Predicting parental Human papillomavirus vaccine hesitancy: development and internal validation of machine learning models
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (PREreview)
Abstract
Objectives
Cervical cancer is Cameroon’s second most common cancer among women. However, high parental Human papillomavirus (HPV) vaccine hesitancy impedes elimination, and no validated predictive tool exists. This study aimed to develop and internally validate machine learning models predicting parental HPV vaccine hesitancy in the Buea Health District and identify key predictors.
Methods
This secondary analysis used cross-sectional survey data from 1,156 parents of children aged 9–18 years. Hesitancy was derived from self-reported vaccination status and intention. Twenty-six predictor groups were screened via group LASSO (17 retained for logistic regression; all 26 for tree-based models). Data were split 80/20 into training (n=925) and test (n=231) sets. Logistic regression, random forest, and XGBoost were tuned via repeated cross-validation while thresholds were set by Youden’s J statistic. Discrimination, calibration, and Brier score were assessed on the test set. Predictions were interpreted using SHAP.
Results
All models showed comparable discrimination (AUC-ROC 0.841–0.870). XGBoost had the highest AUC (0.870, vs 0.869 for logistic regression, DeLong p=0.965); logistic regression had the highest sensitivity (73.1%), F1-score (74.9%), and lowest Brier score (0.1435). Calibration slopes exceeded 1 (1.12–1.46) across models. Insufficient vaccine information, perceived vaccine unsafety, and distrust in the Ministry of Health were the dominant predictors.
Discussion
The three models achieved comparable, clinically meaningful discrimination, and algorithmic complexity did not improve prediction over standard regression. Hesitancy was more strongly predicted by information access and institutional trust than sociodemographic disadvantage.
Conclusion
These findings support trust-building engagement over broad demographic campaigns.
What is already known on this topic
Parental HPV vaccine hesitancy is a major barrier to HPV vaccination, but predictive modelling evidence from sub-Saharan Africa is limited and no internally validated model for parental HPV vaccine hesitancy has been developed in the region.
What this study adds
We developed and internally validated logistic regression, random forest and XGBoost models for predicting parental HPV vaccine hesitancy, with all three showing comparable discrimination (AUC-ROC 0.841–0.870) and logistic regression achieving the lowest Brier score and highest sensitivity and F1-score. Insufficient vaccine information, perceived vaccine unsafety and distrust in the Ministry of Health were the dominant predictors across the models and interpretability analyses.
How this study might affect research, practice or policy
The findings support the development and external validation of simple, interpretable prediction tools to identify parents who may benefit from targeted, trust-building vaccine communication and community engagement to support HPV vaccination and cervical cancer prevention.
Article activity feed
-
This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/23892405.
1. Does the introduction explain the objective of the research presented in the preprint?
Yes.
The introduction establishes the public health importance of HPV vaccination and how high HPV vaccine hesitancy, particularly in sub-Saharan African settings like Cameroon, can make it difficult to get children vaccinated. It identifies a research gap concerning the lack of any locally developed and internally validated predictive models for HPV. The study objective is clearly stated, which is to develop and compare machine learning models, and identify important predictors of parental HPV vaccine hesitancy.
2. Are the methods well-suited for this research?
Yes.
The comparison of logistic …
This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/23892405.
1. Does the introduction explain the objective of the research presented in the preprint?
Yes.
The introduction establishes the public health importance of HPV vaccination and how high HPV vaccine hesitancy, particularly in sub-Saharan African settings like Cameroon, can make it difficult to get children vaccinated. It identifies a research gap concerning the lack of any locally developed and internally validated predictive models for HPV. The study objective is clearly stated, which is to develop and compare machine learning models, and identify important predictors of parental HPV vaccine hesitancy.
2. Are the methods well-suited for this research?
Yes.
The comparison of logistic regression, random forest, and XGBoost is appropriate for assessing whether more complex algorithms offer predictive advantages over a conventional statistical model. The authors also used cross-validation, discrimination, and SHAP-based interpretation. The authors do acknowledge possible issues with the generalizability of their findings due to how the participants were selected using convenience sampling, and that because all participants were from one area West Africa there are limits to the data's representativeness. Additionally, the data and analysis code are not readily available in the current preprint; therefore, sharing the analysis code and any model specifications could improve methodological transparency.
3. Are the conclusions supported by the data?
Somewhat.
The results support the conclusion that all three models utilized demonstrate reasonably strong discrimination within the test sample, and that more complex predictive models like the XGBoost did not show a clear advantage over logistic regression. The finding that information sufficiency, perceived vaccine safety, and institutional trust are important predictors of vaccine hesitancy is also consistent across the 3 model approaches. The proposed use of a targeted screening tool in routine immunization settings is a reasonable future direction, but it is only a potential benefit yet to be supported. The main concern is ensuring that they are not attempting to establish a causal relationship.
4. Are the data presentations, including visualizations, well-suited to represent the data?
Yes.
The use visualizations such as ROC and calibration curves, the table of ranked predictors, another table outlining all predictors used, and a model-performance table, all seemed appropriate for this comparative prediction study.
5. How clearly do the authors discuss, explain, and interpret their findings and potential next steps for the research?
Clearly.
The discussion is well structured and connects the results to the broader literature on HPV vaccination, public health, and machine learning. The observation that logistic regression performed comparably to more complex algorithms indicates that greater model complexity may not always produce better predictions. The authors also identify relevant next steps, including how further research may consider ways to increase external validation and advocating for immunization services. My main concern is that explanations involving information hesitancy predictors may not be fully established by their cross-sectional analyses.
6. Is the preprint likely to advance academic knowledge?
Yes.
The study addresses an important public health problem in their study location, and contributes locally relevant evidence from Cameroon, a setting in which they have identified predictive modelling of parental HPV vaccine hesitancy to be limited. The main contribution of this preprint may be for hypothesis generation rather than changing practices in the field.
7. Would it benefit from language editing?
No.
8. Would you recommend this preprint to others?
Somewhat yes.
I would recommend this preprint to researchers working on vaccine hesitancy, public health prediction modeling, and the application of machine learning in low- and middle-income countries. This is because the comparison of modeling approaches and the emphasis on local evidence make it a useful contribution for further investigation. However, I would recommend readers interpret the findings as preliminary evidence of predictive performance within this one West African district, rather than as evidence that the models are ready for any large scale implementation. Additionally, these findings should not be used to justify individual-level risk or interventions without further research.
9. Is it ready for attention from an editor, publisher, or broader audience?
Somewhat Yes.
The manuscript addresses a relevant public health issue with a clearly stated objective, and reports a well thought comparison of three predictive modelling approaches. I would prioritize further discussion of the sampling limitations, more cautious interpretation of the patterns found, and possible open access to analysis code.
Competing interests
The author declares that they have no competing interests.
Use of Artificial Intelligence (AI)
The author declares that they did not use generative AI to come up with new ideas for their review.
-
Predicting parental Human papillomavirus vaccine hesitancy: development and internal validation of machine learning models https://www.medrxiv.org/content/10.64898/2026.09.18.26363393v1
-
