# Precautionary Health Behaviours as Potential Confounders in COVID-19 Vaccine Effectiveness Studies

## TL;DR

This study shows that precautionary health behaviors can affect how effective COVID-19 vaccines appear in research, especially during the first booster season.

## Contribution

The study identifies precautionary health behaviors as potential confounders in vaccine effectiveness estimates during specific pandemic periods.

## Key findings

- PHBs were positively associated with vaccination status and case/control status during the first booster season.
- Adjusting for PHBs increased vaccine effectiveness estimates by 6 to 9 percentage points in the first booster season.
- No confounding by PHBs was observed in the second booster season.

## Abstract

Background/Objectives: Precautionary health behaviours (PHBs), such as hand-washing or self-isolation, are non-pharmaceutical interventions used to reduce SARS-CoV-2 transmission. We investigated the potential confounding by PHBs of COVID-19 vaccine effectiveness (VE) estimates in a subset of study participants enrolled in id.DRIVE. Methods: The id.DRIVE COVID-19 VE study (formerly COVIDRIVE) is a European multicentre test-negative case–control study estimating COVID-19 VE against hospitalisation due to laboratory-confirmed SARS-CoV-2 in patients with severe acute respiratory infection. All adults (≥18 y) prospectively enrolled between 16 November 2021 and 16 August 2023 at three sites were invited to complete a PHB survey capturing indicators of PHBs in the 3 months preceding admission. Fisher’s exact test with Bonferroni-adjusted threshold was used to measure the level of association between PHB indicators and both COVID-19 vaccine status and SARS-CoV-2 test result. VE estimates were generated with and without adjustment for PHBs. Results: PHBs were modified over time, with higher precautionary attitudes in the first COVID-19 vaccine booster season (2021–2022) compared to the second one (2022–2023). For the first booster season, PHBs were positively associated with exposures (vaccination status) and outcomes (case or control status). Adjusting for PHBs led to a 6 to 9 percentage-point increase in VE estimates. Conversely, no confounding by PHBs was observed in the second booster season. Conclusions: PHBs should be considered a possible confounder of COVID-19 VE studies. Further research is needed to define when PHBs should be integrated into VE models, as the level of confounding may differ according to the study population and the epidemiological context.