Healthcare organizations face a critical challenge: to efficiently analyze large volumes of medical records every day without compromising security and compliance standards. Using language models to analyze this data presents itself as a viable solution, although it often involves restructuring large portions of information, thus increasing operational costs and response times.
Care Access, a leading company in providing healthcare services and clinical research globally, encountered this issue when expanding its health program. Their ability to process between 300 and 500 medical records daily entailed multiple requests for each analysis, leading to the need to reprocess significant portions of these records. As hundreds of new participants decided to share their medical information daily, Care Access faced the urgent need to implement a solution that could scale efficiently while meeting strict privacy and compliance regulations.
The incorporation of Amazon Bedrock’s “caching of prompts” feature proved crucial in this process. By storing the static content of medical records and only varying the analysis questions, Care Access significantly reduced costs and improved processing times. This advancement transformed the handling of medical records from a potential obstacle to a facilitator of program growth.
In their mission to improve healthcare in communities with access barriers, Care Access offers quality research and medical care services, with nearly 15,000 new participants joining monthly worldwide. However, rapid growth brought logistical challenges that required appropriate scalability in response to increasing demand. By adopting a language model solution through Amazon Bedrock, the company was able to analyze diverse medical records while still adhering to compliance and security standards. However, the original implementation involved multiple requests per analysis, resulting in high operational costs.
The “caching of prompts” ability made it possible to reuse parts of a request that would have previously needed to be recomputed for each record. Thus, medical content becomes a static prefix, while analysis questions constitute the dynamic part of each query. This approach proved highly effective, improving performance and significantly reducing costs.
With the integration of Amazon Bedrock, Care Access optimized both speed and costs associated with processing thousands of medical records using a language model. This resulted in an 86% cost reduction and a 66% decrease in processing time per record. Additionally, the company experienced savings of 4 to 8 hours daily in processing time.
This innovative solution not only allowed Care Access to address immediate challenges, but also positioned it for sustained growth, connecting more communities with healthcare opportunities and clinical research resources that can transform lives. This case highlights the importance of adopting suitable technologies to solve business problems while supporting long-term goals.
via: MiMub in Spanish
