Pharmacoepidemiology and Pharmacovigilance Division. Spanish Agency of Medicines and Medical Devices (AEMPS). Madrid, Spain, Spain
Background: The Spanish Pharmacoepidemiological Research Database for Public Health System (BIFAP) was used for a Covid-19 vaccines effectiveness (VE) study against severe Covid-19. Cases were hospitalization ‘with’ Covid-19 identified automatically.
Objectives: Assess the main reason of hospitalization for cases of severe Covid-19 identified in a prior effectiveness study and the potential misclassification of the severity of the SARS-CoV-2 infection (i.e. hospitalised for other reasons) by vaccination status and vaccine brand (Janssen-JA, Moderna-MD, Pfizer-PZ, AstraZeneca-AZ).
Methods: Cases of severe Covid-19 were automatically identified in a cohort of 299,842 pairs of patients completely vaccinated and unvaccinated controls (matched 1:1 on date of vaccination, year of birth, sex and region) aged ≥18 and free of prior SARS-CoV-2 infection, between December 2020-October 2021. Severe Covid-19 was defined as admissions to hospital or Intensive Care Units (ICU) ‘with’ a positive SARS-CoV-2 test result from 30 days before to 120 days after. A manual review of primary care physicians’ free text comments (gold-standard) included in the clinical histories around the hospitalisation episodes records and blinded to vaccination status, was performed to ratify hospitalisations ‘for’ Covid-19 as primary reason. Free text comments could include hospital discharge letters, reasons to hospital referrals as well as descriptions of the general practice consultations.
Results: Overall, 233 cases of severe Covid-19 were automatically identified and reviewed. Admission ‘for’ Covid-19 was confirmed in 44% of vaccinated (by vaccine brand: 61% PF; 30% JA; 0% MD; 0% AZ) and 69% of unvaccinated cases (74% PF; 46% JA; 56% MD; 90% AZ). Admission for other reasons was proved in 29% of vaccinated (15% PF; 42% JA; 75% MD; 0% AZ) and 11% of unvaccinated (9% PF; 23% JA; 11% MD; 0% AZ). Remaining episodes (27% of vaccinated and 19% of unvaccinated) did not have additional information related to admission or stay in hospital or the reason of hospitalization was not clarified.
Conclusions: Overall, the information recorded in the clinical histories suggested differential misclassification of the Covid-19 severity among vaccinated and controls. Thus, correcting the vaccine effectiveness estimates by excluding admissions for other reasons is recommended. Although the sample size was low (mainly among MD and AZ with ≤13 cases in the compared groups) and limits the precision of the predictive values, current validation parameters could be used to adjust vaccine effectiveness estimates when manual validation cannot be performed.
