JCO Clinical Cancer Informatics最新文献

Purpose: The expanding presence of the electronic health record (EHR) underscores the necessity for improved interoperability. To test the interoperability within the field of oncology research, our team at Vanderbilt University Medical Center (VUMC) enabled our Epic-based EHR to be compatible with the Minimal Common Oncology Data Elements (mCODE), which is a Fast Healthcare Interoperability Resources (FHIR)-based consensus data standard created to facilitate the transmission of EHRs for patients with cancer.

Methods: Our approach used an extract, transform, load tool for converting EHR data from the VUMC Epic Clarity database into mCODE-compatible profiles. We established a sandbox environment on Microsoft Azure for data migration, deployed a FHIR server to handle application programming interface (API) requests, and mapped VUMC data to align with mCODE structures. In addition, we constructed a web application to demonstrate the practical use of mCODE profiles in health care.

Results: We developed an end-to-end pipeline that converted EHR data into mCODE-compliant profiles, as well as a web application that visualizes genomic data and provides cancer risk assessments. Despite the complexities of aligning traditional EHR databases with mCODE standards and the limitations of FHIR APIs in supporting advanced statistical methodologies, this project successfully demonstrates the practical integration of mCODE standards into existing health care infrastructures.

Conclusion: This study provides a proof of concept for the interoperability of mCODE within a major health care institution's EHR system, highlighting both the potential and the current limitations of FHIR APIs in supporting complex data analysis for oncology research.

Randomized trials provide high-quality, internally consistent data on selected clinical questions, but lack generalizability for the aging population who are most often diagnosed with cancer and have comorbid conditions that may affect the interpretation of treatment benefit. The need for high-quality, relevant, and timely data is greater than ever. Promising solutions lie in the collection and analysis of real-world data (RWD), which can potentially provide timely insights about the patient's course during and after initial treatment and the outcomes of important subgroups such as the elderly, rural populations, children, and patients with greater social health needs. However, to inform practice and policy, real-world evidence must be created from trustworthy and comprehensive sources of RWD; these may include pragmatic clinical trials, registries, prospective observational studies, electronic health records (EHRs), administrative claims, and digital technologies. There are unique challenges in oncology since key parameters (eg, cancer stage, biomarker status, genomic assays, imaging response, side effects, quality of life) are not recorded, siloed in inaccessible documents, or available only as free text or unstructured reports in the EHR. Advances in analytics, such as artificial intelligence, may greatly enhance the ability to obtain more granular information from EHRs and support integrated diagnostics; however, they will need to be validated purpose by purpose. We recommend a commitment to standardizing data across sources and building infrastructures that can produce fit-for-purpose RWD that will provide timely understanding of the effectiveness of individual interventions.

Purpose: Identification of those at risk of hereditary cancer syndromes using electronic health record (EHR) data sources is important for clinical care, quality improvement, and research. We describe diagnostic processes, previously seldom reported, for a common hereditary cancer syndrome, Lynch syndrome (LS), using EHR data within a community-based, multicenter, demographically diverse health system.

Methods: Within a retrospective cohort enrolled between 2015 and 2020 at Kaiser Permanente Northern California, we assessed electronic diagnostic domains for LS including (1) family history of LS-associated cancer; (2) personal history of LS-associated cancer; (3) LS screening via mismatch repair deficiency (MMRD) testing of newly diagnosed malignancy; (4) germline genetic test results; and (5) clinician-entered diagnostic codes for LS. We calculated proportions and overlap for each diagnostic domain descriptively.

Results: Among 5.8 million individuals, (1) 28,492 (0.49%) had a family history of LS-associated cancer of whom 3,635 (13%) underwent genetic testing; (2) 100,046 (1.7%) had a personal history of a LS-associated cancer; and (3) 8,711 (0.1%) were diagnosed with colorectal cancer of whom 7,533 (86%) underwent MMRD screening and of the positive screens (486), 130 (27%) underwent germline testing. One thousand seven hundred and fifty-seven (0.03%) were diagnosed with endometrial cancer of whom 1,613 (92%) underwent MMRD screening and of the 195 who screened positive, 55 (28%) underwent genetic testing. (4) 30,790 (0.05%) had LS germline genetic testing with 707 (0.01%) testing positive; and (5) 1,273 (0.02%) had a clinician-entered diagnosis of LS.

Conclusion: It is feasible to electronically characterize the diagnostic processes of LS. No single data source comprehensively identifies all LS carriers. There is underutilization of LS genetic testing for those eligible and underdiagnosis of LS. Our work informs similar efforts in other settings for hereditary cancer syndromes.

Purpose: Real-world data (RWD) collected on patients treated as part of routine clinical care form the basis of cancer clinical registries. Capturing accurate death data can be challenging, with inaccurate survival data potentially compromising the integrity of registry-based research. Here, we explore the utility of data linkage (DL) to state-based registries to enhance the capture of survival outcomes.

Methods: We identified consecutive adult patients with brain tumors treated in the state of Victoria from the Brain Tumour Registry Australia: Innovation and Translation (BRAIN) database, who had no recorded date of death and no follow-up within the last 6 months. Full name and date of birth were used to match patients in the BRAIN registry with those in the Victorian Births, Deaths and Marriages (BDM) registry. Overall survival (OS) outcomes were compared pre- and post-DL.

Results: Of the 7,346 clinical registry patients, 5,462 (74%) had no date of death and no follow-up recorded within the last 6 months. Of the 5,462 patients, 1,588 (29%) were matched with a date of death in BDM. Factors associated with an increased number of matches were poor prognosis tumors, older age, and social disadvantage. OS was significantly overestimated pre-DL compared with post-DL for the entire cohort (pre- v post-DL: hazard ratio, 1.43; P < .001; median, 29.9 months v 16.7 months) and for most individual tumor types. This finding was present independent of the tumor prognosis.

Conclusion: As revealed by linkage with BDM, a high proportion of patients in a brain cancer clinical registry had missing death data, contributed to by informative censoring, inflating OS calculations. DL to pertinent registries on an ongoing basis should be considered to ensure accurate reporting of survival data and interpretation of RWD outcomes.

Purpose: Rare cancers constitute over 20% of human neoplasms, often affecting patients with unmet medical needs. The development of effective classification and prognostication systems is crucial to improve the decision-making process and drive innovative treatment strategies. We have created and implemented MOSAIC, an artificial intelligence (AI)-based framework designed for multimodal analysis, classification, and personalized prognostic assessment in rare cancers. Clinical validation was performed on myelodysplastic syndrome (MDS), a rare hematologic cancer with clinical and genomic heterogeneities.

Methods: We analyzed 4,427 patients with MDS divided into training and validation cohorts. Deep learning methods were applied to integrate and impute clinical/genomic features. Clustering was performed by combining Uniform Manifold Approximation and Projection for Dimension Reduction + Hierarchical Density-Based Spatial Clustering of Applications with Noise (UMAP + HDBSCAN) methods, compared with the conventional Hierarchical Dirichlet Process (HDP). Linear and AI-based nonlinear approaches were compared for survival prediction. Explainable AI (Shapley Additive Explanations approach [SHAP]) and federated learning were used to improve the interpretation and the performance of the clinical models, integrating them into distributed infrastructure.

Results: UMAP + HDBSCAN clustering obtained a more granular patient stratification, achieving a higher average silhouette coefficient (0.16) with respect to HDP (0.01) and higher balanced accuracy in cluster classification by Random Forest (92.7% ± 1.3% and 85.8% ± 0.8%). AI methods for survival prediction outperform conventional statistical techniques and the reference prognostic tool for MDS. Nonlinear Gradient Boosting Survival stands in the internal (Concordance-Index [C-Index], 0.77; SD, 0.01) and external validation (C-Index, 0.74; SD, 0.02). SHAP analysis revealed that similar features drove patients' subgroups and outcomes in both training and validation cohorts. Federated implementation improved the accuracy of developed models.

Conclusion: MOSAIC provides an explainable and robust framework to optimize classification and prognostic assessment of rare cancers. AI-based approaches demonstrated superior accuracy in capturing genomic similarities and providing individual prognostic information compared with conventional statistical methods. Its federated implementation ensures broad clinical application, guaranteeing high performance and data protection.

Advancements in variant curation challenges: minority representation and incomplete data reporting.

Purpose: The RECIST guidelines provide a standardized approach for evaluating the response of cancer to treatment, allowing for consistent comparison of treatment efficacy across different therapies and patients. However, collecting such information from electronic health records manually can be extremely labor-intensive and time-consuming because of the complexity and volume of clinical notes. The aim of this study is to apply natural language processing (NLP) techniques to automate this process, minimizing manual data collection efforts, and improving the consistency and reliability of the results.

Methods: We proposed a complex, hybrid NLP system that automates the process of extracting, linking, and summarizing anticancer therapy and associated RECIST-like responses from narrative clinical text. The system consists of multiple machine learning-/deep learning-based and rule-based modules for diverse NLP tasks such as named entity recognition, assertion classification, relation extraction, and text normalization, to address different challenges associated with anticancer therapy and response information extraction. We then evaluated the system performances on two independent test sets from different institutions to demonstrate its effectiveness and generalizability.

Results: The system used domain-specific language models, BioBERT and BioClinicalBERT, for high-performance therapy mentions identification and RECIST responses extraction and categorization. The best-performing model achieved a 0.66 score in linking therapy and RECIST response mentions, with end-to-end performance peaking at 0.74 after relation normalization, indicating substantial efficacy with room for improvement.

Conclusion: We developed, implemented, and tested an information extraction system from clinical notes for cancer treatment and efficacy assessment information. We expect this system will support future cancer research, particularly oncologic studies that focus on efficiently assessing the effectiveness and reliability of cancer therapeutics.

Purpose: Data on lines of therapy (LOTs) for cancer treatment are important for clinical oncology research, but LOTs are not explicitly recorded in electronic health records (EHRs). We present an efficient approach for clinical data abstraction and a flexible algorithm to derive LOTs from EHR-based medication data on patients with glioblastoma multiforme (GBM).

Methods: Nonclinicians were trained to abstract the diagnosis of GBM from EHRs, and their accuracy was compared with abstraction performed by clinicians. The resulting data were used to build a cohort of patients with confirmed GBM diagnosis. An algorithm was developed to derive LOTs using structured medication data, accounting for the addition and discontinuation of therapies and drug class. Descriptive statistics were calculated and time-to-next-treatment (TTNT) analysis was performed using the Kaplan-Meier method.

Results: Treating clinicians as the gold standard, nonclinicians abstracted GBM diagnosis with a sensitivity of 0.98, specificity 1.00, positive predictive value 1.00, and negative predictive value 0.90, suggesting that nonclinician abstraction of GBM diagnosis was comparable with clinician abstraction. Of 693 patients with a confirmed diagnosis of GBM, 246 patients contained structured information about the types of medications received. Of them, 165 (67.1%) received a first-line therapy (1L) of temozolomide, and the median TTNT from the start of 1L was 179 days.

Conclusion: We described a workflow for extracting diagnosis of GBM and LOT from EHR data that combines nonclinician abstraction with algorithmic processing, demonstrating comparable accuracy with clinician abstraction and highlighting the potential for scalable and efficient EHR-based oncology research.