M. Zhao1,2, R. Benson1, I. Friesner3, M. Ho1, W. T. Li1, C. I. Nnadi4, J. Chew5, N. W. Cho5, H. Vasudevan5, S. Sinha5, M. Rabow4, L. Boreta5, S. E. Braunstein5, D. Spiegel6, and J. C. Hong5; 1Bakar Computational Heath Sciences Institute, University of California San Francisco, San Francisco, CA, 2University of California, Berkeley, Berkeley, CA, 3University of California, San Francisco, Bakar Computational Health Sciences Institute, San Francisco, CA, 4University of California San Francisco, San Francisco, CA, 5Department of Radiation Oncology, University of California San Francisco, San Francisco, CA, 6Beth Israel Deaconess Medical Center, Boston, MA
Purpose/Objective(s): Cancer and its treatment carry psychiatric burdens alongside physical symptoms.To determine if higher levels of stress are correlated with increased side effects of cancer radiotherapy (RT), we analyzed the association between the NCCN Distress Thermometer (DT), a routinely administered 0 to 10 scale of distress, and natural language processing (NLP)-extracted symptoms documented in clinical notes. Materials/
Methods: We analyzed a single-institution, retrospective cohort of adult patients who received RT between 2019 to 2021. NCCN DT patient-reported outcomes, routinely collected at first consultation, were tabulated by chart review. Common Terminology Criteria for Adverse Events (CTCAE)-defined patient symptoms 30 days prior to, during, and 30 days following RT were identified from clinical notes in the electronic health record (EHR) utilizing a previously validated clinical Text Analysis and Knowledge Extraction System (cTAKES)-based NLP pipeline. Multiple logistic regression was applied to identify associations between NCCN DT scores and the 20 most commonly documented symptoms in the EHR, controlling for clinical covariates. Results: We identified 1,109 patients who completed the NCCN DT prior to RT. Median time from NCCN DT administration to RT was 14 days (IQR 8-21). More than half (52.4%) of patients identified as male. Median age was 65 (IQR 55-73). Most patients identified as White (59.2%), followed by Asian (17.7%), Other (9.2%), and Black (5.9%). The most common treatment diagnoses were metastasis (22.0%), breast (14.0%), prostate (13.0%), head and neck (12.2%), lung (6.8%), and hematologic (4.4%) cancer. The median NCCN DT score was 2.0 (IQR 0.0-5.0). The most common NLP-identified documented symptoms were pain (80.8% before RT, 77.5% during/after RT), fatigue (65.7%, 71.9%), nausea (66.4%, 63.3%), vomiting (42.7%, 40.2%), and constipation (36.0%, 45.1%). Multiple logistic regression controlling for symptom-specific covariates demonstrated independent associations between NCCN DT and documentation of back pain (OR 1.16 per NCCN DT point, [95% CI 1.10-1.23]), constipation (1.14, [1.08-1.20]), generalized muscle weakness (1.13, [1.07-1.19]), fever (1.12, [1.06-1.19]), and pain (1.12, [1.05-1.20]) before RT, and insomnia (1.14, [1.07-1.22]), back pain (1.13, [1.06-1.20]), constipation (1.12, [1.06-1.18]), cough (1.10, [1.03-1.18]), and nausea (1.09, [1.03-1.15]) during/after RT. Conclusion: Patient distress as measured by the NCCN DT is associated with greater burden of documented symptoms, both before and during/after treatment. Study findings may be impacted by biases in symptom documentation or NCCN DT completion. Further work is needed to establish relationships with formal physician- and patient-reported outcomes and the potential impact of distress management.
