Mental health metrics and number of symptom mentions on Twitter are measured daily using pre-trained machine learning models applied to a random 1% Twitter data. 

Data is available here:  https://github.com/gtsherman/covid19-data/blob/master/map-data-upload.csv 
We update this file everyday. It has data for all states in the United States. 

A part of this dataset is being used by 
1/ Washington State in their behavioral health reports: 
https://www.doh.wa.gov/Emergencies/COVID19/HealthcareProviders/BehavioralHealthResources 
2/ Penn COVID Twitter map: http://bitly.com/penncovidmap 

Columns in the file:
`State`
`Date`
`# COVID Tweets`: Number of tweets in our dataset containing terms related to COVID-19
`# Twitter Users`: Number of users tweeting COVID related tweets

Symptoms: Number of tweets containing words related to any of following symptoms (each in a separate column)
Skin, Body Aches, Fatigue, Flu-like, Seasonal Cold, Trouble Breathing, Nausea or Diarrhea, Loss of Appetite, Fever, Changes in Smell or Taste, Abdominal Pain, Chills, Cough, Headache

Mental health estimates based on random 1% Twitter data: standardized scores compared to January 2020. 
`Relative Twitter Sentiment`
`Relative Twitter Loneliness`
`Relative Twitter Anxiety`

If you use this dataset, please cite the following paper: 
Guntuku, S. C., Sherman, G., Stokes, D. C., Agarwal, A. K., Seltzer, E., Merchant, R. M., & Ungar, L. H. (2020). Tracking mental health and symptom mentions on Twitter during COVID-19. Journal of general internal medicine, 35(9), 2798-2800.