Using Synthetic Data for Machine Learning-based Childhood Vaccination Prediction in Narok, Kenya
Authors: Jimmy Bach, Yang Li, Yaqi Liu, John Sankok, Rose Kimani, Carrie B. Dolan, Julius N. Odhiambo, Haipeng Chen
Organizations: William & Mary Williamsburg, VA, United States · Ignite Global Health Research Lab, William & Mary Global Research Institute Williamsburg, VA, USA · William & Mary Department Williamsburg, VA, USA · Data-Driven Decision Intelligence Lab (D3i) Williamsburg, VA, USA · Community Health Partners Narok, Kenya · William & Mary Health Sciences Williamsburg, VA, USA · William & Mary Department of Data Science Williamsburg, VA, USA
Background: Limited data utilization in low-resource settings poses a barrier to the vaccine delivery ecosystem, undermining efforts to achieve equitable immunization coverage. In nomadic populations, individuals face an increased risk of missing crucial vaccination doses as children. One such population is the Maasai in Narok County, Kenya, where the absence of high-volume, quality data hampers accurate coverage estimates, impedes efficient resource allocation, and weakens the ability to deliver timely interventions. Additionally, data privacy concerns are heightened in groups with limited sensitive data. Objectives: First, we aim to identify children at risk of missing key vaccines across a large population to provide timely, evidence-based interventions that support increased vaccination coverage. Second, we aim to better protect the privacy of sensitive health data in a vulnerable population. Methods: We digitized 8 years of child vaccination records from the MOH 510 registry (n=6,913) and applied machine learning models (Logistic Regression and XGBoost) to identify children at risk. Additionally, we utilize a novel approach to tabular diffusion-based synthetic data generation (TabSyn) to protect patient privacy within the models. Results: Our findings show that classification techniques can reliably and successfully predict children at risk of missing a vaccine, with recall, precision, and F1-scores exceeding 90% for some vaccines modeled. Additionally, training these models with synthetic data rather than real data, thus preserving the privacy of individuals within the original dataset, does not lead to a loss in predictive performance. Conclusion: These results support the use of synthetic data implementation in health informatics strategies for clinics with limited digital infrastructure, enabling privacy-preserving, scalable forecasting for childhood immunization coverage.
Figures & tables
Figure 1. Data Preprocessing Steps
Variable
Categorical Value
Overall
n
6912
Age (yrs), mean (SD)
3.3 (2.1)
First Visit Day, mean (SD)
24.9 (55.2)
Gender, n (%)
Female
3396 (49.1)
Male
3516 (50.9)
Clinic, n (%)
Aitong
1523 (22.0)
Table 1. Summary Statistics of the Study Population
Figure 2. Number of Individuals Within Each Clinic Registry
Figure 3. Distances Traveled by Individuals to the Nearest Health Facility
Vaccine
n = 6913, No / Yes (%)
BCG
301 (4.4) / 6612 (95.6)
POLIO
3310 (47.9) / 3603 (52.1)
OPV1
1476 (21.4) / 5437 (78.6)
OPV2
2177 (31.5) / 4736 (68.5)
OPV3
2894 (41.9) / 4019 (58.1)
IPV
3208 (46.4) / 3705 (53.6)
Table 2. Outcome Distribution for Different Vaccines
Logistic Regression
XGBoost
Vaccine
Training Split
F1
Precision
Recall
F1
Precision
Recall
BCG
Real
0.8348
0.9203
0.7921
0.8853
0.9239
0.863
BCG
Synthetic
0.8334 ± 0.0193
0.9198 ± 0.0012
0.7909 ± 0.0245
0.8866 ± 0.0065
0.9173 ± 0.0015
0.867 ± 0.0102
BCG
Real + Synthetic
0.8234 ± 0.0089
0.9209 ± 0.0011
0.7776 ± 0.0114
0.8807 ± 0.0051
0.9255 ± 0.011
0.857 ± 0.0073
ROTA2
Real
0.746
0.7485
0.7484
0.7501
0.7505
0.7513
ROTA2
Synthetic
0.745 ± 0.006
0.7482 ± 0.0059
0.7478 ± 0.0059
0.7391 ± 0.0067
0.7401 ± 0.0068
0.7412 ± 0.0067
Table 3. Comparison of ML Model Performance (Mean ± Standard Deviation) across Iterations of Synthetic Data.
Column
BCG
DPT3
MR1
All Vaccines
clinic
0.9866 ± 0.0026
0.9854 ± 0.0037
0.9898 ± 0.0045
0.9888 ± 0.0044
age
0.9607 ± 0.0035
0.9716 ± 0.0050
0.9610 ± 0.0053
0.9628 ± 0.0097
gender
0.9919 ± 0.0042
0.9010 ± 0.0067
0.9536 ± 0.0132
0.9833 ± 0.0236
village
0.9457 ± 0.0035
0.9428 ± 0.0043
0.9637 ± 0.0031
0.9529 ± 0.0086
first visit day
0.9876 ± 0.0036
0.9797 ± 0.0034
0.9866 ± 0.0043
0.9846 ± 0.0046
vax
0.9975 ± 0.0027
0.9971 ± 0.0024
0.9945 ± 0.0047
0.9953 ± 0.0034
Table 4. Comparison of Column-Wise Density (Mean ± Standard Deviation) across Vaccines.
Column Pairs
BCG
DPT3
MR1
All Vaccines
clinic/age
0.9347 ± 0.0045
0.9140 ± 0.0068
0.9195 ± 0.0050
0.9244 ± 0.0167
clinic/gender
0.9826 ± 0.0024
0.9003 ± 0.0064
0.9523 ± 0.0114
0.9732 ± 0.0210
clinic/village
0.9246 ± 0.0021
0.9274 ± 0.0049
0.9531 ± 0.0036
0.9392 ± 0.0143
clinic/first visit day
0.9803 ± 0.0018
0.9777 ± 0.0029
0.9820 ± 0.0056
0.9819 ± 0.0043
clinic/vax
0.9835 ± 0.0022
0.9684 ± 0.0058
0.9812 ± 0.0032
0.9803 ± 0.0060
age/gender
0.9574 ± 0.0034
0.8989 ± 0.0049
0.9423 ± 0.0054
0.9532 ± 0.0165
Table 5. Comparison of Pair-Wise Column Correlation Scores (Mean ± Standard Deviation) across Vaccines.
Figure 4. Feature Importance Analysis: DPT3 Real Data
Figure 5. Feature Importance Analysis: DPT3 Synthetic Data
Children with rare genetic diseases often exhibit distinctive facial phenotypes, yet developing computer vision systems for early diagnosis remains challenging due to extreme data scarcity, privacy constraints, and limited data sharing in pediatric settings. These challenges not only hinder automated diagnosis but also restrict the availability of visual resources for clinical genetic counseling. While prior work has shown that synthetic data can augment real datasets and preserve phenotype-level semantics, it remains unclear whether synthetic data alone is sufficient for learning in ultra-low-resource pediatric settings. In this work, we study the synthetic-only regime for pediatric rare disease recognition. Under a controlled experimental setup, models are trained exclusively on phenotype-aware synthetic facial images at increasing scales. We find that synthetic-only training achieves performance comparable to real-data-only baselines at sufficient scale across multiple backbones, suggesting that high-fidelity synthetic data can approximate clinically meaningful distributions. These findings together further enable the use of synthetic pediatric facial images as privacy-preserving resources for genetic education and counseling, supporting clinician training and patient communication. Our results highlight the potential of computer vision to improve data efficiency and expand accessible visual tools in children's healthcare.
Child malnutrition remains a major public health challenge in low- and middle-income countries, particularly in South Asia, where early identification of vulnerable children is critical for timely intervention and resource allocation. This study aims to evaluate the feasibility, fairness, and temporal robustness of using a pretrained large language model (LLM) in a zero-shot setting for child stunting prediction using population health survey data. Using Bangladesh Demographic and Health Survey (BDHS) data collected between 2007 and 2022, we transformed maternal, child, healthcare, and household characteristics into semantically interpretable prompt-based representations and evaluated GPT-4o-mini for zero-shot stunting prediction, comparing its performance against a random forest baseline and assessing fairness across demographic and socioeconomic groups as well as temporal robustness across survey waves. The results demonstrate that zero-shot inference using GPT-4o-mini achieved comparable balanced accuracy to the supervised baseline while exhibiting substantially higher sensitivity for identifying stunting cases, relatively consistent performance across child sex groups, and stable predictive behaviour across BDHS waves; however, important fairness disparities were observed across residence and household wealth categories, highlighting the need for further investigation before deployment of foundation models in public health prediction settings.
Muhammad Ashad Kabir, Md Ahshanul Haque
School of Computing, Mathematics and Engineering, Charles Sturt University, Bathurst, NSW 2795, Australia
Synthetic data is widely used in healthcare to create datasets that preserve statistical properties of real data without exposing sensitive patient information. Generating and evaluating synthetic data across privacy, utility, and fairness dimensions is crucial for enabling high-quality data availability in downstream prediction tasks and clinical decision making. We present Memisis, a tool that orchestrates and evaluates synthetic data by leveraging existing synthesis libraries, large language models (LLMs), and state-of-the-art evaluation metrics. Our tool creates a unified workflow for data generation, validation, and evaluation. Users can control training size, training epochs, and the number of synthetic rows to sample. Beyond manual configuration, an interactive agent mode allows users to specify data generation goals in natural language, and the tool orchestrates the full pipeline by invoking existing synthesizers while performing the requisite evaluation. For the demo, we use an open-source schizophrenia dataset with protected attributes related to race and gender, evaluate six synthesizers spanning GANs, VAEs, diffusion models, and normalizing flows, and use a local LLM to orchestrate the workflow. The system affords users flexibility and control over the data generation and evaluation process.