Comparison between manual and automatic coding of medical record statistical cards at a university hospital

Abstract

The coding error rate of systems for medical record statistical cards (MRSCs) throughout health services is about 30%. A program using automatic coding has been developed at the Institute of Clinical Surgery II, Padua University Hospital, with a view to reducing this percentage. Out of an overall sample of 4776 MRSCs from all departments of the hospital, 54 were automatically coded at our institute. Categories of discrepancy between the discharge diagnosis codes of the 4722 manually coded MRSCs and the other 54 MRSCs were classified as follows: types I-III, diagnosis assigned to an erroneous under-class, class or heading (ICD-9) respectively; type IV, incorrect diagnosis formulation precluding code assignment; type V, two or more discrepancies on MRSC; and type VI, secondary diagnosis not coded. Discrepancy rates were as follows: 22.3% and 0.0% for type I; 21.3% and 0.0% for type II; 17.6% and 0.0% for type III; 1.9% and 0.0% for type IV; 5.8% and 0.0% for type V; 31% and 1.9% for type VI. Code discrepancy rates for surgical procedures, which were also compared, ranged from 7.0 to 12.5% for manual coding, while no discrepancy was observed in automatically-coded MRSCs. The results clearly demonstrate the utility of the system reported on, and it is suggested that it should be used in a modified form in other hospital departments.

Automated Coding of Patient Discharge Summaries Using Conceptual Graphs

“…In Europe, the International Classification of Diseases and Operations for Hospitals (HCIMO) is a translation and an extension of the ICD-9-CM [4]. Despite its large and often mandatory use in many countries, the encoding procedure is still a burdensome task with a significant error rate and inconsistency between human coders [5]. Providing a system which contributes to the design of a generic system that generates classification codes from natural-language PDSs, to support resource management and public health research, is a recommendation of the European AIM office [6].…”