Workshop Structure
📅 Day 0: Software Introduction, Setup & 1:1 Mentor Support (30 mins)
- Introduction to Google Colab, Python notebooks and the workshop’s public health analysis workflow.
- Guidance on Google account setup, notebook access, file uploads and downloading workshop materials.
- Introduction to essential Python libraries: pandas, NumPy, matplotlib, scikit-learn and ipywidgets.
- Accessing selected public datasets from WHO, Our World in Data and CDC.
- Understanding dataset documentation, reporting periods, geographic coverage and population denominators.
- Running setup cells, checking library availability and testing sample visualizations.
- Optional local installation guidance for Python and Jupyter Notebook when browser-based access is unavailable.
🛠️ Hands-on
- Open the prepared notebook and import a small public health CSV dataset.
- Run sample analysis cells, generate a chart and save the notebook.
- Verify dataset access and resolve setup issues before Day 1.
- 1:1 Mentor Support: Individual assistance with account access, software setup, prerequisites and participant queries through allocated support slots.
Hands-on Tools: Google Colab, Python, pandas, NumPy, matplotlib, scikit-learn, ipywidgets; optional local Jupyter Notebook.
📅 Day 1: Public Health Data, Digital Surveillance & Early-Warning Signals
- Foundations of AI in epidemiology, disease surveillance and population health monitoring.
- Indicator-based, syndromic and event-based surveillance: purposes and differences.
- Exploring public health datasets: cases, deaths, hospitalizations, vaccination and population indicators.
- Identifying missing values, duplicate records, reporting delays and changes in testing or reporting.
- Calculating incidence rates, moving averages and changes in disease trends.
- Introduction to multi-source surveillance using clinical reports, wastewater signals and digital indicators.
- Distinguishing an unusual data signal from a verified outbreak.
🛠️ Hands-on
- Import and check a prepared infectious-disease dataset in Google Colab.
- Calculate weekly incidence and moving averages, then visualize trends across selected locations.
- Apply a simple historical-baseline threshold to flag unusual increases and discuss possible reporting artefacts.
Hands-on Tools: Google Colab, Python, pandas, matplotlib; curated WHO, Our World in Data or CDC CSV datasets.
Practical Output: A checked dataset, disease-trend chart and preliminary surveillance-alert table.
📅 Day 2: Machine-Learning Forecasting & Explainable Disease-Trend Analysis
- Understanding forecasting horizons, seasonality and lagged disease indicators.
- Distinguishing disease-count forecasting from outbreak detection and causal explanation.
- Creating simple lag and rolling-average features from historical surveillance data.
- Comparing a persistence baseline with a prepared Random Forest regression model.
- Using chronological training and testing to prevent future-data leakage.
- Evaluating forecasts using mean absolute error and interpreting prediction limitations.
- Introducing explainable AI through feature importance and discussing changing disease patterns.
🛠️ Hands-on
- Use the Day 1 dataset to create lagged features in a prepared notebook.
- Train a small Random Forest model to forecast the next reporting period and compare it with a persistence baseline.
- Plot observed versus predicted counts, calculate forecast error and interpret selected feature contributions.
Hands-on Tools: Google Colab, Python, pandas, scikit-learn, matplotlib.
Practical Output: A baseline-versus-model comparison, forecast plot, error summary and feature-importance chart.
Chronological evaluation is essential: scikit-learn’s forecasting examples demonstrate how inappropriate random splitting can produce overly optimistic performance estimates. scikit-learn
📅 Day 3: Precision Public Health, Interactive Dashboards & Responsible AI
- Connecting surveillance indicators and forecasts with public health investigation and planning.
- Comparing population-standardized disease burden across locations.
- Designing transparent alert thresholds and understanding false alarms versus missed signals.
- Introduction to vaccination planning and healthcare resource prioritization using descriptive indicators.
- Emerging applications: wastewater-informed surveillance, One Health intelligence and AI-assisted analysis of public health reports.
- Generative AI and NLP for report summarization: source verification, hallucinations and human review.
- Responsible AI: privacy, representativeness, reporting bias and communicating uncertainty.
🛠️ Hands-on
- Build a simple interactive surveillance dashboard using the prepared Day 1–2 outputs.
- Select a location and adjust an alert threshold to observe changes in flagged reporting periods.
- Complete a short public health interpretation template covering the signal, supporting evidence, data limitations and recommended verification steps.
Hands-on Tools: Google Colab, Python, pandas, matplotlib, ipywidgets; publicly accessible WHO and CDC reports for interpretation.
Practical Output: An interactive notebook dashboard and a concise, evidence-based surveillance brief.
