Performance evaluation of a reflexive algorithm for suspected heparin-induced thrombocytopenia.

BACKGROUND: The diagnostic workup of heparin-induced thrombocytopenia (HIT) involves a heparin-platelet factor 4 complex-dependent screening test, such as an enzyme-linked immunosorbent assay (ELISA) or a rapid latex immunoturbidimetric assay (LIA), followed by a heparin-dependent platelet activation assay. Because of the high false-positive rates observed with LIA alone in our hospital, our laboratory adopted a reflex algorithm in which LIA-positive results are confirmed by an immunoglobulin G-specific ELISA before evaluation with serotonin release assay (SRA). OBJECTIVES: To evaluate the diagnostic performance and operational impact of an LIA-first reflex ELISA algorithm for suspected HIT compared with ELISA or LIA alone. METHODS: We retrospectively compared ELISA alone, LIA alone, and LIA with reflex ELISA confirmation across 3 diagnostic eras, with performance assessed against SRA-confirmed HIT. Sensitivity, specificity, positive predictive value (PPV), negative predictive value, and accuracy were calculated for each era, and associations between the magnitudes of screening-assay reactivity and SRA positivity were assessed. RESULTS: PPV improved from 31.3% with ELISA alone and 25.3% with LIA alone to 61.2% with reflex testing, while accuracy improved from 39.8% and 33.3% to 63.5%, respectively. Reflex ELISA confirmation reduced the number of patients classified as screening positive from 127 of 804 to 49 of 804, corresponding to a 61.4% reduction in potential SRA send-outs and a decrease from 15.8 to 6.1 SRA send-outs per 100 patients evaluated. Higher combined LIA and ELISA reactivities were associated with higher probability of SRA positivity. CONCLUSION: Reflex ELISA confirmation improved the diagnostic efficiency of HIT evaluation by increasing PPV and accuracy while reducing SRA send-outs. Combined screening-assay strength may help stratify the likelihood of HIT and guide the need for confirmatory testing, potentially improving time to HIT diagnosis and management in certain scenarios.