Phenotyping of heart failure with preserved ejection faction using electronic health records and echocardiography

Q: What is the clinical evidence from this study?

Study design: Cohort. Population: Heart failure with preserved ejection fraction (HFpEF) (n=2500). Intervention: Machine learning-derived clinical and echocardiographic phenotyping (4 clusters) vs. Comparison across four distinct phenogroups (Clusters 1, 2, 3, and 4). Primary outcome: Combined endpoint of death and/or cardiovascular hospitalization.

Q: What does this research mean for the field?

Patients with heart failure with preserved ejection fraction comprise four distinct clinical and echocardiographic phenogroups that exhibit significantly different survival rates and hospitalization risks. Novelty: ClaimNovelty.INCREMENTAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

Morgane Pierre‐Jean; Benjamin Marut; Elizabeth Curtis; Elena Galli; Marc Cuggia; Guillaume Bouzillé; Erwan Donal

doi:10.1093/ehjopen/oead133

December 14, 2023European Heart Journal OpenOpen Access

Phenotyping of heart failure with preserved ejection faction using electronic health records and echocardiography

View Full Paper

Ask AI

Bookmark

Share

Figures

Graphical abstract published by European Heart Journal Open — Graphical abstract · Published by European Heart Journal Open · CC BY-NC

Why the study?

Patients with HFpEF are not a homogenous population, and different phenotypes can differ in prognosis and optimal management strategies.

Does machine learning-based clustering of clinical electronic health record and echocardiographic variables identify distinct phenotypes associated with death and cardiovascular hospitalization in patients with HFpEF?

Study Design

Type

Cohort (n=2,500)

Multicenter

No

Structured PICO

Does machine learning-based clustering of clinical electronic health record and echocardiographic variables identify distinct phenotypes associated with death and cardiovascular hospitalization in patients with HFpEF?

P

Population

Approximately 2,500 patients with HFpEF and echocardiographic data from a single French university hospital clinical data warehouse (training set n=1,281 across four clusters).

E

Exposure

Unsupervised machine learning clustering (two-step cluster analysis using principal component analysis followed by spectral clustering) and supervised predictive modeling (multinomial logistic regression, support vector machine, and random forest) using 18 clinical variables from electronic health records and 8 echocardiographic variables.

O

Outcome

Composite of all-cause death and/or cardiovascular hospitalization (admissions to a cardiology or geriatric department); and machine learning model classification performance (AUC and accuracy).composite

Limitations

Imperfect automatic clinical variable extraction from electronic health records with poor quality or missing data for a non-negligible number of patients
Time-costly manual quality control required by clinicians
Imputation of missing values using the k-nearest neighbors algorithm, which does not represent real data
Frailty was not quantified
Single-center study requiring further external validation
Imperfect clinical variable extraction from electronic health records due to automatic extraction, resulting in poor quality or missing data for a non-negligible number of patients.
Time-costly manual quality control required by clinicians to verify extreme values and data quality.
Imputation of missing values was performed using the k-nearest neighbors algorithm, which has limitations as these are not real data.
Frailty was not quantified in the cohort.

Ask about this paper

Answers are grounded in this paper and cite their sources.

Implication

Machine learning applied to routine electronic health record and echocardiographic data can classify HFpEF patients into four distinct phenotypic clusters with differing prognoses, potentially enabling more personalized follow-up and targeted therapy.

Abstract

Aims: Patients presenting symptoms of heart failure with preserved ejection fraction (HFpEF) are not a homogenous population. Different phenotypes can differ in prognosis and optimal management strategies. We sought to identify phenotypes of HFpEF by using the medical information database from a large university hospital centre using machine learning. Methods and results: We explored the use of clinical variables from electronic health records in addition to echocardiography to identify different phenotypes of patients with HFpEF. The proposed methodology identifies four phenotypic clusters based on both clinical and echocardiographic characteristics, which have differing prognoses (death and cardiovascular hospitalization). Conclusion: This work demonstrated that artificial intelligence-derived phenotypes could be used as a tool for physicians to assess risk and to target therapies that may improve outcomes.

Discussion

No takes yet. Share an insight, caveat, or question.

Cite This Study

Pierre‐Jean et al. (2023) conducted a cohort in Heart failure with preserved ejection fraction (HFpEF) (n=2,500). Machine learning-derived clinical and echocardiographic phenotyping (4 clusters) vs. Comparison across four distinct phenogroups (Clusters 1, 2, 3, and 4) was evaluated on Combined endpoint of death and/or cardiovascular hospitalization.

synapsesocial.com/papers/6a7b1abf123479a499f3918d https://doi.org/10.1093/ehjopen/oead133

View Full Paper

Ask AI

Bookmark

Share

Phenotyping of heart failure with preserved ejection faction using electronic health records and echocardiography

Figures

Study Design

Structured PICO

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Phenotyping of heart failure with preserved ejection faction using electronic health records and echocardiography

Figures

Study Design

Structured PICO

Limitations

Ask about this paper

Implication

Abstract

Discussion

Cite This Study

Keep reading

Keep reading