Beth Israel Lahey Health boosts document search performance with Azure AI services

Like many academic medical centers, Beth Israel Deaconess Medical Center (BIDMC) continually strives to improve patient care and optimize efficiency for providers and staff. The difference maker for this teaching affiliate of Harvard Medical School is the innovative way it embraces leading-edge technology to eliminate barriers and help it deliver exceptional patient care experiences.

BIDMC is part of Beth Israel Lahey Health (BILH), a nationally awarded healthcare system with more than 4,700 physicians and 39,000 employees in the greater Boston area. The medical center wanted to improve its ability to search its digital document repository, which many clinicians found cumbersome to use. BILH used its Innovation Lab team—an internal assembly of data scientists, software developers, and product managers—to find a solution that could unlock fast, accurate access to over 3,800 critical policies, procedures, guidelines, and directives (PPGD) used on a daily basis.

“The previous methodology was limited to keyword search, which was not user-friendly and wasted valuable time to locate information,” says Venkat Jegadeesan, Vice President, Technology and Innovation at BILH. “To compensate, our clinicians would find other ways to get the answers they needed, creating staff inefficiencies.”

Initiating a Microsoft Azure solution

After reviewing several search-based technologies, the Innovation Lab team decided to develop an AI-driven, natural language processing solution. During the discovery phase, the team looked into Azure AI capabilities and found a range of integrated service advantages built into Azure AI Foundry.

“Azure AI Document Intelligence proved especially useful for indexing and searching the high volume of documents we use,” explains John Shang, Director, Data Engineering and EHR Systems at BILH. Shang also emphasizes that Azure OpenAI Service provided BILH with the latest GPT-series models, creating additional advantages in app performance.

Innovating with its first generative AI search app

Using Azure AI Document Intelligence, Azure OpenAI Service, and Azure AI Search, the Innovation Lab team built a mobile-friendly application called ChatPPGD. This generative AI app lets staff ask questions in natural language and quickly get a response, along with a reference to the originating documents and relevant medical, academic, and research citations. Shang notes Microsoft’s safety and filtering options were critical in helping ensure secure and accurate responses.

“The benefits of using these integrated services together allowed the app to read all information stored in documents, extract key details, index it, and create embeddings,” explains Shang. “Azure AI Search helped us provide both vector and hybrid search, which significantly improved the usability of the service and provided the right context for searches.”

Optimizing patient and provider experience

“Since launching in February of 2023, ChatPPGD has shown 98% accuracy in its responses,” says Jegadeesan. “We are consistently seeing upwards of 800 queries per week, improved efficiency and policy compliance, and user feedback has been incredibly positive.” Jegadeesan notes these outcomes are helping the hospital deliver on its goals.

“The internal success of the ChatPPGD app has opened up new possibilities for us to accelerate development of solutions that help optimize the patient and provider experience,” says Jegadeesan. “Our Innovation Lab team is getting many requests for AI-based services and solutions, so we’ve begun meeting twice a week with clinicians, researchers, and staff from across BILH to hear their ideas, collaborate, and create AI solutions together. We see this as a big win.”

Discover more about Beth Israel Deaconess Medical Center on Facebook, Instagram, LinkedIn, X/Twitter, and YouTube.