Smarter Referrals: How AI Could Help Doctors

Doctors are overwhelmed with patient referrals, and delays can harm patients. Can artificial intelligence (AI) and machine learning help? Before implementing AI, we must ensure the system is safe and effective.

The Problem

Currently, doctors write referrals in free text, similar to a long letter. However, computers struggle with unstructured data. They perform best with structured data, like forms with checkboxes, ensuring every patient's information is clear and complete.

Ensuring AI Makes Good Decisions

Different doctors may make different decisions for the same patient. Since humans aren't perfect, we need to train AI on a gold-standard decision set—referrals where experts agree on the best course of action.

A Safe Approach

To keep patients safe, we should start small. Simple decision trees (like flowcharts) can make basic decisions. As we progress, we can introduce more complex machine learning tools, testing and improving the system step by step.

Will This Work?

Some researchers conducted a survey of doctors and analyzed denied referrals. They also created draft decision trees. This research helps us understand the challenges and find the best solutions.