Wemex Uses Azure OpenAI to Create SOAP Text for Medication History Input in 10 Secs

Manually entering medication histories hinders healthcare in shifting its focus of care from medicine to people

When Wemex launched the Medicom MC-1, Japan's first medical-receipt computer, in 1972, it became one of the first companies to offer doctors and administrators the digitalized receipts they had been calling for. The company went on to develop many cutting-edge products for medical institutions while encouraging greater efficiency in the medical field. It strives to support the wellbeing of Japanese citizens through the digital transformation of healthcare.

“We provide many solutions to support a broad spectrum of health and medical care in Japan,” says Keiji Okumura, Manager of Cloud Development in the Wemex Healthcare IT Division. Wemex covers many fields, including medical information systems such as electronic medical records, telemedicine, and comprehensive health solutions to manage health checkup data. “We have Japan’s top market share for electronic medical record systems in clinics^*1 as well as the highest share for medical computers in clinics and hospitals^*2.”

Wemex also provides tailored solutions for pharmacies. In June 2018, it introduced PharnesV-MX, an electronic medication history system for pharmacies, streamlining medication history management and other pharmacy workflows. In November 2023, Wemex expanded its offerings with the PharnesX, which boasts improved information processing capabilities and smoother operations.

*1: Source: Fuji Keizai, Electronic medical record system for clinics based on company share and value for 2022 (PHC findings) from “Current and Future State of DX Markets in Medicine and Healthcare 2023”

*2: Source: Fuji Keizai, medical-receipt computers based on company share and volume for 2020 (PHC findings) from “Current and Future State of Medicine-Adjacent and Medical Platform Markets in 2022”

Medication histories must record the instructions given to patients, the information received from them when issuing the prescription, and be understandable to other pharmacists to provide medications correctly and safely. For this reason, the medical world adopted the SOAP descriptive technique. This framework organizes content into four categories: subjective information (S), objective information (O), assessment (A), and plan (P) and describes them separately.

“We designed these solutions to support the Ministry of Health, Labour and Welfare’s initiative in shifting the focus of pharmacists from medicine to patients,” says Kanji Kitagawa, licensed pharmacist and a senior staff member of Dispensary Products in the Wemex Healthcare IT Division. The main goal, he explains, is to streamline medication history management and other pharmacist work so they have more time to engage with patients.

“Pharmacists must talk to patients to receive information and then manually record it afterwards, which takes a lot of time. Many pharmacists have asked us to simplify this process.”

Azure adopted for full compliance with regulations on medical information systems

“We continued to take on the challenge of saving time and effort in the medication history input process,” says Takuya Sekiguchi, a developer in the Wemex Cloud Development section, Healthcare IT Division. “We started creating a system based on standardized templates, but since medication history details differ from patient to patient, this approach was impractical.”

The launch of ChatGPT in November 2022 paved the way for a new solution.

Sekiguchi explains, “We had the idea of capturing text from conversations between pharmacists and patients, then loading it into generative AI to automatically collate in SOAP format. We tried ChatGPT and other generative AI while considering how to achieve our goal.”

There were two major hurdles to using generative AI to record medication histories. First, noisy pharmacy environments made accurate text capture unexpectedly difficult. Furthermore, as a medical system, the solution had to comply with the 3 Ministries, 2 Guidelines rules established for medical information systems by Japanese ministries. These guidelines include security requirements to fight increasingly diverse and sophisticated cyberattacks and protect personal information.

Sekiguchi says, “We decided to build the system on Azure to convert noise-mixed audio to text while complying with 3 Ministries, 2 Guidelines. Even with audio with audio with background noise, combining Azure AI Speech recognition and Azure OpenAI Service delivered more accurate text output than other solutions. Azure also has stronger security and transparent rules for learning, and knowing we can file any dispute in Japanese courts for security events is reassuring.”

Wemex identified security and privacy as particularly crucial requirements. This allowed the company to focus on developing features instead of spending time on non-functional security requirements. “All of our other cloud-based solutions use Azure because of its compliance with Three Ministries, Two Guidelines,” Sekiguchi says.

In addition, Azure OpenAI Service uses Provisioned Throughput Units (PTUs) to ensure consistent maximum latency and throughput.

“We started by using shared resources on a pay-as-you-go plan, but response times were uneven and there were timeouts and retries,” Sekiguchi continues. “But response is fast and stable now.”

A system to record verbal medication guidance and generate SOAP-formatted text automatically in tens of seconds

Wemex completed the first version of its medication history input service, which combines voice recognition and generative AI, in October 2023. At that time, the company was still using ChatGPT, but later announced a plan to build on Azure to meet requirements.

In the same month, Wemex presented the system as a reference exhibit at the 56th Annual Meeting of the Japan Pharmaceutical Association, where it received positive feedback, with audiences commenting on its practicality and usability. In parallel, proof-of-concept trials began, including tuning prompts to convert text to SOAP format.

Large-scale demonstration tests began in May 2024. Tsuruha Holdings, a pharmacy operator in northern Japan, cooperated with Wemex, using the system to record 400 instances of medication guidance at two stores. Based on the results, further tests began in Saera Pharmacy, which operates in the regions of Kansai, Kanto, Tokai, Hokuriku, and Chugoku in Japan, from July 2024. 

These tests included around 3,000 medication guidance sessions and contributed to further functional improvements for the system.

The commercial version of the system, named AI Medication History Input Service, released in September 2024.

“With our generative AI medication history input service, pharmacists can automatically generate SOAP text in seconds just by recording themselves when explaining medication guidance,” says Sekiguchi. Azure OpenAI Service responds in less than 10 seconds on average and is remarkably stable. “Pharmacists can spend more time interacting with patients since they no longer need to input medication histories manually; they can simply review the SOAP records created by generative AI and make corrections as needed.”

“AI Medication History Input Service does more than streamline pharmacist work,” says Kitagawa.

“Some young pharmacists have trouble recording medication histories because it requires experience, but the ability to view drug histories created by generative AI will make it easier to learn. Veteran pharmacists can review medication histories to see if their instructions are clear and understandable to patients. This increased consciousness will allow us to deliver patient-friendly medication guidance.”

Two weeks after releasing the system commercially, Wemex exhibited AI Medication History Input Service at the 57th Annual Meeting of the Japan Pharmaceutical Association. The system garnered significant acclaim and triggered a wave of inquiries.

However, Wemex hasn’t yet completed the challenge of streamlining medication history entry and management. The company is now working on a system that integrates with existing solutions and is developing a product that incorporates Wemex Online Medication Instruction, which it launched in September 2024.

“We provide many solutions to support a broad spectrum of health and medical care in Japan,” says Okumura. “We can provide even greater value by combining these with generative AI-powered solutions.”