Estimates of underlying health biases in SARS-CoV-2 vaccination recipients: a nationwide study in previously-infected adults

  1. Uwe Riedmann
  2. Alena Chalupka
  3. Lukas Richter
  4. Dirk Werber
  5. Martin Sprenger
  6. Peter Willeit
  7. Marc Rijksen
  8. Julia Lodron
  9. Tracy Beth Høeg
  10. John PA Ioannidis
  11. Stefan Pilz

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Background

Observational studies may over- or under-estimate SARS-CoV-2 vaccine effectiveness (VE) depending on whether healthier (i.e. healthy vaccine effect (HVE)) or more ill individuals are preferentially vaccinated. To evaluate this issue, we compared non-COVID-19, all-cause, cancer and COVID-19 mortality in vaccinated versus unvaccinated individuals.

Methods

This is a nationwide retrospective observational study in the entire adult population in Austria with previously documented SARS-CoV-2 infection with a follow-up from 2021 to 2023. Cox regression analyses were used to calculate hazard ratios (HRs) according to the number of SARS-CoV-2 vaccinations. We also performed matched analyses, where on each day, newly vaccinated individuals were matched with unvaccinated individuals based on age, sex and nursing home residency.

Results

Overall, 4,324,485 individuals (median age (IQR): 46 (33-59) years; 52.56% female) were eligible and 2.23 non-COVID-19 deaths occurred per 100,000 person days. Group differences in non-COVID-19 mortality risk were most prominent in the early periods (e.g., in Q4 2021, adjusted HRs (95% CI) in vaccinated versus unvaccinated were 0.69 (0.59 - 0.81), 0.65 (0.58 - 0.74), and 0.56 (0.48 - 0.66) for 1-, 2-, and 3-vaccinations, respectively) and decreased thereafter. Matched analyses for the first two weeks after vaccination showed HRs below 0.5 for vaccinated versus unvaccinated individuals irrespective of vaccination numbers. Similar findings were retrieved for non-COVID-19, all-cause, and cancer deaths. Overall, COVID-19 deaths were significantly reduced in vaccinated individuals.

Conclusions

HVE for SARS-CoV-2 vaccines was strong early after vaccination and diminished over time. HVE should be considered when estimating VE.

Article activity feed

  1. Version published to 10.1101/2025.02.19.25322515v2 on medRxiv
  2. PREreview

    This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/15048733.

    Summary

    The Health Vaccinee Effect (HVE) is an established bias in observational studies of vaccine effectiveness (VE). People who are otherwise healthier or more integrated into the healthcare system may preferentially receive the vaccine, leading to an inflated estimation of the vaccine's effectiveness.

    This manuscript performs a retrospective cohort study of all adult Austrians who previously contracted Covid-19. The authors estimate hazard ratios for overall and matched analyses of all-cause and non-Covid-19 mortality for vaccinated (stratified by number of vaccines) and unvaccinated groups. They also analyze cancer and motor vehicle deaths for these groups. They find some evidence for HVE in Covid-19 vaccine uptake in Austria and provide a call to consider for HVE in future Covid-19 VE studies.

    This paper is publishable with few revisions.

    Major Issues

    1. There are uncited statements and assertions in the manuscript.

      1. Cancer as an indicator of overall poor health (p.4, paragraph 2)

      2. Vehicle accidents as indicators of risky behavior (p.4, paragraph 2)

      3. Two-week lag-time until vaccine protection (p.4, paragraph 2)

    2. The authors should include a discussion, or mention, of the possible impact of disease status misclassification due to imperfect testing (p.4)

    3. The authors should explicitly explain why they subset the population to only Covid-19 survivors.

    Minor Issues

    1. Including a note within the figure (not the figure title) that the reference group is "unvaccinated" would be helpful in reading figure 1.

    2. One of the strengths of this paper is the large sample size, allowing the authors to subset to people who had Covid-19. While there are many reasonable explanations for this, the paper would benefit from an explicit explanation of this decision.

    3. The word "tentative" is used throughout. This may mean statistically non-significant, but it is unclear. ("In 2022, this trend was also tentatively present in both periods" (p.8), "Cancer mortality reflected the main results, showing a tentatively lower risk" (p.9))

    Suggestions/Extensions

    1. Only a small number of sensitivity analyses are performed and discussed in the body of the paper. The paper would be strengthened by analyses considering geographic location, diagnostic misclassification and Covid-19 death misclassification. Death misclassification is addressed in the discussion, but a quantitative exploration may support the author's argument and the reader's understanding.

    2. This study is focused on Austria, but as a nationwide study it may have broader implications. We would like to see a discussion of social factors driving vaccine access, or possible reasons for why there might be HVE in Austria and whether these factors are universal or an artifact of the Austrian health system and therefore not applicable to other contexts. 

    3. The preprint identifies a waning of HVE, reasons for which are not stated explicitly in the preprint

    4. The authors stratify supplemental analyses by vaccine type – are there any relationships between vaccine type and demographics/outcomes that may be explored in further research? 

    Competing interests

    The authors declare that they have no competing interests.

  3. Version published to 10.1101/2025.02.19.25322515v1 on medRxiv