Unlocking echocardiogram measurements for heart disease research through natural language processing

Natriuretic peptide testing in veterans hospitalized with heart failure: Potential differences by sex

BACKGROUND

Natriuretic peptide testing (NPT) is recommended to assist in diagnosis and prognostication during heart failure hospitalization (HFH). NPT on admission for HFH and sex-based variation in NPT are unknown.

OBJECTIVES

We investigated the utilization of NPT among Veterans with HFH, evaluated for sex-based differences, and examined associations with demographic, clinical, and facility characteristics.

METHODS

Among Veterans with HFH in the Veterans Affairs Healthcare System between October 2015-September 2020, we assessed the rate of NPT on admission and sex-based differences in NPT. We determined associations with demographic, clinical covariates, (comorbidities, laboratory values, loop diuretic use), and facility characteristics using logistic regression.

RESULTS

Of 55,935 patients with HFH (women=1237 (2.2 %)), women were younger (68.3 versus 72.8 years, p < 0.001), less likely to have cardiac comorbidities, and more likely to have ejection fraction >40 %. Admission NPT occurred in 78.3 % of patients (men=78.4 %, women=74.7 %; p = 0.002). In adjusted analyses for clinical and facility-related factors, women were 15 % less likely to receive NPT compared with men [odds ratio =0.85, 95 % CI (0.75, 0.98)]. In sex-stratified models, atrial fibrillation and prior loop diuretic use were associated with increased likelihood of NPT and previous NPT was associated with decreased likelihood in both sexes. Overall associations were similar in both sexes.

CONCLUSIONS

Women were less likely to receive NPT during HFH compared to men, potentially risking greater delays in HF diagnosis and treatment. Further investigation should examine the impact of the absence of admission NPT on clinical outcomes and identify strategies to improve obtaining NPT in all patients.

Rurality, Cardiovascular Risk Factors, and Early Cardiovascular Disease Among Childhood, Adolescent, and Young Adult Cancer Survivors

Purpose: Cardiovascular risk factors (CVRFs) later in life potentiate risk for late cardiovascular disease (CVD) from cardiotoxic treatment among survivors. This study evaluated the association of baseline CVRFs and CVD in the early survivorship period. Methods: This analysis included patients ages 0-29 at initial diagnosis and reported in the institutional cancer registry between 2010 and 2017 (n = 1228). Patients who died within 5 years (n = 168), those not seen in the oncology clinic (n = 312), and those with CVD within one year of diagnosis (n = 17) were excluded. CVRFs (hypertension, diabetes, dyslipidemia, and obesity) within 1 year of initial diagnosis were constructed and extracted from the electronic health record based on discrete observations, ICD9/10 codes, and RxNorm codes for antihypertensives. Results: Among survivors (n = 731), 10 incident cases (1.4%) of CVD were observed between 1 and 5 years after the initial diagnosis. Public health insurance (p = 0.04) and late effects risk strata (p = 0.01) were positively associated with CVD. Among survivors with public insurance (n = 495), two additional cases of CVD were identified from claims data with an incidence of 2.4%. Survivors from rural areas had a 4.1 times greater risk of CVD compared with survivors from urban areas (95% CI: 1.1-15.3), despite adjustment for late effects risk strata. Conclusion: Clinically computable phenotypes for CVRFs among survivors through informatics methods were feasible. Although CVRFs were not associated with CVD in the early survivorship period, survivors from rural areas were more likely to develop CVD.

Diagnosis extraction from unstructured Dutch echocardiogram reports using span- and document-level characteristic classification

BACKGROUND

Clinical machine learning research and artificial intelligence driven clinical decision support models rely on clinically accurate labels. Manually extracting these labels with the help of clinical specialists is often time-consuming and expensive. This study tests the feasibility of automatic span- and document-level diagnosis extraction from unstructured Dutch echocardiogram reports.

METHODS

We included 115,692 unstructured echocardiogram reports from the University Medical Center Utrecht, a large university hospital in the Netherlands. A randomly selected subset was manually annotated for the occurrence and severity of eleven commonly described cardiac characteristics. We developed and tested several automatic labelling techniques at both span and document levels, using weighted and macro F1-score, precision, and recall for performance evaluation. We compared the performance of span labelling against document labelling methods, which included both direct document classifiers and indirect document classifiers that rely on span classification results.

RESULTS

The SpanCategorizer and MedRoBERTa.nl models outperformed all other span and document classifiers, respectively. The weighted F1-score varied between characteristics, ranging from 0.60 to 0.93 in SpanCategorizer and 0.96 to 0.98 in MedRoBERTa.nl. Direct document classification was superior to indirect document classification using span classifiers. SetFit achieved competitive document classification performance using only 10% of the training data. Utilizing a reduced label set yielded near-perfect document classification results.

CONCLUSION

We recommend using our published SpanCategorizer and MedRoBERTa.nl models for span- and document-level diagnosis extraction from Dutch echocardiography reports. For settings with limited training data, SetFit may be a promising alternative for document classification. Future research should be aimed at training a RoBERTa based span classifier and applying English based models on translated echocardiogram reports.

Ontology-guided machine learning outperforms zero-shot foundation models for cardiac ultrasound text reports

Big data can revolutionize research and quality improvement for cardiac ultrasound. Text reports are a critical part of such analyses. Cardiac ultrasound reports include structured and free text and vary across institutions, hampering attempts to mine text for useful insights. Natural language processing (NLP) can help and includes both statistical- and large language model based techniques. We tested whether we could use NLP to map cardiac ultrasound text to a three-level hierarchical ontology. We used statistical machine learning (EchoMap) and zero-shot inference using GPT. We tested eight datasets from 24 different institutions and compared both methods against clinician-scored ground truth. Despite all adhering to clinical guidelines, institutions differed in their structured reporting. EchoMap performed best with validation accuracy of 98% for the first ontology level, 93% for first and second levels, and 79% for all three. EchoMap retained performance across external test datasets and could extrapolate to examples not included in training. EchoMap's accuracy was comparable to zero-shot GPT at the first level of the ontology and outperformed GPT at second and third levels. We show that statistical machine learning can map text to structured ontology and may be especially useful for small, specialized text datasets.

Predicted Mortality and Cardiology Follow-up Following Heart Failure Hospitalizations Among Veterans Health Administration Patients

BACKGROUND

Guidelines recommend timely follow-up with a cardiology specialist for patients hospitalized with heart failure (HF), but it is unknown whether the timeliness of specialty cardiovascular care after discharge correlates with clinical risk. We south to assess the association between estimated mortality risk and post-HF hospitalization cardiology follow-up.

METHODS AND RESULTS

In a cohort of veterans hospitalized with HF in acute care Veterans Health Administration (VA) hospitals between January 1, 2018, and September 15, 2022, we estimated the association of mortality risk at discharge with postdischarge cardiology encounters via logistic regression. We also evaluated the association between cardiology visits and sociodemographic and clinical characteristics, and described variability in postdischarge follow-up rates across VA facilities. We identified a cohort of 84,348 veterans hospitalized with HF with 120,619 hospital admissions. Of a subcohort of 57,554 veterans with 79,866 hospitalizations surviving at least 1 year after discharge, 32.1% of hospitalizations were followed by a cardiology visit within 2 weeks, and 49.3% within 1 month. Marginal probabilities of 2-week and 1-month follow-up were higher for hospitalizations in the highest-risk quintile than those in the lowest-risk quintile (34% vs. 30% and 51% vs. 47%, respectively; P < 0.001 for both intervals). In a time-to-event model in the full cohort, there was a slightly negative association between risk and likelihood of 1-month follow-up (coefficient for MAGGIC score = -0.004, 95% confidence interval [CI] -0.005 to -0.003). Black veterans were less likely to have either 2-week or 1-month follow-up (adjusted odds ratios, 0.93 [95% CI 0.90-0.97] for 2 weeks and 0.93 [95% CI 0.89-0.96] for 1 month). Female veterans were also less likely to have follow-up within 1 month of hospital discharge (adjusted odds ratio 0.90 [95% CI 0.90-0.98]). Conversely, patients with a primary vs secondary hospital diagnosis of HF and those with reduced vs preserved ejection fraction were more likely to have 2-week follow-up (adjusted odds ratios 1.67 [95% CI 1.62-1.73] and 1.72 [95% CI 1.67-1.78], respectively) and 1-month follow-up (adjusted odds ratios 1.83 [95% CI 1.78-1.88] and 1.85 [95% CI 1.80-1.90], respectively). The 1-month follow-up rates varied from 5% to 69% across VA facilities.

CONCLUSIONS

The rate of visits with a cardiologist within 2 weeks or 1 month after HF hospitalization was low overall, was at most modestly associated with estimated mortality risk at discharge, and varied by sex, race/ethnicity, and across VA facilities. Increasing the visit rate after HF hospitalization should be evaluated as a mechanism to improve outcomes after HF hospitalizations, particularly for higher-risk individuals.

Incidence of Right Ventricular Dysfunction in an Echocardiographic Referral Cohort

Introduction

Incidence rates (IRs) of RV dysfunction (RVD) are unknown. We examined the rates, risk factors, and heart failure (HF) hospitalization hazard associated with incident RVD in patients referred for Transthoracic Echocardiogram (TTE).

Methods

In this retrospective cohort study, we extracted tricuspid regurgitant velocity (TRV) and tricuspid annular systolic plane excursion (TAPSE) from TTEs at Vanderbilt (2010-2023). We followed patients from their earliest TTE with normal RV function (TAPSE≥17mm) and a reported TRV. The primary outcome was new RVD (TAPSE<17mm), and the secondary outcome was HF hospitalization after second TTE. Poisson regression and multivariable cox models estimated IRs and hazard ratios, adjusted for demographics, comorbidities, and TTE measures.

Results

Among 45,753 patients (63 years [IQR 50-72], 45% Male, 13% Black) meeting inclusion criteria, 13,735 (30.1%) underwent a follow up TTE and 4,198 (9.2%) developed RVD. The IR of RVD in the full cohort was 3.2/100 person/years (95%CI 3.1-3.3) and 8.2 (95%CI 8.0-8.5) in the repeat TTE cohort. IRs increased with rising RVSP. Risk factors for incident RVD were most prominently HF (HR 1.88; 95%CI 1.75-2.03), left-sided valvular disease (HR 1.68; 95%CI 1.53-1.85), and other cardiovascular comorbidities. Baseline RVSP >35 mmHg associated with TAPSE decline over time. Incident RVD increased hazard of HF hospitalization (HR 2.02; 95%CI 1.85-2.21). Hazard of HF hospitalization increased when TAPSE declined by ≥5mm.

Conclusions

RVD incidence is substantial among patients referred for TTE. Clinical monitoring is warranted if RVSP >35mmHg. Cardiovascular comorbidities drive RVD in this population. Incident RVD associates with increased hazard of HF hospitalization.

Uptake of Sodium-Glucose Cotransporter-2 Inhibitors in Hospitalized Patients With Heart Failure: Insights From the Veterans Affairs Healthcare System

BACKGROUND

The use of sodium-glucose cotransporter-2 inhibitors (SGLT2is) in Veterans Affairs (VA) patients hospitalized with heart failure (HF) has not been reported previously.

METHODS

VA electronic health record data were used to identify patients hospitalized for HF (primary or secondary diagnosis) from 01/2019-11/2022. Patients with SGLT2i allergy, advanced/end-stage chronic kidney disease (CKD) or advanced HF therapies were excluded. We identified factors associated with discharge SGLT2i prescriptions for patients hospitalized due to HF in 2022. We also compared SGLT2i and angiotensin receptor-neprilysin inhibitor (ARNI) prescription rates. Hospital-level variations in SGLT2i prescriptions were assessed via the median odds ratio.

RESULTS

A total of 69,680 patients were hospitalized due to HF; 10.3% were prescribed SGLT2i at discharge (4.4% newly prescribed, 5.9% continued preadmission therapy). SGLT2i prescription increased over time and was higher in patients with HFrEF and primary HF. Among 15,762 patients hospitalized in 2022, SGLT2i prescription was more likely in patients with diabetes (adjusted odds ratio [aOR] 2.27; 95% confidence interval [CI]: 2.09-2.47) and ischemic heart disease (aOR 1.14; 95% CI: 1.03-1.26). Patients with increased age (aOR 0.77 per 10 years; 95% CI: 0.73-0.80) and lower systolic blood pressure (aOR 0.94 per 10 mmHg; 95% CI: 0.92-0.96) were less likely to be prescribed SGLT2i, and SGLT2i prescription was not more likely in patients with CKD (aOR 1.07; 95% CI 0.98-1.16). The adjusted median odds ratio suggested a 1.8-fold variation in the likelihood that similar patients at 2 random VA sites were prescribed SGLT2i (range 0-21.0%). In patients with EF ≤ 40%, 30.9% were prescribed SGLT2i while 26.9% were prescribed ARNI (P < 0.01).

CONCLUSION

One-tenth of VA patients hospitalized for HF were prescribed SGLT2i at discharge. Opportunities exist to reduce variation in SGLT2i prescription rates across hospitals and to promote its use in patients with CKD and older age.

National Trends in Hospital Performance in Guideline-Recommended Pharmacologic Treatment for Heart Failure at Discharge

BACKGROUND

The use of recommended heart failure (HF) medications has improved over time, but opportunities for improvement persist among women and at rural hospitals.

OBJECTIVES

This study aims to characterize national trends in performance in the use of guideline-recommended pharmacologic treatment for HF at U.S. Department of Veterans Affairs (VA) hospitals, at which medication copayments are modest.

METHODS

Among patients discharged from VA hospitals with HF between January 1, 2013, and December 31, 2019, receipt of all guideline-recommended HF pharmacotherapy among eligible patients was assessed, consisting of evidence-based beta-blockers; angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, or angiotensin receptor neprilysin inhibitors; mineralocorticoid receptor antagonists; and oral anticoagulation.

RESULTS

Of 55,560 patients at 122 hospitals, 32,304 (58.1%) received all guideline-recommended HF medications for which they were eligible. The proportion of patients receiving all recommended medications was higher in 2019 relative to 2013 (OR: 1.54; 95% CI: 1.44-1.65). The median of hospital performance was 59.1% (Q1-Q3: 53.2%-66.2%), improving with substantial variation across sites from 2013 (median 56.4%; Q1-Q3: 50.0%-62.0%) to 2019 (median 65.7%; Q1-Q3: 56.3%-73.5%). Women were less likely to receive recommended therapies than men (adjusted OR [aOR]: 0.84; 95% CI: 0.74-0.96). Compared with non-Hispanic White patients, non-Hispanic Black patients were less likely to receive recommended therapies (aOR: 0.83; 95% CI: 0.79-0.87). Urban hospital location was associated with lower likelihood of medication receipt (aOR: 0.73; 95% CI: 0.59-0.92).

CONCLUSIONS

Forty-two percent of patients did not receive all recommended HF medications at discharge, particularly women, minority patients, and those receiving care at urban hospitals. Rates of use increased over time, with variation in performance across hospitals.

Healthcare utilization and left ventricular ejection fraction distribution in methamphetamine use associated heart failure hospitalizations

BACKGROUND

Although methamphetamine use associated heart failure (MU-HF) is increasing, data on its clinical course are limited due to a preponderance of single center studies and significant heterogeneity in the definition of MU-HF in the published literature. Our objective was to evaluate left ventricular ejection fraction (LVEF) distribution, methamphetamine use treatment engagement and postdischarge healthcare utilization among Veterans with heart failure hospitalization in the department of Veterans Affairs (VA) medical centers for MU-HF versus HF not associated with methamphetamine use (other-HF).

METHODS

Observational study including a cohort of Veterans with a first heart failure hospitalization during 2007 - 2020 using data in the VA Corporate Data Warehouse. MU-HF was identified based on the presence of an ICD-code for methamphetmaine use or positive toxicology results within 1-year of heart failure hospitalization. LVEF values entered in the medical record were identified using a validated natural language processing algorithm. Healthcare utilization data was obtained using clinic stop-codes and hosptilaization records.

RESULTS

Of 203,005 first-time heart failure hospitlaizations, 4080 were categorized as MU-HF. Median (interquartile range) of LVEF was 30 (20-45) % for MU-HF versus 40 (25-55)% for other-HF (P < .0001). Eighteen percent of MU-HF had LVEF ≥ 50% compared to 28% in other-HF. Discharge against medical advice was higher in MU-HF (8% vs 2%). Among Veterans with MU-HF, post hospital discharge methamphetamine use treatment engagement was low (18% at 30 days post discharge), with higher follow-up in primary care (76% at 30 days). Post discharge emergency department visits (33% versus 22% at 30 days) and rehospitalizations (24% versus 18% at 30 days) were higher in MU-HF compared to other-HF.

CONCLUSIONS

While the majority of MU-HF hospitalizations are HFrEF, a sizeable minority have HFpEF. This finding has implications for accurate MU-HF classification, treatment, and prognosis. Patients with MU-HF have low addiction treatment receipt and high postdischarge unplanned healthcare utilization. Increasing substance use disorder treatment in this population must be a priority to improve health outcomes. Care-coordination and linkage interventions are urgently needed to increase post-hospitalization addiction treatment and follow-up in an effort to increase evidence-base care and mitigate unplanned healthcare utilization.

Rurality, Cardiovascular Risk Factors, and Early Cardiovascular Disease among Childhood, Adolescent, and Young Adult Cancer Survivors

Background and Aims

Cardiovascular risk factors (CVRFs) later in life potentiate risk for late cardiovascular disease (CVD) from cardiotoxic treatment among survivors. This study evaluated the association of baseline CVRFs and CVD in the early survivorship period.

Methods

This analysis included patients ages 0-29 at initial diagnosis and reported in the institutional cancer registry between 2010 and 2017 (n = 1228). Patients who died within five years (n = 168), those not seen in the oncology clinic (n = 312), and those with CVD within one year of diagnosis (n = 17) were excluded. CVRFs (hypertension, diabetes, dyslipidemia, and obesity) within one year of initial diagnosis, were constructed and extracted from the electronic health record based on discrete observations, ICD9/10 codes, and RxNorm codes for antihypertensives.

Results

Among survivors (n = 731), 10 incident cases (1.4%) of CVD were observed between one year and five years after the initial diagnosis. Public health insurance (p = 0.04) and late effects risk strata (p = 0.01) were positively associated with CVD. Among survivors with public insurance(n = 495), two additional cases of CVD were identified from claims data with an incidence of 2.4%. Survivors from rural areas had a 4.1 times greater risk of CVD compared with survivors from urban areas (95% CI: 1.1-15.3), despite adjustment for late effects risk strata.

Conclusions

Clinically computable phenotypes for CVRFs among survivors through informatics methods were feasible. Although CVRFs were not associated with CVD in the early survivorship period, survivors from rural areas were more likely to develop CVD.

Implications for Survivors

Survivors from non-urban areas and those with public insurance may be particularly vulnerable to CVD.

Development and Optimization of the Veterans Affairs' National Heart Failure Dashboard for Population Health Management

BACKGROUND

In 2020, the Veterans Affairs (VA) health care system deployed a heart failure (HF) dashboard for use nationally. The initial version was notably imprecise and unreliable for the identification of HF subtypes. We describe the development and subsequent optimization of the VA national HF dashboard.

MATERIALS AND METHODS

This study describes the stepwise process for improving the accuracy of the VA national HF dashboard, including defining the initial dashboard, improving case definitions, using natural language processing for patient identification, and incorporating an imaging-quality hierarchy model. Optimization further included evaluating whether to require concurrent ICD-codes for inclusion in the dashboard and assessing various imaging modalities for patient characterization.

RESULTS

Through multiple rounds of optimization, the dashboard accuracy (defined as the proportion of true results to the total population) was improved from 54.1% to 89.2% for the identification of HF with reduced ejection fraction (HFrEF) and from 53.9% to 88.0% for the identification of HF with preserved ejection fraction (HFpEF). To align with current guidelines, HF with mildly reduced ejection fraction (HFmrEF) was added to the dashboard output with 88.0% accuracy.

CONCLUSIONS

The inclusion of an imaging-quality hierarchy model and natural-language processing algorithm improved the accuracy of the VA national HF dashboard. The revised dashboard informatics algorithm has higher use rates and improved reliability for the health management of the population.

Artificial intelligence: revolutionizing cardiology with large language models

Natural language processing techniques are having an increasing impact on clinical care from patient, clinician, administrator, and research perspective. Among others are automated generation of clinical notes and discharge letters, medical term coding for billing, medical chatbots both for patients and clinicians, data enrichment in the identification of disease symptoms or diagnosis, cohort selection for clinical trial, and auditing purposes. In the review, an overview of the history in natural language processing techniques developed with brief technical background is presented. Subsequently, the review will discuss implementation strategies of natural language processing tools, thereby specifically focusing on large language models, and conclude with future opportunities in the application of such techniques in the field of cardiology.

A general text mining method to extract echocardiography measurement results from echocardiography documents

BACKGROUND

In everyday medical practice, the results of cardiac ultrasound examinations are generally recorded in unstructured text, from which extracting relevant information is an important and challenging task. This paper presents a generally applicable language and corpus-independent text mining method for extracting and structuring numerical measurement results and their descriptions from echocardiography reports.

METHOD

The developed method is based on generally applicable text mining preprocessing activities, it automatically identifies and standardizes the descriptions of the cardiac ultrasound measures, and it stores the extracted and standardized measurement descriptions with their measurement results in a structured form for later usage. The method does not contain any regular expression-based search and does not rely on information about the structure of the document.

RESULTS

The method has been tested on a document set containing more than 20,000 echocardiographic reports by examining the efficiency of extracting 12 echocardiography parameters considered important by experts. The method extracted and structured the echocardiography parameters under the study with good sensitivity (lowest value: 0.775, highest value: 1.0, average: 0.904) and excellent specificity (for all cases 1.0). The F1 score ranged between 0.873 and 1.0, and its average value was 0.948.

CONCLUSION

The presented case study has shown that the proposed method can extract measurement results from echocardiography documents with high confidence without performing a direct search or having detailed information about the data recording habits. Furthermore, it effectively handles spelling errors, abbreviations and the highly varied terminology used in descriptions. As it does not rely on any information related to the structure or the language of the documents or data recording habits, it can be applied for processing any free-text written medical texts.

Rural-urban trends in health care utilization, treatment, and mortality among US veterans with congestive heart failure: A retrospective cohort study

PURPOSE

To compare longitudinal rates of health care utilization, evidence-based treatment, and mortality between rural and urban-dwelling patients with congestive heart failure (CHF).

METHODS

We used electronic medical record data from the Veterans Health Administration (VHA) to identify adult patients with CHF from 2012 through 2017. We stratified our cohort using left ventricular ejection fraction percentage at diagnosis (<40% = reduced ejection fraction [HFrEF]; 40%-50% = midrange ejection fraction [HFmrEF]; >50% = preserved ejection fraction [HFpEF]). Within each ejection fraction cohort, we stratified patients into rural or urban groups. We used Poisson regression to estimate annual rates of health care utilization and CHF treatment. We used Fine and Gray regression to estimate annual hazards of CHF and non-CHF mortality.

FINDINGS

One-third of patients with HFrEF (N = 37,928/109,110), HFmrEF (N = 24,447/68,398), and HFpEF (N = 39,298/109,283) resided in a rural area. Rural compared to urban patients used VHA facilities at similar or lower annual rates for outpatient specialty care across all ejection fraction cohorts. Rural patients used VHA facilities at similar or higher rates for primary care and telemedicine-delivered specialty care. They also had lower and declining rates of VHA inpatient and urgent care use over time. There were no meaningful rural-urban differences in treatment receipt among patients with HFrEF. On multivariable analysis, the rate of CHF and non-CHF mortality was similar between rural and urban patients in each ejection fraction cohort.

CONCLUSIONS

Our findings suggest the VHA may have mitigated access and health outcome disparities typically observed for rural patients with CHF.

Dronedarone Versus Sotalol in Antiarrhythmic Drug-Naive Veterans With Atrial Fibrillation

BACKGROUND

Sotalol and dronedarone are both used for maintenance of sinus rhythm for patients with atrial fibrillation. However, while sotalol requires initial monitoring for QT prolongation and proarrhythmia, dronedarone does not. These treatments can be used in comparable patients, but their safety and effectiveness have not been compared head to head. Therefore, we retrospectively evaluated the effectiveness and safety using data from a large health care system.

METHODS

Using Veterans Health Administration data, we identified 11 296 antiarrhythmic drug-naive patients with atrial fibrillation prescribed dronedarone or sotalol in 2012 or later. We excluded patients with prior conduction disease, pacemakers or implantable cardioverter-defibrillators, ventricular arrhythmia, cancer, renal failure, liver disease, or heart failure. We used natural language processing to identify and compare baseline left ventricular ejection fraction between treatment arms. We used 1:1 propensity score matching, based on patient demographics, comorbidities, and medications, and Cox regression to compare strategies. To evaluate residual confounding, we performed falsification analysis with nonplausible outcomes.

RESULTS

The matched cohort comprised 6212 patients (3106 dronedarone and 3106 sotalol; mean [±SD] age, 71±10 years; 2.5% female; mean [±SD] CHA2DS2-VASC, 2±1.3). The mean (±SD) left ventricular ejection fraction was 55±11 and 58±10 for dronedarone and sotalol users, correspondingly. Dronedarone, compared with sotalol, did not demonstrate a significant association with risk of cardiovascular hospitalization (hazard ratio, 1.03 [95% CI, 0.88-1.21]) or all-cause mortality (hazard ratio, 0.89 [95% CI, 0.68-1.16]). However, dronedarone was associated with significantly lower risk of ventricular proarrhythmic events (hazard ratio, 0.53 [95% CI, 0.38-0.74]) and symptomatic bradycardia (hazard ratio, 0.56 [95% CI, 0.37-0.87]). The primary findings were stable across sensitivity analyses. Falsification analyses were not significant.

CONCLUSIONS

Dronedarone, compared with sotalol, was associated with a lower risk of ventricular proarrhythmic events and conduction disorders while having no difference in risk of incident cardiovascular hospitalization and mortality. These observational data provide the basis for prospective efficacy and safety trials.

Primary Occurrence of Cardiovascular Events After Adding Sodium-Glucose Cotransporter-2 Inhibitors or Glucagon-like Peptide-1 Receptor Agonists Compared With Dipeptidyl Peptidase-4 Inhibitors: A Cohort Study in Veterans With Diabetes

BACKGROUND

The effectiveness of glucagon-like peptide-1 receptor agonists (GLP1RA) and sodium-glucose cotransporter-2 inhibitors (SGLT2i) in preventing major adverse cardiac events (MACE) is uncertain for those without preexisting cardiovascular disease.

OBJECTIVE

To test the hypothesis that MACE incidence was lower with the addition of GLP1RA or SGLT2i compared with dipeptidyl peptidase-4 inhibitors (DPP4i) for primary cardiovascular prevention.

DESIGN

Retrospective cohort study of U.S. veterans from 2001 to 2019.

SETTING

Veterans aged 18 years or older receiving care from the Veterans Health Administration, with data linkage to Medicare, Medicaid, and the National Death Index.

PATIENTS

Veterans adding GLP1RA, SGLT2i, or DPP4i onto metformin, sulfonylurea, or insulin treatment alone or in combination. Episodes were stratified by history of cardiovascular disease.

MEASUREMENTS

Study outcomes were MACE (acute myocardial infarction, stroke, or cardiovascular death) and heart failure (HF) hospitalization. Cox models compared the outcome between medication groups using pairwise comparisons in a weighted cohort adjusted for covariates.

RESULTS

The cohort included 28 759 GLP1RA versus 28 628 DPP4i weighted pairs and 21 200 SGLT2i versus 21 170 DPP4i weighted pairs. Median age was 67 years, and diabetes duration was 8.5 years. Glucagon-like peptide-1 receptor agonists were associated with lower MACE and HF versus DPP4i (adjusted hazard ratio [aHR], 0.82 [95% CI, 0.72 to 0.94]), yielding an adjusted risk difference (aRD) of 3.2 events (CI, 1.1 to 5.0) per 1000 person-years. Sodium-glucose cotransporter-2 inhibitors were not associated with MACE and HF (aHR, 0.91 [CI, 0.78 to 1.08]; aRD, 1.28 [-1.12 to 3.32]) compared with DPP4i.

LIMITATION

Residual confounding; use of DPP4i, GLP1RA, and SGLT2i as first-line therapies were not examined.

CONCLUSION

The addition of GLP1RA was associated with primary reductions of MACE and HF hospitalization compared with DPP4i use; SGLT2i addition was not associated with primary MACE prevention.

PRIMARY FUNDING SOURCE

VA Clinical Science Research and Development and supported in part by the Centers for Diabetes Translation Research.

Predictors of Incident Heart Failure Diagnosis Setting: Insights From the Veterans Affairs Healthcare System

BACKGROUND

Early recognition of heart failure (HF) can reduce morbidity, yet HF is often diagnosed only after symptoms require urgent treatment.

OBJECTIVES

The authors sought to describe predictors of HF diagnosis in the acute care vs outpatient setting within the Veterans Health Administration (VHA).

METHODS

The authors estimated whether incident HF diagnoses occurred in acute care (inpatient hospital or emergency department) vs outpatient settings within the VHA between 2014 and 2019. After excluding new-onset HF potentially caused by acute concurrent conditions, they identified sociodemographic and clinical variables associated with diagnosis setting and assessed variation across 130 VHA facilities using multivariable regression analysis.

RESULTS

The authors identified 303,632 patients with new HF, with 160,454 (52.8%) diagnosed in acute care settings. In the prior year, 44% had HF symptoms and 11% had a natriuretic peptide tested, 88% of which were elevated. Patients with housing insecurity and high neighborhood social vulnerability had higher odds of acute care diagnosis (adjusted odds ratio: 1.22 [95% CI: 1.17-1.27] and 1.17 [95% CI: 1.14-1.21], respectively) adjusting for medical comorbidities. Better outpatient quality of care (blood pressure control and cholesterol and diabetes monitoring within the prior 2 years) predicted a lower odds of acute care diagnosis. Likelihood of acute care HF diagnosis varied from 41% to 68% across facilities after adjusting for patient-level risk factors.

CONCLUSIONS

Many first HF diagnoses occur in the acute care setting, especially among socioeconomically vulnerable populations. Better outpatient care was associated with lower rates of an acute care diagnosis. These findings highlight opportunities for timelier HF diagnosis that may improve patient outcomes.

Genetic architecture of heart failure with preserved versus reduced ejection fraction

Pharmacologic clinical trials for heart failure with preserved ejection fraction have been largely unsuccessful as compared to those for heart failure with reduced ejection fraction. Whether differences in the genetic underpinnings of these major heart failure subtypes may provide insights into the disparate outcomes of clinical trials remains unknown. We utilize a large, uniformly phenotyped, single cohort of heart failure sub-classified into heart failure with reduced and with preserved ejection fractions based on current clinical definitions, to conduct detailed genetic analyses of the two heart failure sub-types. We find different genetic architectures and distinct genetic association profiles between heart failure with reduced and with preserved ejection fraction suggesting differences in underlying pathobiology. The modest genetic discovery for heart failure with preserved ejection fraction (one locus) compared to heart failure with reduced ejection fraction (13 loci) despite comparable sample sizes indicates that clinically defined heart failure with preserved ejection fraction likely represents the amalgamation of several, distinct pathobiological entities. Development of consensus sub-phenotyping of heart failure with preserved ejection fraction is paramount to better dissect the underlying genetic signals and contributors to this highly prevalent condition.

Incidence and Trends in the Use of Palliative Care among Patients with Reduced, Middle-Range, and Preserved Ejection Fraction Heart Failure

Background: Clinical practice guidelines recommend integrating palliative care (PC) into the care of patients with heart failure (HF) to address their many palliative needs. However, the incidence rates of PC use among HF subtypes are unknown. Methods: We conducted a retrospective cohort study of patients with the following HF subtypes in the Department of Veterans Affairs: reduced ejection fraction (HFrEF), mid-range ejection fraction (HFmEF), and preserved ejection fraction (HFpEF). Patients were included at the time of HF diagnosis from 2011 to 2015 and followed until a minimum of five years or death. Incidence rates of receipt of PC (primary outcome) were calculated using generalized estimating equations. We evaluated the time to incident PC by HF subtype with Kaplan-Meier analyses and with adjusted restricted mean survival time. Results: Of the 113,555 patients, 69% were ≥65 years, 98% were male, 73% White, and 18% Black; 58% had HFrEF, 7% HFmEF, and 34% HFpEF. Twenty percent received PC during follow-up, and 66% died. Adjusted PC incidence rates were higher among patients with HFrEF (47 per 1000 person-years, confidence interval [95% CI] 43-52) than for HFmEF and HFpEF (42 per 1000 person-years, CI 38-47 for both). Restricting follow-up to five years, patients with HFrEF received PC six weeks earlier than patients with HFpEF. There was no significant difference in time to PC between patients with HFmEF versus HFpEF. Conclusion: About 1 in 20 patients with HFrEF and 1 in 25 patients with HFmEF and HFpEF receive PC annually. Patients with HFrEF receive PC sooner than patients with HFmEF and HFpEF.

Impact of Pharmacist-Led Heart Failure Clinic on Optimization of Guideline-Directed Medical Therapy (PHARM-HF)

This prospective study included patients with heart failure (HF) with reduced ejection fraction (HFrEF) with LVEF < = 40% to evaluate the impact of pharmacist on guideline directed medical therapy (GDMT). The primary outcome was to compare proportion of triple GDMT achieved for Angiotensin-Converting-Enzyme-Inhibitors (ACEI)/Angiotensin-Receptor-Blockers (ARB)/Angiotensin-Receptor-Neprilysin-Inhibitors (ARNI), beta-blockers, aldosterone antagonists (AA), and quadruple GDMT which in additional to triple therapy, included Sodium glucose co-transporter 2 inhibitor (SGLT2i) at 90-day post-enrollment compared to baseline. Secondary endpoints included achieving target and/or maximally tolerated ACEI/ARB/ARNI and beta-blockers combined and individually as well as SGLT2i and AA GDMT at 90-day post-enrollment compared to baseline. We also compared combined and individual HF-related hospitalization/emergency room (ER) visits 90 days pre-/post-enrollment. Of the total 974 patients screened, 80 patients seen at least once in the heart failure medication titration clinic (HMTC) were included in the analysis. Median (IQR) age was 71 (57-69) years with majority white male. There was a significant improvement in the proportion of patients who achieved quadruple GDMT (p = 0.001) and triple GDMT (p-value = 0.020) at 90-day post-enrollment compared to baseline. The secondary GDMT outcomes were also significantly increased at 90 days post-enrollment compared to baseline. Significant difference in mean as well as proportion of combined HF-related hospitalization/ER-visits was found 90 days pre-/post-enrollment (p = 0.047). Our study found that pharmacist's intervention increased the proportion of patients who achieved GDMT at 90 days.

Graphical analysis of guideline adherence to detect systemwide anomalies in HIV diagnostic testing

Background

Analyses of electronic medical databases often compare clinical practice to guideline recommendations. These analyses have a limited ability to simultaneously evaluate many interconnected medical decisions. We aimed to overcome this limitation with an alternative method and apply it to the diagnostic workup of HIV, where misuse can contribute to HIV transmission, delay care, and incur unnecessary costs.

Methods

We used graph theory to assess patterns of HIV diagnostic testing in a national healthcare system. We modeled the HIV diagnostic testing guidelines as a directed graph. Each node in the graph represented a test, and the edges pointed from one test to the next in chronological order. We then graphed each patient’s HIV testing. This set of patient-level graphs was aggregated into a single graph. Finally, we compared the two graphs, the first representing the recommended approach to HIV diagnostic testing and the second representing the observed patterns of HIV testing, to assess for clinical practice deviations.

Results

The HIV diagnostic testing of 1.643 million patients provided 8.790 million HIV diagnostic test results for analysis. Significant deviations from recommended practice were found including the use of HIV resistance tests (n = 3,007) and HIV nucleic acid tests (n = 16,567) instead of the recommended HIV screen.

Conclusions

We developed a method that modeled a complex medical scenario as a directed graph. When applied to HIV diagnostic testing, we identified deviations in clinical practice from guideline recommendations. The model enabled the identification of intervention targets and prompted systemwide policy changes to enhance HIV detection.

Implantable cardioverter-defibrillator placement among patients with left ventricular ejection fraction ≤35 % at least 40 days after acute myocardial infarction

Background

Implantable cardioverter-defibrillators (ICDs) reduce the risk of sudden cardiac death among patients with persistently reduced (≤35 %) left ventricular ejection fraction (LVEF) at least 40 days following acute myocardial infarction (AMI). Few prior studies have used LVEF measured after the 40-day waiting period to examine primary prevention ICD placement.

Methods

We sought to determine factors associated with ICD placement among patients who met LVEF criteria post-MI within a large integrated health care system in the U.S by conducting a retrospective cohort study of Veteran patients hospitalized for AMI from 2004 to 2017 who had documented LVEF ≤35 % from echocardiograms performed between 40 and 455 (90 days +1 year) days post-MI. We used multivariable logistic regression to examine factors associated with ICD placement.

Results

Of 12,893 patients with LVEF ≤35 % at least 40 days post-MI, 2176 (16.9 %) received an ICD between 91- and 455-days post-MI. Younger age, fewer comorbidities, revascularization with PCI, and greater use of GDMT were associated with increased odds of receiving an ICD. However, half of patients treated with a beta-blocker, ACE inhibitor or angiotensin receptor blocker, and mineralocorticoid receptor antagonist prior to LVEF assessment did not receive an ICD. Eligible Black patients were less likely (odds ratio 0.80, 95 % confidence interval 0.69-0.92) to receive an ICD than White patients.

Conclusion

Many factors affect ICD placement among Veteran patients with a confirmed LVEF ≤35 % at least 40 days post-MI. Greater understanding of factors influencing ICD placement would help clinicians ensure guideline-concordant care.

Clinical predictors of improvement in left ventricular ejection fraction in U.S. veterans with heart failure

Background

Our understanding of the factors associated with improvement of LVEF and a heart failure with improved EF (HFimpEF) phenotype remains incomplete.

Methods

We conducted a retrospective study using a national database of patients followed in the Veterans Affairs (VA) health system with serial assessment of left ventricular ejection fraction (LVEF) by echocardiography. We identified US veterans with a new diagnosis of heart failure with: (i) LVEF of <40 % in the 12 months prior to diagnosis, and (ii) follow-up LVEF assessment at least 6 months after their diagnosis. We defined HFimpEF as a final LVEF of ≥40 %.

Results

Among the 106,414 US veterans with an initial LVEF of <40 % in this analysis, 39,994 (37.6 %) had a final EF of >40 % after a median follow up of 5 years. Multivariate regression analysis identified several factors that were independently associated with LVEF improvement including female sex, younger age, higher BMI, and a history of specific comorbid conditions such as hypertension, valve disease, atrial fibrillation, connective tissue disease, liver disease, and malignancy (p < 0.001). Conversely, a history of ischemic heart disease and peripheral arterial disease, as well as specific racial backgrounds (Black and Hispanic) were associated with lower rates of LVEF improvement. The model c-statistic for predicting LVEF improvement was 0.70.

Conclusions

This large, detailed dataset facilitated an analysis of a large number of variables that significantly associated with HFimpEF; however, their combined discriminatory value for LVEF improvement remained modest, underscoring the complexity of the gene-environment-treatment interactions that govern LV function.

Change in Left Ventricular Ejection Fraction With Coronary Artery Revascularization and Subsequent Risk for Adverse Cardiovascular Outcomes

BACKGROUND

Coronary revascularization is recommended to treat ischemic cardiomyopathy. However, the relations of revascularization-associated ejection fraction (EF) change to subsequent outcomes have not been elucidated.

METHODS

In 10 071 veterans (mean age 67 years; 1% women; 15% non-White) who underwent a first percutaneous coronary intervention (PCI) or coronary artery bypass grafting between January 1, 1995, and December 31, 2010, and had prerevascularization and postrevascularization EF measured, we calculated delta-EF (postprocedure EF-preprocedure EF). We related delta-EF as a continuous measure and as categories (≤-5, -5<delta-EF<0, delta-EF=0, 0<delta-EF<5, and delta-EF≥5) to death (using Cox regression) and heart failure hospitalization days (using negative binomial regression) in multivariable-adjusted models, for total sample, and PCI and coronary artery bypass grafting strata.

RESULTS

Over follow-up (mean/maximum 5/14 years) 56% died. Each 5% improvement in delta-EF was associated with statistically significant reductions in death and heart failure hospitalization days of 5% (95% CI, 3%-7%) and 10% (95% CI, 5%-15%), respectively, in the total sample and 6% (95% CI, 4%-8%) and 10% (95% CI, 5%-16%), respectively, in the PCI subgroup. Patients in the highest delta-EF category had 27% (95% CI, 19%-34%) lower mortality (30% [95% CI, 21%-37%] lower in PCI stratum) and ≈40% lower heart failure hospitalization days in total sample and PCI stratum, compared with those in the lowest category. Relations of delta-EF and outcomes in coronary artery bypass grafting subgroup did not reach statistical significance.

CONCLUSIONS

Revascularization-associated EF improvement was associated with significant reductions in mortality and heart failure hospitalization burden, particularly in the PCI subgroup.

An architecture for research computing in health to support clinical and translational investigators with electronic patient data

OBJECTIVE

Obtaining electronic patient data, especially from electronic health record (EHR) systems, for clinical and translational research is difficult. Multiple research informatics systems exist but navigating the numerous applications can be challenging for scientists. This article describes Architecture for Research Computing in Health (ARCH), our institution's approach for matching investigators with tools and services for obtaining electronic patient data.

MATERIALS AND METHODS

Supporting the spectrum of studies from populations to individuals, ARCH delivers a breadth of scientific functions-including but not limited to cohort discovery, electronic data capture, and multi-institutional data sharing-that manifest in specific systems-such as i2b2, REDCap, and PCORnet. Through a consultative process, ARCH staff align investigators with tools with respect to study design, data sources, and cost. Although most ARCH services are available free of charge, advanced engagements require fee for service.

RESULTS

Since 2016 at Weill Cornell Medicine, ARCH has supported over 1200 unique investigators through more than 4177 consultations. Notably, ARCH infrastructure enabled critical coronavirus disease 2019 response activities for research and patient care.

DISCUSSION

ARCH has provided a technical, regulatory, financial, and educational framework to support the biomedical research enterprise with electronic patient data. Collaboration among informaticians, biostatisticians, and clinicians has been critical to rapid generation and analysis of EHR data.

CONCLUSION

A suite of tools and services, ARCH helps match investigators with informatics systems to reduce time to science. ARCH has facilitated research at Weill Cornell Medicine and may provide a model for informatics and research leaders to support scientists elsewhere.

Automatic extraction of 12 cardiovascular concepts from German discharge letters using pre-trained language models

Objective

A vast amount of medical data is still stored in unstructured text documents. We present an automated method of information extraction from German unstructured clinical routine data from the cardiology domain enabling their usage in state-of-the-art data-driven deep learning projects.

Methods

We evaluated pre-trained language models to extract a set of 12 cardiovascular concepts in German discharge letters. We compared three bidirectional encoder representations from transformers pre-trained on different corpora and fine-tuned them on the task of cardiovascular concept extraction using 204 discharge letters manually annotated by cardiologists at the University Hospital Heidelberg. We compared our results with traditional machine learning methods based on a long short-term memory network and a conditional random field.

Results

Our best performing model, based on publicly available German pre-trained bidirectional encoder representations from the transformer model, achieved a token-wise micro-average F1-score of 86% and outperformed the baseline by at least 6%. Moreover, this approach achieved the best trade-off between precision (positive predictive value) and recall (sensitivity).

Conclusion

Our results show the applicability of state-of-the-art deep learning methods using pre-trained language models for the task of cardiovascular concept extraction using limited training data. This minimizes annotation efforts, which are currently the bottleneck of any application of data-driven deep learning projects in the clinical domain for German and many other European languages.

Risk factors and prediction models for incident heart failure with reduced and preserved ejection fraction

AIMS

This study aims to develop the first race-specific and sex-specific risk prediction models for heart failure with preserved (HFpEF) and reduced ejection fraction (HFrEF).

METHODS AND RESULTS

We created a cohort of 1.8 million individuals who had an outpatient clinic visit between 2002 and 2007 within the Veterans Affairs (VA) Healthcare System and obtained information on HFpEF, HFrEF, and several risk factors from electronic health records (EHR). Variables were selected for the risk prediction models in a 'derivation cohort' that consisted of individuals with baseline date in 2002, 2003, or 2004 using a forward stepwise selection based on a change in C-index threshold. Discrimination and calibration were assessed in the remaining participants (internal 'validation cohort'). A total of 66 831 individuals developed HFpEF, and 92 233 developed HFrEF (52 679 and 71 463 in the derivation cohort) over a median of 11.1 years of follow-up. The HFpEF risk prediction model included age, diabetes, BMI, COPD, previous MI, antihypertensive treatment, SBP, smoking status, atrial fibrillation, and estimated glomerular filtration rate (eGFR), while the HFrEF model additionally included previous CAD. For the HFpEF model, C-indices were 0.74 (SE = 0.002) for white men, 0.76 (0.005) for black men, 0.79 (0.015) for white women, and 0.77 (0.026) for black women, compared with 0.72 (0.002), 0.72 (0.004), 0.77 (0.017), and 0.75 (0.028), respectively, for the HFrEF model. These risk prediction models were generally well calibrated in each race-specific and sex-specific stratum of the validation cohort.

CONCLUSIONS

Our race-specific and sex-specific risk prediction models, which used easily obtainable clinical variables, can be a useful tool to implement preventive strategies or subtype-specific prevention trials in the nine million users of the VA healthcare system and the general population after external validation.

Evaluation of Quality of Care for US Veterans With Recent-Onset Heart Failure With Reduced Ejection Fraction

Importance

Multiple guideline-recommended therapies for heart failure with reduced ejection fraction (HFrEF) are available and promoted by performance measures. However, contemporary data on the use of these therapies are limited.

Objective

To evaluate trends in guideline-directed medical therapy, implantable cardioverter-defibrillator (ICD) use, and risk-adjusted mortality among patients with recent-onset HFrEF.

Design, Setting, and Participants

This cohort study analyzed claims and electronic health record data of patients with recent-onset HFrEF diagnosed at US Department of Veterans Affairs (VA) health care system facilities from July 1, 2013, through June 30, 2019. Veterans who had a history of heart transplant or used a ventricular assist device were among the patients who were excluded.

Exposures

Guideline-directed medical therapy (any β-blocker, guideline-recommended β-blocker [bisoprolol, carvedilol, or metoprolol succinate], angiotensin-converting enzyme inhibitor, angiotensin receptor blocker, angiotensin receptor-neprilysin inhibitor, mineralocorticoid receptor antagonist, and hydralazine plus nitrate) and ICD.

Main Outcomes and Measures

Treatment rates for guideline-directed medical therapies and ICDs were calculated within 6 months of the index HFrEF date using medication fills, procedural codes for implantation and monitoring, and diagnosis codes. Risk-adjusted mortality was calculated after adjusting for baseline patient characteristics. For both treatment rates and risk-adjusted mortality, we evaluated the change over 3 periods (period 1: July 1, 2013, to June 30, 2015; period 2: July 1, 2015, to June 30, 2017; and period 3: July 1, 2017, to June 30, 2019) and variation across VA facilities.

Results

The final cohort comprised 144 074 eligible patients with incident HFrEF that was diagnosed between July 1, 2013, and June 30, 2019. The cohort had a mean (SD) age of 71.0 (11.4) years and was mostly composed of men (140 765 [97.7%]). Overall, changes in medical therapy rates were minimal over time, with the use of a guideline-recommended β-blocker increasing from 64.2% in 2013 to 72.0% in 2019. Rates for mineralocorticoid receptor antagonist therapy increased from 23.9% in 2013 to 26.9% in 2019, and rates for hydralazine plus nitrate therapy remained stable at 24.2% over the study period. Rates for angiotensin receptor-neprilysin inhibitor therapy increased since its introduction in 2015 but only to 22.6% in 2019. Among patients with an ICD indication, early use rates decreased over time. Substantial variation in medical therapy rates persisted across VA facilities. Risk-adjusted mortality decreased over the study period from 19.9% (95% CI, 19.6%-20.2%) in July 1, 2013, to June 30, 2015, to 18.4% (95% CI, 18.0%-18.7%) in July 1, 2017, to June 30, 2019 (OR, 0.96 per additional year; 95% CI, 0.96-0.97).

Conclusions and Relevance

This study found only marginal improvement between 2013 and 2019 in the guideline-recommended therapy and mortality rates among patients with recent-onset HFrEF. New approaches to increase the uptake of evidence-based HFrEF treatment are urgently needed and could lead to larger reductions in mortality.

Systematic review of current natural language processing methods and applications in cardiology

Natural language processing (NLP) is a set of automated methods to organise and evaluate the information contained in unstructured clinical notes, which are a rich source of real-world data from clinical care that may be used to improve outcomes and understanding of disease in cardiology. The purpose of this systematic review is to provide an understanding of NLP, review how it has been used to date within cardiology and illustrate the opportunities that this approach provides for both research and clinical care. We systematically searched six scholarly databases (ACM Digital Library, Arxiv, Embase, IEEE Explore, PubMed and Scopus) for studies published in 2015-2020 describing the development or application of NLP methods for clinical text focused on cardiac disease. Studies not published in English, lacking a description of NLP methods, non-cardiac focused and duplicates were excluded. Two independent reviewers extracted general study information, clinical details and NLP details and appraised quality using a checklist of quality indicators for NLP studies. We identified 37 studies developing and applying NLP in heart failure, imaging, coronary artery disease, electrophysiology, general cardiology and valvular heart disease. Most studies used NLP to identify patients with a specific diagnosis and extract disease severity using rule-based NLP methods. Some used NLP algorithms to predict clinical outcomes. A major limitation is the inability to aggregate findings across studies due to vastly different NLP methods, evaluation and reporting. This review reveals numerous opportunities for future NLP work in cardiology with more diverse patient samples, cardiac diseases, datasets, methods and applications.

Sacubitril/Valsartan Initiation Among Veterans Who Are Renin-Angiotensin-Aldosterone System Inhibitor Naïve With Heart Failure and Reduced Ejection Fraction

Background

Sacubitril/valsartan, a first‐in‐class angiotensin receptor neprilysin inhibitor, received US Food and Drug Administration approval in 2015 for heart failure with reduced ejection fraction (HFrEF). Our objective was to describe the sacubitril/valsartan initiation rate, associated characteristics, and 6‐month follow‐up dosing among veterans with HFrEF who are renin‐angiotensin‐aldosterone system inhibitor (RAASi) naïve.

Methods and Results

Retrospective cohort study of veterans with HFrEF who are RAASi naïve defined as left ventricular ejection fraction (LVEF) ≤40%; ≥1 in/outpatient heart failure visit, first RAASi (sacubitril/valsartan, angiotensin‐converting enzyme inhibitor [ACEI]), or angiotensin‐II receptor blocker [ARB]) fill from July 2015 to June 2019. Characteristics associated with sacubitril/valsartan initiation were identified using Poisson regression models. From July 2015 to June 2019, we identified 3458 sacubitril/valsartan and 29 367 ACEI or ARB initiators among veterans with HFrEF who are RAASi naïve. Sacubitril/valsartan initiation increased from 0% to 26.5%. Sacubitril/valsartan (versus ACEI or ARB) initiators were less likely to have histories of stroke, myocardial infarction, or hypertension and more likely to be older and have diabetes mellitus and lower LVEF. At 6‐month follow‐up, the prevalence of ≥50% target daily dose for sacubitril/valsartan, ACEI, and ARB initiators was 23.5%, 43.2%, and 47.1%, respectively.

Conclusions

Sacubitril/valsartan initiation for HFrEF in the Veterans Administration increased in the 4 years immediately following Food and Drug Administration approval. Sacubitril/valsartan (versus ACEI or ARB) initiators had fewer baseline cardiovascular comorbidities and the lowest proportion on ≥50% target daily dose at 6‐month follow‐up. Identifying the reasons for lower follow‐up dosing of sacubitril/valsartan could support guideline recommendations and quality improvement strategies for patients with HFrEF.

Natural Language Mapping of Electrocardiogram Interpretations to a Standardized Ontology

BACKGROUND

Interpretations of the electrocardiogram (ECG) are often prepared using software outside the electronic health record (EHR) and imported via an interface as a narrative note. Thus, natural language processing is required to create a computable representation of the findings. Challenges include misspellings, nonstandard abbreviations, jargon, and equivocation in diagnostic interpretations.

OBJECTIVES

Our objective was to develop an algorithm to reliably and efficiently extract such information and map it to the standardized ECG ontology developed jointly by the American Heart Association, the American College of Cardiology Foundation, and the Heart Rhythm Society. The algorithm was to be designed to be easily modifiable for use with EHRs and ECG reporting systems other than the ones studied.

METHODS

An algorithm using natural language processing techniques was developed in structured query language to extract and map quantitative and diagnostic information from ECG narrative reports to the cardiology societies' standardized ECG ontology. The algorithm was developed using a training dataset of 43,861 ECG reports and applied to a test dataset of 46,873 reports.

RESULTS

Accuracy, precision, recall, and the F1-measure were all 100% in the test dataset for the extraction of quantitative data (e.g., PR and QTc interval, atrial and ventricular heart rate). Performances for matches in each diagnostic category in the standardized ECG ontology were all above 99% in the test dataset. The processing speed was approximately 20,000 reports per minute. We externally validated the algorithm from another institution that used a different ECG reporting system and found similar performance.

CONCLUSION

The developed algorithm had high performance for creating a computable representation of ECG interpretations. Software and lookup tables are provided that can easily be modified for local customization and for use with other EHR and ECG reporting systems. This algorithm has utility for research and in clinical decision-support where incorporation of ECG findings is desired.

A review of risk prediction models in cardiovascular disease: conventional approach vs. artificial intelligent approach

Cardiovascular disease (CVD) is the leading cause of death worldwide and is a global health issue. Traditionally, statistical models are used commonly in the risk prediction and assessment of CVD. However, the adoption of artificial intelligent (AI) approach is rapidly taking hold in the current era of technology to evaluate patient risks and predict the outcome of CVD. In this review, we outline various conventional risk scores and prediction models and do a comparison with the AI approach. The strengths and limitations of both conventional and AI approaches are discussed. Besides that, biomarker discovery related to CVD are also elucidated as the biomarkers can be used in the risk stratification as well as early detection of the disease. Moreover, problems and challenges involved in current CVD studies are explored. Lastly, future prospects of CVD risk prediction and assessment in the multi-modality of big data integrative approaches are proposed.

Association between HIV and incident pulmonary hypertension in US Veterans: a retrospective cohort study

Background

Pulmonary hypertension incidence based on echocardiographic estimates of pulmonary artery systolic pressure in people living with HIV remains unstudied. We aimed to determine whether people living with HIV have higher incidence and risk of pulmonary hypertension than uninfected individuals.

Methods

In this retrospective cohort study, we evaluated data from participants in the Veterans Aging Cohort Study (VACS) referred for echocardiography with baseline pulmonary artery systolic pressure measures of 35 mm Hg or less. Incident pulmonary hypertension was defined as pulmonary artery systolic pressure higher than 35 mm Hg on subsequent echocardiogram. We used Poisson regression to estimate incidence rates (IRs) of pulmonary hypertension by HIV status. We then estimated hazard ratios (HRs) by HIV status using Cox proportional hazards regression. We further categorised veterans with HIV by CD4 count or HIV viral load to assess the association between pulmonary hypertension risk and HIV severity. Models included age, sex, race or ethnicity, prevalent heart failure, chronic obstructive pulmonary disease, hypertension, smoking status, diabetes, body-mass index, estimated glomerular filtration rate, hepatitis C virus infection, liver cirrhosis, and drug use as covariates.

Findings

Of 21 314 VACS participants with at least one measured PASP on or after April 1, 2003, 13 028 VACS participants were included in the analytic sample (4174 [32%] with HIV and 8854 [68%] without HIV). Median age was 58 years and 12 657 (97%) were male. Median follow-up time was 3·1 years (IQR 0·9-6·8) spanning from April 1, 2003, to Sept 30, 2017. Unadjusted IRs per 1000 person-years were higher in veterans with HIV (IR 28·6 [95% CI 26·1-31·3]) than in veterans without HIV (IR 23·4 [21·9-24·9]; p=0·0004). The risk of incident pulmonary hypertension was higher among veterans with HIV than among veterans without HIV (unadjusted HR 1·25 [95% CI 1·12-1·40], p<0·0001). After multivariable adjustment, this association was slightly attenuated but remained significant (HR 1·18 [1·05-1·34], p=0·0062). Veterans with HIV who had a CD4 count lower than 200 cells per μL or of 200-499 cells per μL had a higher risk of pulmonary hypertension than did veterans without HIV (HR 1·94 [1·49-2·54], p<0·0001, for those with <200 cell μL and HR 1·29 [1·08-1·53], p=0·0048, for those with 200-499 cells per μL). Similarly, veterans with HIV who had HIV viral loads of 500 copies per mL or more had a higher risk of pulmonary hypertension than did veterans without HIV (HR 1·88 [1·46-2·42], p<0·0001).

Interpretation

HIV is associated with pulmonary hypertension incidence, adjusting for risk factors. Low CD4 cell count and high HIV viral load contribute to increased pulmonary hypertension risk among veterans with HIV. Thus, as with other cardiopulmonary diseases, suppression of HIV should be prioritised to lessen the burden of pulmonary hypertension in people living with HIV.

Funding

National Heart, Lung, and Blood Institute (National Institutes of Health, USA); National Institute on Alcohol Abuse and Alcoholism (National Institutes of Health, USA).

A systematic review of natural language processing applied to radiology reports

BACKGROUND

Natural language processing (NLP) has a significant role in advancing healthcare and has been found to be key in extracting structured information from radiology reports. Understanding recent developments in NLP application to radiology is of significance but recent reviews on this are limited. This study systematically assesses and quantifies recent literature in NLP applied to radiology reports.

METHODS

We conduct an automated literature search yielding 4836 results using automated filtering, metadata enriching steps and citation search combined with manual review. Our analysis is based on 21 variables including radiology characteristics, NLP methodology, performance, study, and clinical application characteristics.

RESULTS

We present a comprehensive analysis of the 164 publications retrieved with publications in 2019 almost triple those in 2015. Each publication is categorised into one of 6 clinical application categories. Deep learning use increases in the period but conventional machine learning approaches are still prevalent. Deep learning remains challenged when data is scarce and there is little evidence of adoption into clinical practice. Despite 17% of studies reporting greater than 0.85 F1 scores, it is hard to comparatively evaluate these approaches given that most of them use different datasets. Only 14 studies made their data and 15 their code available with 10 externally validating results.

CONCLUSIONS

Automated understanding of clinical narratives of the radiology reports has the potential to enhance the healthcare process and we show that research in this field continues to grow. Reproducibility and explainability of models are important if the domain is to move applications into clinical use. More could be done to share code enabling validation of methods on different institutional data and to reduce heterogeneity in reporting of study properties allowing inter-study comparisons. Our results have significance for researchers in the field providing a systematic synthesis of existing work to build on, identify gaps, opportunities for collaboration and avoid duplication.

Hospitalization for Heart Failure Among Patients With Diabetes Mellitus and Reduced Kidney Function Treated With Metformin Versus Sulfonylureas: A Retrospective Cohort Study

Background Metformin and sulfonylurea are commonly prescribed oral medications for type 2 diabetes mellitus. The association of metformin and sulfonylureas on heart failure outcomes in patients with reduced estimated glomerular filtration rate remains poorly understood. Methods and Results This retrospective cohort combined data from National Veterans Health Administration, Medicare, Medicaid, and the National Death Index. New users of metformin or sulfonylurea who reached an estimated glomerular filtration rate of 60 mL/min per 1.73 m2 or serum creatinine of 1.5 mg/dL and continued metformin or sulfonylurea were included. The primary outcome was hospitalization for heart failure. Echocardiogram reports were obtained to determine each patient's ejection fraction (EF) (reduced EF <40%; midrange EF 40%-49%; ≥50%). The primary analysis estimated the cause-specific hazard ratios for metformin versus sulfonylurea and estimated the cumulative incidence functions for heart failure hospitalization and competing events. The weighted cohort included 24 685 metformin users and 24 805 sulfonylurea users with reduced kidney function (median age 70 years, estimated glomerular filtration rate 55.8 mL/min per 1.73 m2). The prevalence of underlying heart failure (12.1%) and cardiovascular disease (31.7%) was similar between groups. There were 16.9 (95% CI, 15.8-18.1) versus 20.7 (95% CI, 19.5-22.0) heart failure hospitalizations per 1000 person-years for metformin and sulfonylurea users, respectively, yielding a cause-specific hazard of 0.85 (95% CI, 0.78-0.93). Among heart failure hospitalizations, 44.5% did not have echocardiogram information available; 29.3% were categorized as reduced EF, 8.9% as midrange EF, and 17.2% as preserved EF. Heart failure hospitalization with reduced EF (hazard ratio, 0.79; 95% CI, 0.67-0.93) and unknown EF (hazard ratio, 0.84; 95% CI 0.74-96) were significantly lower in metformin versus sulfonylurea users. Conclusions Among patients with type 2 diabetes mellitus who developed worsening kidney function, persistent metformin compared with sulfonylurea use was associated with reduced heart failure hospitalization.

Automating Clinical Chart Review: An Open-Source Natural Language Processing Pipeline Developed on Free-Text Radiology Reports From Patients With Glioblastoma

PURPOSE

The aim of this study was to develop an open-source natural language processing (NLP) pipeline for text mining of medical information from clinical reports. We also aimed to provide insight into why certain variables or reports are more suitable for clinical text mining than others.

MATERIALS AND METHODS

Various NLP models were developed to extract 15 radiologic characteristics from free-text radiology reports for patients with glioblastoma. Ten-fold cross-validation was used to optimize the hyperparameter settings and estimate model performance. We examined how model performance was associated with quantitative attributes of the radiologic characteristics and reports.

RESULTS

In total, 562 unique brain magnetic resonance imaging reports were retrieved. NLP extracted 15 radiologic characteristics with high to excellent discrimination (area under the curve, 0.82 to 0.98) and accuracy (78.6% to 96.6%). Model performance was correlated with the inter-rater agreement of the manually provided labels (ρ = 0.904; P < .001) but not with the frequency distribution of the variables of interest (ρ = 0.179; P = .52). All variables labeled with a near perfect inter-rater agreement were classified with excellent performance (area under the curve > 0.95). Excellent performance could be achieved for variables with only 50 to 100 observations in the minority group and class imbalances up to a 9:1 ratio. Report-level classification accuracy was not associated with the number of words or the vocabulary size in the distinct text documents.

CONCLUSION

This study provides an open-source NLP pipeline that allows for text mining of narratively written clinical reports. Small sample sizes and class imbalance should not be considered as absolute contraindications for text mining in clinical research. However, future studies should report measures of inter-rater agreement whenever ground truth is based on a consensus label and use this measure to identify clinical variables eligible for text mining.

Cardiovascular Informatics: building a bridge to data harmony

The search for new strategies for better understanding cardiovascular disease is a constant one, spanning multitudinous types of observations and studies. A comprehensive characterization of each disease state and its biomolecular underpinnings relies upon insights gleaned from extensive information collection of various types of data. Researchers and clinicians in cardiovascular biomedicine repeatedly face questions regarding which types of data may best answer their questions, how to integrate information from multiple datasets of various types, and how to adapt emerging advances in machine learning and/or artificial intelligence to their needs in data processing. Frequently lauded as a field with great practical and translational potential, the interface between biomedical informatics and cardiovascular medicine is challenged with staggeringly massive datasets. Successful application of computational approaches to decode these complex and gigantic amounts of information becomes an essential step toward realizing the desired benefits. In this review, we examine recent efforts to adapt informatics strategies to cardiovascular biomedical research: automated information extraction and unification of multifaceted -omics data. We discuss how and why this interdisciplinary space of Cardiovascular Informatics is particularly relevant to and supportive of current experimental and clinical research. We describe in detail how open data sources and methods can drive discovery while demanding few initial resources, an advantage afforded by widespread availability of cloud computing-driven platforms. Subsequently, we provide examples of how interoperable computational systems facilitate exploration of data from multiple sources, including both consistently-formatted structured data and unstructured data. Taken together, these approaches for achieving data harmony enable molecular phenotyping of cardiovascular (CV) diseases and unification of cardiovascular knowledge.

Trends in Readmission and Mortality Rates Following Heart Failure Hospitalization in the Veterans Affairs Health Care System From 2007 to 2017

Importance

The Centers for Medicare & Medicaid Services and the Veterans Affairs Health Care System provide incentives for hospitals to reduce 30-day readmission and mortality rates. In contrast with the large body of evidence describing readmission and mortality in the Medicare system, it is unclear how heart failure readmission and mortality rates have changed during this period in the Veterans Affairs Health Care System.

Objectives

To evaluate trends in readmission and mortality after heart failure admission in the Veterans Affairs Health Care System, which had no financial penalties, in a decade involving focus on heart failure readmission reduction (2007-2017).

Design, Setting, and Participants

This cohort study used data from all Veterans Affairs-paid heart failure admissions from January 2007 to September 2017. All Veterans Affairs-paid hospital admissions to Veterans Affairs and non-Veterans Affairs facilities for a primary diagnosis of heart failure were included, when the admission was paid for by the Veterans Affairs. Data analyses were conducted from October 2018 to March 2020.

Exposures

Admission for a primary diagnosis of heart failure at discharge.

Main Outcomes and Measures

Thirty-day all-cause readmission and mortality rates.

Results

A total of 164 566 patients with 304 374 hospital admissions were included. Among the 304 374 hospital admissions between 2007 and 2017, 298 260 (98.0%) were for male patients, and 195 205 (64.4%) were for white patients. The mean (SD) age was 70.8 (11.5) years. The adjusted odds ratio of 30-day readmission declined throughout the study period to 0.85 (95% CI, 0.83-0.88) in 2015 to 2017 compared with 2007 to 2008. The adjusted odds ratio of 30-day mortality remained stable, with an adjusted odds ratio of 1.01 (95% CI, 0.96-1.06) in 2015 to 2017 compared with 2007 to 2008. Stratification by left ventricular ejection fraction showed similar readmission reduction trends and no significant change in mortality, regardless of strata.

Conclusions and Relevance

In this analysis of an integrated health care system that provided guidance and nonfinancial incentives for reducing readmissions, such as public reporting of readmission rates, risk-adjusted 30-day readmission declined despite inclusion of clinical variables in risk adjustment, but mortality did not decline. Future investigations should focus on evaluating the effectiveness of specific approaches to readmission reduction to inform efficient and effective application in individual health systems, hospitals, and practices.

Underserved populations with missing race ethnicity data differ significantly from those with structured race/ethnicity documentation

OBJECTIVE

We aimed to address deficiencies in structured electronic health record (EHR) data for race and ethnicity by identifying black and Hispanic patients from unstructured clinical notes and assessing differences between patients with or without structured race/ethnicity data.

MATERIALS AND METHODS

Using EHR notes for 16 665 patients with encounters at a primary care practice, we developed rule-based natural language processing (NLP) algorithms to classify patients as black/Hispanic. We evaluated performance of the method against an annotated gold standard, compared race and ethnicity between NLP-derived and structured EHR data, and compared characteristics of patients identified as black or Hispanic using only NLP vs patients identified as such only in structured EHR data.

RESULTS

For the sample of 16 665 patients, NLP identified 948 additional patients as black, a 26%increase, and 665 additional patients as Hispanic, a 20% increase. Compared with the patients identified as black or Hispanic in structured EHR data, patients identified as black or Hispanic via NLP only were older, more likely to be male, less likely to have commercial insurance, and more likely to have higher comorbidity.

DISCUSSION

Structured EHR data for race and ethnicity are subject to data quality issues. Supplementing structured EHR race data with NLP-derived race and ethnicity may allow researchers to better assess the demographic makeup of populations and draw more accurate conclusions about intergroup differences in health outcomes.

CONCLUSIONS

Black or Hispanic patients who are not documented as such in structured EHR race/ethnicity fields differ significantly from those who are. Relatively simple NLP can help address this limitation.

Transformation of Pathology Reports Into the Common Data Model With Oncology Module: Use Case for Colon Cancer

BACKGROUND

Common data models (CDMs) help standardize electronic health record data and facilitate outcome analysis for observational and longitudinal research. An analysis of pathology reports is required to establish fundamental information infrastructure for data-driven colon cancer research. The Observational Medical Outcomes Partnership (OMOP) CDM is used in distributed research networks for clinical data; however, it requires conversion of free text-based pathology reports into the CDM's format. There are few use cases of representing cancer data in CDM.

OBJECTIVE

In this study, we aimed to construct a CDM database of colon cancer-related pathology with natural language processing (NLP) for a research platform that can utilize both clinical and omics data. The essential text entities from the pathology reports are extracted, standardized, and converted to the OMOP CDM format in order to utilize the pathology data in cancer research.

METHODS

We extracted clinical text entities, mapped them to the standard concepts in the Observational Health Data Sciences and Informatics vocabularies, and built databases and defined relations for the CDM tables. Major clinical entities were extracted through NLP on pathology reports of surgical specimens, immunohistochemical studies, and molecular studies of colon cancer patients at a tertiary general hospital in South Korea. Items were extracted from each report using regular expressions in Python. Unstructured data, such as text that does not have a pattern, were handled with expert advice by adding regular expression rules. Our own dictionary was used for normalization and standardization to deal with biomarker and gene names and other ungrammatical expressions. The extracted clinical and genetic information was mapped to the Logical Observation Identifiers Names and Codes databases and the Systematized Nomenclature of Medicine (SNOMED) standard terminologies recommended by the OMOP CDM. The database-table relationships were newly defined through SNOMED standard terminology concepts. The standardized data were inserted into the CDM tables. For evaluation, 100 reports were randomly selected and independently annotated by a medical informatics expert and a nurse.

RESULTS

We examined and standardized 1848 immunohistochemical study reports, 3890 molecular study reports, and 12,352 pathology reports of surgical specimens (from 2017 to 2018). The constructed and updated database contained the following extracted colorectal entities: (1) NOTE_NLP, (2) MEASUREMENT, (3) CONDITION_OCCURRENCE, (4) SPECIMEN, and (5) FACT_RELATIONSHIP of specimen with condition and measurement.

CONCLUSIONS

This study aimed to prepare CDM data for a research platform to take advantage of all omics clinical and patient data at Seoul National University Bundang Hospital for colon cancer pathology. A more sophisticated preparation of the pathology data is needed for further research on cancer genomics, and various types of text narratives are the next target for additional research on the use of data in the CDM.

The Utility of Clinical Notes for Sexual Minority Health Research

INTRODUCTION

Despite improvements in electronic medical record capability to collect data on sexual orientation, not all healthcare systems have adopted this practice. This can limit the usability of systemwide electronic medical record data for sexual minority research. One viable resource might be the documentation of sexual orientation within clinical notes. The authors developed an approach to identify sexual orientation documentation and subsequently derived a cohort of sexual minority patients using clinical notes from the Veterans Health Administration electronic medical record.

METHODS

A hybrid natural language processing approach was developed and used to identify and categorize instances of terms and phrases related to sexual orientation in Veterans Health Administration clinical notes from 2000 to 2019. System performance was assessed with positive predictive value and sensitivity. Data were analyzed in 2019.

RESULTS

A total of 2,413,584 sexual minority terms/phrases were found within clinical notes, of which 439,039 (18%) were found to be related to patient sexual orientation with a positive predictive value of 85.9%. Documentation of sexual orientation was found for 115,312 patients. When compared with 2,262 patients with a record of administrative coding for homosexuality, the system found mentions of sexual orientation for 1,808 patients (79.9% sensitivity).

CONCLUSIONS

When systemwide structured data are unavailable or inconsistent, deriving a cohort of sexual minority patients in electronic medical records for research is possible and permits longitudinal analysis across multiple clinical domains. Although limitations and challenges to the approach were identified, this study makes an important step forward for the Veterans Health Administration sexual minority research, and the methodology can be applied in other healthcare organizations.

HIV- and HCV-specific markers and echocardiographic pulmonary artery systolic pressure among United States veterans

Hepatitis C virus (HCV) may increase pulmonary hypertension (PH) risk among people living with HIV (PLWH). Prior studies on this topic have been relatively small and examined selected populations. We determine whether HIV/HCV coinfection is associated with higher pulmonary artery systolic pressure (PASP) and prevalent echocardiographic PH. We performed a cross-sectional analysis of 6032 (16% HIV/HCV coinfected) Veterans Aging Cohort Study participants enrolled 4/1/2003-9/30/2012 with echocardiographic PASP measures. We performed multiple linear and logistic regression analyses to determine whether HIV/HCV mono- or co-infection were associated with PASP and PH compared to uninfected individuals. Individuals with HIV/HCV coinfection displayed a higher PASP than uninfected individuals ([Formula: see text]=1.10, 95% CI 0.01, 2.20) but there was no association between HIV/HCV coinfection and prevalent PH. Subset analyses examined HIV and HCV disease severity markers separately and jointly. Among PLWH, HCV coinfection ([Formula: see text]=1.47, 95% CI 0.26, 2.67) and CD4 + cell count ([Formula: see text]= - 0.68, 95% CI - 1.10, - 0.27), but not HIV viral load nor ART regimen, were associated with PASP. Among people with HCV, neither HIV coinfection nor HCV biomarkers were associated with PASP. Among US veterans referred for echocardiography, HIV/HCV coinfection was not associated with a clinically significant elevation in pulmonary pressure. Lower absolute CD4 + T-cell count was inversely associated with PASP which warrants further investigation in prospective studies.

Trends in Left Ventricular Ejection Fraction for Patients With a New Diagnosis of Heart Failure

BACKGROUND

The left ventricular ejection fraction (LVEF) guides treatment of heart failure, yet this data has not been systematically collected in large data sets. We sought to characterize the epidemiology of incident heart failure using the initial LVEF.

METHODS

We identified 219 537 patients in the Veterans Affairs system between 2011 and 2017 who had an LVEF documented within 365 days before and 30 days after the heart failure diagnosis date. LVEF was obtained from natural language processing from imaging and provider notes. In multivariate analysis, we assessed characteristics associated with having an initial LVEF <40%.

RESULTS

Most patients were male and White; a plurality were within the 60 to 69 year age decile. A majority of patients had ischemic heart disease and a high burden of co-morbidities. Over time, presentation with an LVEF <40% became slightly less common, with a nadir in 2015. Presentation with an initial LVEF <40% was more common in younger patients, men, Black and Hispanic patients, an inpatient presentation, lower systolic blood pressure, lower pulse pressure, and higher heart rate. Ischemic heart disease, alcohol use disorder, peripheral arterial disease, and ventricular arrhythmias were associated with an initial LVEF <40%, while most other comorbid conditions (eg, atrial fibrillation, chronic obstructive pulmonary disease, malignancy) were more strongly associated with an initial LVEF >50%.

CONCLUSIONS

For patients with heart failure, particularly at the extremes of age, an initial preserved LVEF is common. In addition to clinical characteristics, certain races (Black and Hispanic) were more likely to present with a reduced LVEF. Further studies are needed to determine if racial differences are due to patient or health systems issues such as access to care.

Association Between Heart Failure and Postoperative Mortality Among Patients Undergoing Ambulatory Noncardiac Surgery

Importance

Heart failure is an established risk factor for postoperative mortality, but how heart failure is associated with operative outcomes specifically in the ambulatory surgical setting is not well characterized.

Objective

To assess the risk of postoperative mortality and complications in patients with vs without heart failure at various levels of echocardiographic (left ventricular systolic dysfunction) and clinical (symptoms) severity who were undergoing ambulatory surgery.

Design, Setting, and Participants

In this US multisite retrospective cohort study of all adult patients undergoing ambulatory, elective, noncardiac surgery in the Veterans Affairs Surgical Quality Improvement Project database during fiscal years 2009 to 2016, a total of 355 121 patient records were identified and analyzed with 1 year of follow-up after surgery (final date of follow-up September 1, 2017).

Exposures

Heart failure, left ventricular ejection fraction, and presence of signs or symptoms of heart failure within 30 days of surgery.

Main Outcomes and Measures

The primary outcomes were postoperative mortality at 90 days and any postoperative complication at 30 days.

Results

Among 355 121 total patients, outcome data from 19 353 patients with heart failure (5.5%; mean [SD] age, 67.9 [10.1] years; 18 841 [96.9%] male) and 334 768 patients without heart failure (94.5%; mean [SD] age, 57.2 [14.0] years; 301 198 [90.0%] male) were analyzed. Compared with patients without heart failure, patients with heart failure had a higher risk of 90-day postoperative mortality (crude mortality risk, 2.00% vs 0.39%; adjusted odds ratio [aOR], 1.95; 95% CI, 1.69-2.44), and risk of mortality progressively increased with decreasing systolic function. Compared with patients without heart failure, symptomatic patients with heart failure had a greater risk of mortality (crude mortality risk, 3.57%; aOR, 2.76; 95% CI, 2.07-3.70), as did asymptomatic patients with heart failure (crude mortality risk, 1.85%; aOR, 1.85; 95% CI, 1.60-2.15). Patients with heart failure had a higher risk of experiencing a 30-day postoperative complication than did patients without heart failure (crude risk, 5.65% vs 2.65%; aOR, 1.10; 95% CI, 1.02-1.19).

Conclusions and Relevance

In this study, among patients undergoing elective, ambulatory surgery, heart failure with or without symptoms was significantly associated with 90-day mortality and 30-day postoperative complications. These data may be helpful in preoperative discussions with patients with heart failure undergoing ambulatory surgery.

Open Agile text mining for bioinformatics: the PubAnnotation ecosystem

MOTIVATION

Most currently available text mining tools share two characteristics that make them less than optimal for use by biomedical researchers: they require extensive specialist skills in natural language processing and they were built on the assumption that they should optimize global performance metrics on representative datasets. This is a problem because most end-users are not natural language processing specialists and because biomedical researchers often care less about global metrics like F-measure or representative datasets than they do about more granular metrics such as precision and recall on their own specialized datasets. Thus, there are fundamental mismatches between the assumptions of much text mining work and the preferences of potential end-users.

RESULTS

This article introduces the concept of Agile text mining, and presents the PubAnnotation ecosystem as an example implementation. The system approaches the problems from two perspectives: it allows the reformulation of text mining by biomedical researchers from the task of assembling a complete system to the task of retrieving warehoused annotations, and it makes it possible to do very targeted customization of the pre-existing system to address specific end-user requirements. Two use cases are presented: assisted curation of the GlycoEpitope database, and assessing coverage in the literature of pre-eclampsia-associated genes.

AVAILABILITY AND IMPLEMENTATION

The three tools that make up the ecosystem, PubAnnotation, PubDictionaries and TextAE are publicly available as web services, and also as open source projects. The dictionaries and the annotation datasets associated with the use cases are all publicly available through PubDictionaries and PubAnnotation, respectively.

Trends and Features of the Applications of Natural Language Processing Techniques for Clinical Trials Text Analysis

Natural language processing (NLP) is an effective tool for generating structured information from unstructured data, the one that is commonly found in clinical trial texts. Such interdisciplinary research has gradually grown into a flourishing research field with accumulated scientific outputs available. In this study, bibliographical data collected from Web of Science, PubMed, and Scopus databases from 2001 to 2018 had been investigated with the use of three prominent methods, including performance analysis, science mapping, and, particularly, an automatic text analysis approach named structural topic modeling. Topical trend visualization and test analysis were further employed to quantify the effects of the year of publication on topic proportions. Topical diverse distributions across prolific countries/regions and institutions were also visualized and compared. In addition, scientific collaborations between countries/regions, institutions, and authors were also explored using social network analysis. The findings obtained were essential for facilitating the development of the NLP-enhanced clinical trial texts processing, boosting scientific and technological NLP-enhanced clinical trial research, and facilitating inter-country/region and inter-institution collaborations.

Evaluating the Portability of an NLP System for Processing Echocardiograms: A Retrospective, Multi-site Observational Study

While natural language processing (NLP) of unstructured clinical narratives holds the potential for patient care and clinical research, portability of NLP approaches across multiple sites remains a major challenge. This study investigated the portability of an NLP system developed initially at the Department of Veterans Affairs (VA) to extract 27 key cardiac concepts from free-text or semi-structured echocardiograms from three academic edical centers: Weill Cornell Medicine, Mayo Clinic and Northwestern Medicine. While the NLP system showed high precision and recall easurements for four target concepts (aortic valve regurgitation, left atrium size at end systole, mitral valve regurgitation, tricuspid valve regurgitation) across all sites, we found moderate or poor results for the remaining concepts and the NLP system performance varied between individual sites.