Real-world data and EHR-derived endpoints in oncology

Definition

Real-world data (RWD) in oncology is clinical evidence generated during routine cancer care: the information recorded in electronic health records, insurance claims, disease registries, and connected devices, rather than under the controlled conditions of a clinical trial. Real-world evidence (RWE) is the clinical evidence about a therapy's use, benefits, or risks that is produced by analyzing that data.

EHR-derived endpoints are outcome measures computed from data that already lives in the electronic health record, rather than collected prospectively for a study. Common examples are overall survival, time to treatment discontinuation, real-world progression, and treatment-related toxicity, each defined from what clinicians actually recorded during care. The endpoint is only as good as the documentation and the logic used to derive it, which is why the data-engineering step - turning messy routine records into research grade variables - is a discipline in its own right.

• Source: routine care - EHRs, claims, registries, devices - not a trial protocol.
• Endpoints: overall survival, time to treatment discontinuation, real-world progression, toxicity, computed from recorded data.
• Strength: scale and generalizability to the patients seen in everyday practice.
• Caveat: missingness, free-text documentation, inconsistent coding, and no randomization must be handled deliberately.

Why it matters

Randomized trials remain the standard for establishing that a therapy works, but they enroll selected populations under controlled conditions. The patients who end up in routine oncology practice are frequently older, have more comorbidities, and are more diverse than a trial cohort, so trial results do not always describe what happens at the bedside. Real-world data reflects those patients directly, and it exists at a scale no prospective study can match, because every clinical encounter already produces it.

That scale and generalizability come with a cost. Real-world data is recorded for care, not for research: key variables are missing, important findings are buried in free text, coding is inconsistent between clinicians and systems, and there is no randomization to balance confounders. Turning it into a trustworthy endpoint requires careful curation, validation, and - most durably - structured data standards that capture the right elements at the point of care. The quality of the data underneath is the whole game. This is where the concept meets the standards work: a standardized, FHIR-shaped record is what makes real-world endpoints reproducible instead of bespoke.

Travis Osterman's work in real-world data

Dr. Travis Osterman is a practicing medical oncologist and Associate Vice President for Research Informatics at Vanderbilt Health, board certified in both medical oncology and clinical informatics. Much of his research program is built on exactly this question: how much clinically useful signal can be recovered from data that already exists in the EHR.

The clearest example is the GE HealthCare Digital Precision Oncology collaboration (2019-2024), on which he served as principal investigator of the flagship study. The team built a machine-learning framework from routine structured data already living in the Vanderbilt EHR - baseline laboratory measurements aggregated over 60-to-365-day windows, comorbidities, prior treatments, and demographics - to predict both the effectiveness and the toxicity of immune checkpoint inhibitors before the first dose. Models forecast one-year overall survival and three major immune-related toxicities (hepatitis, colitis, and pneumonitis) using only pre-treatment data. The flagship paper, Prediction of Effectiveness and Toxicities of Immune Checkpoint Inhibitors Using Real-World Patient Data (Lippenszky et al., JCO Clinical Cancer Informatics, 2024), drew on more than 2,200 patients treated through the end of 2018, trained and tested the models on an internal 80/20 split at Vanderbilt, and reported reasonably strong discrimination across all four outcomes (areas under the curve of roughly 0.73 to 0.76). To the authors' knowledge it was the first machine-learning solution to assess individual ICI risk-benefit profiles based predominantly on routine structured EHR data, which means it required no additional data collection beyond what clinicians already record.

Making that approach credible beyond a single institution took a separate step. External validation of the efficacy and toxicity models on a real-world German pan-cancer cohort of roughly 4,250 patients was reported at the Society for Immunotherapy of Cancer meeting in 2023 (Kiss, Lippenszky et al., SITC 2023, abstract 1294) and again in Dr. Osterman's ESMO Immuno-Oncology talk in Geneva in December 2024, where the models retained a substantial fraction of their training-cohort performance on the external cohort. That external result, reported apart from the original flagship paper, is what moved the work from a promising local model toward generalizable real-world evidence.

A separate companion paper carried the same real-world approach into imaging, a data source outside the structured-EHR flagship model: Radiomics and Deep Learning Prediction of Immunotherapy-Induced Pneumonitis From Computed Tomography (Smith et al., JCO Clinical Cancer Informatics, 2025) used routinely acquired pre-treatment CT volumes to predict which patients would later develop ICI-induced pneumonitis. The unglamorous but essential half of the work - converting raw EHR records into research-grade cohorts - is documented in Accelerated curation of checkpoint inhibitor-induced colitis cases from electronic health records (Rahman et al., JAMIA Open, 2023; Osterman a co-author), which speeds the identification of toxicity cases hidden in routine documentation. The full peer-reviewed record is on the research page, and the applied-AI framing sits on the AI-in-oncology expertise page.

The structured-data caveat

Real-world endpoints are only as reliable as the structure of the data they are derived from, and this is the through-line of Dr. Osterman's leadership work. Under his direction of clinical informatics at the Vanderbilt-Ingram Cancer Center, the Vanderbilt Health EHR holds more structured genomic data than any other institution in the United States - the kind of curated substrate that makes real-world analysis tractable in the first place. As Chair of the mCODE Executive Committee, he leads the FHIR-based oncology data standard implemented at more than 70 institutions across six countries. mCODE is also the only method of submitting data to the CMS Enhancing Oncology Model, which means real-world oncology data has already become regulatory infrastructure: a value-based care program that runs on standardized, EHR-derived elements. Standardize the input and the endpoints computed from it stop being bespoke and start being comparable across institutions.

Key works

1. Lippenszky L, Mittendorf KF, Kiss Z, LeNoue-Newton ML, Napan-Molina P, Rahman P, Ye C, Laczi B, Csernai E, Jain NM, Holt ME, Maxwell CN, Ball M, Ma Y, Mitchell MB, Johnson DB, Smith DS, Park BH, Micheel CM, Fabbri D, Wolber J, Osterman TJ. Prediction of Effectiveness and Toxicities of Immune Checkpoint Inhibitors Using Real-World Patient Data. JCO Clinical Cancer Informatics 2024.
2. Kiss Z, Lippenszky L, Laczi B, Napan-Molina P, Csernai E, Brehmer A, Kim M, Keyl J, Siveke J, Meyer M, Grünwald V, Kasper S, Roesch A, Schuler M, Osterman T, Wolber J, Kleesiek J. External validation of machine learning models to predict efficacy and toxicity of immune checkpoint inhibitors using real-world pan cancer cohorts. SITC 2023, abstract 1294.
3. Smith DS, Lippenszky L, LeNoue-Newton ML, Jain NM, Mittendorf KF, Micheel CM, Cella PA, Wolber J, Osterman TJ. Radiomics and Deep Learning Prediction of Immunotherapy-Induced Pneumonitis From Computed Tomography. JCO Clinical Cancer Informatics 2025.
4. Rahman P, Ye C, Mittendorf KF, LeNoue-Newton M, Micheel C, Wolber J, Osterman T, Fabbri D. Accelerated curation of checkpoint inhibitor-induced colitis cases from electronic health records. JAMIA Open 2023.

Related concepts: AI in oncology · mCODE · Clinical genomics in the EHR · CMS Enhancing Oncology Model.

On this site: Digital Precision Oncology (case study) · AI in oncology (expertise) · peer-reviewed record · leadership and governance.

Back to top