Learn how integrated real-world data, AI-powered analytics, and unstructured EHR insights helped identify early-stage patients, uncover disease screening trends, and inform more effective commercialization strategies.
A global biopharma company with a diabetes therapy needed to quantify the national rate of early disease screenings and gauge physician sentiment. It wanted to assess how screening rates varied geographically and over time, plus physician intention, to see where screening rate changes could shape commercialization efforts.
Identifying early-stage patients within real-world data is complex. Claims and lab data often lack detail to differentiate between certain diagnoses, understand clinical reasoning behind diagnostic tests, and detect early disease onset that precedes formal ICD-10 coding. This diagnostic ambiguity made it difficult to build a high-confidence patient cohort and evaluate the effectiveness of early detection efforts.
The company partnered with Norstella, which used NorstellaLinQ, biopharma’s first fully integrated data asset, to find a solution that combined clinical expertise, AI-powered insights, and multimodal data integration.
Over a six-week period, the teams defined precise identification logic and clinical signals. During an eight-week model development phase, advanced machine learning classified patient stage and type from unstructured EHR notes, incorporating autoantibody (Aab) testing patterns, physician sentiment, and behavioral trends to build HCPlevel triggers. Large language model-driven insights validated directional hypotheses drawn from structured data.
This surfaced nuanced signals like physician intent, test-ordering behavior, and specialty-specific diagnostic trends, plus physician sentiment and decision-making factors, giving a more accurate view of national Aab test ordering patterns.
With NorstellaLinQ, the client merged the depth of unstructured EHR data with the breadth of structured data to gain comprehensive insights into disease screening activity, including geographic and specialtyspecific trends and how they changed over time, as well as patient demographics and HCP specialties responsible for early detection.
Actionable insights and KPIs from the project drove commercialization, engagement strategies, and real-world evidence development, including targeted support for field deployment, educational outreach, and alignment of field team incentives.
With these insights, the client has a repeatable, scalable model to identify early diagnosis in this and other hard-to-detect disease states—ushering in a new era of data-driven precision in early patient identification.
Get Started
Frequently asked questions
Integrated real-world data combines structured and unstructured healthcare information to identify early-stage patients, reveal screening trends, and provide deeper insights into disease progression before formal diagnosis.
Unstructured EHR notes capture clinical context—such as physician intent, diagnostic reasoning, and testing behavior—that is often unavailable in claims or laboratory data alone, enabling more accurate early patient identification.
AI-powered analytics and machine learning analyze multimodal healthcare data to classify patient stage, identify disease screening patterns, and uncover physician behaviors that help improve early detection and commercialization strategies.
Norstella helped the biopharma company build a scalable model for early disease detection, uncover geographic and specialty-specific screening trends, and generate actionable insights that supported commercialization, field engagement, and real-world evidence development.
NorstellaLinQ integrates structured and unstructured healthcare data with AI-powered analytics to identify early-stage patients, reveal physician screening behaviors, and create repeatable models that improve early diagnosis across difficult-to-detect diseases.
