GMS Medizinische Informatik, Biometrie und Epidemiologie (Apr 2005)
Coding training for medical students: How good is diagnoses coding with ICD-10 by novices?
Abstract
Teaching of knowledge and competence in documentation and coding is an essential part of medical education. Therefore, coding training had been placed within the course of epidemiology, medical biometry, and medical informatics. From this, we can draw conclusions about the quality of coding by novices. One hundred and eighteen students coded diagnoses from 15 nephrological cases in homework. In addition to interrater reliability, validity was calculated by comparison with a reference coding. On the level of terminal codes, 59.3% of the students' results were correct. The completeness was calculated as 58.0%. The results on the chapter level increased up to 91.5% and 87.7% respectively. For the calculation of reliability a new, simple measure was developed that leads to values of 0.46 on the level of terminal codes and 0.87 on the chapter level for interrater reliability. The figures of concordance with the reference coding are quite similar. In contrary, routine data show considerably lower results with 0.34 and 0.63 respectively. Interrater reliability and validity of coding by novices is as good as coding by experts. The missing advantage of experts could be explained by the workload of documentation and a negative attitude to coding on the one hand. On the other hand, coding in a DRG-system is handicapped by a large number of detailed coding rules, which do not end in uniform results but rather lead to wrong and random codes. Anyway, students left the course well prepared for coding.
