Clinical effectiveness of a Bayesian algorithm for the diagnosis and management of heparin‐induced thrombocytopenia

Abstract

Background Heparin-induced thrombocytopenia (HIT) is a life-threatening drug reaction caused by antiplatelet factor 4/heparin (anti-PF4/H) antibodies. Commercial tests to detect these antibodies have suboptimal operating characteristics. We previously developed a diagnostic algorithm for HIT that incorporated 'four Ts' (4Ts) scoring and a stratified interpretation of an anti-PF4/H enzyme-linked immunosorbent assay (ELISA) and yielded a discriminant accuracy of 0.97 (95% confidence interval [CI], 0.93-1.00). Objectives The purpose of this study was to validate the algorithm in an independent patient population and quantitate effects that algorithm adherence could have on clinical care. Methods A retrospective cohort comprised patients who had undergone anti-PF4/H ELISA and serotonin release assay (SRA) testing in our healthcare system from 2010 to 2014. We determined the algorithm recommendation for each patient, compared recommendations with the clinical care received, and enumerated consequences of discrepancies. Operating characteristics were calculated for algorithm recommendations using SRA as the reference standard. Results Analysis was performed on 181 patients, 10 of whom were ruled in for HIT. The algorithm accurately stratified 98% of patients (95% CI, 95-99%), ruling out HIT in 158, ruling in HIT in 10 and recommending an SRA in 13 patients. Algorithm adherence would have obviated 165 SRAs and prevented 30 courses of unnecessary antithrombotic therapy for HIT. Diagnostic sensitivity was 0.82 (95% CI, 0.48-0.98), specificity 0.99 (95% CI, 0.97-1.00), PPV 0.90 (95% CI, 0.56-0.99) and NPV 0.99 (95% CI, 0.96-1.00). Conclusions An algorithm incorporating 4Ts scoring and a stratified interpretation of the anti-PF4/H ELISA has good operating characteristics and the potential to improve management of suspected HIT patients.

Computerised risk scores to guide recognition and diagnosis in patients with possible heparin‐induced thrombocytopenia

“…Not all patients underwent an SRA, and therefore, the definition of HIT in some patients was based only on the combination of an ELISA result and a calculated 4Ts score. However, this approach has been previously shown to have good diagnostic accuracy 11,12 . The NPV and PPV calculated for patients suspected of having HIT cannot be confidently extrapolated to a wider population of unselected patients for whom we envision undergoing HIT‐CR surveillance, and our present results should be considered hypothesis‐generating in this respect.…”

“…However, this approach has been previously shown to have good diagnostic accuracy. 11,12 The NPV and PPV calculated for patients suspected of having HIT cannot be confidently extrapolated to a wider population of unselected patients for whom we envision undergoing HIT-CR surveillance, and our present results should be considered hypothesis-generating in this respect. For example, there could be differences in clinical characteristics such as the likelihood of other causes of thrombocytopenia that could have led clinicians not to test for HIT, even in cases with a HIT-CR score of 4.…”

“…In cases in which an SRA was not performed, we used a previously validated algorithm that combines ELISA results and a calculated 4Ts score to determine the post-test probability. 11 A post-test probability of HIT >50% according to the algorithm was considered positive for HIT.…”

“…The ‘gold standard’ for the diagnosis of HIT was based on a positive ELISA [optical density (OD) >0·6) and positive serotonin release assay (SRA, >20% release at low concentration and ≥10% decrease in release at high concentration). In cases in which an SRA was not performed, we used a previously validated algorithm that combines ELISA results and a calculated 4Ts score to determine the post‐test probability 11 . A post‐test probability of HIT >50% according to the algorithm was considered positive for HIT.…”

A novel diagnostic algorithm for heparin‐induced thrombocytopenia

“…Others have proposed combining clinical pretest probability assessments using the 4T score and PF4-heparin ELISA OD values to improve HIT diagnosis. [11][12][13] Using a "combined criteria" (4T score = 8, or 4T score = 6 or 7 and ELISA OD ≥ 1.0, or 4T score = 4 or 5 and ELISA OD ≥ 2.0) similar to one used by others 11 (Table 2), we tested its diagnostic performance in 62 consecutive patients with positive PF4-heparin ELISA testing (training set). As shown in Figure 2A, the combined criteria had a sensitivity with 75% (12/16) and specificity of 98% (45/46) for HIT diagnosis.…”