No abstract

No abstract

IntroductionResearch using electronic health care databases continues to grow given widespread adoption of electronic health records (EHR). As manual record review is often not feasible with the increasing amount of data, use of International Statistical Classification of Diseases and Related Health Problems‐Clinical Modification (ICD‐CM) codes are a method of capturing the study population or abstracting outcomes. The accuracy of ICD‐CM coding and methodological validity of its use in medical research has been variable.ObjectivesTo compare the performance of ICD‐CM code data to pharmacist EHR review to identify patients that developed an in‐hospital thromboembolic event (TE) following four‐factor prothrombin complex concentrate (4F‐PCC) administration for anticoagulation reversal.MethodsA total of 337 patients were evaluated for TE occurrence following 4F‐PCC administration by two separate methods; discharge ICD‐CM code capture and pharmacist EHR review. To adjudicate ICD‐CM code and/or pharmacist‐identified TE cases, a blinded physician investigator reviewed the EHR and categorized the patient as having a TE or not based on extensive study definitions. To quantify the accuracy of ICD‐CM coding for patient outcome event and pharmacist investigator EHR review, positive predictive values (PPV) were calculated using physician review as the gold standard.ResultsA total of 76 cases were identified: 57 cases were identified by ICD‐CM codes only, 13 by pharmacist review only, and six by both ICD‐CM code and pharmacist review. Upon review by the physician investigator, 27 (35.5%) were identified as having an in‐hospital TE. There was a high degree of agreement between physician and pharmacist identification (88.2%). ICD‐CM code data demonstrated low PPV (22.2%) while pharmacist EHR review demonstrated high PPV (94.7%) for the identification of TE.ConclusionICD‐CM code data may be helpful in identifying study patients in large datasets, however low PPV for identification of TE requires clinician validation before acceptance as outcome data in clinical research.