---
title: "Frailty prediction in heart failure with acute infection: role of thiazide diuretics and an interp"
id: "plos-one-9-frailty-prediction-in-heart-failure-patients-with-acute-infections-the"
canonical_url: "https://medichelpline.com/clinical-feed/plos-one-9-frailty-prediction-in-heart-failure-patients-with-acute-infections-the"
content_type: "clinical_feed_article"
specialty: "Cardiology"
source_name: "PLOS ONE (Medicine)"
source_url: "https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0355848"
published_at: "2026-08-12T14:00:00.000Z"
evidence_level: "Journal Feed"
license: "CC-BY-NC-4.0 / Informational Use"
---
# Frailty prediction in heart failure with acute infection: role of thiazide diuretics and an interp
## Provenance & Clinical Metadata
- **Canonical URL:** https://medichelpline.com/clinical-feed/plos-one-9-frailty-prediction-in-heart-failure-patients-with-acute-infections-the
- **Specialty:** [Cardiology](https://medichelpline.com/clinical-feed/cardiology.md)
- **Primary Source:** PLOS ONE (Medicine)
- **Source URL:** [Original Journal Publication](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0355848)
- **Published At:** 2026-08-12T14:00:00.000Z
- **Evidence Rating:** Journal Feed
## Executive GIST (TL;DR)
- This study analyzed 1,498 hospitalized patients aged ≥65 years with heart failure (HF) and acute infections from Nanjing First Hospital (2023) to identify predictors of frailty and build an interpretable prediction model. - Frailty prevalence in the cohort was high (80.3%) based on the Clinical Frailty Scale (CFS). - Candidate predictors were screened using univariate analysis and LASSO regression; selected variables included medication, laboratory, functional, demographic, and cardiac measures. - Key independent predictors associated with frailty probability included use of **thiazide diuretics** (linked to lower frailty probability), serum **albumin**, estimated glomerular filtration rate (**eGFR**), lymphocyte percentage, mean corpuscular hemoglobin concentration (MCHC), capacity for action, age, left ventricular ejection fraction (LVEF), NYHA functional class, history of cerebral infarction, and smoking. - Eight machine learning algorithms were trained and compared; **XGBoost** performed best with AUROC 0.872 and AUPRC 0.969 on the test set. - Model interpretability was provided using the **SHAP** (SHapley Additive exPlanations) framework to quantify each feature’s contribution to individual predictions. - An online calculator was created as a proof-of-concept for real-time risk estimation and potential clinical use. - Study design: retrospective cohort from electronic records, inclusion: age ≥65, HF per Chinese Guidelines 2024 and evidence of acute infection; exclusions included incomplete records and certain severe comorbidities. - Ethical approval was obtained (Nanjing First Hospital Ethics Committee No. KY20250120-KS-03); consent was waived due to retrospective design. - Data underlying the findings are publicly available via Figshare (DOI provided in the paper).
## Clinical Analysis & Structured Key Points
Frailty prediction in heart failure patients with acute infections: the potential role of thiazide diuretics? | PLOS One Browse Subject Areas ? Click through the PLOS taxonomy to find articles in your field. For more information about PLOS Subject Areas, click here . Article Authors Metrics Comments Media Coverage Peer Review Reader Comments Figures Figures Abstract Background Frailty remains a significant risk factor for adverse health outcomes in hospitalized patients. Few have evaluated frailty risk and its influencing factors in heart failure (HF) patients with acute infections, and previous machine learning models have predominantly overlooked the incorporation of visualization techniques. This study aims to investigate frailty risk factors in this population and develop an interpretable prediction model for frailty. Methods This study enrolled 1498 patients hospitalized for HF with acute infections at Nanjing First Hospital in 2023. Participants were randomly divided into training and testing sets at a 7:3 ratio. Potential predictors were screened through univariate analysis and the least absolute shrinkage and selection operator (LASSO) regression. Eight machine learning algorithms were evaluated to determine the optimal predictive model. Model interpretability was enhanced using the SHapley Additive exPlanations (SHAP) method. Results Frailty was prevalent in 80.3% of the cohort. Key predictors included the use of thiazide diuretics, serum albumin, estimated glomerular filtration rate (eGFR), lymphocyte percentage, mean corpuscular hemoglobin concentration (MCHC), capacity for action, age, left ventricular ejection fraction (LVEF), New York Heart Association (NYHA) functional class, history of cerebral infarction, and smoking. Comparative analysis of the eight models revealed that eXtreme Gradient Boosting (XGBoost) achieved superior performance, with the highest area under the receiver operating characteristic curve (AUROC: 0.872) and precision-recall curve (AUPRC: 0.969). Conclusions This study identified the use of thiazide diuretics as an independent predictor associated with lower frailty probability. We developed an online calculator as a proof-of-concept tool to demonstrate the potential application of the predictive model and facilitate real-time risk estimation. Citation: Huang T, Liu S, Zhang S, Song X, Xu M, Wu H, et al. (2026) Frailty prediction in heart failure patients with acute infections: the potential role of thiazide diuretics? PLoS One 21(8): e0355848. https://doi.org/10.1371/journal.pone.0355848 Editor: Francesco Curcio, University of Naples Federico II, ITALY Received: April 28, 2026; Accepted: July 27, 2026; Published: August 12, 2026 Copyright: © 2026 Huang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: All data underlying the findings described in this manuscript are fully available without restriction. The minimal anonymised dataset necessary to replicate the study results has been deposited in the public repository Figshare, and can be accessed via the following DOI: 10.6084/m9.figshare.3311188 1. Funding: The author(s) declare that financial support was received for the research and/or publication of this article. This work was supported by the National Natural Science Foundation of China [82173899], the Jiangsu Pharmaceutical Association [H202108, A2021024, Q202202, JY202207, Z04JKM2023E040], and the Hospital Management Innovation Research Project of Jiangsu Provincial Hospital Association (JSYGY-3-2023-264). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. There was no additional external funding received for this study. Competing interests: The authors have declared that no competing interests exist. Abbreviations: HF, heart failure; CHARLS, China Health and Retirement Longitudinal Study; COPD, chronic obstructive pulmonary disease; NYHA functional class, New York Heart Association functional class; CFS, Clinical Frailty Scale; ADL, Activities of Daily Living; IADL, Instrumental Activities of Daily Living; BMI, body mass index; ARNI, angiotensin receptor neprilysin inhibitor; ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; SGLT-2 inhibitors, sodium-glucose cotransporter protein-2 inhibitors; MRA, mineralocorticoid receptor antagonist; CCB, calcium channel blocker; sGC stimulators, soluble guanylate cyclase stimulators; MCHC, mean corpuscular hemoglobin concentration; eGFR, estimated glomerular filtration rate; BNP, B-type natriuretic peptide; NT-proBNP, N-terminal pro-BNP; LVEF, left ventricular ejection fraction; AOD, aortic diameter; LAD, left atrial diameter; IVSD, interventricular septum diameter; LVEDD, left ventricular end diastolic diameter; LVPWD, left ventricular posterior wall diameter; LVESD, left ventricular end systolic diameter; KNN, K-nearest neighbors; LASSO, least absolute shrinkage and selection operator; VIF, variance inflation factor; AUROC, area under the receiver operating characteristic curve; AUPRC, area under the precision-recall curve; DCA, decision curve analysis; SHAP, SHapley Additive exPlanations; LR, logistic regression; RF, random forest; XGBoost, eXtreme Gradient Boosting; LightGBM, light gradient boosting machine; SVM, support vector machine; CatBoost, categorical boosting; NB, naive bayes; MLP, multilayer perceptron; RAAS, renin-angiotensin-aldosterone system; ATP, adenosine triphosphate Introduction Frailty, a multidimensional geriatric syndrome closely associated with aging, manifests as an excessive decline in reserve and function across multiple physiological systems, leading to diminished responsiveness to minor stressors [ 1 ]. It is highly prevalent among hospitalized patients and significantly increases the risk of adverse health outcomes, including hospitalization, falls, disability, and mortality [ 2 , 3 ]. It is noteworthy that heart failure (HF), similar to frailty, represents a complex clinical syndrome with a persistently increasing hospitalization prevalence, which has become a significant global public health concern [ 4 – 6 ]. However, previous research on frailty has primarily focused on stable chronic disease populations [ 7 ]. However, frailty-related risk factors in patients with coexisting acute and chronic conditions—a common yet understudied clinical scenario—remain largely unexplored. In HF with acute infection, the overlapping pathophysiology of fluid retention (HF) and fluid exudation (infection) may synergistically exacerbate fluid overload, manifesting as bilateral lower extremity edema and secondary muscle strength decline, thereby increasing the risk of prolonged bed rest and adverse outcomes [ 8 ]. These complications not only worsen prognosis but also impose substantial challenges to clinical management [ 9 , 10 ]. Moreover, existing frailty prediction models, whether based on conventional regression or machine learning approaches, have notable limitations. Logistic regression-based nomograms, while achieving acceptable discrimination (e.g., AUCs of 0.912 and 0.881 in diabetic patients [ 11 ]; AUC of 0.847 in elderly coronary heart disease patients [ 12 ]), often suffer from limited practical utility or small sample sizes. Machine learning-based models, such as an XGBoost model for chronic obstructive pulmonary disease (COPD) patients using China Health and Retirement Longitudinal Study (CHARLS) data (test AUC 0.942) [ 13 ], have demonstrated superior performance but frequently lack objective laboratory parameters. Moreover, most existing models overlook model interpretability—a critical barrier to clinical adoption, as clinicians require intuitive understanding of how individual predictors contribute to frailty risk at the patient level. SHAP (SHapley Additive exPlanations), a game theory-based approach, quantifies each feature’s marginal contribution to individual predictions, effectively bridging the gap between predictive accuracy and clinical transparency [ 14 , 15 ]. To solve these disadvantages, this study aimed to investigate frailty-associated risk factors in patients with concurrent HF and acute infections, and to develop a frailty prediction model utilizing multiple machine learning algorithms. Additionally, a clinically accessible online calculator was developed and deployed to accurately identify high-risk individuals and tailor appropriate interventions based on individual risk factors. Ultimately, this approach sought to reduce the incidence of frailty or even reverse its progression in affected patients. Methods Design and participants This study collected clinical data from patients with HF complicated by acute infections at Nanjing First Hospital between January 1 and December 31, 2023, and employed various machine learning algorithms to predict frailty in this patient population. The data used for this study were accessed for research purposes on 15/04/2025 from the hospital’s electronic medical record system. The inclusion criteria were as follows: (1) age ≥ 65 years; (2) diagnosed with HF according to the Chinese Guidelines for the Diagnosis and Treatment of Heart Failure 2024 [ 16 ]; (3) evidence of acute infections, including fever, tachycardia, elevated inflammatory markers, or imaging findings suggestive of infection. The exclusion criteria were as follows: (1) incomplete medical records; (2) uninfected patients; (3) comorbidities such as acute myocardial infarction, advanced malignancy, psychiatric disorders, or severe trauma; (4) New York Heart Association (NYHA) functional class I. This study was approved by the Ethics Committee of Nanjing First Hospital (Approval No. KY20250120-KS-03). As this was a retrospective study, the Ethics Committee waived the requirement for informed consent. This study was conducted by the ethical standards outlined in the Declaration of Helsinki. Frailty status assessment All patients were assessed for frailty using the Clinical Frailty Scale (CFS), with scores ranging from 1 (very fit) to 9 (terminally ill). According to the CFS scoring criteria, a score≤4 indicates a non-frail state, while a score≥5 is classified as frail [ 17 ]. The CFS assessment for all patients was performed by a nurse who received standardized training. To ensure consistency and standardization in the assessments, all participating nurses underwent specialized training before data collection. The training covered the theoretical background of the CFS, detailed definitions for each level of the scale, and specific assessment procedures. A senior geriatrician conducted the entire training process, assessment supervision, and quality control. The assessments strictly adhered to the internationally recognized CFS scoring form, which has been validated in Chinese populations [ 18 ]. Furthermore, an outcome-blinding design was implemented to mitigate potential bias. Nurses conducting CFS assessments were excluded from clinical decision-making for the patients. Assessments were finalized upon resolution of the acute infection phase and at the time of hospital discharge, with all results systematically documented in the electronic medical records. The CFS includes the Activities of Daily Living (ADL) scale (eating, bathing, grooming (toothbrushing, face washing, shaving, hair combing), dressing (fastening shoes, buttoning clothes), bowel control, bladder control and toilet use, transfers (bed-to-chair mobility), ambulation (walking 45 meters on level ground), stair climbing) and the Instrumental Activities of Daily Living (IADL) scale (telephone usage, shopping, meal preparation, housekeeping, laundry, transportation, medication management, financial handling). Clinical data collection All clinical data were extracted from the electronic medical record system, including: (1) demographic characteristics: age, sex, literacy, marital status, Body Mass Index (BMI), hospital days, capacity for action, smoking, drinking, and NYHA functional class; (2) medication use: angiotensin receptor neprilysin inhibitor (ARNI), angiotensin-converting enzyme inhibitor (ACEI) angiotensin receptor blocker (ARB), sodium-glucose cotransporter protein-2 (SGLT-2) inhibitors, beta receptor blockers, mineralocorticoid receptor antagonist (MRA), calcium channel blocker (CCB), loop diuretics, thiazide diuretics, nitrates, statins, antiplatelet drugs, anticoagulants, cardiac glycosides, soluble guanylate cyclase (sGC) stimulators; (3) comorbidities: hypertension, diabetes mellitus, hyperlipidemia, coronary heart disease, atrial fibrillation, fatty liver disease, cirrhosis, COPD, chronic cor pulmonale, renal insufficiency, osteoporosis, cerebral infarction, malignant tumor; (4) infection characteristics: pulmonary infection, urinary tract infection, abdominal infection, skin and soft tissue infection, bloodstream infection, sepsis, septic shock; (5) admission vital signs: first recorded blood pressure (systolic blood pressure, diastolic blood pressure) and heart rate; (6) biomarkers: white blood cell, lymphocyte percentage, neutrophil percentage, red blood cell, hemoglobin, mean corpuscular hemoglobin concentration (MCHC), platelet, D-dimer, alanine aminotransferase, aspartate aminotransferase, serum albumin, total bilirubin, urea, creatinine, uric acid, serum potassium, serum sodium, total cholesterol, triglycerides, high-density lipoprotein, low-density lipoprotein, estimated glomerular filtration rate (eGFR), interleukin-6, procalcitonin, c-reactive protein, B-type natriuretic peptide (BNP), N-terminal pro-BNP (NT-proBNP), elevated natriuretic peptide levels were defined as BNP>100pgmL or NT-proBNP>300pgmL [ 19 ]; (7) echocardiography: left ventricular ejection fraction (LVEF), aortic diameter (AOD), left atrial diameter (LAD), interventricular septum diameter (IVSD), left ventricular end diastolic diameter (LVEDD), left ventricular posterior wall diameter (LVPWD), left ventricular end systolic diameter (LVESD). Ethics statement This study was approved by the Ethics Committee of Nanjing First Hospital (Approval No. KY20250120-KS-03). As this was a retrospective study, the Ethics Committee waived the requirement for informed consent. This study was conducted by the ethical standards outlined in the Declaration of Helsinki. Sample size justification Prior to model development, we estimated the minimum required sample size using the events per variable (EPV) criterion [ 20 ]. Based on a review of similar prediction models in the literature [ 21 – 23 ] and our preliminary data, we anticipated that the final model might contain up to 15 predictors. Adhering to the EPV criterion with a threshold of at least 20 EPV, we required a minimum of 300 outcome events. In our training set, we observed 834 frail cases, yielding an actual EPV of 55.6 (834/15), which met our pre-specified requirement. Data preprocessing All analyses began with a rigorous cleaning protocol applied to the raw dataset. First, duplicate records were screened using unique participant identifiers; none were found. Next, potential outliers in variables were evaluated through manual screening of the original data. Any values deemed clinically unusual were cross-verified against original source records and discussed with relevant clinical staff to assess their validity. No observations were excluded based on this review process. Of the 81 variables included in the study, six contained missing values, each with a missingness rate below 18% ( S1 Table ). Missing data were imputed using the K-nearest neighbors (KNN) algorithm after splitting the dataset into training and testing sets. The optimal k value (k = 5) was determined by minimizing the root mean square error (RMSE) of imputation performance, assessed via internal cross-validation within the training set. The KNN imputation model was fitted exclusively on the training set, and the learned imputation strategy was subsequently applied to the testing set. The KNN algorithm estimates missing entries based on Euclidean distances in the feature space among the k most similar observations [ 24 ]. This approach preserves underlying data structure and inter-variable relationships while avoiding listwise deletion. KNN imputation was implemented using the KNNImputer class from the scikit-learn library (version 1.2.2). Data analysis The normality of continuous variables was assessed by the Shapiro-Wilk test: normally distributed variables were presented as mean ± SD, and group comparisons were made using independent samples t-tests; non-normally distributed variables were summarized as median (interquartile range) [M (IQR)], and the Mann-Whitney U test was used for group comparisons. Categorical variables were described as frequencies (percentages) [n (%)], and the chi-square test or Fisher’s exact test was selected for analysis of differences between groups according to sample size characteristics. All statistical tests were two-tailed, and P < 0.05 was considered statistically significant. Model development All analyses described in this section were performed in R (v4.3.3). The dataset was randomly split(set.seed(42)) into training and testing sets in a 7:3 ratio. The training set was used for variable selection and model development, while the testing set evaluated model performance. Continuous variables were standardized using Z-score normalization based on the mean and standard deviation calculated from the training dataset, and the same scaling parameters were subsequently applied to the testing dataset. Categorical variables were encoded via one-hot encoding [ 25 , 26 ]. All feature selection procedures were performed exclusively within the training dataset. Variables significantly associated with frailty in univariate analysis (p < 0.05) were retained for subsequent LASSO regression. Variables significantly associated with frailty (p < 0.05) in univariate analysis were retained for further analysis. The descriptive statistics and univariate comparisons were generated using the tableone package (version 0.13.2). Significant variables were further analyzed using least absolute shrinkage and selection operator (LASSO) regression, implemented in R with the glmnet package (version 4.1.8). LASSO applies L1 regularization to perform automatic feature selection by shrinking the coefficients of less informative predictors toward zero. The regularization parameter λ was tuned via 10-fold cross-validation on the training set, and the optimal λ was selected using the “1-standard-error” rule to favor a more parsimonious model while maintaining predictive performance [ 27 ]. Only variables with non-zero coefficients at the chosen λ were retained for subsequent modeling. To assess multicollinearity, variables with a variance inflation factor (VIF) ≥ 5 were excluded to ensure feature independence [ 28 ]. VIF was calculated in R using the car package (version 3.1.2). Given the high frailty prevalence (80.3%), all models that natively support class weighting—including logistic regression (LR), random forest (RF), eXtreme Gradient Boosting (XGBoost), light gradient boosting machines (LightGBM), support vector machine (SVM), categorical boosting (CatBoost) —were trained with automatic adjustment (class weight = ’balanced’ or equivalent) to reduce bias toward the majority class. Naive Bayes (NB) and multilayer perceptron (MLP), which do not support this functionality, were included as unadjusted benchmark models. Meanwhile, we prioritized AUROC and AUPRC as core evaluation metrics, as these indicators are more robust to skewed class distributions compared to raw accuracy, ensuring reliable assessment of the model’s discriminative power across both frail and non-frail subgroups. All analyses
## Related Clinical Research

- [Multimodal extraction of heart disease risk factors from free-text and structured records using Li](https://medichelpline.com/clinical-feed/plos-one-0-leveraging-free-text-clinical-records-for-heart-disease-classification-through.md)
- [Intensive blood pressure control linked to 15% lower 7-year dementia risk](https://medichelpline.com/clinical-feed/medical-news-today-0-intensive-blood-pressure-control-cuts-7-year-dementia-risk-by-15-study-finds.md)
- [Infection Does Not Impair Decongestion or Short-Term Kidney Outcomes in Cardiorenal Syndrome Type 1](https://medichelpline.com/clinical-feed/plos-one-11-impact-of-infection-on-decongestion-and-kidney-outcomes-in-patients-with.md)
- [Cardiometabolic pathways mediating the effect of educational attainment on cardiovascular disease:](https://medichelpline.com/clinical-feed/medrxiv-13-cardiometabolic-pathways-linking-genetically-proxied-educational-attainment-to.md)
- [RACOVIR registry: Oral anticoagulation in primary care patients with non‑valvular atrial fibrillat](https://medichelpline.com/clinical-feed/medrxiv-1-oral-anticoagulation-registry-in-patients-with-atrial-fibrillation-treated-in.md)

## Navigation
- [← Back to Cardiology Feed](https://medichelpline.com/clinical-feed/cardiology.md)
- [← All Clinical Specialties](https://medichelpline.com/clinical-feed.md)
## Medical & Regulatory Disclaimer

> [!CAUTION]
> MedicHelpline content is structured for research, educational, and professional discovery purposes. It does not constitute individual medical advice, clinical diagnosis, or treatment recommendations.
> Always verify dosing, contraindications, and regulatory alerts against official product labeling and primary regulatory sources before clinical decision-making.