How this healthcare company cracked the code on making AI actionable — and safe

Article By: Eric Schrock

Blog Source From : https://www.healthcaredive.com/

We’re living at a crossroads in healthcare, with AI set to revolutionize personalized medicine and improve patient care. This transformative moment mirrors previous technological medical revolutions, but with unprecedented potential scale and impact.

AI systems can now analyze complex patient data across genomics, medical imaging, electronic health records, wearable devices, and more. However, this promise remains contingent on solving the fundamental challenge underneath: managing the scale and complexity of healthcare data.

Healthcare data is unique in its complexity. It includes diverse types of information like clinical notes, medical imaging, lab results, vital signs, and genomic data. These come in different formats and standards that must be handled under strict regulatory guidelines.

When a patient visits a doctor for the first time, for example, they might bring years of medical records. Doctors don’t have time to comb through thousands of pages of charts. Multiply that one patient by all the patients these doctors see daily and the real challenge becomes clear.

But it’s not just a volume problem — it’s understanding each patient’s unique journey. It’s not enough to know someone is a 40-year-old male from Atlanta. They need to understand their specific clinical history, symptoms, and treatment responses.

As providers work to build these comprehensive patient profiles, they face an additional challenge of mounting cyberthreats. The last two years have been record-breaking for patient privacy breaches, according to a recent report. Healthcare providers must balance protecting sensitive patient information with maintaining efficient data accessibility.

OM1 is a leading data and technology company leveraging big clinical data and AI to better understand, compare and predict patient outcomes. The company is at the forefront of balancing the unique complexities of healthcare data, security and governance, and massive scale, while unlocking the benefits of AI.

How did they crack the code on AI-driven personalized patient insights? By transforming their data strategy.

OM1 started with a homegrown data processing environment that presented significant challenges as they scaled. The complexity of managing patient data, particularly unstructured information from clinicians’ notes, was compounded by reliability issues in their data processing systems. The company’s managed Spark-based model struggled with stability, job scaling, and costly maintenance. Data engineers struggled with a disjointed process that required complex pipeline setups, manual Spark cluster management, and execution across different environments.

OM1’s move to Snowflake started with data storage in 2017 by standardizing on SQL and dbt for processing structured data. But they quickly realized it could evolve it into a comprehensive solution through Snowpark and its ecosystem of partners. By integrating tools like Modelbit for machine learning engineering and Hex for Python code authoring, OM1 streamlined its data processing workflow. The transition eliminated the need for external frameworks and simplified its entire data infrastructure, making it more accessible to data analysts, clinical informaticists, and data scientists who primarily work with SQL.

By investing in its data foundation, OM1 saw a 75% cost reduction in data processing operations while being able to analyze over 100 million records in under 30 minutes. The company adopted Snowpark Container Services for seamless integration of tools like Private AI for text redaction into their dbt pipelines, transforming complex multi-step processes into simple SQL functions. The new system demonstrated enhanced reliability and performance while maintaining strict HIPAA compliance and strengthening the company’s security posture.

This modernized infrastructure positioned OM1 perfectly for AI implementation, particularly in their PhenOM® platform, an AI-powered digital phenotyping system. The simplified model deployment enabled rapid iteration of AI models and easier delivery as services through SQL or APIs. This foundation supports OM1’s mission to leverage AI for personalized medicine, allowing them to process complex patient data more effectively and securely.

Looking ahead, there is tremendous potential for what OM1 is doing and AI-driven personalized medicine. Healthcare providers are moving toward a future where they can better identify undertreated conditions, guide treatment decisions, and provide clinicians with AI-powered insights.

Imagine a PhenOM model combing through a decade of data, identifying signals within your unique ‘fingerprint’ of disease progression, treatments, and symptoms to inform doctors of the best approach for your treatment resistant depression – before you ever meet. What if we could identify patients at elevated risk of abdominal aortic aneurysms and get twice the number of patients the screening they need over traditional guidelines?

The goal isn’t to replace clinical judgment, but to enhance it. By providing clinicians with better tools to understand patient data, we can help them make more informed decisions and improve patient outcomes, without requiring them to search through hundreds of hours of records.

The journey from paper to electronic health records to a new era of AI has taught us that while technology is important, its impact on human lives is what truly matters. As we continue to integrate AI into healthcare, we must remember that we’re not just solving technical challenges, but working on a future where every patient receives the personalized care they deserve for treatment, research, and innovation.