ML course Project

COVID-19 in Mexico: Predicting severe disease outcomes using health and socioeconomic variables.

"AI applications present opportunities for the future of healthcare and can be harnessed at this time, as clinicians take on the complexities of responding to COVID-19" Xiangao Jiang et al(2020)

Team Members:

• Roberto Barroso-Luque (barrosoluquer)
• Luz Stephanie Ramos Gomez (stephanieramos)
• Oscar Enrique Noriega Villarreal (onoriega)
• Jesica Maria Ramirez Toscano (jramireztoscano)

Table of contents

• Motivation and Brief Summary
• Setup
• Data Collection and Processing
• Notebooks Structure
• Final Report and Acknowledgements

Motivation and Brief Summary

The current pandemic caused by the SARS-CoV-2 virus represents an unprecedented challenge to the global economy. As of June 2020, about 7 million people have tested positive and more than 400,000 have died as a result of the virus. To provide data-driven policy advice to individual countries analysis and modeling should be done on national-level data. With under-funded healthcare systems, institutional-weakness, and large informal economic sectors, Mexico faces a gigantic challenge to contain the virus. We developed classification algorithms to predict death and hospitalization among individuals who tested positive for the virus in Mexican soil using a mix of health and socioeconomic variables at the individual and municipality levels. Using ten-fold cross-validation and synthetic minority over-sampling support vector machine, logistic regression, complement naive Bayes, decision trees, and random forest models were trained to predict patient outcomes (survival vs death and hospitalization vs home recovery).

Setup

If you want to replicate the presented analysis, some modules must be installed. Particularly, the following:

pandas
geopandas
numpy
matplotlib
seaborn
mpl_toolkits
requests
zipfile
rarfile
imblearn
sklearn

Data Collection and Processing

All the data used in this project is stored in the data folder. To get daily updated data, we wrote:

• covid_data.py
  This module serves to download the most recent COVID-19 data from Mexico. The data is wrangled, cleaned, and transformed into a usable data frame. It also provides a function to aggregating all information to state level.
  • daily_covid() returns a pandas dataframe in which each row represents an individual tested for COVID-19
  • convert_to_state_date()converts COVID individual level data to counts by state and date

If you run the following, a csv file with daily COVID-19 individual cases will be downloaded to the data folder.

$ python3 covid_data.py

• health_data.py
  This module obtains data from ENSANUT, CONAPO, CONEVAL, and the Health Ministry at municipality and state levels. Particularly, with the mentioned data sources, we obtain:
  • The number of hospital beds, doctors, and nurses from the Health Resources Data provided by the Secretary of Health.
  • Percentage of people living in poverty from CONEVAL
  • Demographic data (population density) from CONAPO

Calling the merge_data_mun() function returns a merged data frame in which each row represents a municipality.

• pre_ml.py
  This model merges individual COVID-19 data with health municipality level data. It is the final processing of data before applying several techniques of sampling and machine learning models.

Notebooks Structure

To predict death and hospitalizations, we used several ML classifications models. Particularly, we applied: logistic regression, naive Bayes, support vector machine, decision trees, and random forest models. In these notebooks, we also used a custom module mlpipeline.py to split, normalize, impute values of the data, apply SMOTE in the learning process, and get the metrics of predicted values. For each model, we provide a notebook that includes the cross-validation approach we used (with imbalanced sample techniques), testing results of the model, and feature importance visualizations:

• Logistic Regression.ipynb
• Naive_Bayes.ipynb
• SVM DecisionTrees.ipynb
• Random Forest Model.ipynb
  Includes additional visualizations about the model's performance in each state of Mexico

Finally, to assess the relative risk of each state, we used the predicted values of the Balanced Random Forest and Support Vector Machine models to construct a simple risk index on COVID death and hospitalizations to compare the risk between states.

• See Relative Risk.ipynb

Final Report and Acknowledgements

The overall analysis and results of this project are documented in ML_paper.pdf.

The CAPP 30254 - Spring 2020 Machine Learning for Public Policy Course motivated this project. We want to thank professor Nick Feamster and the Teaching Assistants: Felipe Alamos, Erika Tyagi, Jonathan Tan, Tammy Glazer, and Alec Macmillen for their great support and comments. 😊

All errors are ours.