Healthcare Accessibility Shifts Analysis During COVID-19 Using Mobility Data

This repository contains code and notebooks for the study:

"Dissecting Healthcare Accessibility Shifts during COVID-19 using Mobility Data"

We integrate large-scale mobility data with interpretable machine learning to examine year-over-year changes in healthcare accessibility between 2019 and 2022, focusing on four service types:

• Adult Primary Care
• Pediatric Primary Care
• Emergency Room Visits
• Urgent Care Visits

Repository Structure

• main.ipynb
  Jupyter notebook that reproduces the main analyses and visualizations for the study.

  • Loads accessibility data for each Census Block Group (CBG) and healthcare category for years 2019–2023.
  • Uses access_cal.py to generate accessibility data.
  • Visualizes year-over-year accessibility changes for the four healthcare categories.
  • Applies CatBoost regressors and SHAP values for model interpretation.

• access_cal.py
  Python script to generate accessibility metrics for each CBG and healthcare category for each year (2019–2023).

  • Reads weekly mobility data (in Parquet format) and filters for healthcare-related NAICS codes.
  • Processes and aggregates visit data for each healthcare service type.
  • Loads demographic and geographic data for CBGs.
  • Calculates a distance-decay weighted accessibility score for each CBG.
  • Outputs results as CSV files (e.g., 2019_Adult Primary Care_access.csv).

• catboost0727.py
  Python script for advanced statistical and machine learning analysis of healthcare accessibility changes.

  • Loads precomputed accessibility difference data and demographic features.
  • Cleans and preprocesses the data, including log transformations and feature engineering (e.g., metro area indicator).
  • Merges accessibility and demographic data for each CBG.
  • For each healthcare category, fits a CatBoost regressor to model the log-percentage change in accessibility between 2020 and 2022.
  • Performs hyperparameter tuning and evaluates model performance using R², MSE, RMSE, and MAE.
  • Uses SHAP (SHapley Additive exPlanations) to interpret feature importance and visualize the drivers of accessibility change.
  • Outputs summary statistics and saves SHAP plots for each category.

Data Requirements

• Mobility Data:
  Weekly patterns data in Parquet format for each year (e.g., /combined_GA_weekly_patterns_2019.parquet). The data description can be found at: https://docs.deweydata.io/docs/advan-research-weekly-patterns The data column headers are: ['placekey', 'parent_placekey', 'safegraph_brand_ids', 'location_name', 'brands', 'store_id', 'top_category', 'sub_category', 'naics_code', 'latitude', 'longitude', 'street_address', 'city', 'region', 'postal_code', 'open_hours', 'category_tags', 'opened_on', 'closed_on', 'tracking_closed_since', 'websites', 'geometry_type', 'polygon_wkt', 'polygon_class', 'enclosed', 'phone_number', 'is_synthetic', 'includes_parking_lot', 'iso_country_code', 'wkt_area_sq_meters', 'date_range_start', 'date_range_end', 'raw_visit_counts', 'raw_visitor_counts', 'visits_by_day', 'visits_by_each_hour', 'poi_cbg', 'visitor_home_cbgs', 'visitor_home_aggregation', 'visitor_daytime_cbgs', 'visitor_country_of_origin', 'distance_from_home', 'median_dwell', 'bucketed_dwell_times', 'related_same_day_brand', 'related_same_week_brand', 'device_type', 'normalized_visits_by_state_scaling', 'normalized_visits_by_region_naics_visits', 'normalized_visits_by_region_naics_visitors', 'normalized_visits_by_total_visits', 'normalized_visits_by_total_visitors']
• Demographic Data:
  Yearly demographic CSVs in dempgraphic_DCA/ (e.g., dempgraphic_DCA/2019.csv). The data is collected from the U.S. Census Bureau, American Community Survey 5-Year Estimates, 2019 - 2023. The example headers are:
• Geographic Data:
  Shapefile for CBGs (e.g., tl_2019_13_bg.shp). The 2010-2019 Census Block Group geometries data is from free data provided by safegraph: https://www.safegraph.com/free-data/open-census-data.

Usage

1. Generate Accessibility Data:

   • Run access_cal.py to process raw data and output accessibility CSVs for each year and category.

2. Reproduce Analysis and Visualizations:

   • Open main.ipynb in Jupyter.
   • Follow the notebook to load, analyze, and visualize accessibility changes.

3. Run Machine Learning Analysis:

• Execute catboost0727.py to perform regression analysis and generate SHAP plots for feature interpretation.

Dependencies

• Python 3.x
• pandas
• geopandas
• numpy
• matplotlib
• seaborn
• catboost
• shap
• scikit-learn
• scipy
• (and other standard scientific Python libraries)

Install dependencies with:

pip install pandas geopandas numpy matplotlib catboost shap

Outputs

• Accessibility CSVs for each year and healthcare category (e.g., 2019_Adult Primary Care_access.csv).
• Visualizations and model interpretation plots in the notebook.
• SHAP summary and bar plots for each healthcare category (saved in the shap_plots/ directory).