(136) Validation of an Algorithm for Unusual Site Thrombosis with Thrombocytopenia Syndrome (TTS) Within Electronic Health Records (EHRs)

Publication Number: 467

Background: Passive surveillance systems have reported cases of TTS involving intracranial or intraabdominal venous thromboses (unusual site TTS) following vaccination with adenoviral vector-based COVID 19 vaccines. No validation study has been published to evaluate the performance of International Classification of Diseases-10-Clinical Modification (ICD-10-CM) based algorithm for unusual site TTS as a composite outcome. It is infeasible to validate the algorithm within linked claims-EHR databases without access to imaging results.

Objectives: To identify unusual site TTS within administrative claims or EHR databases using ICD-10-CM diagnosis codes, and to validate the algorithm using clinical information in EHRs within the network of US Food and Drug Administration (FDA) Biologics Effectiveness and Safety (BEST) Initiative.

Methods: Based on literature review and clinical input, we developed an algorithm based on ICD-10-CM codes for unusual site TTS as a composite outcome—first recorded ICD 10-CM final or discharge diagnosis in the inpatient (IP) or emergency department (ED) setting indicating an unusual site thrombosis, with a co-occurring final or discharge diagnosis for thrombocytopenia in the IP, ED, or outpatient setting within 14 days of the thrombosis diagnosis. Patients with potential TTS were identified using the claims-based algorithm in the EHR database of an academic health network. Validation occurred for up to 50 cases at each thrombosis site. Positive predictive value (PPV) and 95% confidence interval (CI) were calculated using pathology or imaging results in EHRs as the gold standard, per an interim Brighton Collaboration definition.

Results: A random sample of 50 of 211 cases of portal vein TTS was selected for validation and all cases of other unusual site TTS were validated, for a total of 117 events. Among 113 patient records with sufficient information for validation of both components of the diagnosis, the overall algorithm PPV was 76% (95% CI 67–83%; 86 of 113 records with imaging and platelet count results). Overall, the PPV for the diagnosis of the thrombocytopenia component (98%, 95% CI: 92–99%; 112 of 115 records with platelet count results) was higher than for the diagnosis of the thrombosis component (76%, 95% CI: 67–83%; 87 of 115 records with imaging results).

Conclusions: To our knowledge, this is the first report of a validated ICD-10-CM-based algorithm for unusual site TTS. This study demonstrates good performance of an ICD-10-CM-based algorithm to identify unusual site TTS for background rates estimation and near real-time surveillance of COVID-19 vaccines in administrative claims databases.