Representing and utilizing clinical textual data for real world studies: An OHDSI approach.

Clinical documentation in electronic health records contains crucial narratives and details about patients and their care. Natural language processing (NLP) can unlock the information conveyed in clinical notes and reports, and thus plays a critical role in real-world studies. The NLP Working Group at the Observational Health Data Sciences and Informatics (OHDSI) consortium was established to develop methods and tools to promote the use of textual data and NLP in real-world observational studies. In this paper, we describe a framework for representing and utilizing textual data in real-world evidence generation, including representations of information from clinical text in the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM), the workflow and tools that were developed to extract, transform and load (ETL) data from clinical notes into tables in OMOP CDM, as well as current applications and specific use cases of the proposed OHDSI NLP solution at large consortia and individual institutions with English textual data. Challenges faced and lessons learned during the process are also discussed to provide valuable insights for researchers who are planning to implement NLP solutions in real-world studies.

Characterizing the Anticancer Treatment Trajectory and Pattern in Patients Receiving Chemotherapy for Cancer Using Harmonized Observational Databases: Retrospective Study.

BACKGROUND: Accurate and rapid clinical decisions based on real-world evidence are essential for patients with cancer. However, the complexity of chemotherapy regimens for cancer impedes retrospective research that uses observational health databases. OBJECTIVE: The aim of this study is to compare the anticancer treatment trajectories and patterns of clinical events according to regimen type using the chemotherapy episodes determined by an algorithm. METHODS: We developed an algorithm to extract the regimen-level abstracted chemotherapy episodes from medication records in a conventional Observational Medical Outcomes Partnership (OMOP) common data model (CDM) database. The algorithm was validated on the Ajou University School Of Medicine (AUSOM) database by manual review of clinical notes. Using the algorithm, we extracted episodes of chemotherapy from patients in the EHR database and the claims database. We also developed an application software for visualizing the chemotherapy treatment patterns based on the treatment episodes in the OMOP-CDM database. Using this software, we generated the trends in the types of regimen used in the institutions, the patterns of the iterative chemotherapy use, and the trajectories of cancer treatment in two EHR-based OMOP-CDM databases. As a pilot study, the time of onset of chemotherapy-induced neutropenia according to regimen was measured using the AUSOM database. The anticancer treatment trajectories for patients with COVID-19 were also visualized based on the nationwide claims database. RESULTS: We generated 178,360 treatment episodes for patients with colorectal, breast, and lung cancer for 85 different regimens. The algorithm precisely identified the type of chemotherapy regimen in 400 patients (average positive predictive value >98%). The trends in the use of routine clinical chemotherapy regimens from 2008-2018 were identified for 8236 patients. For a total of 12 regimens (those administered to the largest proportion of patients), the number of repeated treatments was concordant with the protocols for standard chemotherapy regimens for certain cases. In addition, the anticancer treatment trajectories for 8315 patients were shown, including 62 patients with COVID-19. A comparative analysis of neutropenia showed that its onset in colorectal cancer regimens tended to cluster between days 9-15, whereas it tended to cluster between days 2-8 for certain regimens for breast cancer or lung cancer. CONCLUSIONS: We propose a method for generating chemotherapy episodes for introduction into the oncology extension module of the OMOP-CDM databases. These proof-of-concept studies demonstrated the usability, scalability, and interoperability of the proposed framework through a distributed research network.

Chemotherapy and COVID-19 Outcomes in Patients With Cancer.

PURPOSE: Coronavirus-2019 (COVID-19) mortality is higher in patients with cancer than in the general population, yet the cancer-associated risk factors for COVID-19 adverse outcomes are not fully characterized. PATIENTS AND METHODS: We reviewed clinical characteristics and outcomes from patients with cancer and concurrent COVID-19 at Memorial Sloan Kettering Cancer Center until March 31, 2020 (n = 309), and observed clinical end points until April 13, 2020. We hypothesized that cytotoxic chemotherapy administered within 35 days of a COVID-19 diagnosis is associated with an increased hazard ratio (HR) of severe or critical COVID-19. In secondary analyses, we estimated associations between specific clinical and laboratory variables and the incidence of a severe or critical COVID-19 event. RESULTS: Cytotoxic chemotherapy administration was not significantly associated with a severe or critical COVID-19 event (HR, 1.10; 95% CI, 0.73 to 1.60). Hematologic malignancy was associated with increased COVID-19 severity (HR, 1.90; 95% CI, 1.30 to 2.80). Patients with lung cancer also demonstrated higher rates of severe or critical COVID-19 events (HR, 2.0; 95% CI, 1.20 to 3.30). Lymphopenia at COVID-19 diagnosis was associated with higher rates of severe or critical illness (HR, 2.10; 95% CI, 1.50 to 3.10). Patients with baseline neutropenia 14-90 days before COVID-19 diagnosis had worse outcomes (HR, 4.20; 95% CI, 1.70 to 11.00). Findings from these analyses remained consistent in a multivariable model and in multiple sensitivity analyses. The rate of adverse events was lower in a time-matched population of patients with cancer without COVID-19. CONCLUSION: Recent cytotoxic chemotherapy treatment was not associated with adverse COVID-19 outcomes. Patients with active hematologic or lung malignancies, peri-COVID-19 lymphopenia, or baseline neutropenia had worse COVID-19 outcomes. Interactions among antineoplastic therapy, cancer type, and COVID-19 are complex and warrant further investigation.

Standardized Observational Cancer Research Using the OMOP CDM Oncology Module.

Observational research in cancer requires substantially more detail than most other therapeutic areas. Cancer conditions are defined through histology, affected anatomical structures, staging and grading, and biomarkers, and are treated with complex therapies. Here, we show a new cancer module as part of the OMOP CDM, allowing manual and automated abstraction and standardized analytics. We tested the model in EHR and registry data against a number of typical use cases.

HemOnc: A new standard vocabulary for chemotherapy regimen representation in the OMOP common data model.

Systematic application of observational data to the understanding of impacts of cancer treatments requires detailed information models allowing meaningful comparisons between treatment regimens. Unfortunately, details of systemic therapies are scarce in registries and data warehouses, primarily due to the complex nature of the protocols and a lack of standardization. Since 2011, we have been creating a curated and semi-structured website of chemotherapy regimens, HemOnc.org. In coordination with the Observational Health Data Sciences and Informatics (OHDSI) Oncology Subgroup, we have transformed a substantial subset of this content into the OMOP common data model, with bindings to multiple external vocabularies, e.g., RxNorm and the National Cancer Institute Thesaurus. Currently, there are >73,000 concepts and >177,000 relationships in the full vocabulary. Content related to the definition and composition of chemotherapy regimens has been released within the ATHENA tool (athena.ohdsi.org) for widespread utilization by the OHDSI membership. Here, we describe the rationale, data model, and initial contents of the HemOnc vocabulary along with several use cases for which it may be valuable.

Deceased-donor kidney transplantation: improvement in long-term survival.

BACKGROUND: Despite marked improvement in short-term renal allograft survival rates (GSR) in recent years, improvement in long-term GSR remained elusive. METHODS: We analysed the kidney transplant experience at our centre accrued over four decades to evaluate how short-term and long-term GSR had changed and to identify risk factors affecting graft survival. The study included 1476 adult recipients of a deceased-donor kidney transplant who were transplanted between 1963 and 2006 and who had received one of five distinct immunosuppressive protocols. RESULTS: Five-year actual GSR steadily improved over the years as immunosuppressive therapy evolved (22-86%, P < 0.001) in spite of an increasing trend in the transplantation of higher-risk donor-recipient pairings. For those whose grafts functioned for the first year, subsequent 4-year GSR (5-year conditional GSR) also improved significantly (63-92%, P < 0.001). Acute rejection and delayed graft function (DGF) were the most significant risk factors for actual graft survival, while acute rejection was the only significant risk factor for conditional GSR. Use of kidneys from expanded-criteria donors (ECD) was not a risk factor, compared to the use of standard-criteria donor kidneys for either 5-year actual or conditional GSR. There was an impressive decline in the incidence of acute rejection events (77.4-5.8%, P < 0.001). While the DGF rate had decreased, it still remained high (68.7-38.5%, P < 0.001). CONCLUSIONS: We found a significant improvement in both short-term and long-term GSR of deceased-donor kidney transplants over the last four decades. These improvements are most likely related to the decreased incidence of acute rejection episodes. Minimizing acute rejection events and preventing DGF could result in further improvement in the GSR. Our experience in the judicious use of ECD kidneys suggests that this source of kidneys could be expanded further.

Novel method for genomic analysis of PKD1 and PKD2 mutations in autosomal dominant polycystic kidney disease.

Genetic testing of PKD1 and PKD2 is useful for diagnosis and prognosis of autosomal dominant polycystic kidney disease (ADPKD), particularly in asymptomatic individuals or those without a family history. PKD1 testing is complicated by the large transcript size, complexity of the gene region, and the extent of gene variations. A molecular assay was developed using Transgenomic's SURVEYOR Nuclease and WAVE Nucleic Acid High Sensitivity Fragment Analysis System to screen for PKD1 and PKD2 variants, followed by sequencing of variant gene segments, thereby reducing the sequencing reactions by 80%. This method was compared to complete DNA sequencing performed by a reference laboratory for 25 ADPKD patients from 22 families. The pathogenic potential of gene variations of unknown significance was examined by evolutionary comparison, effects of amino acid substitutions on protein structure, and effects of splice-site alterations. A total of 90 variations were identified, including all 82 reported by the reference laboratory (100% sensitivity). A total of 76 variations (84.4%) were in PKD1 and 14 (15.6%) in PKD2. Definite pathogenic mutations (seven nonsense, four truncation, and three splicing defects) were detected in 64% (14/22) of families. The remaining 76 variants included 26 missense, 33 silent, and 17 intronic changes. Two heterozygous nonsense mutations were incorrectly determined by the reference laboratory as homozygous. "Probably pathogenic" mutations were identified in an additional five families (overall detection rate 86%). In conclusion, the SURVEYOR nuclease method was comparable to direct sequencing for detecting ADPKD mutations, achieving high sensitivity with lower cost, providing an important tool for genetic analysis of complex genes.