Personalized Medicine for Cancer Prevention: New Study Finds Electronic Health Records Crucial for Early Detection

Strategic Insights -

A groundbreaking study on personalized medicine for cancer prevention has been published, highlighting the importance of electronic health records (EHRs) in early detection. Researchers at Harvard College developed a prediction model using data from Mass General Brigham's EHR to identify high-risk individuals for non-small cell lung cancer (NSCLC). The study found that incorporating EHR-derived features significantly improved the model's performance compared to a baseline model relying on demographic and smoking information. The prediction model achieved an area under the receiver operating characteristic (ROC) curve of 0.801, indicating a high level of accuracy for predicting 1-year NSCLC risk in the population aged 18 and above.

Key Takeaways:

  • Researchers at Harvard College developed a prediction model using electronic health records (EHRs) to identify high-risk individuals for non-small cell lung cancer (NSCLC).
  • The model incorporated 127 EHR-derived features, including smoking, lab test results, and chronic lung diseases, which were found to be predictive of early NSCLC diagnosis.
  • The model achieved an area under the ROC curve (AUC) of 0.801 for predicting 1-year NSCLC risk in a population aged 18 and above.
  • The model demonstrated superior performance compared to a baseline model that only relied on demographic and smoking information.
  • The study emphasized the importance of incorporating EHR-derived features for personalized cancer screening recommendations and early detection.
  • The researchers identified EHR-derived features that are predictive of early NSCLC diagnosis, including smoking and relevant lab test results.
  • The study found that the predictive model improved the early detection of NSCLC compared to the baseline model.
  • The researchers developed a three-stage ensemble learning approach to build the prediction model.
  • The study highlighted the potential of incorporating EHR-derived features for personalized cancer screening recommendations and early detection.

Statistics:

  • The prediction model achieved an area under the ROC curve (AUC) of 0.801 for predicting 1-year NSCLC risk in a population aged 18 and above.
  • The model achieved an AUC of 0.757 for predicting 1-year NSCLC risk in a population aged 40 and above.
  • The predictive model had a positive predictive value (PPV) of 0.0173 and specificity of 0.02 for predicting NSCLC risk in a population aged 18 and above.
  • The model incorporated 127 EHR-derived features, which were found to be predictive of early NSCLC diagnosis.

Sources:

  • Early detection of non-small cell lung cancer: an electronic health record data-driven approach. BMC Medicine, 2025;23(1):551.
  • Bmc, Campus, 4 Crinan St, London N1 9XW, England. (BioMed Central - www.biomedcentral.com/; BMC Medicine - www.biomedcentral.com/bmcmed/)
  • NewsRx. Researchers at Harvard College Target Personalized Medicine (Early detection of non-small cell lung cancer: an electronic health record data-driven approach). Cancer Weekly. October 21, 2025; p 739.