Inter-Rater Reliability of the 4Ts Scoring System for Diagnosing Heparin-Induced Thrombocytopenia in Critically Ill Patients.

Abstract 3995

Poster Board III-931

Heparin-induced thrombocytopenia (HIT) is commonly suspected, but rarely confirmed, in the critically ill. The lack of readily available rapid diagnostic tests makes the timely diagnosis of HIT challenging. As a result, when HIT is suspected, heparin is often stopped and an alternative anticoagulant is initiated until diagnostic test results become available. A valid scoring system that accurately predicts the likelihood of HIT would be very valuable since it would allow rapid risk stratification that then might allow treatment to be modified in patients at moderate or high risk of harboring HIT, while other strategies might be employed for those at low risk. The 4Ts scoring system has been proposed as such a classification. 4Ts classifies patients on a 4 component, 8 point linear scale. The final score is assigned to 1 of 3 categories representing low, moderate and hand high pretest probability (scores of 0 to 3, 4 to 5 and 6 or more, respectively). We describe results of a pilot study using the 4Ts scoring system within the PROTECT trial (a 3650 patient randomized controlled trial comparing low molecular weight with unfractionated heparin for primary prevention of venous thromboembolism (VTE) in medical-surgical ICU patients). As per the PROTECT protocol, patients are evaluated for HIT if their platelet count falls more than 50% from their initial platelet count (study day 1), if their platelet count falls to less than 50 × 10⁹/l, if VTE occurs, or if HIT is clinically suspected. As of August 6^th, 2009, 2657 patients had been enrolled in PROTECT, of whom 523 (19.7%) met criteria for investigation of HIT (170 due to a platelet count fall; 106 due to a platelet count of less than 50×10⁹/l; 343 due to VTE; and 39 due to clinical suspicion (categories not mutually exclusive)). All suspected cases of HIT are to be evaluated using the serotonin release assay (SRA), with HIT confirmed if the HIT test is positive, or HIT refuted if the SRA is negative. To evaluate the 4Ts score, patients are adjudicated independently by 2 experts blinded to each other's scores (MC, hematologist and DC, intensivist). The adjudication occurs in 2 stages: in Stage 1 the clinical case is reviewed and a 4Ts score assigned, masked to SRA testing; in Stage 2 the presence or absence of HIT is ascribed based on review of the patient's laboratory data. Discordant cases are then discussed and consensus achieved. To August 6^th, 2009 100 cases have been adjudicated and consensus achieved. Stage 1 raw agreement on the 4Ts score was poor (raw agreement 47.0%). Stage 1 agreement on risk category (low, moderate or high) was better (77%, weighted kappa 0.36 (0.10 to 0.61) reflecting fair agreement). Discordance was largely due to differences in the timing component of the 4Ts score (raw agreement 77%) and the “other” potential cause of thrombocytopenia component (raw agreement 61%). Agreement for the degree of thrombocytopenia and thrombosis components was excellent (92% and 93%, respectively). We conclude that inter-rater reliability of the 4Ts score in medical-surgical ICU patients is fair overall. Additional calibration work or modification of the scoring system is required if 4Ts is to be widely used to accurately predict HIT in critically ill patients. This poster is p