Interobserver variability in calculated 4T score in the evaluation of heparin-induced thrombocytopenia Free

Introduction: Heparin-induced thrombocytopenia (HIT) is a potentially fatal complication of heparin therapy, resulting in thrombocytopenia and thrombosis. The 4T score is often used to determine the pre-test probability of HIT and guide testing, but calculation of this score requires knowledge of the platelet trend, timeline of heparin exposure, and presence or absence of thrombosis, and this information may not be available. The 4T score also includes a subjective component regarding alternative explanations for thrombocytopenia that relies upon the user's judgement. This study utilized our institution's HIT data to evaluate concordance between 4T scores found in the patient chart and those calculated by the senior author, an experienced hematologist, as well as performance of the 4T score compared with platelet count and platelet fall in predicting HIT.

Methods: Patients from 2010-2024 were identified by performance of platelet factor 4 (PF4) or HIT antibody test followed by, if positive, serotonin release assay (SRA) testing, and were categorized into two groups: negative PF4 or positive PF4 and negative SRA, and positive PF4 and positive SRA. Only those with both a positive PF4 and positive SRA were considered to have HIT. Inclusion criteria included presence of additional laboratory data at or around the time of HIT testing as well as information on patient circumstances surrounding testing and the presence of a documented 4T score in the patient's chart. All patients with a positive PF4 were initially included as were 100 randomly selected patients with a negative PF4. Change in platelet count was defined as the difference between the maximum platelet count in the past 30 days and platelet count at or around time of HIT testing.

Results: Of 76 patients with HIT and 200 without (100 with positive PF4 and negative SRA; 100 with negative PF4), 50% and 24.5% respectively had 4T scores documented in the chart. There was no significant difference between chart-recorded and hematologist-calculated 4T scores (HIT 4.9 chart vs 5.3 hematologist, P = 0.25; no HIT 3.8 vs 4.0, P = 0.53). However, there were low rates of concordance between chart-recorded and hematologist-calculated scores (% agreement: HIT 23.7%, no HIT 34.7%), with the hematologist more frequently giving a higher score in HIT cases (47.4% higher, 28.9% lower), and a fairly even split between the hematologist giving higher (34.7%) or lower (30.6%) scores in no HIT cases. The author specialties of the chart-recorded 4T scores were most commonly hematology (55.2%) followed by critical care (32.2%). There was no difference in likelihood of agreement with the hematologist between hematology and non-hematology chart-recorded 4T scores (31.3% hematology vs 28.2% non-hematology, P = 0.76).

The sensitivity of predicting a HIT diagnosis using chart-recorded and hematologist-calculated 4T scores with a cut-off of 4 were both 86.8%, with a specificity of 30.6%, positive predictive value (PPV) of 49.3% and negative predictive value (NPV) of 38.5%. Prediction of a HIT diagnosis utilizing a platelet fall of 30% or 50% from maximum platelet count both had a sensitivity of 97.1% (four patients were excluded due to absence of prior platelet values), with (respectively) specificities of 12.5% and 25%, PPV of 44.7% and 48.6%, and NPV of 87.5% and 92.3%. Utilizing a platelet count at time of HIT testing as 100 or lower yielded a sensitivity of 92.1% with a specificity of 12.2%, PPV of 45.5%, and NPV of 70.0%.

Discussion: This data shows the low utilization or documentation of the 4T score as well as the low concordance between recorded 4T scores and those determined by an experienced hematologist, even when the recorded score was performed by another hematologist. This illustrates the subjective nature of the 4T score and/or its reliance on expertise to be appropriately utilized. Utilization of platelet fall of 30% or 50%, or platelet count of 100 or less yield superior sensitivity for the diagnosis of HIT, though specificity and PPV are lower. These findings suggest that the 4T score lacks objectivity and should be utilized with caution when judging the likelihood of HIT, and that the easily performed platelet fall or the platelet count itself may be alternative predictive methods, though this will require testing on a larger population.