To compare the diagnostic performance of 4 clinical prediction scores for heparin-induced thrombocytopenia (HIT) in patients undergoing cardiac surgery with cardiopulmonary bypass (CPB).

Bicentric retrospective observational study.

Two tertiary university hospitals in France.

Adult patients who underwent cardiac surgery with CPB between 2014 and 2021 and for whom HIT testing was requested during the postoperative period.

None.

HIT diagnosis was established using a standardized approach combining anti–platelet factor 4/heparin IgG enzyme-linked immunosorbent assay, a functional platelet activation test, and multidisciplinary clinical adjudication. Among 283 patients investigated for suspected HIT, 55 (19%) were classified as HIT-positive. The diagnostic performance of 4 clinical probability scores, the 4Ts score, the HIT Expert Probability score, the cardiopulmonary bypass score, and the Groupe Français d’Étude sur l’Hémostase et la Thrombose score, was assessed using receiver operating characteristic (ROC) curves and formally compared using the DeLong nonparametric test for correlated ROC curves. Areas under the ROC curve ranged from 0.79 (cardiopulmonary bypass) to 0.86 (HIT Expert Probability), with no statistically significant differences observed between scores. Using optimized thresholds, negative predictive values ranged from 94% to 96%, whereas positive predictive values remained modest (31%-49%). HIT-positive patients exhibited a characteristic biphasic platelet count pattern, with an initial postoperative decline followed by a delayed second nadir around postoperative day 10. Among HIT-positive patients, the 30-day and 1-year mortality rates were 9.1% and 14.5%, respectively. Median intensive care unit and hospital lengths of stay were 11 [6-18] and 23 [18-32] days.

In patients undergoing cardiac surgery with CPB, no clinical probability score has demonstrated clear superiority for the diagnosis of HIT. Their principal clinical value lies in their high negative predictive performance, supporting their use as rule-out tools. Diagnostic strategies specifically tailored to the cardiac surgery setting remain needed and warrant prospective validation.