The Global Health Innovators Seminar #19 - Dr. Michiel Niesen (Nference)
This episode focuses on the collaboration work between Nference and the Mayo Clinic entitled "Surveillance for Covid-19 Vaccine Using Electronic Health Record Data."
Dr. Michiel Niesen began by discussing "The Clinical Data Analytics Platform'' developed by Nference, followed by "Surveillance of the Safety of Three Doses of COVID-19 mRNA Vaccination Using Electronic Health Records." This research intends to address the critical question, "Is a third dose of FDA-approved COVID-19 mRNA vaccines safe?"
Recent reports of diminishing COVID-19 vaccine immunity led to the approval and distribution of additional doses and booster vaccines. Those who are at a greater risk of getting SARS-CoV-2 receive extra vaccination doses in addition to the clinical trial regimen. The risks and profiles of side effects associated with further vaccination doses are poorly known at present.
Dr. Michiel Niesen and his colleagues analyzed electronic health record data from 47,999 patients who received three doses of the mRNA COVID-19 vaccination to address this problem. Then, Dr. Niesen explained how BERT (A bidirectional encoder representation from transformers) was used to extract adverse event emotions from EHR Data. Among the 19 adverse event phenotypes assessed were anaphylaxis, arthralgia, cerebral venous sinus thrombosis, chills, diarrhea, erythema, facial paralysis, fatigue, fever, headache, local pain, local swelling, lymphadenopathy, myalgia, myocarditis, nausea, pericarditis, soreness, and vomiting. A statistical analysis was conducted to determine the risk of an adverse event after immunization.
Dr. Niesen concluded by discussing the results and conclusions of the research, adding that cohort analysis indicates that a third dose of the same kind of immunization after a primary series of BNT162b2 or mRNA-1273 is related to safe outcomes. Although there was an increase in early post-vaccination adverse events after the third dosage compared to previous doses, these results were insignificant (ie, fatigue, lymphadenopathy, nausea, and diarrhea). There was no statistically significant increase in EHR reporting of serious adverse events after the third dosage compared to the second dose, and the incidence was similar to that documented in the prior trial.
In this presentation, Dr. Niesen also emphasized the need for data de-identification in order to protect patient privacy. Check out the talk to learn more about the amazing work done by Dr. Nieson and the team!
Q&A highlights:
Question: What were the challenges in extracting information such as symptoms from electronic health records?
Answer: Every clinical note has a structure consisting of multiple sections or segments, therefore the model has extracted information from the relevant sections/segments.
Question: What kind of data is used for this analysis?
Answer: Data includes like Vitals data, Lab data, Physical Exam data
Question: How was the de-identification done?
Answer: Patient ids were de-Identified by random replacement.