Machine learning-based natural language processing to extract PD-L1 expression levels from clinical notes

Introduction: PD-L1 expression is used to determine oncology patients' response to and eligibility for immunologic treatments; however, PD-L1 expression status often only exists in unstructured clinical notes, limiting ability to use it in population-level studies. Methods: We developed and evaluated a machine learning based natural language processing (NLP) tool to extract PD-L1 expression values from the nationwide Veterans Affairs electronic health record system. Results: The model demonstrated strong evaluation performance across multiple levels of label granularity. Mean precision of the overall PD-L1 positive label was 0.859 (sd, 0.039), recall 0.994 (sd, 0.013), and F1 0.921 (0.024). When a numeric PD-L1 value was identified, the mean absolute error of the value was 0.537 on a scale of 0 to 100. Conclusion: We presented an accurate NLP method for deriving PD-L1 status from clinical notes. By reducing the time and manual effort needed to review medical records, our work will enable future population-level studies in cancer immunotherapy.

A mass-balance approach to evaluate arsenic intake and excretion in different populations

Unless a toxicant builds up in a deep compartment, intake by the human body must on average balance the amount that is lost. We apply this idea to assess arsenic (As) exposure misclassification in three previously studied populations in rural Bangladesh (n = 11,224), Navajo Nation in the Southwestern United States (n = 619), and northern Chile (n = 630), under varying assumptions about As sources. Relationships between As intake and excretion were simulated by taking into account additional sources, as well as variability in urine dilution inferred from urinary creatinine. The simulations bring As intake closer to As excretion but also indicate that some exposure misclassification remains. In rural Bangladesh, accounting for intake from more than one well and rice improved the alignment of intake and excretion, especially at low exposure. In Navajo Nation, comparing intake and excretion revealed home dust as an important source. Finally, in northern Chile, while food-frequency questionnaires and urinary As speciation indicate fish and shellfish sources, persistent imbalance of intake and excretion suggests imprecise measures of drinking water arsenic as a major cause of exposure misclassification. The mass-balance approach could prove to be useful for evaluating sources of exposure to toxicants in other settings.