Assistance in identifying hospital stay description files with coding errors

摘要

Assistance process for automated identification of hospital stay records likely to include coding errors. The method includes a learning phase (A) and a prediction phase (B). The learning phase (A) includes the generation of a cartography (C) in a K2-dimensional space comprising reference points representing respectively reference stays. The prediction phase includes: a projection of a vector representing a stay to be evaluated in K2-dimensional space; an identification (B220) in space at K2 dimensions of reference points closest to the projected point; a calculation (B230) of an estimated score for the stay considered by comparison between the severity levels of the closest reference stays identified and the level of severity of the stay considered; a weighting (B240) of the score estimated by at least one weighting coefficient obtained by applying at least one business rule to the data describing the stay in question so as to obtain a weighted score (SP); an identification (B250) of the stay as being or not likely to include coding errors on the basis of the weighted score (SP). Figure for the abstract: fig. 2b