Phenogrouping heart failure with preserved or mildly reduced ejection fraction using electronic health record data

Fardad Soltani; David A Jenkins; Amit Kaura; Joshua Bradley; Nicholas Black; John P Farrant; Simon G Williams; Abdulrahim Mulla; Benjamin Glampson; Jim Davies; Dimitri Papadimitriou; Kerrie Woods; Anoop D Shah; Mark R Thursz; Bryan Williams; Folkert W Asselbergs; Erik K Mayer; Christopher Herbert; Stuart Grant; Nick Curzen; Iain Squire; Thomas Johnson; Kevin O'Gallagher; Ajay M Shah; Divaka Perera; Rajesh Kharbanda; Riyaz S Patel; Keith M Channon; Richard Lee; Niels Peek; Jamil Mayet; Christopher A Miller

doi:10.1186/s12872-024-03987-9

Phenogrouping heart failure with preserved or mildly reduced ejection fraction using electronic health record data

Fardad Soltani, David A Jenkins, Amit Kaura, Joshua Bradley, Nicholas Black, John P Farrant, Simon G Williams, Abdulrahim Mulla, Benjamin Glampson, Jim Davies, Dimitri Papadimitriou, Kerrie Woods, Anoop D Shah, Mark R Thursz, Bryan Williams, Folkert W Asselbergs, Erik K Mayer, Christopher Herbert, Stuart Grant, Nick CurzenIain Squire, Thomas Johnson, Kevin O'Gallagher, Ajay M Shah, Divaka Perera, Rajesh Kharbanda, Riyaz S Patel, Keith M Channon, Richard Lee, Niels Peek, Jamil Mayet, Christopher A Miller^*

^*Corresponding author for this work

Research output: Contribution to journal › Article (Academic Journal) › peer-review

Abstract

BACKGROUND: Heart failure (HF) with preserved or mildly reduced ejection fraction includes a heterogenous group of patients. Reclassification into distinct phenogroups to enable targeted interventions is a priority. This study aimed to identify distinct phenogroups, and compare phenogroup characteristics and outcomes, from electronic health record data.

METHODS: 2,187 patients admitted to five UK hospitals with a diagnosis of HF and a left ventricular ejection fraction ≥ 40% were identified from the NIHR Health Informatics Collaborative database. Partition-based, model-based, and density-based machine learning clustering techniques were applied. Cox Proportional Hazards and Fine-Gray competing risks models were used to compare outcomes (all-cause mortality and hospitalisation for HF) across phenogroups.

RESULTS: Three phenogroups were identified: (1) Younger, predominantly female patients with high prevalence of cardiometabolic and coronary disease; (2) More frail patients, with higher rates of lung disease and atrial fibrillation; (3) Patients characterised by systemic inflammation and high rates of diabetes and renal dysfunction. Survival profiles were distinct, with an increasing risk of all-cause mortality from phenogroups 1 to 3 (p < 0.001). Phenogroup membership significantly improved survival prediction compared to conventional factors. Phenogroups were not predictive of hospitalisation for HF.

CONCLUSIONS: Applying unsupervised machine learning to routinely collected electronic health record data identified phenogroups with distinct clinical characteristics and unique survival profiles.

Original language	English
Article number	343
Journal	BMC Cardiovascular Disorders
Volume	24
DOIs	https://doi.org/10.1186/s12872-024-03987-9
Publication status	Published - 5 Jul 2024

Bibliographical note

Publisher Copyright:
© The Author(s) 2024.

Keywords

Humans
Heart Failure/physiopathology
Electronic Health Records
Female
Male
Aged
Stroke Volume
Middle Aged
Risk Assessment
Ventricular Function, Left
United Kingdom/epidemiology
Risk Factors
Prognosis
Aged, 80 and over
Databases, Factual
Unsupervised Machine Learning
Hospitalization
Time Factors
Comorbidity
Cause of Death
Phenotype
Data Mining

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1186/s12872-024-03987-9Licence: CC BY

Handle.net

Persistent link

Cite this

Soltani, F., Jenkins, D. A., Kaura, A., Bradley, J., Black, N., Farrant, J. P., Williams, S. G., Mulla, A., Glampson, B., Davies, J., Papadimitriou, D., Woods, K., Shah, A. D., Thursz, M. R., Williams, B., Asselbergs, F. W., Mayer, E. K., Herbert, C., Grant, S., ... Miller, C. A. (2024). Phenogrouping heart failure with preserved or mildly reduced ejection fraction using electronic health record data. BMC Cardiovascular Disorders, 24, Article 343. https://doi.org/10.1186/s12872-024-03987-9

@article{333c9e806b514be39c3d38152370e466,

title = "Phenogrouping heart failure with preserved or mildly reduced ejection fraction using electronic health record data",

abstract = "BACKGROUND: Heart failure (HF) with preserved or mildly reduced ejection fraction includes a heterogenous group of patients. Reclassification into distinct phenogroups to enable targeted interventions is a priority. This study aimed to identify distinct phenogroups, and compare phenogroup characteristics and outcomes, from electronic health record data.METHODS: 2,187 patients admitted to five UK hospitals with a diagnosis of HF and a left ventricular ejection fraction ≥ 40% were identified from the NIHR Health Informatics Collaborative database. Partition-based, model-based, and density-based machine learning clustering techniques were applied. Cox Proportional Hazards and Fine-Gray competing risks models were used to compare outcomes (all-cause mortality and hospitalisation for HF) across phenogroups.RESULTS: Three phenogroups were identified: (1) Younger, predominantly female patients with high prevalence of cardiometabolic and coronary disease; (2) More frail patients, with higher rates of lung disease and atrial fibrillation; (3) Patients characterised by systemic inflammation and high rates of diabetes and renal dysfunction. Survival profiles were distinct, with an increasing risk of all-cause mortality from phenogroups 1 to 3 (p < 0.001). Phenogroup membership significantly improved survival prediction compared to conventional factors. Phenogroups were not predictive of hospitalisation for HF.CONCLUSIONS: Applying unsupervised machine learning to routinely collected electronic health record data identified phenogroups with distinct clinical characteristics and unique survival profiles.",

keywords = "Humans, Heart Failure/physiopathology, Electronic Health Records, Female, Male, Aged, Stroke Volume, Middle Aged, Risk Assessment, Ventricular Function, Left, United Kingdom/epidemiology, Risk Factors, Prognosis, Aged, 80 and over, Databases, Factual, Unsupervised Machine Learning, Hospitalization, Time Factors, Comorbidity, Cause of Death, Phenotype, Data Mining",

author = "Fardad Soltani and Jenkins, {David A} and Amit Kaura and Joshua Bradley and Nicholas Black and Farrant, {John P} and Williams, {Simon G} and Abdulrahim Mulla and Benjamin Glampson and Jim Davies and Dimitri Papadimitriou and Kerrie Woods and Shah, {Anoop D} and Thursz, {Mark R} and Bryan Williams and Asselbergs, {Folkert W} and Mayer, {Erik K} and Christopher Herbert and Stuart Grant and Nick Curzen and Iain Squire and Thomas Johnson and Kevin O'Gallagher and Shah, {Ajay M} and Divaka Perera and Rajesh Kharbanda and Patel, {Riyaz S} and Channon, {Keith M} and Richard Lee and Niels Peek and Jamil Mayet and Miller, {Christopher A}",

note = "Publisher Copyright: {\textcopyright} The Author(s) 2024.",

year = "2024",

month = jul,

day = "5",

doi = "10.1186/s12872-024-03987-9",

language = "English",

volume = "24",

journal = "BMC Cardiovascular Disorders",

issn = "1471-2261",

publisher = "BioMed Central",

}

Soltani, F, Jenkins, DA, Kaura, A, Bradley, J, Black, N, Farrant, JP, Williams, SG, Mulla, A, Glampson, B, Davies, J, Papadimitriou, D, Woods, K, Shah, AD, Thursz, MR, Williams, B, Asselbergs, FW, Mayer, EK, Herbert, C, Grant, S, Curzen, N, Squire, I, Johnson, T, O'Gallagher, K, Shah, AM, Perera, D, Kharbanda, R, Patel, RS, Channon, KM, Lee, R, Peek, N, Mayet, J & Miller, CA 2024, 'Phenogrouping heart failure with preserved or mildly reduced ejection fraction using electronic health record data', BMC Cardiovascular Disorders, vol. 24, 343. https://doi.org/10.1186/s12872-024-03987-9

TY - JOUR

T1 - Phenogrouping heart failure with preserved or mildly reduced ejection fraction using electronic health record data

AU - Soltani, Fardad

AU - Jenkins, David A

AU - Kaura, Amit

AU - Bradley, Joshua

AU - Black, Nicholas

AU - Farrant, John P

AU - Williams, Simon G

AU - Mulla, Abdulrahim

AU - Glampson, Benjamin

AU - Davies, Jim

AU - Papadimitriou, Dimitri

AU - Woods, Kerrie

AU - Shah, Anoop D

AU - Thursz, Mark R

AU - Williams, Bryan

AU - Asselbergs, Folkert W

AU - Mayer, Erik K

AU - Herbert, Christopher

AU - Grant, Stuart

AU - Curzen, Nick

AU - Squire, Iain

AU - Johnson, Thomas

AU - O'Gallagher, Kevin

AU - Shah, Ajay M

AU - Perera, Divaka

AU - Kharbanda, Rajesh

AU - Patel, Riyaz S

AU - Channon, Keith M

AU - Lee, Richard

AU - Peek, Niels

AU - Mayet, Jamil

AU - Miller, Christopher A

N1 - Publisher Copyright: © The Author(s) 2024.

PY - 2024/7/5

Y1 - 2024/7/5

N2 - BACKGROUND: Heart failure (HF) with preserved or mildly reduced ejection fraction includes a heterogenous group of patients. Reclassification into distinct phenogroups to enable targeted interventions is a priority. This study aimed to identify distinct phenogroups, and compare phenogroup characteristics and outcomes, from electronic health record data.METHODS: 2,187 patients admitted to five UK hospitals with a diagnosis of HF and a left ventricular ejection fraction ≥ 40% were identified from the NIHR Health Informatics Collaborative database. Partition-based, model-based, and density-based machine learning clustering techniques were applied. Cox Proportional Hazards and Fine-Gray competing risks models were used to compare outcomes (all-cause mortality and hospitalisation for HF) across phenogroups.RESULTS: Three phenogroups were identified: (1) Younger, predominantly female patients with high prevalence of cardiometabolic and coronary disease; (2) More frail patients, with higher rates of lung disease and atrial fibrillation; (3) Patients characterised by systemic inflammation and high rates of diabetes and renal dysfunction. Survival profiles were distinct, with an increasing risk of all-cause mortality from phenogroups 1 to 3 (p < 0.001). Phenogroup membership significantly improved survival prediction compared to conventional factors. Phenogroups were not predictive of hospitalisation for HF.CONCLUSIONS: Applying unsupervised machine learning to routinely collected electronic health record data identified phenogroups with distinct clinical characteristics and unique survival profiles.

AB - BACKGROUND: Heart failure (HF) with preserved or mildly reduced ejection fraction includes a heterogenous group of patients. Reclassification into distinct phenogroups to enable targeted interventions is a priority. This study aimed to identify distinct phenogroups, and compare phenogroup characteristics and outcomes, from electronic health record data.METHODS: 2,187 patients admitted to five UK hospitals with a diagnosis of HF and a left ventricular ejection fraction ≥ 40% were identified from the NIHR Health Informatics Collaborative database. Partition-based, model-based, and density-based machine learning clustering techniques were applied. Cox Proportional Hazards and Fine-Gray competing risks models were used to compare outcomes (all-cause mortality and hospitalisation for HF) across phenogroups.RESULTS: Three phenogroups were identified: (1) Younger, predominantly female patients with high prevalence of cardiometabolic and coronary disease; (2) More frail patients, with higher rates of lung disease and atrial fibrillation; (3) Patients characterised by systemic inflammation and high rates of diabetes and renal dysfunction. Survival profiles were distinct, with an increasing risk of all-cause mortality from phenogroups 1 to 3 (p < 0.001). Phenogroup membership significantly improved survival prediction compared to conventional factors. Phenogroups were not predictive of hospitalisation for HF.CONCLUSIONS: Applying unsupervised machine learning to routinely collected electronic health record data identified phenogroups with distinct clinical characteristics and unique survival profiles.

KW - Humans

KW - Heart Failure/physiopathology

KW - Electronic Health Records

KW - Female

KW - Male

KW - Aged

KW - Stroke Volume

KW - Middle Aged

KW - Risk Assessment

KW - Ventricular Function, Left

KW - United Kingdom/epidemiology

KW - Risk Factors

KW - Prognosis

KW - Aged, 80 and over

KW - Databases, Factual

KW - Unsupervised Machine Learning

KW - Hospitalization

KW - Time Factors

KW - Comorbidity

KW - Cause of Death

KW - Phenotype

KW - Data Mining

U2 - 10.1186/s12872-024-03987-9

DO - 10.1186/s12872-024-03987-9

M3 - Article (Academic Journal)

C2 - 38969974

SN - 1471-2261

VL - 24

JO - BMC Cardiovascular Disorders

JF - BMC Cardiovascular Disorders

M1 - 343

ER -

Phenogrouping heart failure with preserved or mildly reduced ejection fraction using electronic health record data

Abstract

Bibliographical note

Keywords

UN SDGs

Access to Document

Handle.net

Fingerprint

Cite this