Announcing the 2018 BHI & BSN Data Challenge

In collaboration with the IEEE Conference on Biomedical and Health Informatics (BHI) 2018 and the IEEE Conference on Body Sensor Networks (BSN), we are hosting a challenge to explore real clinical questions in critically ill patients using the MIMIC-III database. Participants in the challenge will be invited the present at the BHI & BSN Annual Conference in Las Vegas, USA (4-7 March 2018): https://bhi-bsn.embs.org/2018/

Will there be prizes?

Yes, prizes are sponsored by Google Cloud and Georgia Tech.

Sponsored by Google Cloud and Georgia Tech

What is the timeline?

The Challenge is now open for registration! Participating teams have one month to carry out a research project using the MIMIC-III database. Final submissions must be sent for consideration by noon (PST) on 6 March 2018.

  • Challenge announced: 5 February 2018
  • Deadline for registration of teams: 1 March 2018 or as space permits (updated: 6 March 2018)
  • Office hours and help with projects: 6 March 2018 (2.15pm-3.45pm) in Treasure Island F.
  • Deadline for final submission: 6 March 2018 (updated: 7am on Wednesday 7 March!)
  • Challenge Presentations: 7 March 2018 (8.50am-11:35am) in Treasure Island F.
  • Awards: 7 March 2018: 5.50pm.

How can I enter the challenge?

The BHI & BSN Data Challenge asks participants to carry out a retrospective study using the MIMIC-III Database. This year’s research question is focused on a controversial and important logistical issue hospitals encounter every week:

Are patients admitted to the intensive care unit on a weekend more likely to die in the hospital than those admitted on a weekday?

Teams must register for the challenge using the following form before midnight (PST) on 1 March 2018: http://goo.gl/yQUazq.

Challenge entries must be submitted in the form of a 7-minute research presentation, which should be emailed to data-challenge@physionet.org before 7am (PST) Wednesday 7 March 2018.

Guidelines for entering the challenge and preparing the presentation are outlined in the “Registration checklist” and “Submission checklist” sections below.

Is there sample code to help our team get started?

Sample code is available at: https://github.com/MIT-LCP/bhi-bsn-challenge. For example, see these Jupyter Notebooks in Python and R.

Who is eligible?

Anyone with an interest in the reuse of routinely collected clinical data to improve healthcare is invited to participate. This is an opportunity for participants to learn about the benefits and challenges of working with complex healthcare data.

While individual submissions are welcomed, we encourage the formation of teams with participants of different backgrounds. Often successful projects consider a broad range of perspectives, including for example the data scientist, statistician, carer, and patient.

At least one member of a team must be able to attend the BHI & BSN Annual Conference in Las Vegas, USA 4-7 March 2018 to give a short presentation at the event on 7 March 2018. Registration at the conference is not a requirement, and not all team members need be present.

How can I access the challenge data?

Participants will use the MIMIC-III database, which comprises routinely collected data for critical care patients for over 50,000 hospital stays. There is a formal process for requesting access to MIMIC-III, which is described at: https://mimic.physionet.org/gettingstarted/access/

Gaining access to the full MIMIC-III dataset may take several days, so we suggest using the MIMIC-III demo while waiting to gain full access: https://mimic.physionet.org/gettingstarted/demo/

If you need help getting set up with access to MIMIC-III for the Challenge, please contact us at: data-challenge@physionet.org.

How are entries assessed?

Project entries will be assessed by a judging panel on the following criteria:

  • Reproducibility
  • Scientific methods
  • Presentation
  • Novelty

Registration checklist

  1. Register your team using the following form: http://goo.gl/yQUazq.
  2. Get access to the data by following the instructions at: https://mimic.physionet.org/gettingstarted/access/
    • Participants should complete human subjects ethics training.
      • Select ‘Massachusetts Institute of Technology Affiliates’ as your organization affiliation.
      • Please complete the “Data or Specimens Only Research” training program.
    • Participants need to agree to the MIMIC-III data use agreement.
      • Please submit the completion report (not the certificate)
      • For the ‘General research area for which the data will be used’ field, please use: ‘BHI & BSN Data Challenge 2018’.
    • Participants need to download the MIMIC-III dataset.
  3. Mark your calendar for the presentation session at the BHI & BSN Annual Conference in Las Vegas on 7 March 2018.

Submission checklist

  1. Prepare your slides in PDF or PPT format:
  2. Is your project reproducible?
    • Is your code publicly available?
    • Could someone replicate your study cohort and analysis easily?
    • Have you clearly cited your data sources?
  3. Have you practiced and timed your presentation?
  4. Have you emailed your presentation to the challenge organizers at data-challenge@physionet.org? Deadline: 7am on Wednesday 7 March 2018.
  5. Please arrive the conference room of the presentation session at least 15 minutes in advance of the start of the session and introduce yourself to the session chair.
MIMIC-III v1.4 documentation