Natural Language Processing for the Ascertainment and Phenotyping of Left Ventricular Hypertrophy and Hypertrophic Cardiomyopathy on Echocardiogram Reports

Extracting and accurately phenotyping electronic health documentation is critical for medical research and clinical care. We sought to develop a highly accurate and open-source natural language processing (NLP) module to ascertain and phenotype left ventricular hypertrophy (LVH) and hypertrophic cardiomyopathy (HCM) diagnoses from echocardiogram reports within a diverse hospital network. After the initial development on 17,250 echocardiogram reports, 700 unique reports from 6 hospitals were randomly selected from data repositories within the Mass General Brigham healthcare system and manually adjudicated by physicians for 10 subtypes of LVH and diagnoses of HCM. Using an open-source NLP system, the module was formally tested on 300 training set reports and validated on 400 reports. The sensitivity, specificity, positive predictive value, and negative predictive value were calculated to assess the discriminative accuracy of the NLP module. The NLP demonstrated robust performance across the 10 LVH subtypes, with the overall sensitivity and specificity exceeding 96%. In addition, the NLP module demonstrated excellent performance in detecting HCM diagnoses, with sensitivity and specificity exceeding 93%. In conclusion, we designed a highly accurate NLP module to determine the presence of LVH and HCM on echocardiogram reports. Our work demonstrates the feasibility and accuracy of NLP to detect diagnoses on imaging reports, even when described in free text. This module has been placed in the public domain to advance research, trial recruitment, and population health management for patients with LVH-associated conditions.

Impact of Moderate Aortic Stenosis in Patients With Heart Failure With Reduced Ejection Fraction

BACKGROUND

Afterload from moderate aortic stenosis (AS) may contribute to adverse outcomes in patients with heart failure with reduced ejection fraction (HFrEF).

OBJECTIVES

The authors evaluated clinical outcomes in patients with HFrEF and moderate AS relative to those without AS and with severe AS.

METHODS

Patients with HFrEF, defined by left ventricular ejection fraction (LVEF) <50% and no, moderate, or severe AS were retrospectively identified. The primary endpoint, defined as a composite of all-cause mortality and heart failure (HF) hospitalization, was compared across groups and within a propensity score-matched cohort.

RESULTS

We included 9,133 patients with HFrEF, of whom 374 and 362 had moderate and severe AS, respectively. Over a median follow-up time of 3.1 years, the primary outcome occurred in 62.7% of patients with moderate AS vs 45.9% with no AS (P < 0.0001); rates were similar with severe and moderate AS (62.0% vs 62.7%; P = 0.68). Patients with severe AS had a lower incidence of HF hospitalization (36.2% vs 43.6%; P < 0.05) and were more likely to undergo AVR within the follow-up period. Within a propensity score-matched cohort, moderate AS was associated with an increased risk of HF hospitalization and mortality (HR: 1.24; 95% CI: 1.04-1.49; P = 0.01) and fewer days alive outside of the hospital (P < 0.0001). Aortic valve replacement (AVR) was associated with improved survival (HR: 0.60; CI: 0.36-0.99; P < 0.05).

CONCLUSIONS

In patients with HFrEF, moderate AS is associated with increased rates of HF hospitalization and mortality. Further investigation is warranted to determine whether AVR in this population improves clinical outcomes.

Development and validation of an echocardiographic algorithm to predict long-term mitral and tricuspid regurgitation progression

AIMS

Prediction of mitral (MR) and tricuspid (TR) regurgitation progression on transthoracic echocardiography (TTE) is needed to personalize valvular surveillance intervals and prognostication.

METHODS AND RESULTS

Structured TTE report data at Beth Israel Deaconess Medical Center, 26 January 2000-31 December 2017, were used to determine time to progression (≥1+ increase in severity). TTE predictors of progression were used to create a progression score, externally validated at Massachusetts General Hospital, 1 January 2002-31 December 2019. In the derivation sample (MR, N = 34 933; TR, N = 27 526), only 5379 (15.4%) individuals with MR and 3630 (13.2%) with TR had progression during a median interquartile range) 9.0 (4.1-13.4) years of follow-up. Despite wide inter-individual variability in progression rates, a score based solely on demographics and TTE variables identified individuals with a five- to six-fold higher rate of MR/TR progression over 10 years (high- vs. low-score tertile, rate of progression; MR 20.1% vs. 3.3%; TR 21.2% vs. 4.4%). Compared to those in the lowest score tertile, those in the highest tertile of progression had a four-fold increased risk of mortality. On external validation, the score demonstrated similar performance to other algorithms commonly in use.

CONCLUSION

Four-fifths of individuals had no progression of MR or TR over two decades. Despite wide interindividual variability in progression rates, a score, based solely on TTE parameters, identified individuals with a five- to six-fold higher rate of MR/TR progression. Compared to the lowest tertile, individuals in the highest score tertile had a four-fold increased risk of mortality. Prediction of long-term MR/TR progression is not only feasible but prognostically important.

Abstract 047: Variation by Provider in Echocardiographic Surveillance of Mitral Regurgitation

Clinical outcomes after surgical treatment of mitral regurgitation (MR) are worse if intervention occurs after deterioration of left ventricular (LV) size and function. Echocardiographic surveillance of patients with MR is therefore indicated to avoid ventricular remodeling that could worsen patient outcomes. However, overly frequent trans-thoracic echocardiograms (TTEs) can impair access for other patients and reduce value in the delivery of care. Given the balance between timeliness of surveillance and possible over-utilization of TTE in valvular disease, we sought to investigate patient and provider factors contributing to variation in TTE utilization. We hypothesized that there was variation attributable to provider practice even after adjustment for patient characteristics.

We obtained records of all TTEs from 2001-2016 ordered at a large echocardiography laboratory. For each TTE, we linked to patient demographic data from hospital administrative records. To control for both clinical and demographic predictors of frequency of echocardiography, we constructed a hierarchical mixed-effects linear regression model with the individual physician as the random effect in the model. The outcome variable was time interval between TTEs. Intra-class correlation coefficient (ICC) was used to assess the proportion of total variation in the outcome variable due to provider practice, and shrinkage estimates were used to measure the contribution of individual providers.

After application of exclusion criteria, 79,194 TTEs corresponding to 55,663 TTE intervals remained. The mean interval between TTEs was 11.9 months for severe MR, 15.4 months for moderate MR, and 17.6 and 17.7 months for mild and trace MR respectively. After multivariate adjustment, male gender (Rate Ratio (RR) 0.96; 95% CI 0.94-0.98) was associated with shorter follow up, and Hispanic race (RR 1.11; 95% CI 1.01-1.21) was associated with longer follow-up intervals. Eight hundred and sixty-seven physicians were included in the analysis. After adjustment for patient factors, 31% of the variation in intervals was associated with provider practice and 19% of providers (161 of 867) were over-utilizers of TTEs and 24% (210 of 867) were under-utilizers.

We conclude that substantial variation exists in follow up intervals for TTE assessment of MR even after risk-adjustment for clinical and demographic variables, likely due to provider factors including specialty and experience. The association of TTE interval with race and gender warrants further investigation. Improving standardization of follow-up intervals may offer opportunity to reduce both overutilization and underutilization of echocardiography.