We designed an NLP model that could quickly identify and assign treatments to patients based on the seriousness of their medical conditions.

As one of the leading healthcare solutions and treatment providers in the country, our client and their employed doctors and care administrators were finding themselves time-stricken to evaluate the ever-increasing number of patient files in order to assign treatment based on urgency. All the extra resource hours being spent on the review of patient files was adding to the lack of healthcare support staff which was becoming a growing problem in light of the COVID-19 situation.

Problems That Needed To Be Addressed

The client was facing issues with its healthcare staff being overworked and unable to adequately focus attention on patients in urgent need of care. There needed to be a system that would help identify critical care patients and assign them immediate attention.

The Solution We Provided

Our team employed natural language processing techniques on a sample of ~5,000 medical dictations to build a classification model, determining how urgently a patient needed treatment.

The Impact and Long-term Results

The artificial intelligence solution we created advised the healthcare practitioners and helped them to identify urgent care requirements, with over 95% accuracy. The model ensured that the client’s healthcare staff was not over-burdened with the task of identifying the most critical patient case files.

The model performs with greater than 95% accuracy, automatically identifying urgent cases in the vast majority of instances and augmenting our client’s ability to rapidly and accurately perform triage.

So, Here’s How We Did It

Natural Language Processing: We used Azure Machine Learning services to quickly train the model using a sample size of over 5,000 medical dictations. Through this exercise that was monitored closely for 100% accuracy in diagnosis, our model was able to scale quickly and expand effectively to review and identify critical cases on a much broader sample in a very short period of time. In order to reduce risks and situations where patient cases get identified wrongly as non-critical, the model was trained to be very conservative while drawing conclusions.

The model also allowed for quick deployment into a production environment.

Artificial Intelligence: The fully developed model now works with greater than 95% accuracy and uses our Artificial Intelligence solution to advise healthcare practitioners about the urgency associated with each case. Our solution has eliminated the need for additional time being spent by the already overburdened healthcare staff.