Language and translation
Since our team is based in Brazil and our potential users are based here too, we decided to build our prototype in Portuguese. You can see the translation for the terms used on the bot here.
Inspiration
Glaucoma is the leading cause of irreversible blindness in the world. It is estimated that there are 79.6 million [1] people with glaucoma worldwide and the American Academy of Ophthalmology estimates that this number will increase to 111.8 million in 2040 [2]. The main clinical treatment to prevent blindness caused by it demands daily usage of eye drops correctly and commitment with the follow-up. But in a lot of cases, because the disease is initially asymptomatic, they don’t feel motivated to adhere to the treatment. In some studies [3], it is reported a treatment dropout rate of 60.5% within 1 year of follow-up. We believe that a virtual assistant that helps patients to get reliable information on the medication and disease could help to improve adherence to the treatment. The information that Nery gives was validated by our team member John, who is an Ophthalmology resident in a university hospital in the major city of São Paulo and by team member Juliana, who is a pharmacist. Nery is the name of the first Brazilian nurse, so our virtual assistant is named after her.
What it does
Since we had a limited time to build it to the hackathon, right now Nery helps glaucoma patients to get information on their medication such as side effects and instructions on how to use it in an interactive way. We want to add features that will help the patients to buy their medicine, to remind them to use the medication and help them to schedule regular appointments with their ophthalmologists.
How we built it
For the natural language processing, we created an application on wit.ai and trained it with common questions on medication and common medicine names. We stored desired answers for those questions on a database (DynamoDb) and created an AWS lambda function (with Python) to access those answers according to Wit.ai results sent through Facebook messenger to ours API gateway. We chose to do it though Facebook messenger because we don’t want the patient to download any extra app to their phone or access any other site that they are not used to. We respect user’s privacy and for this prototype, no user information is stored
Challenges we ran into
The main challenge was to understand the core of patient’s problems and match with the features of Nery. Regarding their problems, our validation process pointed out that a lot of them don’t know how to use the eye drops properly (technically, fail to recall)
Accomplishments that we are proud of
We received an award of 4000 brazilian reais (something between USD 500 and USD 1000) for our project from the Medicine department of the University of São Paulo, Brazil. (there wasn’t any prototype by them, the prototype was built entirely during this hackathon)
What we learned
Among other things, we learned how to use Natural Language Processing (NLP) to provide an interaction with the user as seamless as possible. We didn’t want to use buttons, we want to use free text because we believe this way we can give a better user experience for our users. So we learned a lot on how to deal with natural language variation on Wit.ai platform.
What's next for Nery
We want to implement the missing features and validate our final prototype with patients. There is a list of 50+ patients that are interested right now on testing it. We would like to implement it on Whatsapp as well, if it’s possible. In the long term, if we intend to use our tool to help patients with other health conditions.
Log in or sign up for Devpost to join the conversation.