25 October 2021 in JCO Clinical Cancer Informatics
PURPOSE: Typically stored as unstructured notes, surgical pathology reports contain data elements valuable to cancer research that require labor-intensive manual extraction. Although studies have described natural language processing (NLP) of surgical pathology reports to automate information extraction, efforts have focused on specific cancer subtypes rather than across multiple oncologic domains. To address this gap, we developed and evaluated an NLP method to extract tumor staging and diagnosis information across multiple cancer subtypes. METHODS: The NLP pipeline was implemented on an open-source framework called Leo. We used a total of 555,681 surgical pathology reports of 329,076 patients to develop the pipeline and evaluated our approach on subsets of reports from patients with breast, prostate, colorectal, and randomly selected cancer subtypes. RESULTS: Averaged across all four cancer subtypes, the NLP pipeline achieved an accuracy of 1.00 for International Classification of Diseases, Tenth Revision codes, 0.89 for T staging, 0.90 for N staging, and 0.97 for M staging. It achieved an F1 score of 1.00 for International Classification of Diseases, Tenth Revision codes, 0.88 for T staging, 0.90 for N staging, and 0.24 for M staging. CONCLUSION: The NLP pipeline was developed to extract tumor staging and diagnosis information across multiple cancer subtypes to support the research enterprise in our institution. Although it was not possible to demonstrate generalizability of our NLP pipeline to other institutions, other institutions may find value in adopting a similar NLP approach—and reusing code available at GitHub—to support the oncology research enterprise with elements extracted from surgical pathology reports.
14 May 2021 in Annals of Pharmacotherapy
Background: Propofol is commonly used to achieve ventilator synchrony in critically ill patients with coronavirus disease 2019 (COVID-19), yet its safety in this patient population is unknown. Objective: To evaluate the safety, in particular the incidence of hypertriglyceridemia, of continuous infusion propofol in patients with COVID-19. Methods: This was a retrospective study at 1 academic medical center and 1 affiliated teaching hospital in New York City. Adult, critically ill patients with COVID-19 who received continuous infusion propofol were included. Patients who received propofol for <12 hours, were transferred from an outside hospital while on mechanical ventilation, or did not have a triglyceride concentration obtained during the infusion were excluded. Results: A total of 252 patients were included. Hypertriglyceridemia (serum triglyceride concentration ≥ 400 mg/dL) occurred in 38.9% of patients after a median cumulative dose of 4307 mg (interquartile range [IQR], 2448-9431 mg). The median time to triglyceride elevation was 3.8 days (IQR, 1.9-9.1 days). In the multivariable regression analysis, obese patients had a significantly greater odds of hypertriglyceridemia (odds ratio = 1.87; 95% CI = 1.10, 3.21). There was no occurrence of acute pancreatitis. The incidence of possible propofol-related infusion syndrome was 3.2%. Conclusion and Relevance: Hypertriglyceridemia occurred frequently in patients with COVID-19 who received propofol but did not lead to acute pancreatitis. Elevated triglyceride concentrations occurred more often and at lower cumulative doses than previously reported in patients without COVID-19. Application of these data may aid in optimal monitoring for serious adverse effects of propofol in patients with COVID-19.
5 September 2021 in Annals of Pharmacotherapy
31 October 2025 in Cancer Epidemiology, Biomarkers & Prevention
Background: The proportion of lung cancers affecting individuals who have never smoked is growing, with these cancers being prone to harbor mutations in the EGFR gene. Little is known about risk factors and prognostic indicators for EGFR-mutant cancers, with current research limited by the scarcity of datasets integrating genomic, clinical, and environmental data. Methods: We created the Meyer Cancer Center Molecularly Enhanced Lung Cancer Database (MCC-MELD), including lung cancer cases from a large catchment area in New York City. We identified cases through linkage to our institution’s cancer registry and a clinician-initiated, manually curated database. We linked all cases to the electronic health record and in-house tumor genomic testing results. We used natural language processing (NLP) to extract unstructured genomic testing results and detailed smoking history. We linked geocoded addresses to detailed area-level measures. Results: MCC-MELD contains 9,573 patients with lung cancer diagnosed from 1988 to 2024, of whom 20% were non-Hispanic Asian, 14% were non-Hispanic Black, and 8% were Hispanic. We identified 1,092 (11.4%) EGFR-mutant cancers, with NLP identifying 397 cases not identified by structured data. NLP showed high accuracy in ascertaining EGFR status (97%) and quantitative smoking history variables (90%–98%). Never smokers made up 16% of the cases in MCC-MELD. Conclusions: MCC-MELD is an NLP-enhanced database containing clinical information, genomic testing results, and linkages to area-level data for patients with lung cancer from a diverse urban setting. Impact: This resource can facilitate studies on lung cancer risk factors, treatment patterns, and outcomes by EGFR and other driver mutation status.