August, 1 2023

Drug Data Import Component Explained

Drug Data Import Component Explained
Synapse Medicine

At Synapse Medicine, our goal is to continually innovate, shaping a better future for healthcare providers and patients alike. We aim to design and create intuitive solutions that streamline healthcare professionals' day-to-day operations and improve patient care efficiency.

Identifying the need for an automated process for transcribing patients' treatments into medical software solutions, we have developed the Data Import component using the latest technological advancements in AI.

Faster Drug Data Entry with Synapse’s Data Import Component

The main objective of our Data Import component is to simplify and enhance the process of recording patient treatments. It offers an array of options to detect prescriptions in diverse formats, including images of prescriptions, document uploads in different formats (such as pdf, jpeg, png, docx, .txt), or even manually entered text.

It has been designed to automatically scan the list of medications, meticulously extracting vital information like the type of medication, its form, the route of administration, and more. This component offers users the flexibility to edit the retrieved data manually, thereby allowing healthcare providers the freedom to add any patient-specific information regarding medication management. This makes the process of inputting data more accurate, personalized, and efficient.

Advanced Technology Driving Seamless Experience

At the heart of our automated prescription recognition module are two cutting-edge technologies: Natural Language Processing (NLP) and Optical Character Recognition (OCR). These innovative technologies work in tandem to ensure seamless, efficient, and error-free transcription of medication data.

Leveraging Natural Language Processing (NLP)

As Mickael Rey, our Data Scientist explains, "We utilize different NLP techniques. For instance, in the first step, which is to detect potentially relevant information, we use deep learning algorithms. These neural networks will discern what is valuable to analyze and what is considered noise. For the data matching part, we use more conventional methods, such as word proximity rules. Given our comprehensive database, this method works exceptionally well."

Harnessing the Power of Optical Character Recognition (OCR)

Optical Character Recognition, or OCR, is another crucial technology that the Data Import component leverages. OCR reads text on images, converting it into a digital format that can be read and utilized by software. This technology plays an essential role when prescriptions are directly uploaded as images.

Drug data import component using nlp and ocr graphic

Enhancing User Experience and Reducing Transcription Errors

By automating the transcription process, our Data Import component doesn't merely save valuable time for healthcare providers. It also significantly reduces the risk of drug-related errors, thereby enhancing the accuracy of patient treatment data. This tool has been designed specifically to meet the unique needs of healthcare professionals and aims to automate the process of entering treatments, thereby improving its reliability.

The component is now an integral part of our extensive library of clinical decision support components, ready for software publishers to enhance their users' experience.

Evaluating the Impact of the Data Import Component

To understand the effectiveness and efficiency of the Data Import component, we conducted a comparative study between the traditional manual method of drug identification and input and our automated solution. This study involved eight healthcare professionals, including doctors and pharmacists. Each participant transcribed 74 prescriptions using both methods, and the time taken for transcription was meticulously recorded.

The results of this study were compelling, with a significant time reduction observed when using the Data Import component. The automated system helped healthcare professionals save 50% more time compared to manually entering treatments, dramatically increasing the efficiency of their workflow.


Analyzing the Accuracy of Drug Treatment Recognition

In addition to time-saving, the accuracy of the transcription is also a critical factor in evaluating the performance of the Data Import component. We used the F1-Score, a widely accepted performance measure, to assess the quality of our drug recognition model.

The F1-Score combines the precision and recall of a classification model into a single value between 0 and 1. A score closer to 1 indicates a superior model performance, pointing to high precision and recall values.

Our component was evaluated based on two different criteria - the detection of drug names only ("F1-Score Shortname") and the detection of a complete entity including the name, dose, and dosage form of the drug ("F1-Score Complete").

accuracy drug data import component

The Results

With the Data Import component, manual transcription time was noticeably reduced, making the process more efficient and streamlined. It also drastically improved the accuracy of drug information transcription, thereby mitigating the risk of medication errors. In addition to these measurable improvements, our Data Import component also offers an intuitive and optimized user experience.

Dr. Colin Bui, our Product Manager, emphasizes, "We developed the Data Import component to allow healthcare professionals to save time during prescription transcription. It can be integrated into any healthcare software."