebanon, NH 03766 | (603) 653-0800

The Dartmouth Atlas of Health Care is based at The Dartmouth Institute for Health Policy and Clinical Practice and is supported by a coalition of funders led by the Robert Wood Johnson Foundation, including the WellPoint Foundation, the United Health Foundation, the California HealthCare Foundation, and the Charles H. Hood Foundation.