Ward-E

Reduce administrative work in patient examination during ward rounds and enable patients to understand procedures and outcomes based on data.

Philipp Bode  Tom Martensen

Desirability

• Documentation in hospitals is a mostly manual task for the care personnel
• Patients and medical professionals do not share the same language, which can lead to misdiagnoses and non-adherence to therapies
• Patients and supporting caregivers want to be involved in the care process
• Health care providers want to make informed decisions

Viability

• Misdiagnoses affect up to 12 million Americans/annum, half of those can be considered harmful [1]
• Unneccesary tests and procedures create costs of up to 700 billion dollars/annum [2]
• Low treatment adherence leads to follow-up costs of up to 300 billion dollars/annum [3]

Feasibility

• Given advancement in personal assistants such as Siri, Alexa or Google Home, we consider speech-to-text a solved problem. However, such language models may have to be re-trained for medical purposes. We propose a semi-supervised approach here.
• Named-Entity recognition in particular in the medical field is the topic of ongoing research. However, as hyperscalers such as AWS tap into the field of natural language processing for medical texts, productive systems seem to be on the horizon.
• Any solution must have an interface to the clinical information system used at the hospital to insert and retrieve information. While there is a multide of different vendors, co-innovating with one of them may pioneer the solution.

Reservations

• How will practitioner react to the automated documentation? Will they feel controlled or will they see the benefits on patient outcomes?
• What kind of medical content is relevant for the patients and how can it be created?
• What about data privacy and security concerns?