2

Call for Expertise — Help Improve the COVID-19 Knowledge Base

Please join our volunteer group of infectious disease specialists, clinicians, epidemiologists, biostatisticians, and researchers from many related fields to help improve the knowledge base for COVID-19 diagnostics.  Register here to join the effort.

Collaborative Knowledge Engineering

While there are no experts yet on COVID-19, many of you have clinical experience with all kinds of lung diseases. To help improve the COVID-19 knowledge base, we need to consolidate all your expertise quickly and efficiently.  

Web-Based Expert Knowledge Elicitation

We are currently hosting a web-based knowledge elicitation session using the BEKEE framework. Upon registration, you will receive the credentials to log into the BEKEE server and can start sharing your expertise. 

From Your Expertise to Worldwide Clinical Practice 

Our objective is to further improve the Bayesian network that serves as the knowledge base for our COVID-19 Adaptive Questionnaire, i.e., an expert system that can differentiate COVID-19 from other flu-like diseases. This web-based this tool is available to clinicians around the world for immediate use.*

All models are wrong, but some are useful

This is a work in progress that improves on a daily basis thanks to the knowledge contributions of many volunteers. To learn about the background of this initiative, please see our recent webinar: 
https://forum.bayesia.us/t/q6hhachk9h1qm/reasoning-under-uncertainty-part-2-differential-diagnosis-of-covid-19

Spread the Word!

Also, if you know someone whose expertise could help us with this effort, please invite them to join our knowledge engineering group. Register here to join the team.

You may also want to share the current status of this project by forwarding a link to the latest iteration of the COVID-19 Adaptive Questionnaire

*Disclaimer

Please note that the COVID-19 WebSimulator is strictly intended for informational and experimental use by medical professionals. This tool is not intended to be a substitute for professional medical advice or diagnosis. Always seek the advice of your physician or another qualified healthcare provider with any questions you may have regarding a medical condition. 

Any questions? Email us at covid@bayesia.us, or post your comment below.

9replies Oldest first
  • Oldest first
  • Newest first
  • Active threads
  • Popular
  • Hello,

    My name is Jack. I am an MD/PhD and would love to get involved. Please let me know how I can help. Thank you in advance.

    Like 2
    • Akop Seksenyan Hi Jack, you can register here to be part of the expert team!

      You'll then receive the emails that will allow you helping us getting the likelihoods for all the pairs of disease/symptom. 

      Many thanks in advance for your help!

      Like
  • Here is public data with symptoms+diagnosis from the Israel Ministry of Health:

    https://data.gov.il/dataset/covid-19/resource/d337959a-020a-4ed3-84f7-fca182292308

    I can help with translation, if someone finds this interesting.

    Translation TL;DR: Browse different datasets in the bottom left. You'll get a table with headers in English.

    Like 2
    • Yonatan Cale Many thanks for sharing this. This is excellent!

      Like
    • Stefan Conrady :)

      I want to translate some disclaimers (biases) about the data:

      1. The symptoms are based on personal reports (people said what symptoms they have).
      2. There is a reporting bias: The information was collected differently from people who were tested as positive or negative.
      For positive results: A full epidemiological investigation was done, that asked specifically about symptoms.
      For negative results: The information was not collected in an organized way and not using direct questions.
      3. The indications for "who needs to be tested" changed over time, and clinical symptoms were not always an indication. (For example, at the beginning, coming back from certain countries or coming in contact with a confirmed case did not require symptoms  in order to get tested).
      4. There were "random-tests" done on specific subsets of the population (for example health workers) during which infected people were found. In these tests, there's no record of clinical symptoms.
      5. There is evidence in the literature that some of the infected people don't have symptoms.

      In Hebrew:
      הבהרה חשובה לגבי התסמינים:
      1. התסמינים מבוססים על סמך דיווח אישי.
      2. ישנה הטיית דיווח מובנית בתסמינים: המידע נאסף בצורה שונה בקרב החיוביים והשליליים.
      בקרב החיוביים – נעשתה חקירה אפידמיולוגית מקיפה, ששאלה באופן ספציפי לגבי התסמינים.
      בקרב השלילים – המידע לא נאסף בצורה עקבית ולא על ידי שאלות ישירות.
      3. האינדיקציות לבדיקה השתנו לאורך הזמן, וסימנים קליניים לא תמיד היוו אינדיקציה לביצוע הבדיקה (לדוגמא, בתחילת ההתמודדות עם הנגיף, חזרה מארצות מסוימות או מגע עם מודבק מאומת לא דרשו הופעת סממנים לשם ביצוע הבדיקה).
      4. ישנם סקרים שבוצעו בקרב אוכלוסיות נבחרות (כמו למשל עובדי בריאות) ובמהלכם נמצאו מודבקים. בסקרים אלו אין תיעוד לתסמינים הקלינים.
      5. יש לציין שקיימת עדות בספרות לכך שחלק מהנבדקים הם א-סימפטומטיים.

      The disclaimers are from an article[0] that analyzes the data. But the article doesn't seem to take these biases into account in their analysis (other than mentioning them).

      [0] (Hebrew link) : https://blog.mafatchallenge.com/2020/04/13/covid-19-testing-with-ml/

      Like
    • Yonatan Cale Super! Many thanks.

      Here is the network learned from this data set.

      Like
    • Lionel Jouffe WOW!!

      That was really fast.

      I just got around to translating known biases in the data. Posted about 1 minute before you replied with the network.

      Like
  • Hello, I am an Engineer, but my wife is a nurse practitioner and got infected while working in a residence and most of our family got the COVID-19. She was the only diagnosed due to limited tests. Nobody tested us, the family. If I can be of assistance, I am available

    Like
Like2 Follow
  • 2 Likes
  • 5 mths agoLast active
  • 9Replies
  • 231Views
  • 7 Following