Development and validation of a clinical prediction tool for non-receipt of updated COVID-19 vaccines

Vaccination with updated COVID-19 vaccines is important to maintain protection against circulating SARS-CoV-2 variants. We developed and validated a prediction model for non-receipt of updated COVID-19 vaccines at Kaiser Permanente Southern California. Of 3,287,287 adults, 20.2 % received 2023–2024 COVID-19 vaccine. We assessed the association of 15 variables available in electronic health records with non-receipt of 2023–2024 updated COVID-19 vaccine. The cohort was split into development and validation samples. Performance of the 15-variable model in the development sample was moderate, with a scaled Brier score of 35.6 %, R 2 of 29.3 %, C-statistic of 0.882, discrimination slope of 0.354, and calibration slope of 1; performance was similar for the validation sample and simplified 6-variable and 2-variable models. The strongest predictors were prior non-receipt of bivalent COVID-19 vaccine or influenza vaccine. These results suggest that a simple model using variables available to healthcare providers can be optimized to guide intervention strategies for updated COVID-19 vaccines.