
What You Should Know:
– PicnicHealth, a company focused on empowering health and research through patient data, announced PicnicAI, a technology platform unlocking the long-awaited Universal Patient Record.
– The innovative platform leverages the power of artificial intelligence (AI) to collect and interpret medical records from across the United States, transforming them into valuable insights for both patients and researchers.
PicnicAI: Powering Personalized Care and Streamlined Research
PicnicAI sits at the core of PicnicHealth’s offerings, driving two key products:
- Patient application: By analyzing individual health data, PicnicAI empowers patients with personalized care tailored to their specific needs.
- Life sciences product: This product streamlines observational research for life sciences companies, enabling them to achieve study goals faster. By harnessing the power of AI, PicnicHealth offers a responsible and efficient approach to healthcare advancement.
Unifying Fragmented Data: A Seamless Patient Experience
One of PicnicAI’s key strengths lies in its ability to bridge the gap between fragmented healthcare systems. By connecting to 100% of U.S. care sites, PicnicAI eliminates the traditional hurdles of inaccessible or scattered patient data. Patients simply sign up, and PicnicAI grants them near-instant access to most records, with over 95% of key records available within five days.
Accuracy and Fairness: Core Principles of PicnicAI’s LLM
PicnicAI utilizes a Large Language Model (LLM) specifically trained on a massive dataset of medical records. This data encompasses over 350 million annotations across records from more than 100,000 care sites. This extensive and diverse training approach allows PicnicAI to achieve exceptional accuracy in structuring and summarizing a patient’s medical history.
Furthermore, PicnicAI’s training data is designed to mitigate biases. Unlike models trained on limited datasets, PicnicAI avoids potential biases related to factors like race, ethnicity, or socioeconomic background, ensuring fair and equitable analysis.
Outperforming the Competition: Benchmarking PicnicAI’s LLM
Performance tests demonstrate PicnicAI’s LLM as a leader in the field. Compared to GPT-4, PicnicAI achieves three times better performance in recognizing clinical entities within medical records. Additionally, it surpasses leading medical LLMs by 47% in record abstraction accuracy. Notably, PicnicAI delivers this superior performance with 90% more cost-effective computing resources per record analyzed.
Continuous Improvement: Optimizing Performance for the Future
PicnicHealth prioritizes continuous improvement for PicnicAI. Through daily sampling, the platform constantly validates and fine-tunes its performance, ensuring the highest level of accuracy even on results initially categorized as high-confidence.
“Personalized medicine requires technology like PicnicAI,” says Troy Astorino, CTO and co-founder of PicnicHealth. “It has taken us a decade to build, but the results feel magical. We can now instantly construct and understand any patient’s complete care journey, pulling records together from every hospital and doctor’s office where that patient has been seen. This allows us to provide clinical insights to patients and our research partners with speed, efficiency, and scale not possible anywhere else. It’s how patient records should work.”