Identification of the Most Appropriate ICD Version for Epidemiologic Studies: Description of a Statistical Method

• Activity Number: 440
• Type: Contributed
• Date/Time: Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ja K. Gu*+ and Luenda E. Charles and Cecil M. Curchfiel and Michael E. Andrew and John M. Violanti
• Companies: National Institute for Occupational Safety and Health and National Institute for Occupational Safety and Health and National Institute for Occupational Safety and Health and National Institute for Occupational Safety and Health and State University of New York at Buffalo
• Keywords: ICD ; Comparability Ratio ; Cause of Death

Abstract

In mortality studies, underlying cause of death has been coded using various International Classification of Diseases (ICD) revisions. Our objective was to identify the most appropriate ICD version for the Buffalo Police Mortality Data (BPMD), 1968-2005. Underlying causes of death were coded using either ICD8 (n=286), ICD9 (n=547), or ICD10 (n=221). All causes were converted into each ICD version using National Center for Health Statistics (NCHS) ICD-comparability ratios. Means, standard errors, and 95% confidence intervals (CI) were obtained for the difference between the number of deaths from the original and new ICD codes. Results were: ICD8, 2.5±3.3, 95% CI (-4.4, 9.4); ICD9, 0.6±0.8 (-1.1, 2.2); and ICD10, 0.7±0.8 (-1.0, 2.4). ICD9 is most appropriate for these data. This statistical method could be useful in studies to identify an appropriate ICD version for coding purposes.