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

Abstract

The heparin-induced thrombocytopenia computerised risk (HIT-CR) score is designed to aid in the diagnosis of HIT. We sought to evaluate its potential clinical utility. In this retrospective cohort study, we collected HIT-CR scores on all inpatients receiving heparin over a 4-month period and performed chart reviews on the subset who independently underwent clinical diagnostic testing for HIT to identify patients with HIT. In all, 34 342 patients received heparin, 1744 had high-risk HIT-CR scores of ≥3 and 220 had the maximal risk score of 4. Only 6% of high-risk and 10% of maximal-risk patients underwent testing for HIT. Conversely, among all 317 patients who underwent independent testing for HIT, 67% had low-risk HIT-CR scores (<3). Among patients independently tested, the positive predictive value (PPV) was 6·6% [95% confidence interval (CI) 4·9–8·8%] and the negative predictive value (NPV) was 99·5% (95% CI 97·1–99·9%) at a HIT-CR score cut-off of 3, and the PPV was 22·7% (95% CI 12·7–37·4%) and NPV was 99·0% (95% CI 97·6–99·6%) at a cut-off of 4. This study suggests clinicians fail to test most high-risk patients and unnecessarily test many low-risk patients for HIT. A reasonable approach to clinical application of HIT-CR scores would be recommending no testing for patients with a score of <3 and recommend testing for patients with a score of 4.