The Musketeer System

The Musketeer System

Inspiration

We were inspired by the stories of people who suffer from Munchausen syndrome by proxy (MSBP), a rare form of child abuse where a caregiver deliberately makes a child sick or injured to gain attention or sympathy. We wanted to create a solution that could help detect and prevent this kind of abuse, as well as other forms of medical resource misuse, such as drug-seeking or insurance fraud.

What it does

The Musketeer System is a middleman software that collects and analyses medical reports from multiple hospitals and clinics. It uses machine learning to flag out suspicious patterns of health care utilisation, such as frequent visits, multiple providers, inconsistent diagnoses, or excessive tests and treatments. The software then alerts the medical staff and provides them with recommended questions to ask the patients to confirm the diagnosis of abuse.

How it should be built

Privacy

The data that are sent by the healthcare body shall first pass through an encryption service. The encryption service will encrypt sensitive data through a one-way hash to ensure the patient's sensitive data are not reverse engineered. The encryption service shall be located at the respective healthcare service to prevent any sensitive data from being sent over to the cloud/external site without it first being encrypted.

Data Extraction

As the data sent by the hospitals may come in different format as they may not have a standardised form field, a data parser is required to extract out relevant parameters to the machine learning (ML) model. Thus, the data are then passed through a data parser with a Large Language Model (LLM) as its foundation. The LLM, such as ChatGPT, are capable of capturing the context/data from the different data formats and parse our relevant parameters for the ML model.

ML Model

The ML model is trained on the data extracted by the data parser. The model itself shall be made based on transparent models. This is to ensure that the decision that the model comes down to are transparent and explainable to human reason. Additionally, a transparent and explainable model ensure that it can be check for bias in its reasoning.

AI Chatbot

AI Chatbot is necessary to consume output of the ML model along with its path to judgement. This is so that healthcare personnel are able to clarify on why would the ML "thinks" that the person is a drug abuser. This will prevent baseless accusations and allow for healthcare professional to exercise their human judgement.

Governance

The Musketeer System should be governed to maintain trust in the system. The governing body shall be made up of all the participating healthcare provider and the developer of the system as they are the stakeholders of the system. The governing body is responsible of implementing data governance policies and periodic audit of the ML model's effectiveness.

Challenges

Some of the potential challenges:

Finding reliable and realistic data sets of medical records that could be used to train and test the machine learning models.
Convincing healthcare provider and patients to provide their data for TMS.
Convincing healthcare provider and patients that the privacy and security of the medical data, especially when dealing with sensitive information such as personal details, diagnoses, or treatments are maintained.
Government approval on handling of privacy data.

What's next for The Musketeers

Some of the things that we plan to do next are:

Assuring privacies to stakeholders by explaining how the system maintain patient's privacy
Expand our target audience by adapting our software to different contexts and scenarios, such as other countries, regions, or health care systems.
Conduct user testing and evaluation to validate our solution and measure its impact.