Background: Using hospital databases to retrieve information, doing epidemiologic researches and making administrative decisions extremely depend on an accurate classification based on clinical coding. The researchers aim was to determine the validity of diagnostic coding based on ICD10 .

Materials and Methods: 370 medical records were selected from KAUMS teaching hospitals in 2008. We abstracted and recoded the first - sequenced diagnostic codes based on ICD10. The validity was determined by the agreement between original codes and recodes as gold standard. The coding errors were classified into major and minor. The major code errors were those happened in nature and topography. Others were considered as minor. Possible determinants were studied through a checklist and observation methods. The data were analyzed through X², fisher test, OR, and CI 95% for OR.

Results: Diagnostic codes accuracy was 77.3%. There were 84 (22.7%) errors in diagnostic codes so that 28 (33.3%) of them were major and 56 (66.7%) were minor. Using coding book and not using abbreviation reduced errors significantly. Complete records review reduced errors . Documenting more information especially diseases etiology increased errors. In addition, the relationship between readability of records and code accuracy was not significant.

Conclusion: Majority�of diagnostic codes were accurate. to reduce current errors, Coders’ factors such as�more attention to available information and better documentation (e.g. not using abbreviation) can increase the quality of diagnostic coding and its databases.