Ebohub is a cloud-based application to assist health-care workers, patients, and administrators coordinate disease surveillance and management on the frontlines of the battle against Ebola through SMS and web interfaces.

(demo) (video)

The Ebola outbreak that started in 2014 was the deadliest outbreak in the history of the disease. The disease spread like wildfire, and the region lacked the infrastructure required to tackle it rapidly and effectively. Well over 11,000 people died.

Major problems with the epidemic included:

• Health care workers had to personally identify remote infections, suspects, and exposures
• Patients could only be safely handled at suitably equipped Ebola response centers
• There were few ways to disseminate information or to reach out to local communities
• There was no central coordinated surveillance system that connected to the public and workers
• Healthcare workers were recording patient, disease, and test data on paper
• Mobile telecomm infrastructure was for feature phones only (SMS-based) with no web access

I developed EboHub as a cloud-based service with a centralized disease database containing surveillance, case and contact-tracing data. It uses 2-way SMS interactions with health-care workers and the public (with call-in and browser-based interfaces as well), to provide real-time access to case status, to-do lists for workers, contact-tracing, symptom and care instructions, outbreak maps, and allows authorized workers to perform SMS-driven updates to the disease database.

EboHub uses Twilio as a gateway for SMS-Http and Voice-Http bridge services, Python as the implementation language, Flask as the web-server framework, PostGresql as a database, and Heroku as a scalable deployment platform. I used ngrok during development to route public HTTP requests (originating from SMS or browser interactions) to my laptop.

EboHub is deployed as a robust functioning prototype with case surveillance, worker, patient, and facility coordination, and community education, which allows:

• 2-way SMS-based interaction with members of the public: #help (responds with menu of options), #info (provides disease, symptom, and care information), loc and name (updates the database with new patient info), and i’m sick (updates patient as a suspect in database).
• 2-way SMS-based interaction with health-care workers: #loc (updates health-care worker’s location), todo (provides worker with assigned list of suspect, infected, and contact cases for worker’s location), update case (updates patient status for a given case), contact (updates contact-tracing information for an infected case), i’m sick (marks health-care worker as a suspect), and #help (provides menu of options).
• Outbreak map with suspects, infections, and exposures (people who may have been exposed and should be tracked).
• Voice and IVR interaction for the public (currently just a placeholder message).
• Initial SMS-blast capability to send SMS messages to a selected set of workers or the public.
• Initial login and access controls for the web-based interface (disabled for demo purposes).

EboHub is unique in its holistic scope focusing on both health workers and locals, and in using SMS as a primary communication channel. EboHub directly addresses key objectives in USAID's original BAA for Fighting Ebola. It strengthens healthcare capacities by providing healthcare workers tools to improve quality and timeliness of patient care thereby preventing the virus from spreading.

One crucial piece in the fight on Ebola is access to real-time information for health-care workers, health-care administrators, patients, and the public. Many mobile-phone based solutions come up short because vast areas in the region lack access to smart-phones and mobile internet access. EboHub will have a significant positive operational impact on current Ebola surveillance, prevention, and treatment efforts.

By building first and foremost atop a foundation of 2-way SMS communication, with voice and IVR as a backup and a web-browser interface, it provides vital outbreak, individual case, contact-tracing, and public-awareness and education services to those on the frontline in this fight.

By continuously building up a shared cloud database, and providing systematic access to workers and the public to key parts of this data, both upon request and via “push”, EboHub will remove one important barrier to better managing the outbreak. Entire populations in the affected communities will benefit. Having clear, current, and consistent information available to both communities and health workers will reduce errors, miscommunications, rumors, and mistrust while increasing effective resource mobilization.

The easiest way is to use the system as currently deployed in the cloud, since it is already configured with a PostGres database and with a Twilio account and phone number for the SMS-http gateway.

Note that aaccess controls are currently disabled to make it easy for the public to run demos. The home page has a link to reset the database if necessary.

To use the currently deployed system:

• Gather 2 or more cell phones. Let's call 2 of these phones Patient and Nurse.
• Point your web browser at EboHub.7insights.com.
• Note the primary contact number for all Ebohub SMS and Voice interactions. Let's call this EboNumber.
• Browse the list of current patients, chiefdoms (regions), and health care workers.
• Patient sends #help to EboNumber to see SMS options as a patient.
• Patient sends #info and receives information to self-check symptoms etc.
• Patient places a voice call to EboNumber and discusses his issues. Voice-reponse is just a placeholder for now.
• Patient registers as a Patient by sending i'm sick to EboNumber.
• Patient registers his name by sending name <some-name>.
• Nurse registers as a health-care worker in Yorkville by:
  • Create an entry on the H C Worker page
  • Register the phone number in this exact form: +1 followed by area-code and number, with no extra characters.
• Nurse asks for her assigned tasks by sending todo. She receives assigned patient information in her current location.
• Patient registers moves into Yorkville by sending loc Yorkville. Nurse is notified since she is in Yorkville.
• Nurse visits Patient and sends update with patient-id and status.
• Nurse gets infected and sends in i'm sick.
• An administrator sorts the tables by different columns to get overviews.
• The administrator visits the Logs and Map page to look at the current outbreak map.

A full local install to run the program or its executable tests involves these main steps (although it is easier to directly use the version I have deployed).

1. Unzip the zip file of the entire program, or git clone from github
2. Install Python 3.4, recommended source: anaconda
3. Set up a Python virtual env using pyvenv in the app folder, and activate it.
4. Run pip install -r requirements.txt to install all the needed libraries.
5. Install a PostGresql database and start the db-server. Run python db_init.py from the src folder.
6. Setup a Twilio account, configure the Twilio phone number to route SMS and Voice requests to your running app. Consider using something like ngrok to connect the public URL to your running app.
7. Start the web server with: python ebohub_app.py.
8. Run pytest tests/test_any.py from within the app folder to run all the tests.
9. Start using the app through your browser and cell phones.

EboHub began as a idea at Columbia University's Ebola Design Challenge in early September, 2014. After attending a talk about the ravages of the (then) young epidemic, I emerged very deeply moved. I started looking around to form a team that could work on a solution. My team had a current Sierra Leone healthcare worker, back from his first round of service; a business school alum; a microbiology Ph.D. candidate; a Biotech masters student; and a sophomore in my school. I was the sole designer and developer of the EboHub application.

I started developing EboHub in mid-September 2014. As it progressed, I got feedback from my team, from other groups including the NYU hackathon, and from OpenIdeo.

It was built under time constraint from Columbia and NYU competitions, with an intense sense of racing against the epidemic, while trying my best to keep up with Freshman course work. I have worked on it for about 8 months now, through several iterations with feedback from healthcare workers (including the one on my team) and from the There Is No Limit foundation, an NGO based in Guinea. As of Summer 2015, EboHub was being prepared for piloting in Guinea by There Is No Limit. This NGO and I applied for USAID funding but we only made it to the 2nd round.

Here are some of my sketches of how the EboHub concept evolved over time, starting from that first Columbia meeting.

A few notes about the technical challenges, absence of mobile web access, transportation. I was too shocked by the heartbreak of the disease itself to make notes about those.

The basic architecture emerges of a Python app leveraging SMS communication. HW (Healthcare workers) get and update patient and suspect information about the outbreak, and the general public has a different interface to the app.

Many more ideas of infrastructure to make the development more effective. Twilio as an SMS bridge, ngrok to help with local app development, the beginings of a web interface.

Discussions with Aissata Camara as well as potential users in Guinea further inform the next steps for EboHub. TODO lists assignment and optimization becomes a thing. Outbreak visualization via Google Maps is implemented. We might also need to connect to existing healthcare IT systems.

It is currently not clear if EboHub will be developed further. My NGO partner has contacts in Guinea healthcare and telecommunications who are interested in taking it further. However, there are now other solutions deployed in the region, and the health-care infrastructure for Ebola has been improved considerably. I have not decided whether to evolve it in the direction of a general disease surveillance platform for under-developed countries, but will willingly work further if the need arises.

Some of the top current items as of Summer 2015 include:

• Improved / customizable TODO case assignment to workers. I have previously used linear programming in PyStemm and would use this approach to optimize TODOs, if only I could install the libraries on Heroku!
• Full support for contact-tracing: where a patient went, who was exposed, where they went, etc.
• Integrate with iHRIS health-facility databases (e.g. where are beds available?) with 2-way data feeds
• Internationalize the application (both SMS and Web interfaces)
• Abbreviated message formats
• More robust security
• Hardware interface to a GSM modem for regions that do not even have SMS-infrastructure

It has been, overall, an awesome experience.

And I plan to do more. In my current university Database course (Fall 2016), I chose in my course project to extend the database schema to improve its design as well as start covering some of the top TODOs. Here is a schema snapshot:

And a running demo, at least as long as the Azure machine hosting all the project databases for that class is kept running: (demo)