Phenotyping of AMI patients using machine learning: integration of clinical, metabolic, and behavioral features to identify high-risk “frailty–inflammation” phenotype
- 1The First Affiliated Hospital of Harbin Medical University , , ,
- 2The Second Affiliated Hospital of Harbin Medical University , , ,
Abstract
Background: Prognosis among patients with acute myocardial infarction (AMI) shows substantial heterogeneity, making precise identification of high-risk phenotypes essential for improving long-term outcomes. Machine learning techniques were used to identify such phenotypes by integrating multidimensional data and to examine their association with major adverse cardiovascular events (MACE).
Methods: A total of 1,139 patients with AMI were enrolled from the First Affiliated Hospital of Harbin Medical University between March 2023 and June 2024. Multidimensional data, including clinical features, metabolic characteristics, and behavioral patterns, were collected. Phenotyping was performed using a hierarchical clustering algorithm. Associations between derived phenotypes and 1-year MACE were evaluated using Kaplan–Meier survival analysis and multivariable Cox regression models. Shapley Additive Explanations (SHAP) values derived from an XGBoost model were used to quantify the contributions of individual variables to cluster assignment.
Results: Hierarchical clustering identified two clinical phenotypes with differing pathophysiological characteristics. Cluster 2 ( n = 243, 21.3%) exhibited a high-risk “frailty–inflammation” phenotype, characterized by impaired activities of daily living, elevated risk of falls and pressure ulcers, increased high-sensitivity C-reactive protein (hs-CRP) levels, and poor health behaviors. Compared with Cluster 1, patients in Cluster 2 showed a higher hazard of 1-year MACE (unadjusted HR = 4.17; 95% CI: 3.237–5.371; P < 0.001) and a markedly increased risk of all-cause mortality (HR = 12.162; 95% CI: 7.445–19.868; P < 0.001). After adjustment using the minimal sufficient confounding model (M3, incorporating 17 prespecified confounders), Cluster 2 membership remained significantly associated with an increased risk of 1-year MACE (adjusted HR = 2.726; 95% CI 1.986–3.743; P < 0.001) and all-cause mortality (adjusted HR = 4.35; 95% CI 2.393–7.905; P < 0.001). In multivariable Cox regression analysis, Cluster 2 membership was identified as an independent predictor for MACE.
Conclusions: Integration of multidimensional data using machine learning identifies a high-risk “frailty–inflammation” phenotype in patients with AMI. This phenotype offers a comprehensive biological–functional–behavioral framework for risk stratification and may support early and precise clinical interventions to reduce long-term adverse outcomes. However, despite its potential clinical value, the current evidence is insufficient to support direct translation into routine clinical practice until independent cohort validation is performed. Future multi-center studies are warranted to confirm the robustness and generalizability of this approach, and to prospectively evaluate whether phenotype-guided intervention strategies can improve long-term outcomes.
Related
- Cluster-Based Phenotyping in an HFpEF Population Identifies a Biomarker-High, Atrial-Dysfunction Phenotype with Impaired Quality of Life, Reduced Exercise Capacity and Higher Healthcare Utilization
- Passive Smartphone-Based Digital Phenotyping for the Identification of PTSD Symptoms: Initial Validation of a Digital Behavioral Marker
- Postoperative malnutrition predicts severe chemotherapy toxicity after gastrectomy: evaluation of an inflammation-anemia phenotype
- Eosinophilic and type 2 inflammation in bronchiectasis: from phenotype recognition toward endotype-informed therapy
- Continuous Glucose Monitoring for Metabolic Phenotyping and Precision Nutrition: Current Evidence, Clinical Applications and Future Perspectives
BibTeX
@article{zhao2026phenotyping,
title = {{Phenotyping of AMI patients using machine learning: integration of clinical, metabolic, and behavioral features to identify high-risk “frailty–inflammation” phenotype}},
author = {Yisha Zhao and Wendi Ding and Hongyi Li and Songqi Guo and Yuying Fan},
journal = {Frontiers in cardiovascular medicine},
year = {2026},
month = oct,
doi = {10.3389/fcvm.2026.1881280},
url = {https://europepmc.org/article/PMC/PMC13690802},
note = {CC BY},
}
RIS
TY - JOUR
TI - Phenotyping of AMI patients using machine learning: integration of clinical, metabolic, and behavioral features to identify high-risk “frailty–inflammation” phenotype
AU - Yisha Zhao
AU - Wendi Ding
AU - Hongyi Li
AU - Songqi Guo
AU - Yuying Fan
PY - 2026
DA - 2026/10/09
JO - Frontiers in cardiovascular medicine
DO - 10.3389/fcvm.2026.1881280
UR - https://europepmc.org/article/PMC/PMC13690802
AB - Background: Prognosis among patients with acute myocardial infarction (AMI) shows substantial heterogeneity, making precise identification of high-risk phenotypes essential for improving long-term outcomes. Machine learning techniques were used to identify such phenotypes by integrating multidimensional data and to examine their association with major adverse cardiovascular events (MACE). Methods: A total of 1,139 patients with AMI were enrolled from the First Affiliated Hospital of Harbin Medical University between March 2023 and June 2024. Multidimensional data, including clinical features, metabolic characteristics, and behavioral patterns, were collected. Phenotyping was performed using a hierarchical clustering algorithm. Associations between derived phenotypes and 1-year MACE were evaluated using Kaplan–Meier survival analysis and multivariable Cox regression models. Shapley Additive Explanations (SHAP) values derived from an XGBoost model were used to quantify the contributions of individual variables to cluster assignment. Results: Hierarchical clustering identified two clinical phenotypes with differing pathophysiological characteristics. Cluster 2 ( n = 243, 21.3%) exhibited a high-risk “frailty–inflammation” phenotype, characterized by impaired activities of daily living, elevated risk of falls and pressure ulcers, increased high-sensitivity C-reactive protein (hs-CRP) levels, and poor health behaviors. Compared with Cluster 1, patients in Cluster 2 showed a higher hazard of 1-year MACE (unadjusted HR = 4.17; 95% CI: 3.237–5.371; P < 0.001) and a markedly increased risk of all-cause mortality (HR = 12.162; 95% CI: 7.445–19.868; P < 0.001). After adjustment using the minimal sufficient confounding model (M3, incorporating 17 prespecified confounders), Cluster 2 membership remained significantly associated with an increased risk of 1-year MACE (adjusted HR = 2.726; 95% CI 1.986–3.743; P < 0.001) and all-cause mortality (adjusted HR = 4.35; 95% CI 2.393–7.905; P < 0.001). In multivariable Cox regression analysis, Cluster 2 membership was identified as an independent predictor for MACE. Conclusions: Integration of multidimensional data using machine learning identifies a high-risk “frailty–inflammation” phenotype in patients with AMI. This phenotype offers a comprehensive biological–functional–behavioral framework for risk stratification and may support early and precise clinical interventions to reduce long-term adverse outcomes. However, despite its potential clinical value, the current evidence is insufficient to support direct translation into routine clinical practice until independent cohort validation is performed. Future multi-center studies are warranted to confirm the robustness and generalizability of this approach, and to prospectively evaluate whether phenotype-guided intervention strategies can improve long-term outcomes.
SN - 2297-055X
ER -