Zuverlässigkeit und Fehlerquellen in einer klinischen Basisdokumentation - ein kritischer Erfahrungsbericht

Abstract

Since 1. 1. 1985, computer-aided basic medical documentation has been routine at the University Orthopedic Clinic in Friedrichsheim, near Frankfurt. In addition to data on patient's histories, all data needed to satisfy the criteria of the Federal Directive on Operating Cost Rates are gathered. The diagnoses are stored in clear text, in a modified Eichler code, and according to ICD 9. Conversion from the Eichler code to ICD 9 is almost fully automated. In a study covering 100 hospitalized cases the following findings were obtained relating to sources of error and reliability: Without any additional in-house plausibility checks, the rate of error in the ID code, created by coding family name, date of birth, and sex, was 7%. In clear text all diagnoses except one and all forms of therapy were correctly reproduced as contained in the medical report. On the other hand, 7% of the conversions into the Eichler code contained errors. The reason for the difference in the quality of data is pointed out. In some of the other surveys, e.g., of infection rates, the rates of error were very high; most errors had been caused by the ward physicians. Data quality is enhanced by exploitation of routine process data when these control administrative procedures or are used for communication between physicians, since they then become relevant to actions and decisions and hence have to be reliable, regardless of documentation purposes.