Volker Winkler1, Jördis J Ott, Heiko Becher. 1. Unit of Epidemiology and Biostatistics, Institute of Public Health, University of Heidelberg, Im Neuenheimer Feld 324, 69120 Heidelberg, Germany.
Abstract
OBJECTIVES: The international classification of diseases (ICD) is used to code death worldwide uniformly and comparably. This study investigates the reliability of national ICD-10 coding practice by assessing agreement of two official coding offices in Germany. METHODS: Inter-observer agreement for coding of 372 quasi-randomly selected death certificates is measured by percentages of agreement and simple kappa coefficients. RESULTS: Only 209 (56%) deaths were coded with the same 3-digit ICD-10 code. Agreement of the main chapters according to ICD-10 is higher with 78.2% and a kappa statistic of 0.69 (CI 95%, 0.63-0.75). Examples show that the coding rules correctly applied to the information given on the death certificates may lead to different conclusions. CONCLUSIONS: Data show good agreement in the marginal distribution, and thus population frequencies of causes of death may be more reliable despite limited agreement between the two coding offices.
OBJECTIVES: The international classification of diseases (ICD) is used to code death worldwide uniformly and comparably. This study investigates the reliability of national ICD-10 coding practice by assessing agreement of two official coding offices in Germany. METHODS: Inter-observer agreement for coding of 372 quasi-randomly selected death certificates is measured by percentages of agreement and simple kappa coefficients. RESULTS: Only 209 (56%) deaths were coded with the same 3-digit ICD-10 code. Agreement of the main chapters according to ICD-10 is higher with 78.2% and a kappa statistic of 0.69 (CI 95%, 0.63-0.75). Examples show that the coding rules correctly applied to the information given on the death certificates may lead to different conclusions. CONCLUSIONS: Data show good agreement in the marginal distribution, and thus population frequencies of causes of death may be more reliable despite limited agreement between the two coding offices.
Authors: Yejin Mok; Kunihiro Matsushita; Yingying Sang; Shoshana H Ballew; Morgan Grams; Sang Yop Shin; Sun Ha Jee; Josef Coresh Journal: PLoS One Date: 2016-04-19 Impact factor: 3.240