A data-driven framework for introducing predictive analytics into expanded program on immunization in Pakistan.

Area of Science:

• Public Health
• Machine Learning
• Epidemiology

Background:

• Pakistan's Expanded Program on Immunization (EPI) faces challenges with low vaccination coverage.
• Previous models for identifying vaccine defaulters had limitations, including a binary classification and lack of area-specific risk categorization.

Purpose of the Study:

• To propose an advanced prediction framework for accurate identification of children likely to default on immunizations.
• To refine the classification of defaulters and introduce risk stratification for targeted interventions.

Main Methods:

• Utilized data from the Pakistan Demographic and Health Survey (PDHS, 2017-2018) with 7153 records.
• Employed demographic and socioeconomic attributes for defaulter prediction and association rule mining.
• Applied a multilayer perceptron (MLP) classifier for prediction.

Main Results:

• Achieved 98% accuracy in identifying children likely to default from immunization series.
• Obtained an Area Under the Curve (AUC) of 0.994, indicating high predictive performance.
• The model effectively identified children at different risk stages of defaulting.

Conclusions:

• The proposed framework represents a data-driven advancement for immunization programs.
• Machine learning techniques and predictive analytics can be leveraged to reduce vaccination drop-outs.
• This approach enables targeted actions to reinforce immunization coverage in Pakistan.