Bots in Healthcare

Overview

With the rise in popularity of chat-based collaboration software such as Microsoft Teams and Slack, bots have become more and more prominent. Using natural language as a user interface medium allows for a more expressive, "human" aspect to computing. While bot technology is somewhat new, modern frameworks and open, standardized APIs allow customers to rapidly create intelligent bots that understand human intent and perform tasks on behalf of a user. We at Microsoft Health & Life Sciences (HLS) have been working with customers to see how bot technology can empower healthcare workers to achieve more.

What is a "Bot"?

A bot is a computer program that can understand a human's intent in an intelligent way. Bots differ from traditional computer interfaces (links, buttons, etc.) in that you can speak natural language commands to control system behavior. A well-designed bot will then understand the user's intent, and perform tasks (often tedious, repetitious tasks) on behalf of the user.

You've probably already interacted with a bot... ever get a "pop-up" on a shopping site that shows an avatar of a "real" person asking you if you need help? While sometimes annoying, these bots perform a valuable service for the retailer: They can monitor and triage user issues. They can determine when you may be having trouble finding something or need help with the site. The bot can ask you questions, then route you to a live human if needed.

The language and intent understanding capabilities of a bot make it a perfect addition to text, email, and message-based channels such as email, Microsoft Teams, Slack, and Facebook Messenger. For example, I have a Bot that scans my emails for me looking for follow up items that I need to get done. When the Bot finds an email where I promise a customer to follow up with something, the bot will notify me and give me the option to create a reminder. When I purchase things online, some sites have a bot that allows me to ask questions about my order and get shipping updates -- all through Facebook Messenger!

Using a bot to look up codes for infections of the central nervous system.

Bots have been used to understand user intent and automate tasks in collaborative scenarios, but how can Bots be used in a healthcare context? The next sections will explore use case scenarios where Bots can assist healthcare workers with their critical duties and focus on two scenarios that we have customers using Bots for today.

ICD10 Code Lookup

When you visit a health care provider, your ailment is typically classified by a code from the International Classification of Diseases (ICD, 10th revision). These codes are attached to everything from diagnosis to treatment to billing. The challenge is that there are over 90,000 of these codes, and health care staff cannot remember them all. Depending on the user-friendliness of the system, looking up these codes can be very time-consuming and error prone process. 

But, what if the clinician could simply ask for a code using natural language? For example, asking "Give me a list of codes for edema of orbit" would return:

• H0522: Edema of orbit

• H05221: Edema of right orbit

• H05222: Edema of left orbit

• H05223: Edema of bilateral orbit

• H05229: Edema of unspecified orbit

Actual ICD10 bot (aptly named "ICD2") running in Microsoft Teams.

Want to make it real? The source code for just such a bot can be found here.

Care Team Huddles

A huddle is a short, stand-up meeting (10 minutes or less) that is typically used once at the start of each workday or shift change. Patient status is shared, processes are monitored, checklists are created, and tasks are assigned -- it's an important practice followed by most healthcare providers. A huddle mixes collaboration with process management, and is a great use case for a Microsoft Teams and a Bot.

An effective huddle solution allows your staff to:

• Measure and identify best practices across your organization.
• Identify patient care issues and potential causes.
• Monitor checklists and assign tasks.
• Share ideas with your health team using natural conversations.

With Microsoft Teams, all the huddle information is kept in one place, and is readily available for the next huddle.  

Want to make it real? The source code for just such a solution can be found here.

Additional Bot Use Cases

There are several additional scenarios where Bots can assist healthcare workers:

• Bots can assist clinicians with finding out who is on staff or who knows a particular language.
• Bots can help staff trade and swap shift schedules.
• Bots can help researchers automate testing and assist with the analysis of test results.
• Combine a Bot with voice input and allow your healthcare staff to work practically "hands free".
• Bots can poll staff for meal choices - and even order the food!

Conclusion

While Bot technology is in it's infancy, capabilities that were thought out of reach a few years ago are available right now. The open and connected nature of the cloud has never made it easier to connect systems and services, and Bots provide the next evolution in user interface design. Platforms like Microsoft Teams, Microsoft Bot Framework, and Microsoft Cognitive Services make it easy for customers to build solutions that connect platforms and automate tasks in an intuitive way. 

Resources

• How is AI Empowering Healthcare Today?
• Introduction to Bots
• Microsoft Bot Framework
• Language Understanding (User Intent)
• HLS West ICD2 Sample Bot
• OfficeDev/Office-365-Huddle-Templates