The National Center for Health Statistics (NCHS) has recently developed a new cause of death coding system, MedCoder. It replaces the Mortality Medical Data System (MMDS) that has been in use for several decades.
MedCoder is a subsystem of NCHS’ National Vital Statistics System for coding causes of death indicated on the death certificates to the International Classification of Diseases 10th Revision (ICD-10) codes. It is based on a fusion of new technologies such as machine learning and natural language processing with well-established technologies such as rules-based programming and regression testing.
The pre-processing component contains rules for analyzing the sequence of causes reported on the death certificate. This component is focused on the dictionary and natural language processing and handles spelling and terminology variants.
The machine learning model is based on multiple years of coded mortality data where the primary focus is on high frequency causes of death. It produces ICD-10 multiple cause codes for high frequency causes based on the coding rules, as captured in the model.
The post-processing component is based on coding rules to produce record-axis codes, which are a processed version of all coded entities reported on the death certificate.
MedCoder currently relies on the rule-based subsystem ACME to generate ICD-10 Underlying Cause codes. ACME is a component of the former system, MMDS. For more information about ACME and how it works, please visit About the Mortality Medical Data System.
MedCoder was put in production on June 6, 2022. NCHS will complete the year 2021 using MMDS and the entire 2022 (and beyond) will be coded using MedCoder.
By design, MedCoder codes causes of death one record at a time and positions us well to handle FHIR (Fast Healthcare Interoperability Resources) transactions of mortality data.