Validation of Guillain-Barré syndrome case identification in three heterogeneous VAC4EU real-world data sources in Spain using the Brighton Collaboration criteria

Abstract

INTRODUCTION: In July 2021, the European Medicines Agency issued a warning about reports of Guillain-Barré syndrome (GBS) following COVID-19 vaccination. As part of the Global COVID Vaccine Safety (GCoVS) project, we aimed to estimate the Positive Predictive Value (PPV) of GBS diagnostic codes across multiple sources to explore differential misclassification. METHODS: We extracted GBS cases who received COVID-19 vaccination from three European data sources (VID, SIDIAP, VHUH in Spain) occurring between 1st December 2020 and 30th November 2022, using ICD-10-CM codes. Cases were validated using the Brighton Collaboration criteria to establish level of certainty (LOC) ranging from LOC1 (definite case) to LOC4 (insufficient evidence to meet case definition), and LOC5 (non-case). PPV was calculated and results stratified by risk window (within or outside 42 days post-vaccination) and before or after the regulatory alert. RESULTS: Overall, 180 GBS cases were identified. Most validated cases were classified with LOC4 (n = 78), except in VHUH. PPVs for LOC1-2 (n = 40) varied from 4.8% (SIDIAP) to 90% (VHUH). The PPV shows small changes after the alert, but these shifts are minor, and the confidence intervals stay essentially the same. Risk-window analyses showed higher PPVs outside the risk window for VID (36.4% vs 73.9%). When using LOC1-4 as true cases, PPVs increased to 65% (SIDIAP) and 77.8% (VID), and stratified results varied. CONCLUSIONS: Access to in-hospital medical records was key to achieving high diagnostic certainty for GBS. PPVs increased when LOC4 were considered true cases. The impact of risk window timing and regulatory alerts was inconsistent. Differences in PPVs stem from data source type and access to information from the setting where the diagnosis is made. Our findings emphasise the importance of GBS cases' validation when conducting pharmacoepidemiological studies using medical charts to identify GBS cases across multiple real-world healthcare data sources.