PixelDr was an ‘E-Simulation’ platform for medical education and development. PixelDr supported medical students and professionals to excel in their field, teaching them key skills and protocols in a manner that they would be encountered in a hospital. PixelDr addressed an acknowledged lack of innovation in higher education by combining the latest methodologies of Simulation-Based Learning, E-Learning and Gamification of education.

Need and challenge

Medical education programs are increasingly adopting physical simulation-based learning. Reasons often include:

• Low student and professional satisfaction with medical education and poor adoption of modern technology.
• Limited practical clinical placement positions for students and professionals.
• Numerous reports supporting the introduction of simulation-based medical education as an important step in curriculum development.
• Positive results in relation to enhancing knowledge, skills, and attitude toward technology-enhanced simulation have been reported in many studies.

Physical simulation-based learning is being used more widely, however, is limited by funding, time and access to expertise.

Though online learning is not new, current attempts to create engaging online medical education platforms have fallen exceptionally short of technical possibilities. Current offerings focus on question and answer testing that does not provide an authentic experience of medical practice and have poor end-user engagement. PixelDr aimed to bring the benefits of simulation-based learning to an online environment.

PixelDr addressed the needs and challenges of its users by offering a gamified approach that adapts to each user's unique learning requirements. The PixelDr approach attempted to improve engagement and outcomes through three key product pillars.

1. Gamified environment
2. Adaptive learning
3. Building a loyal community

Gamified environment

Current online education for medical students comprises question and answer testing. Such education practices fail to present problems in a way encountered by doctors within the hospital, and so it’s effectiveness is highly limited. PixelDr provides medical scenarios in a 2D isometric gamified environment, each scenario challenged the user's medical knowledge and situational judgment. Interfaces that match those of real hospitals were created for the best possible learning experience.

Each scenario was built on the medical dogma of history, examine, investigation, and management:

History

The image below shows the user taking a history of a patient using PixelDr’s conversational bot interface. This dialogue was entirely adaptive and self-improving, altering based on questions the user asks. A log is kept of the questions asked and marked against the appropriateness of those questions to the scenario presented. The order in which the questions are asked can also be marked to assess the flow of the user's history taking ability. As most diagnoses are made from the patient’s history this is an incredibly powerful tool and a key component of the PixelDr Engine.

From a technical perspective, the intent of each question was first analysed, and then any key entities from each question asked were extracted. So for example, say the used asked the following question; Do you take propranolol? The intent of this question is do they take something, in the context of a medical history, this is almost always referring to a medication of some sort. The entity is then the medication in question, i.e. propranolol in this case. Both of these were extracted, and then the appropriate response returned depending on the particular scenario. Each scenario was stored in a JSON format, storing all important medical information about each patient within a scenario.

Examination

Below we can see the examinations interface. Examinations are laid out in accordance with the different interactions a doctor could perform within an examination. Therefore, the interface helps to cement clinical knowledge on practical skills, which are typically more difficult to teach through online tools. As well as text fields for visual and palpable findings, these examinations also provide audio for auscultation and palpation. This means that the players must learn how to assess for themselves as opposed to being supplied with information – as is currently the case with available online tools.

Investigation

Below you can see the computer interface of the hospital. Players could order investigations as they would in a hospital. However, they were also able to look up information for that investigation, allowing them to expand their learning while playing the game. There was a time delay placed on each investigation to better simulate hospital conditions as well as continuing the gamification theme.

Once the investigation returned you could view the results in the patient’s information screen. These investigations are laid out as you would expect on hospital systems, providing an additional aid of preparing tomorrows doctors for the new computer-based prescribing and note taking which has begun to be rolled out across the UK. When viewing imaging the screen appears with the associated image. You may choose whether to view the radiographer's report or assess the X-ray yourself. From this screen you are also able to see the corrent review status for the patient in the information panel.

Management

Finally, we can see the prescribing interface that is based on current electronic prescribing protocols. Clearly, there is far more to managing a patients condition than just prescribing medications, but we decided it was best to focus on a particular quantifiable aspect of management, and then grow from there.

All quantities, medicines and intervals are akin to a real hospital environment and marks are gained for the correct selection. Relevant doses for each medication is supplied throughout, as is the case on prescriptions in a hospital. Once the user is happy with their management the user can choose to discharge or admit the patient. This is the end of the scenario and the point at which a user’s performance data is compiled for feedback.

Scenario Feedback

At the end of each scenario, the player would get a summary of their performance. This feedback was also split by the essential dogma. For each section, you would get an individual score for each section, as well as an overall score which would be used for ranking and competitive scoring. Below is an example of the feedback from an example scenario (sorry for the poor quality).

Adaptive Learning

A great deal of work went into the finer background workings of PixelDr’s learning platform.

As a student progresses in the game, code is collecting performance data and collating it into a score using bespoke algorithms. This data is then used to tailor the learning experience to the user, supplying them with scenarios in areas they may be lacking in knowledge. As is so often the case, students practice that which they are most comfortable with, but using our algorithms PixelDr ensures that students achieve a comprehensive medical understanding of all areas – not just a select few.

Below is an overview graphic of how this is done. Scenario, performance and game data is stored in a Azure CosmosDB secure servers. Scenario content can be edited and added to using the PixelDr editor programme. The pixel bot, to the right of the graphic, is where conversation data is stored to allow our bot to continuously improve the history-taking process using machine learning.

All of this data is brought together and assimilated within the PixelBrain. This is where user data from across all areas is brought together to calculate exactly what scenarios the player should receive and how best to accelerate learning. The PixelBrain is adaptable - we can tweak algorithms so that content can be returned to fit with modules that the student is currently studying.

Admitedly, the PixelBrain was just a marketable name for the underlying algorithms, which comprised a number of serverless functions to analyse the data returning from the game about the user and their performance. Data such as age, medical school and current rotation was also collated to attempt to match scenarios to users curriculums.

Building a Loyal Community

The ambition was to be the largest E-Simulation platform for medical learning. A key enabler of this is having the best learning scenarios and content creation. We intended to have community contributions so that users could build and share their own scenarios with the world through the medium of PixelDr. In return for PixelDr using these community-built scenarios in-game, users benefited from:

• Access to a collective of medical students and professionals for peer-to-peer discussion and networking.
• If a users’ scenarios meet the high standards of the PixelDr vetting process, then official PixelDr Editor accreditation will be given with an accompanying certificate acknowledging their aid in producing a revolutionary medical education platform.
• Public exposure as a certified content creator on PixelDr’s webpage and social media presence as well as accreditation as the scenario creator within the game.

This would allow students to meet and challenge one another via PixelDr creating the compassionate and engaged community we were hoping to nurture.

Scenarios could be submitted by users through the editor program that provides a simple way to create a medical scenario for submission to the PixelDr game engine. Each scenario would be certified as medically correct by a GMC qualified doctor before being used in production. This program will sit within the community section of the website allowing people to build and share their medical creativity with the community. Users would be able to share these publicly (with the whole PixelDr community) or locally (inviting individual friends to attempt their purpose-built scenario).

How PixelDr Changed

Once we had built our game and spent all too long building it, we seemed to have overlooked a key part of our development process, and that was whether PixelDr was an appropriate market fit.

When we took PixelDr to the market, we found that while people enjoyed playing the game, in this format, it is not necessarily something they would turn to to revise. Medical students didn't associate this form of education with serious learning, however interesting they may have found the game.

There was also an element that we may have bitten off more than we could chew, at least as such a small team. The number of variables to plan for was exceptional, and the level of data required to create a realistic game with a sufficient number of scenarios became difficult to fill, even with community-driven scenario creation.

However, the overwhelming good feedback we got for the game came from aspects of the history component. Students found the ability to take a completely unguided history to be incredibly useful, and given that history taking is such an important aspect of being a good doctor, we decided this was the element we had to focus on.

Students also liked the other components of the game but would prefer for them not to be in the format of a game. That's not to say they didn't like aspects of gamification, just not gamification to the level that we had implemented.

PixelDr 2.0

For the reformed version of PixelDr, we extracted the essential components of the game, being the dogma of a medical assessment, but removed much of the gamified view, attempting to meet the users desire for slightly less gamification.

History was still processed in the same way, but we were now placing an essential focus on its accuracy, as we found that this was the essential component that brought users back to the platform.

The platform was now built on angular and had a familiar dashboard appearance. Below are some product shots of how PixelDr looked now.