Natural language processor as a tool to assess heparin induced thrombocytopenia awareness

Abstract

The life-threatening consequences of heparin induced thrombocytopenia (HIT) may be prevented with early recognition, prompt heparin withdrawal and direct thrombin inhibitor use. To determine the level of HIT awareness, electronic term recognition software can be used to query the electronic medical record (EMR) to assess the thought process and test ordering behavior of health care providers confronted with falling platelet counts. We sought to assess the awareness of HIT in a large teaching institution using these tools. Mayo Clinic databases were queried to identify a cohort of hospitalized adults receiving heparin (06/1/08-06/1/09). Serial platelet counts for each patient were scrutinized for a 50% decrement from baseline. "Clinician awareness" was defined by mention of HIT (determined by electronic term recognition software) within the hospital record by any member of the healthcare team or requisition of platelet factor 4/heparin antibody testing. During this time period, 34,694 adults were hospitalized and 24,956 received heparin. Only 3,239 (13%) patients had more than 1 platelet count during the hospital stay. Of 199 patients (6.1%) with ≥50% platelet count drop, clinician awareness was 36%. The absolute platelet count was the only independent variable associated with HIT awareness (P<0.001). Both appropriate platelet count monitoring and HIT awareness are low at this large teaching institution. Software tools for monitoring awareness and providing realtime alerts of significant platelet count decrements may be useful.