Air Pollution & Sociodemographic Impact on COVID-19 Outcomes by Chaya ChaipitakpornAir Pollution & Sociodemographic Impact on COVID-19 Outcomes by Chaya Chaipitakporn

🧪 Air Pollution & Sociodemographic Impact on COVID-19 Outcomes in the U.S.

📌 Project Type: Academic Research – Forecasting & Risk Segmentation

📍 Published in: Frontiers in Epidemiology

👨‍🔬 Role: Co-Author – ARIMA Forecasting, Data Cleaning & Visualization

📄 Read Full Paper →

🧩 Project Summary

This nationwide study investigated how sociodemographic and environmental factors impacted COVID-19 infections and fatalities across large U.S. counties during the pre-vaccination phase (Jan–Dec 2020).

It revealed that the effect of risk factors like education, pollution, and poverty changed over time, and were often only significant in high-infection zones, highlighting the importance of dynamic, region-based health strategies.

🧪 My Technical Contributions

✅ Time-Series Forecasting with ARIMA

Built ARIMA models to predict pollution levels (PM2.5, NO₂, SO₂, O₃) over time

Generated 95% confidence intervals to visualize uncertainty across different regions

Created zone-specific forecasts aligned with the COVID-19 wave timeline (Phases 1 & 2)

✅ Data Cleaning & Structuring

Processed and merged large-scale public datasets (EPA air quality, population, COVID-19 rates)

Structured time-series data by pollutant and county; ensured temporal alignment

Resolved missing or inconsistent values across thousands of time points

✅ Scientific Visualization

Created time-series plots with confidence bands for inclusion in published figures

Visualized regional trends across infection zones for insight into environmental shifts

🔧 Techniques Used

🔍 Key Insights from the Study

Phase 1: Population density, poverty, and education were stronger predictors

Phase 2: Those same variables lost significance, while age and O₃ exposure became more important

NO₂ correlated with fatality only in high-infection zones

Forecasting helped visualize pollution recovery and spikes in high-activity areas

✅ Relevance to Freelance / Consulting Work

This project proves my ability to:

Build forecasting models using Python and real-world environmental data

Segment and compare trends across geographies and time periods

Visualize uncertainty for decision support or policy design

Handle large, messy public datasets with scientific rigor

Applicable to:

Demand forecasting, trend monitoring, retention prediction, sustainability analytics

🛠️ Tools & Skills Demonstrated

Python • pandas • statsmodels (ARIMA) • data wrangling • scientific visualization • public health data • multi-source dataset integration