Search

PURPOSE Although the bulk of patient-generated health data are increasing exponentially, their use is impeded because most data come in unstructured format, namely as free-text clinical reports. A variety of natural language processing (NLP) methods have emerged to automate the processing of free text ranging from statistical to deep learning–based models; however, the optimal approach for medical text analysis remains to be determined. The aim of this study was to provide a head-to-head comparison of novel NLP techniques and inform future studies about their utility for automated medical text analysis. PATIENTS AND METHODS Magnetic resonance imaging reports of patients with brain metastases treated in two tertiary centers were retrieved and manually annotated using a binary classification (single metastasis v two or more metastases). Multiple bag-of-words and sequence-based NLP models were developed and compared after randomly splitting the annotated reports into training and test sets in an 80:20 ratio. RESULTS A total of 1,479 radiology reports of patients diagnosed with brain metastases were retrieved. The least absolute shrinkage and selection operator (LASSO) regression model demonstrated the best overall performance on the hold-out test set with an area under the receiver operating characteristic curve of 0.92 (95% CI, 0.89 to 0.94), accuracy of 83% (95% CI, 80% to 87%), calibration intercept of –0.06 (95% CI, –0.14 to 0.01), and calibration slope of 1.06 (95% CI, 0.95 to 1.17). CONCLUSION Among various NLP techniques, the bag-of-words approach combined with a LASSO regression model demonstrated the best overall performance in extracting binary outcomes from free-text clinical reports. This study provides a framework for the development of machine learning-based NLP models as well as a clinical vignette of patients diagnosed with brain metastases.

7029 Background: MGMT promoter methylation holds important prognostic and predictive implications for glioblastoma (GBM) patients. Herein we evaluate whether any barriers faced MGMT testing. Methods: Adults with newly-diagnosed GBMs were identified from the U.S. National Cancer Database (2010-2016). Patient socioeconomic, tumor, and cancer program features were evaluated for association with MGMT testing by multivariable logistic regression. Results: Of 12,380 GBM patients, only 57% had MGMT testing–a rate that increased to 74% by 2016. Testing was independent of patients' sex or race/ethnicity (all p > 0.05); and was largely independent of GBM histology, size, and location. Older patients were less likely to receive testing (54% of ≥70-years-olds vs 62% of those in their 40s, p = 0.01). Insurance status was an independent predictor of testing: only 43% of uninsured patients had testing, vs 61% of privately-insured or 55% of Medicare patients (p≤0.01). The uninsured rate dropped from 3.6% in 2010 to 2.5% in 2016 and was associated with Medicaid expansion under the Affordable Care Act. Patients from non-expansion states had the lowest rates of testing (55%), vs patients from expansion states (58%, p < 0.001). Additionally, household incomes were independently associated with testing: only 50% of patients living in the poorest quartile of households had testing vs 61% in the richest quartile (p = 0.001). Diagnosing hospital type was also an independent predictor of testing: patients diagnosed at academic/NCI-designated cancer programs were most likely to have testing (65%), vs only 45% of community and 44% of comprehensive community cancer program patients were provided testing (all p < 0.001). By 2016, testing rates had improved in all program types–particularly at comprehensive community (67%) and academic (80%) programs, but lagged at community programs (56%, p < 0.001). Patients without MGMT testing received less chemotherapy (mOR = 0.66, p < 0.001) and were associated with worse overall survival (mHR = 1.08, p = 0.002). Conclusions: Newly-diagnosed GBM patients in the U.S. who were uninsured, from the poorest quartile of households, or diagnosed at community cancer programs faced significant barriers to receiving MGMT testing. Medicaid expansion under the ACA was associated with reduced uninsurance, increased Medicaid insurance, and improved rates of testing. Reduced testing rates were associated with less chemotherapy and worse OS. Together our results indicate that substantial socioeconomic and care setting disparities exist in MGMT testing for GBM patients.

2025Background: There is currently a lack of well-established meta-analyses examining the association between depression and patient survival in glioma patients. The aim of this meta-analysis was t...