Abstract
Based on computer linkage of death records and hospital discharge abstracts, underlying cause of death and discharge diagnoses are compared for 9,724 Vermont resident in-hospital deaths occurring between 1969 and 1975. The agreement between the diagnoses recorded in the two data systems provides a measure of the reproducibility of recording, abstracting, and coding practices. Using the first three digits of the International Classification of Diseases, the agreement between cause and closest medical record diagnosis was 72 per cent. Concordance declined by patient age and length of hospital stay and varied significantly by coded cause of death. A major source of variation was the hospital of death where agreement levels ranged between 45 and 84 per cent. The latter finding is regarded as a potential starting point for targeting investigation of sources of discrepancy and initiating efforts to improve diagnosis recording and coding in the two record systems. The value of both depends on continuing efforts to improve and maintain data quality.