SHINGLES vaccine non-vaccination fell amongst older U.S. adults, however prediction fashions misplaced calibration as uptake modified.
Shingles Vaccine Non-vaccination Stays Frequent
The proportion of U.S. adults aged 50 years and older who had not obtained a shingles vaccine declined from 73.7% in 2019 to 56.1% in 2024, in response to an evaluation of Nationwide Well being Interview Survey information.
Though this development signifies bettering shingles vaccine uptake, greater than half of eligible adults remained unvaccinated in 2024. Non-vaccination was highest amongst adults aged 50–59 years, at 71.4%, in contrast with 59.6% amongst these aged 60–64 years and 44.6% amongst adults aged 65 years and older.
Researchers analyzed information from 96,663 adults aged 50 years and older. Fashions had been developed utilizing 2019–2023 information from 79,245 contributors and examined towards a separate 2024 pattern of 17,418 adults.
Machine Studying Offers Restricted Enchancment
The research in contrast survey-weighted logistic regression with XGBoost, a machine-learning method able to figuring out nonlinear relationships and interactions. Each fashions used the identical 15 predictors to estimate the likelihood that an grownup would stay unvaccinated.
Within the 2024 check pattern, logistic regression achieved an space below the receiver working attribute curve of 0.764, whereas XGBoost achieved 0.771. Though the distinction was statistically important, absolutely the enchancment was solely 0.0065. Brier scores had been additionally related, at 0.210 and 0.207, respectively.
Influenza vaccination, age, schooling, and race and ethnicity had been the 4 most influential predictors in each fashions. These shared predictors counsel {that a} extra complicated algorithm added comparatively little predictive worth over standard regression.
Altering Uptake Impacts Mannequin Reliability
Each fashions systematically overestimated the likelihood of shingles vaccine non-vaccination in 2024, regardless that their calibration slopes remained near 1. This sample mirrored temporal calibration drift as vaccination prevalence modified.
Proscribing mannequin growth to newer 2022–2023 information produced related discrimination however considerably lowered overprediction. The findings point out that fashions supposed to information population-level outreach might require common updating as vaccine uptake, entry, and healthcare use evolve.
For clinicians and public well being groups, prediction instruments may assist establish teams extra more likely to stay unvaccinated. Nonetheless, the outcomes warning towards deploying older fashions with out temporal and exterior validation. Latest information could also be significantly vital when fashions inform vaccine schooling, preventive care discussions, or focused outreach.
Reference
Costa NS et al. Temporal validation of machine-learning fashions for shingles vaccine non-vaccination amongst U.S. adults aged 50 years and older. Vaccine. 2026;92:129099.
Featured Picture: bernardbodo on Adobe Inventory.
