Vaccine Effects on In-hospital COVID-19 Outcomes.

Area of Science:

• Epidemiology
• Biostatistics
• Vaccinology

Background:

• Studies comparing COVID-19 outcomes by vaccination status are common.
• These studies often compare hospitalized patients, but the groups may not be causally comparable.
• Standard analyses may misinterpret vaccine effects due to confounding factors.

Purpose of the Study:

• To address the limitations in causal inference for observational studies on COVID-19 vaccine effectiveness.
• To introduce and apply the principal stratification framework for a more accurate estimation of vaccine effects.
• To clarify the interpretation of studies comparing hospitalized COVID-19 patients by vaccination status.

Main Methods:

• Utilized the principal stratification framework from causal inference.
• Defined a specific subgroup of individuals comparable across vaccination and hospitalization status.
• Conducted sensitivity analyses to assess the robustness of the estimated causal effect.

Main Results:

• Standard analyses comparing all hospitalized COVID-19 patients by vaccination status do not yield a valid causal vaccine effect.
• The principal stratification framework allows for the estimation of a causal vaccine effect in a specific subgroup.
• Sensitivity analyses were performed to evaluate the impact of unobserved confounding.

Conclusions:

• Observational studies comparing outcomes of hospitalized COVID-19 patients by vaccination status provide descriptive, not causal, evidence.
• The principal stratification approach is crucial for accurately estimating the causal effect of vaccination in this context.
• Re-interpretation of existing studies is necessary to avoid drawing incorrect causal conclusions about vaccine effectiveness.