Identification of antithrombotic drugs related to total joint replacement using anonymised free text notes

Background: Intensive antithrombotic treatment is recommended to prevent venous thromboembolic events after total joint replacement (TJR) surgery. Antithrombotic drugs are predominantly dispensed at the hospital. Unfortunately, hospital prescription data is often lacking in general practitioner databases, thereby limiting the possibility to use these rich databases to study the longer term effects in real practice. Un-coded anonymised free text from hospital discharge letters may be used to collect additional information on anticoagulant drug use related to TJR. Objectives: To design and test a method to extract additional information on anticoagulation therapy in patients undergoing TJR from anonymised free text notes in the Clinical Practice Research Datalink (CPRD). Methods: Anticoagulant drug use related to total hip (THR) or total knee replacement (TKR) from January 2008 until October 2012 was identified using both anonymised free text and prescription data. Internal validity of our newly designed method was determined by calculating positive predictive values (PPVs) of positive hits for predefined keywords in a random sample anonymised free text notes. To confirm external validity of our method, the proportions of patients using new oral anticoagulants (NOACs) or low molecular weight heparins (LMWHs) were compared with the proportions that have been reported by the UK National Joint Registry. In order to determine the usefulness of our method, TJR patients were then compared with regards to their status of exposure to antithrombotics. Results: PPVs ranging between 97% - 99% for NOAC or LMWH exposure related to TJR were obtained with our anonymised free text search method, using manual analysis of the surrounding free text as the gold standard. Aspirin users were identified with PPVs ranging from 91% - 95%. Our anonymised free text algorithm increased detection rates by 57%, yielding a total proportion of 18.5% of all THR and 18.6% of all TKR surgeries. Identified users of NOACs and LMWHs were largely similar with regards to age, sex, lifestyle and disease and drug history compared to patients without identified drug use. In contrast, Aspirin users were different as compared to unknown users in particular with regards to a history of ischaemic heart disease. Conclusions: We have developed a useful method to identify additional exposure to NOACs or LMWHs with TJR surgery.