Search

ObjectiveExtracting and accurately phenotyping electronic health documentation is critical for medical research and clinical care. While there are a variety of techniques to accomplish this task, natural language processing (NLP) has been developed for numerous domains to transform clinical documentation into data available for computational work. Accordingly, we sought to develop a highly accurate and open-source NLP module to ascertain and phenotype left ventricular hypertrophy (LVH) and hypertrophic cardiomyopathy (HCM) diagnoses on echocardiogram reports from a diverse hospital network.Methods700 echocardiogram reports from six hospitals were randomly selected from data repositories within the Mass General Brigham healthcare system and manually adjudicated by physicians for 10 subtypes of LVH and diagnoses of HCM. Using an open-source NLP system, the module was developed on 300 training set reports and validated on 400 reports. The sensitivity, specificity, positive predictive value, and negative predictive value were calculated to assess the discriminative accuracy of the NLP module.ResultsThe NLP demonstrated robust performance across the 10 LVH subtypes with overall sensitivity and specificity exceeding 96%. Additionally, the NLP module demonstrated excellent performance detecting HCM diagnoses, with sensitivity and specificity exceeding 93%.ConclusionWe designed a highly accurate NLP module to determine the presence of LVH and HCM on echocardiogram reports. Our work demonstrates the feasibility of NLP to detect diagnoses on imaging reports, even when described in free-text. These modules have been placed in the public domain to advance research, trial recruitment, and population health management for individuals with LVH-associated conditions.

Objective: Accurate ascertainment of comorbidities is paramount in clinical research. While manual adjudication is labor‐intensive and expensive, the adoption of electronic health records enables computational analysis of free‐text documentation using natural language processing (NLP) tools. Hypothesis: We sought to develop highly accurate NLP modules to assess for the presence of five key cardiovascular comorbidities in a large electronic health record system. Methods: One‐thousand clinical notes were randomly selected from a cardiovascular registry at Mass General Brigham. Trained physicians manually adjudicated these notes for the following five diagnostic comorbidities: hypertension, dyslipidemia, diabetes, coronary artery disease, and stroke/transient ischemic attack. Using the open‐source Canary NLP system, five separate NLP modules were designed based on 800 “training‐set” notes and validated on 200 “test‐set” notes. Results: Across the five NLP modules, the sentence‐level and note‐level sensitivity, specificity, and positive predictive value was always greater than 85% and was most often greater than 90%. Accuracy tended to be highest for conditions with greater diagnostic clarity (e.g. diabetes and hypertension) and slightly lower for conditions whose greater diagnostic challenges (e.g. myocardial infarction and embolic stroke) may lead to less definitive documentation. Conclusion: We designed five open‐source and highly accurate NLP modules that can be used to assess for the presence of important cardiovascular comorbidities in free‐text health records. These modules have been placed in the public domain and can be used for clinical research, trial recruitment and population management at any institution as well as serve as the basis for further development of cardiovascular NLP tools.

Lipoprotein(a) [Lp(a)] is independently associated with atherosclerotic cardiovascular disease and calcific aortic valve stenosis. Elevated Lp(a) affects approximately one in five individuals and meaningfully contributes to the residual cardiovascular risk in individuals with otherwise well‐controlled risk factors. With targeted therapies in the therapeutic pipeline, there is a need to further characterize the clinical phenotypes and outcomes of individuals with elevated levels of this unique biomarker. The Mass General Brigham Lp(a) Registry will be built from the longitudinal electronic health record of two large academic medical centers in Boston, Massachusetts, to develop a detailed cohort of patients who have had their Lp(a) measured. In combination with structured data sources, clinical documentation will be analyzed using natural language processing techniques to accurately characterize baseline characteristics. Important outcome measures including all‐cause mortality, cardiovascular mortality, and cardiovascular events will be available for analysis. Approximately 30 000 patients who have had their Lp(a) tested within the Mass General Brigham system from January 2000 to July 2019 will be included in the registry. This large Lp(a) cohort will provide meaningful observational data regarding the differential risk associated with Lp(a) values and cardiovascular disease. With a new frontier of targeted Lp(a) therapies on the horizon, the Mass General Brigham Lp(a) Registry will help provide a deeper understanding of Lp(a)'s role in long term cardiovascular outcomes.

Importance Socioeconomic disadvantage is associated with poor health outcomes. However, whether socioeconomic factors are associated with post–myocardial infarction (MI) outcomes in younger patient populations is unknown. Objective To evaluate the association of neighborhood-level socioeconomic disadvantage with long-term outcomes among patients who experienced an MI at a young age. Design, Setting, and Participants This cohort study analyzed patients in the Mass General Brigham YOUNG-MI Registry (at Brigham and Women’s Hospital and Massachusetts General Hospital in Boston, Massachusetts) who experienced an MI at or before 50 years of age between January 1, 2000, and April 30, 2016. Each patient’s home address was mapped to the Area Deprivation Index (ADI) to capture higher rates of socioeconomic disadvantage. The median follow-up duration was 11.3 years. The dates of analysis were May 1, 2020, to June 30, 2020. Exposures Patients were assigned an ADI ranking according to their home address and then stratified into 3 groups (least disadvantaged group, middle group, and most disadvantaged group). Main Outcomes and Measures The outcomes of interest were all-cause and cardiovascular mortality. Cause of death was adjudicated from national registries and electronic medical records. Cox proportional hazards regression modeling was used to evaluate the association of ADI with all-cause and cardiovascular mortality. Results The cohort consisted of 2097 patients, of whom 2002 (95.5%) with an ADI ranking were included (median [interquartile range] age, 45 [42-48] years; 1607 male individuals [80.3%]). Patients in the most disadvantaged neighborhoods were more likely to be Black or Hispanic, have public insurance or no insurance, and have higher rates of traditional cardiovascular risk factors such as hypertension and diabetes. Among the 1964 patients who survived to hospital discharge, 74 (13.6%) in the most disadvantaged group compared with 88 (12.6%) in the middle group and 41 (5.7%) in the least disadvantaged group died. Even after adjusting for a comprehensive set of clinical covariates, higher neighborhood disadvantage was associated with a 32% higher all-cause mortality (hazard ratio, 1.32; 95% CI, 1.10-1.60;P = .004) and a 57% higher cardiovascular mortality (hazard ratio, 1.57; 95% CI, 1.17-2.10;P = .003). Conclusions and Relevance This study found that, among patients who experienced an MI at or before age 50 years, socioeconomic disadvantage was associated with higher all-cause and cardiovascular mortality even after adjusting for clinical comorbidities. These findings suggest that neighborhood and socioeconomic factors have an important role in long-term post-MI survival.