Is the Parsonnet's score a good predictive score of mortality in adult cardiac surgery: Assessment by a French multicentre study

Objective: Parsonnet proposed a preoperative score (''initial Parsonnet's score'', which predicts the hospital mortality of adult cardiac surgery. This score was then modified by including several risk factors used in the 'SUMMIT' system (''modified Parsonnet's score'', 44 variables). We wanted to assess the predictive value of these two scores in a French surgical population. Methods: From December 1992 to April 1993, in France, we organised a prospective multicentre study on adult cardiac surgery mortality and morbidity. Data on 6649 patients were included. We analysed statistically the predictive value of each risk factor and of the two scores on mortality and morbidity at one month. Results: Only 6 of the 15 variables of the initial Parsonnet's score and 19 of the 44 variables of the modified Parsonnet's score significantly influence hospital mortality. Both scores are able to predict hospital mortality and severe morbidity, but the modified Parsonnet's score has the best predictive value (initial Parsonnet's score: odds ratio by point of score = 1.01, area under the roc curve = 0.64; modified Parsonnet's score: odds ratio by point of score = 1.05, area under the roc curve = 0.70). Conclusions: This study shows that the Parsonnet's scores are predictive, but that these scores remain imperfect: many risk factors are non significant, the initial Parsonnet's score has a moderate predictive value, and the modified Parsonnet's score is too complex (44 variables). Thus, we have built a new score for cardiac surgery in French adults. (C) 1997 Elsevier Science B.V.