Ninety percent of any business is effective administration and medicine is no different. You need to be treating the right patient for the right thing.

Problem

Over 20%* of entries in electronic patient records contain errors or omissions. Inaccurate notes can result in delayed or wrong treatment.

Examples

• A consultant may order an “Abdominal scan” rather than a “Pelvic scan”. Radiologists tell me that this is a very common mistake.
• A respiratory condition is wrongly named. Medical coders tell me that this is complex area and staff frequently enter an erroneous value.
• A consultant fails to note that a scan ordered at the last appointment has not been actioned within expected timescales. 
• An entry contains only 10 words and is insufficient for use by nurses or ML/AI.

Cause

The health system recognises over 10,000 medical conditions each which may each have many different treatment pathways. It is not possible for medical professionals who are making notes to be know the correct name for everything or to look it up in the time allowed.

Solution

An AI assistant should validate each entry in the patient notes when the medical professional clicks “Enter”. It should respond to the entry with a Green, Amber or Red indicator with a hint if necessary. For example,

“Red: This patient has prostate issues and you have ordered an abdominal scan. In 95% of similar cases a pelvic scan was ordered.”

Ease of implementation

This is technically easy to implement because the AI has access to the entire patient history and can easily spot errors of commission and omission. I believe that it would be relatively easy for a team of medical professionals to tune the output so that it supported their work effectively. I think that medical staff would welcome this assistance. Therefore the cost of implementation should be low.