Safe Hands patient engagement platform to reduce no shows

Our solution helps to predict no-show appointments for clinical trials, primary care and specialty medical appointments powered by Pega self learning adaptive AI models with contactless collaboration.

Comment 84

Inspiration

In today's world, patients are far too occupied to have time to wait around for the services that they don’t need. There are significant number of patients who would no-show for their scheduled visits or do not respond to the automated telephone call to cancel their appointments. Based on a recent study, one out of every five patients who schedule an appointment don’t show up. Every time a person skips their appointment, the healthcare industry endures a financial loss and patients miss out on receiving the treatment they need.  If we could predict the no-show appointments, it would create a slot for others who are in actual need. Life is precious and every life is important.

In just a matter of four months as the COVID-19 pandemic spread across the world, it has stretched healthcare infrastructure of even the most developed countries. The rapidly increasing demand on health facilities and health care workers threatens to leave some health systems overstretched and unable to operate effectively. Our whole idea is to improve the health care utilization during this pandemic situation.

What it does

To reduce the repercussions of patient no-shows, we have implemented a new patient engagement platform built on Pega Infinity. This solution helps to predict no-show appointments for clinical trials, primary care and even specialty medical appointments powered by Pega self-learning adaptive AI models. This platform also helps to reduce physical interactions complemented with digital contact less collaboration and personalize all patient communications based on individuals' health conditions, medications with auto reminders

Key features of our solution:

1)Predict No-Show Appointments: AI is becoming a critical component for every industry. We have leveraged Pega Self-learning adaptive models that can suggest No-Show propensity by taking decisions in real-time. We have defined set of predictors based on data sets and designed an interactive mobile app to feed the data to train our models.

2)Contactless Virtual Appointments: Virtual consultation through skype that take place between patients and doctors via communications technology — the video and audio connectivity that allows “virtual” meetings to occur in real time, from virtually any location.

3)Patient Centric Safe Hands Mobile App: Now a days, Mobile apps become a primary way through which people patients contact hospitals. We have leveraged Pega Low Code mobile app builder to develop this Pega Mobile app.

4)Automated Personalized Reminders: Designed a framework on top of Pega Notification Framework to send daily Prescription reminders and appointment notifications to accomplish critical health tasks. These reminders are typically sent to individual patients to independently perform health tasks.

5)Instant Collaboration Platform: This feature facilitates collaboration and conversation among patients, doctors, and health care assistants. We have leveraged the out of the box Pega Pulse and Spaces to create patient groups to enable patients to post health related questions.

6)Data Sharing: Data sharing is an option to send videos, files between hospital and their patient. Patient data may include information relating to their past and current health or illness, their treatment history, lifestyle choices and diet details. We have leveraged the Pega out of the box documents gadget to build a better alternative to emails as the data might contain PHI.

https://safehandsapp.github.io/SafeHands/

How I built it

We prepared our product backlog with three sprint user stories with two epics and few features. Later we formed a team and started our development using Pega low code app studio. During the course of our agile discussions, we have improved our solution capabilities. We have used the below three major technologies to implement this solution.

Pega 8.4 Community Edition: Used most of the out of the box platform capabilities to build our micro journeys. Low code mobile development platform to build our patient centric app. Self-learning adaptive AI models to predict appointment cancellation or no-show probability using Pega Customer Decision Hub.

DistanceMatrix.AI (Free Trial): To compute the distance and travel time between points on a map with reliable and accurate APIs. This is used as a predictor for our self-leaning AI model which predicts appointment no-show.

Font Awesome: Icons are an essential part of many user interfaces, visually expressing objects, actions, and ideas. We believe that great icons can also affect in user experience. To leverage hospital or patient specific icons we have leveraged free version of font awesome to meet our requirements.

Leverage AI models in appointments:

• Adaptive AI models starts to learn based on the case response of every appointment micro-journey. • Used Pega OOTB AI predictors to create a case Predictions. • If the appointment is successfully completed, it will increase the model propensity.
• For this solution, we are showing this score to the Healthcare staff in the "All Appointments" landing page as a column. • If the Probability of appointment no-show is high, Healthcare staff can either cancel, allow overbooking, add no-show fee, change appointment type, reschedule based on the organization policies. • Right-now, we are allowing Healthcare staff to cancel the appointment. • The above scenario scenarios are different for different clinics. Ex: There are lot many institutes like MHCD which can't even charge for no-shows. • This solution can be used not only in Healthcare but also in early Clinical Trials, Airlines, Banking and many other industries. Based on the Kaggle dataset we realized that once the models perform well, its ROI will be high. • We can also pass Patient past appointment interaction history to add as one the predictors to our models. This will be part of our future roadmap.

Reasons for choosing Pega to solve this real time problem:

• Based on few studies, we came to that September month has more healthcare no-shows and the lowest will be in October. • Men has lower no-show compared to women. And there are many factors like hospital distance, time, day, patient age, gender etc. • Based on research, each dataset will be different as it changes with hospital services. Then, how do we solve this?? We made an attempt using a new approach of Pega industry-standard Adaptive decisioning which captures and analyzes data to deliver predictions in situations where historical information is not available and continuously increases the accuracy of its No-Shows by learning from each appointment response.... #NoShowPattern #ModelDrivenResults #NaiveBayesClassifier

Challenges I ran into

We faced challenges to get the datasets to define our predictors. We spoke to few doctors in this field and spent most of the time to define our core requirements. We came to know that the hospital distance from home and travel time are two major real-time predictors in No-Show appointments and it was a challenge to calculate the distance on the appointment day considering traffic and distance. We are not able to find any libraries to get these values but after exploring we found distance matrix which compute the distance and travel time between points on a map in less than 1 second and 5 times cheaper.

Accomplishments that I'm proud of

Once we started this solution, we came to know that there is lot research going on this area. Even there are no major products(except few) or patents which can solve this problem. We have leveraged most of the capabilities of Pega which includes CDH Pega Mobile, Multichannel bots, Collaboration Platforms, OOTB gadgets to build something which is useful for the current situation.

The entire solution was built in less than 3-4 weeks. Initial 2-3 weeks of time was spent to design our idea.

What I learned

We learned to solve real world problems using Pega AI. Our intention is to apply AI for ethical and a good use case. Thanks to Pega for giving this wonderful opportunity to explore multiple use cases with this Hackathon. This Hackathon journey won't stop here as we wanted to apply Pega AI into many real time use cases.

What's next for Safe Hands patient engagement platform to reduce no shows

We would like use this application by extending as "Appointment prediction as a service" with any of the existing appointment platforms or tools used by Health Care applications. This can be used across the industries. Once our models becomes matured this will definitely reduce appointment no-show loss and create a new way of handling appointments across industries. This not only applies for Healthcare but it also applies for all the businesses which deals with appointment or booking process.

Features planned for Safe Hands 2.0 :

  1. Schedule group appointments through mobile app and desktop.

  2. Pass group appointments as inputs to self learning adaptive models.

  3. Voice based reminders using Amazon Alexa Skills.

  4. Leverage appointment booking interaction history into adaptive models.

  5. Telemedicine enabled consultation/clinical trials using biomarkers and other native mobile features.

  6. Predict events through stream data passed through wearable devices.

  7. AI enabled personalized nutrition diet based on the health conditions.

  8. Exploring Amazon Comprehend Medical to detect unstructured information and link to medical ontologies such as ICD-10-CM or RxNorm so it can be used to migrate the existing applications to Safehands.

  9. Enable video calling facilities like Amazon Chime which is a HIPAA compliance to enable virtual appointments.

  10. Build a generic appointment as a service which can be used across other industries. Ex: Overbooking in airlines. https://www.youtube.com/watch?v=ZFNstNKgEDI&feature=youtu.be

  11. Customize this solution for veterinary hospitals as well.

  12. View all the previous appointments with observations both from internal and external applications using interfaces.

  13. Enable localization to support multiple languages.

  14. Store the patient health care records in block-chain for emergency support.

  15. Survey based feedback capture and post-service follow-up care.

Built With

  • distancematrixai
  • fontawesome
  • google-maps
  • multichannelwebbots
  • pega8.4
  • pegacdh
  • pegamobile
+ 169 more
Share this project:

Updates

posted an update —

There was a question from many who has seen this solution. How did we got the predictor details for our self learning models? https://www.kaggle.com/joniarroba/noshowappointments (Added kaggle dataset link) and based on few research articles https://www.insidetucsonbusiness.com/news/ua-researcher-develops-system-to-reduce-doctor-appointment-no-shows/article_f7c3a3fc-b1ab-11ea-b496-7b07179630d6.html

Log in or sign up for Devpost to join the conversation.