Introduction: Administrative data are attractive for research, policy and quality improvement initiatives as large amounts of data can often be obtained quickly and at low cost. Unfortunately, administrative data often have significant limitations owing to how they were collected and coded. In many cases, free text, often hand written, diagnoses provided by physicians are converted into ICD-10 (International Statistical Classification of Diseases and Related Health Problems, 10th Revision) codes by trained nosologists for administrative purposes. However, because of the large data sets often obtained from administrative sources, it is difficult to verify the accuracy of the data, which may lead researchers to misleading or false conclusions. The objective of this study was to evaluate the accuracy of ICD-10 codes for the diagnosis of pulmonary embolism (PE) in emergency department (ED) patients. Methods: As part of a larger study examining the effectiveness of a clinical decision support intervention on CT utilization and diagnostic yield for ED patients with suspected PE, all patients with an ICD-10 code corresponding to PE (126.0 and 126.9) on ED discharge were obtained from four adult urban EDs and one urgent care center from August 2016 to March 2017. PE diagnosis was confirmed by reviewing electronic medical records and imaging reports for all patients. Discrepancies between coded ICD-10 diagnoses and actual imaging findings were quantified. This study was REB approved. Results: Of 584 ED patients with ICD-10 codes identifying PE as a discharge diagnosis, 535 had imaging that could be reviewed. Of these, 225 (42.1%) did not have clinical diagnoses of PE, and thus were incorrectly coded, resulting in false positive ICD-10 codes. Common coding errors included physician free text diagnoses of rule out PE or query PE being coded as positive for PE. Conclusion: Administrative data are subject to errors in coding. In this study ICD-10 codes were not reliable for the diagnosis of PE, with 42.1% of PE diagnoses being false positives. Similar coding errors are likely for other diagnoses that require waits for confirmatory imaging (e.g. appendicitis). Nosologist coding of physician free text diagnoses is challenging and prone to errors. Consequently, validation of ICD-10 coding prior to analysis of administrative datasets is crucial for meaningful results.