Evaluating a computerized tool for coding patient information

Computerized tools may be useful in speeding up and facilitating the laborious task of coding patient information. This paper describes a method of objectively evaluating their efficiency. 38 study subjects were randomly assigned to a manual coding group or an automated coding group, with stratification according to two variables (used to coding yes/no, physician yes/no). Subjects then coded the same standardized set of diagnoses in a limited time. The numbers of exact codes retrieved were compared using a global analysis of variance model. The two groups were not significantly different with regard to the number of physicians (p = 0.74) and the number of usual coders (p = 0.52) they included. Significantly more exact codes were achieved in the group using automated coding than among the manual group (p = 0.04). Physicians were significantly more efficient at coding than non-physicians (p = 0.02). This study describes an objective means of evaluating the performance of an automated coding tool. It shows that better results were achieved with the computerised compared to the manual method, even when the superior abilities of physicians were taken into account.