PrivyCare

More than half of the world now uses social media (58.4%), 4.62 billion people around the world either comment or express feelings online.  An unconventional way to keep tabs on someones well being is through their social media messages from either: Meta (Facebook), Twitter, or for the older generations, Whatsapp.

Research shows that PEOPLE IN POORER COUNTRIES ARE JUST AS LIKELY TO USE SOCIAL MEDIA FOR NEWS AS THOSE IN WEALTHIER COUNTRIES.

Our prototype is a machine algorithm that has been trained to analyze social media messages to determine medical distress.  The machine starts to analyze up to 5 posts and produces a probability of how likely the person who is posting the message might be going through some sort of medical distress.

This probability percentage would be provided to healthcare facilities, and case workers to engage the person for a medical assessment.  Below is an actual screenshot of our machine assessing Twitter feeds to determine a person's suicidal ideation.  PrivyCare can work to do the same for medical problems.

Our solution is provided to healthcare organizations to get ahead of the medical problem.  The potential patient is the benefactor that the medical system can intervene based on messages online and receiving bad advice from other social media users.  In fact, Almost 90% of social media influencers are sharing inaccurate health information, according to a new study. Ref: https://www.businessinsider.co...

This can be offered as a community solution to healthcare systems in poorer communities that want to reduce hospital admissions or potential environmental infections/cross-contaminations.

PrivyCare was recognized at the 2022 MIT Hacking Medicine challenge and came in 2nd place for the ability to demonstrate patient monitoring.  Since the challenge, the team decided to create a machine that could read messages to suggest proper medical advice.

The PrivyCare team consists of healthcare professionals ranging from clinicians to clinical informaticists.  The PrivyCare team also has members from around the world: the United States, Canada, Ecuador, and Colombia.

The team was founded at the MIT event and has explored multiple ways healthcare systems can monitor at-risk patients.

PrivyCare hopes that by the use of artificial intelligence and machine learning that we are able to quickly identify areas that may be highly contaminated by disease or social determinants for proper healthcare.  Typically this requires a verbal conversation or assessment at the healthcare facility - by allowing the machine to tap into the day-to-day banter of social media users, this eliminates the need to actually be at the facility in order to receive help.

Its more unconventional in our opinion than innovative.  The innovation is the machine certainly, and teaching the machine how to understand the way people post about various topics.  Using social media to determine outcomes have been used to market products, it's just a matter of time that someone uses it to triage health problems and then has a probability of what could be wrong.  The innovation would be a healthcare system that offers a virtual triage page on Meta (Facebook) for example and through members (the community) that follow the page and submit problems they are experiencing.. the site can determine a proper medical solution - come on-site OR follow up with PCP etc.. 

PrivyCare's goals are to be recognized as an out-of-the-box solution in which MIT can offer guidance to enhance features and test in real-world settings.  We hope to pilot our mobile application and our clinical dashboard that would populate these trends from social media posts.  But overall, we hope within five years to have people write to us and tell us how our solution prevented a major health issue from occuring.

Our key performance indicators are 1) the number of reduced hospital admissions; 2) the number of end-users that the hospital reached out to that as able to resolve their issue straight from social media; 3) the number of accuracies from our machine and of course the number of healthcare systems that are willing to utilize our program.

PrivyCare expects this unconventional method to determine medical problems will have an impact because social media is used all over the world and has no barrier to poverty based on the data we collected.  The convenience of simply expressing your potential healthcare issues online and having a machine that is tied to your local healthcare organization is something that is moving in the right step of "at the elbow" care.  We now see that you can purchase items straight from social media, we believe this trend will cross into healthcare where maybe triage starts from a simple post vs. travel and waiting at the emergency room.

PrivyCare is developed on an AWS HIPAA server and consists of FHIR APIs that could be linked directly to the end-users electronic health record.  The patient monitoring assessments "check-ins" are clinical evidence-based assessments used to gauge how the patient might be doing throughout the day.  The algorithm to read and determine medical problems has been trained on 105,236 tweets so far.   We used Vectorizer (Convert Words to Numbers For Machine Learning Algorithms).

Here is a screenshot of the clinical dashboard that would populate the results of the notifications as well as the probability of the social media posts.  This is in the event that the clinician would like to follow up with the patient in a remote setting:

The dashboard is embedded into the EHR and used to review at-risk patients in real-time.

Here are screenshots of the PrivyCare mobile application, again given to the patient for follow-up and monitoring:

The app is given to patients who are considered at risk for readmissions and is used to monitor patients' activities

That is determined by the healthcare system which we offer the application/ machine learning algorithm. 

The PrivyCare team is pretty diverse from a professional background and also cultural background.  The team consists of members from 4 different counties and 4 languages.  With each use case and solution, we have incorporated the communities in which we grew up and are living.  The machine doesn't know any bias and is trained to understand all common and local terms related to expressing medical problems.