pendragon.online

Cura-T

earthware ltd internetship project | Facebook Chatbot | C# Bot Framework | Support from Cancer Research UK

Define

Finding clinical trials online will often land you on two websites – both of which use language which is arguably ‘medicinal’ and the interfaces of each are arguably ‘clunky’. How can finding clinical trials be made a simpler experience? Whilst interning at Earthware this was the project I worked on.

Discover

Research online showed that finding out about clinical trials often came directly from a doctor, but you could find themselves by using several websites that allowed you to find clinical trials for cancer based on several factors such as age, sex, location and type of cancer. However, there were very few apps or alternative digital methods of finding a clinical trial.

Creating a user journey that got a user to find a clinical trial for a specific case, it was soon clear to see that the websites online suffer from poor user interfaces and often have so much technical language that it was difficult for the test candidates to truly understand the trials that they were being asked to find.

What users said

Comments from those who tested some existing solutions mentioned.

'It's got the right information - it's just poorly laid-out, poorly accessed and poorly given across.'

'It's a wall of text - it's quite distracting.'

'Even though it was only loading 50 results, it was still taking a good number of seconds.'

'The medical language used was a big barrier. I'm one of those people that was soon as I start reading words that I don't understand, I just turn off but keep reading.'

'I'd like to be able to put my details in and it filter results for me, rather than me have to look at each trial and find which are suitable for me.'

'The website looks clincial - whites and blues everywhere. It didn't feel like it was made for members of the public.'

A group of people graphic.

Develop

A chatbot was chosen as a new and exciting platform to create a clinical trial finder on. The chatbot was written in C# using the Microsoft Bot Framework and being hosted on Microsoft Azure meant that it could be deployed to Facebook Messenger, Slack, Microsoft Teams, Kik and several other platforms easily. The chatbot was designed to work best with Facebook Messenger since this is the messaging platform that I expected the most people to use.

Personas

To help empathise with the target audience of this chatbot, user personas were created which would be used to shape the direction of the project.

Samuel

64 year old married retied plumber with colorectal cancer

Stock image of an elderly man indoors wearing a blue coat looking at the camera (without glasses).

I am married with my lovely wife; Brenda and we have three children.

I suffer from colorectal cancer and when I heard the news, I thought of all the things I would have to stop doing and things I’d need to start doing such as changing my diet.

I have talked with my doctor about the options I have to treat my condition and he talked me through clinical trials.

I am really worried about the possible side effects of the drug being tested and whether that can lead to other type of illness or end up being worse than the traditional treatment. I still need to clarify many of the risks associated with clinical trials to assess whether it is worth participating or not. But if my wife tells me to go, I won’t think twice.

Likes

  • Meals out and good food
  • Playing tennis every now and then
  • Reading journals and newspapers
  • Watching TV with his wife

Dislikes

  • Using the internet
  • Wasting time looking for things
  • Being in crowded places
  • Eating healthy foods

Amelia

36 year old single English teacher with a daughter with leukaemia

I am a mom desperate to help my daughter, Carrie that has leukaemia. It was diagnosed a few months ago, right before she turned 2.

I have done an extensive research on her cancer and have looked for different options that will make her suffer the least.

I found some clinical trials that could relief her pain and possibly treat her condition. I showed them to my doctor and he explained them to me. I want to make sure I choose the best option, so I don’t want to miss any detail or have questions unanswered. I am willing to go anywhere to give my child the best treatment available so that I can see her grow as a healthy child.

Likes

  • Working with children
  • Jogging
  • Shopping and new trends
  • Participating in school events

Dislikes

  • People with no charisma
  • Messiness
  • Maths
  • Waiting for others
Stock image of a young mother looking directly at her child. The mother has straight, long brunette/blonde hair and is wearing a white top. The child is 2 years old, looking at his mother and is wearing an orange top.

Kayleigh

23 year old single designer

Stock image of a young woman with straight brunette/dark hair wearing a camo jacket outside. She is wearing a grey top.

I recently started working at this amazing multinational company and have met wonderful people with whom I learned a lot.

I was diagnosed with breast cancer last month and, even though I knew I was more likely to have it because my nanny had it, I wasn’t expecting to have it, and so soon!? I accompanied her while she was on chemo and I really don’t think I am able to go through everything she’s gone through. I discussed with my doctor other options and he advised me to participate in a clinical trial as I could really benefit from it and help others.

I just want to find the most suitable one now!

Likes

  • Drawing and being creative
  • Travelling
  • Nights out with friends
  • Cooking and baking

Dislikes

  • Monotonous life
  • Ceiling fans
  • Stuck-up people
  • Tight clothes

Conversational flow diagrams

Diagram showing the original flow of the chatbot. It is a complex diagram that shows that the user is asked for a lot of information which is processed to find clinical trials.

The first step to creating the chatbot was to consider the kind of information that the chatbot would need to collect and how it would process this information. Initial ideas produced this complex diagram that showed the chatbot taking a lot of information about the user’s condition, the stage of treatment that they were currently receiving and the type of trial that they were looking for. ‘Help’ buttons would be present throughout the chatbot to help the user answer technical questions.

However, after showing this diagram and the questions on it to some users, it was soon apparent that it was too difficult to answer, the answers had too many variables and it would not be quick to complete.

Diagram showing the a version of the chatbot flow, close to the final version.

Using the feedback received from testing current solutions, it was obvious that users wanted a solution that was simple, quick and had language that anybody could understand. Therefore, the first thing that was considered were the questions that the chatbot would ask and the order in which they needed to be asked.

The core information required to start finding clinical trials is:

  • The user’s age
  • The user's location
  • Whether the user is willing to travel
  • The type of cancer that the user needs to find a trial for

With this information it is possible to begin finding a clinical trial.

With that in mind, the diagram on the right was produced to show the in order in which information is collected by Cura-T.

Diagram showing the flow of the chatbot. The user is first asked if they'd like to find a trial, then they are asked for their sex, postcode, whether they are willing to travel or not and finally the cancer type before Cura-T finds a trial.

The next thing to consider was the language used in the chatbot. I realised that I was communicating with cancer patients, so language from the chatbot needed to be friendly and pleasant, but not too ‘jokey’ or informal. The diagram above shows a revised conversation flow, now featuring options for the user to select as opposed to natural language (not only does this make the chatbot easier to develop, but also quicker to use) and some of the language that the chatbot would use. Occasional compliments such as saying ‘that’s a great place!’ when the user selects a location makes the chatbot more personal and appear more human as opposed to ‘robotic’.

Chatbot persona

A full user persona was developed just for the chatbot. This defines how the chatbot acts and what its traits are - since the chatbot is effectively like talking to a human, this was required.

Cura-T logo: 'Cura T' is written in an off-white text on a pale green background. The 'U' in 'Cura' is in the shape of a stephoscope (it looks like a happy face), the tail of the stephoscope travels beneath the wording and finishes between the 'A' and 'T' to form a dot. The words 'A bridge to a cure' is written underneath 'Cura T'.

I know everything about clinical trials and I help others learn more about them. I know how people suffer, so I just want to ease their lives by clarifying their worries and concerns, and find the best trials for them.

Goals

  • Be adaptable to questions received
  • Present the most up-to-date and accurate information in a manner that users can understand
  • Expand my own knowledge about clinical trials
  • Be able to comfort my users through the use of the language I use
  • Make sure I am clear and helpful at all times

Likes

  • Simple things
  • Learning new things
  • Helping others
  • Being needed and sharing knowledge

Traits

  • Professional
  • Always available
  • Friendy
  • Adaptable

User journeys

Diagram showing the user journey of a persona called Kayleigh, who is 23 years old, single and working as a designer in Southampton. She wants to find a clinical trial for breast cancer, which she suffers from.

User journeys formed an important part of the development process. Shown is a user journey showing the different stages of interaction that one user may have with the chatbot in order to find a clinical trial for breast cancer. The journey is relatively short and Kayleigh (the persona) is able to find a clinical trial for breast cancer relatively quickly using the chatbot.

Visual design was a little restrictive because Facebook dictate the style and appearance of chatbots that run on Messenger for the most part, but I was able to design a basic visual representation of the chatbot and its screens in Adobe Illustrator and also produced visual representations of Kayleigh’s user journey.

Diagram showing Kayleigh's user journey as she uses the chatbot in Facebook Messenger. The screens that she sees and interacts with as she uses the chatbot are shown in this diagram with arrows pointing to the individual buttons that she interacts with.
Illustration of how the chatbot looks whilst running in Facebook Messenger on a phone. Certain elements such as short and long messages and carousels containing information are shown.

Test

In addition to formative testing where users were asked to complete a user journey on an existing website, testing for the chatbot conversation was also carried out.

The chatbot would ask questions and the user would need to respond to them, so to do this a series of questions were composed and I asked various people to empathise with a cancer victim and answer the questions. This helped to determine the type of answers users would give and the conversation flows that the chatbot should have. This testing did prove that the user could give any number of answers to a single question, so it was decided that the chatbot should give the user a selection of answers to choose from.

The chatbot can ask the questions about what they require and display a list of clinical trials, soured from Cancer Research UK data, which the user can scroll through and view more information about on the Cancer Research UK website.

Outcome

The final prototype is a working Facebook Chatbot (that can also work on Kik, Slack, Microsoft Teams and a variety of other platforms) that can find clinical trials for its users in a quick, easy and mobile manner. The user simply answers a few questions from the chatbot and it is able to find trials from Cancer Research UK that fit the user's criteria by querying a database.

The user is able to easily scroll through the trials in the carousel message.

The Cura-T prototype chatbot displayed on a Samsung phone in amongst a box of medicines.
The prototype chatbot running on a Samsung phone in a box of medicines.

Cura-T chatbot displaying the clinical trials that it found and how the user can easily find out more about them.

Evolve

Eventually, incorporating natural language answers and also clinical trials from a wider range of sources could have been good options for developing this chatbot.

However, during the chatbot's development it was proposed that there could be 'two routes' for the user to explore. One route would be the route that was created (finding a trial), the other would allow the user to find out more about clinical trials by asking the chatbot a number of pre-set questions. This would make the chatbot do what users wanted from the current solutions, which was to easily and quickly find out about clinical trials.

Shown are some user journeys that show how this could have worked.

The proposed user journey for Samuel wherehe uses the chatbot to tell him more about clinical trials and ask it some of his own questions.

The user would have the option to ask the chatbot some pre-set questions and it would be able to give some more information about clinical trials themselves as opposed to just information on current trials.

Proposed user journey for Amelia, showing how the chatbout would be able to answer personal questions before finding a trial, extending its functionality.

The chatbot was presented to Cancer Research UK by earthware after I finished interning there and periodically still receives messages, presumably from those still testing the chatbot.

Related

Enjoyed this case study? Have a look at some similar work!

Broads Authority Web App

How can you encourage families in Great Yarmouth and Lowestoft to visit The Broads using technology? Much of my time in Year 2 at university was spent on this live client brief that rendered interesting results.

Helpdesk Windows 10 Universal App

My final piece for A level Computer Science was a helpdesk app which runs on Windows 10. The app allowed users of Wymondham High Academy to report IT issues from their Windows 10 device.

Helping The Visually-Impaired Read

'Expensive, difficult to use and incompatible with some platforms'. This is what visually-impaired users said about using technology to read. How can this be improved?