Shuoyan Ning, MD; Rebecca Barty, MSc; Yang Liu, MMath; Nancy M. Heddle, MSc; Bram Rochwerg, MD; Donald M. Arnold, MD CHEST Sep 2016; 150(3): 516-523
Background: Platelet transfusions are commonly used in critically ill patients, but transfusion thresholds, count increments, and predictors of ineffectual transfusions remain unclear.
Methods: This retrospective study included consecutive adult nononcology patients who received platelet transfusions in ICUs at three Canadian academic hospitals between 2006 and 2015. Data were collected from a validated transfusion database. We determined independent predictors of ineffectual platelet transfusions, defined as transfusions that raised platelet counts by < 5 × 109/L. Reasons for transfusion were adjudicated in a subgroup of patients who underwent transfusion despite normal platelet counts.
Results: We identified 7,320 ICU admissions (n = 7,073 patients) during which 15,879 platelet transfusions were administered. Most admissions (78.7%) were for cardiac surgery. Based on 5,700 analyzable transfusions, the median pretransfusion platelet count was 87 × 109/L (interquartile range [IQR], 57-130). The pretransfusion platelet count was ≥ 50 × 109/L and ≥ 150 × 109/L for 79.6% and 17.8% of transfusions, respectively. Reasons for transfusion despite a normal platelet count were active bleeding or surgery in patients receiving antiplatelet agents or anticoagulants. The median platelet count increment was 23 × 109/L (IQR, 7-44), and 21.8% of transfusions were ineffectual. ABO incompatibility, sepsis, liver disease, and red cell and cryoprecipitate transfusions were associated with a poor platelet count increment.
Conclusions: Platelet transfusions were commonly used in the ICU when platelet counts were ≥ 50 × 109/L. One platelet transfusion increased platelet count by 23 × 109/L. One in five transfusions was ineffectual, and ABO incompatibility was identified as a modifiable risk factor. These data can help direct efforts to reduce platelet overuse and improve transfusion quality.
Weblink here