Abstract
Background
The prognostic impact of red blood cell distribution width (RDW) on adverse outcomes in patients with heart failure with preserved ejection fraction (HFpEF) has been established. However, its non-linear relationship and incremental prognostic value in critically ill HFpEF patients remain insufficiently explored.
Methods
A retrospective cohort study was conducted involving 3802 HFpEF patients from the Medical Information Mart for Intensive Care IV (MIMIC-IV, v3.1) database. Multivariable Cox regression and restricted cubic spline (RCS) analyses were performed to evaluate the association between RDW and 28-day mortality, adjusting for potential confounders. A prediction model was constructed by the least absolute shrinkage and selection operator (Lasso)-Cox regression, and receiver operating characteristic (ROC) curve was utilized to assessed the incremental prognostic effect of RDW on 28-day mortality prediction.
Results
Within the final cohort of 1962 HFpEF patients, the incidence of 28-day mortality showed a significant increase across RDW tertiles (T1 9.4% vs T2 13.1% vs T3 22.9%, P < 0.001). The Kaplan-Meier analysis demonstrated significantly higher all-cause mortality risk among patients with higher RDW levels (log-rank P < 0.001). Multivariable Cox regression analysis revealed that RDW levels were independently correlated with all-cause mortality after adjusting for multiple variables for both continuous (HR 1.13, 95% CI: 1.07–1.19; P < 0.001) and categorical comparisons (T3 vs T1: HR 1.66, 95% CI: 1.20–2.29; P = 0.002). RCS analysis indicated a significant linear relationship between RDW levels and the risk of mortality (P < 0.001 for overall association; P = 0.222 for non-linearity). When RDW was incorporated into the basic prediction model constructed by Lasso-Cox regression (AUC = 0.814), it slightly improved the predictive ability for 28-day mortality (AUC = 0.819), as evidenced by an IDI of 0.017 (95% CI: 0.006–0.036) and an NRI of 0.140 (95% CI: 0.041–0.203).
Conclusions
This study established a significant linear association between RDW at admission and poor prognosis in critically ill patients with HFpEF. Higher RDW levels were associated with increased 28-day mortality risk, suggesting that it may be a relevant biomarker for mortality risk stratification in HFpEF patients.
Keywords: Heart failure with preserved ejection fraction (HFpEF), Red blood cell distribution width (RDW), All-cause mortality, Incremental prognostic effect
1. Introduction
Heart failure with preserved ejection fraction (HFpEF) is a distinct type of heart failure characterized by the presence of heart failure symptoms or signs and elevated levels of natriuretic peptides, despite a left ventricular ejection fraction (LVEF) within the normal range (typically >50%) [1], [2]. Epidemiological surveys have revealed that HFpEF accounts for over 50% of all heart failure cases, with a higher prevalence among women, and its incidence continues to rise [3]. The pathogenesis of HFpEF is highly complex, often involving comorbidities such as hypertension, obesity, diabetes mellitus, and renal dysfunction [4], [5]. Traditional management strategies targeting aberrant activation of the sympathetic nervous and neurohumoral systems, including beta-blockers and renin-angiotensin system inhibitors (RASI), have failed to demonstrate significant clinical benefits in patients with HFpEF [3]. Therefore, it is clinically valuable to stratify the risk for prognosis separately in these types of heart failure.
Red blood cell distribution width (RDW) quantifies the heterogeneity in circulating erythrocyte size and, combined with mean corpuscular volume, serves as a key parameter for the differential diagnosis and classification of anemia [6]. Elevations in RDW may reflect impaired erythropoiesis resulting from deficiencies in essential substrates (e.g., vitamin B12, iron, folate), increased erythrocyte destruction, blood transfusion, or underlying pathologies including hepatic disease, malnutrition, cancer, and metastatic bone marrow tumors [7]. Recent studies have reported strong association between RDW and various disorders, such as cardiovascular diseases, malignancy, and infectious diseases [8], [9], [10], [11]. Emerging evidence further indicated that elevated RDW predicts adverse outcomes in HFpEF cohorts [12], [13]. However, whether RDW exhibits a non-linear dose–response relationship with mortality and its incremental prognostic value beyond conventional risk factors in critically ill HFpEF patients remain insufficiently explored. In this study, we undertook an investigation into the association between RDW levels and adverse clinical outcomes in critically ill patients with HFpEF admitted to the intensive care unit (ICU).
2. Methods
2.1. Study population
This was a retrospective cohort study based on the Medical Information Mart for Intensive Care IV (MIMIC-IV, version 3.1) database. Maintained by Beth Israel Deaconess Medical Center (BIDMC) in Boston, Massachusetts, this openly available, de-identified repository provides extensive patient demographics, laboratory results, medication histories, and comprehensive clinical documentation, serving as a vital asset for detailed healthcare investigations [14]. Author Jing Xiao gained full authorization to the MIMIC-IV resource and retrieved pertinent study data. Ethical oversight by the BIDMC Institutional Review Board approved MIMIC-IV utilization; individual consent was waived due to the dataset's strict compliance with Health Insurance Portability and Accountability Act (HIPAA) standards. Our study was reported adhering to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for observational research [15].
Study inclusion criteria (Fig. 1) comprised: 1) Adults aged 18 years or older. 2) Patients diagnosed with HFpEF, ascertained via International Classification of Diseases (ICD) codes as previously reported [16]: ICD-9 code 42830, 42831, 42832, 42833 and ICD-10 code I503, I5030, I5031, I5032, 15033; 3) First time ICU admission. Exclusion criteria encompassed: 1) RDW Missed or outliers; 2) lacking of BMI values; 3) or patients with missing data (<5% of the total sample sizes). Following these criteria, 1962 patients constituted the analytical cohort. Patients were subsequently stratified into RDW tertiles: T1 (<14.2), T2 (14.2–15.8), and T3 (>15.8). Data extraction employed Structured Query Language (SQL). To mitigate selection bias, only the first ICU admission was analyzed for individuals with multiple hospitalizations.
Fig. 1.
The flowchart of this study.
2.2. Study variables
To mitigate confounding effects from post-admission therapeutic interventions, the first available blood measurements following ICU admission were utilized, except for ICU first glucose, which was replaced by the ICU mean glucose owing to extensive missing values. The covariables extracted for analysis included demographics (age, gender, race and body mass index [BMI]); clinical comorbidities (myocardial infarction [MI], hypertension, diabetes mellitus [DM], cerebrovascular disease, peripheral vascular disease, chronic pulmonary disease [CPD], and chronic kidney disease [CKD]; initial ICU laboratory results (hemoglobin, red blood cell count [RBC], white blood cell count [WBC], platelet count, RDW, creatinine, blood urea nitrogen [BUN], albumin, glucose, sodium, chloride, potassium, International Normalized Ratio [INR] an prothrombin time [PT]); medications (aspirin, clopidogrel, beta blocker, angiotensin-converting enzyme inhibitor [ACEI]/angiotensin-receptor blocker [ARB], diuretic, heparin, warfarin, amiodarone, digitalis, calcium channel blockers [CCB], and statin); vital signs (systolic and diastolic blood pressure [SBP/DBP], oxygen saturation [SpO₂], respiratory rate [RR], and temperature); and clinical severity indexes, including the Systemic Inflammatory Response Syndrome (SIRS) score, Sequential Organ Failure Assessment (SOFA) score, Glasgow Coma Scale (GCS) score, Acute physiology score III (APS III) and Simplified Acute Physiology Score II (SAPS II).
The primary endpoint of the study was all-cause mortality within 28 days following ICU admission.
2.3. Statistical methods
Missing data were handled using multiple imputation by chained equations (MICE) under the missing-at-random (MAR) assumption. We generated 5 imputed datasets including all candidate variables regardless of the proportion of missingness. Estimates from each imputed dataset were pooled using Rubin's rules [17], [18].
Continuous variables are shown as the mean and standard deviation or median and interquartile range and were compared with one-way analysis of variance or Kruskal-Wallis one-way analysis of variance according to the distribution of data. Categorical variables, which are expressed as counts and percentages, were compared using chi-square testing. Kaplan-Meier curve was used to explore the cumulative incidence of mortality across RDW tertiles within 28 days post-ICU admission.
Multivariable Cox proportional hazards models were applied to examine the association between RDW levels and mortality both continuously and categorically. The proportional hazards assumption was assessed using Schoenfeld residuals test. No significant violation was observed, as shown in Supplementary Fig. 2. The selection of confounders was based on clinical relevance and existing scientific literature [12], [13]. Four models were constructed: Model 1 adjusted for age, sex, race, and BMI; Model 2 further included HR, Spo2, Temperaturef, WBC, BUN, creatinine, glucose, lactate, potassium, chloride, INR; Model 3 further adjusted for SIRI, SOFA, APSIII, and comorbidities including hypertension, diabetes, CKD, CPD; Model 4 further adjusted for medication during ICU admission, including aspirin, clopidogrel, beta blocker, ACEI/ARB, amiodarone, statin. Multicollinearity was quantified via variance inflation factors (VIF), where values exceeding 5 indicated substantive collinearity. To explore potential nonlinear dose-response relationships between RDW and all-cause mortality, restricted cubic spline (RCS) regression was executed specifying three knots at the 10th, 50th and 95th percentiles.
Additionally, we further developed a basic prognostic model by the Least absolute shrinkage and selection operator (Lasso)-Cox regression, which incorporated age, race, BUN, lactate, potassium, chloride, SOFA, APS III, therapy with aspirin, beta blockers, and ACEI/ARB. LASSO was used solely for variable selection and shrinkage, and the internal validation of the prediction model was performed using 500 bootstrap resamples. The optimism-corrected C-statistic, calibration slope, and Brier score were calculated. Receiver operating characteristic (ROC) curve was utilized to assessed the incremental prognostic effect of RDW on 28-day mortality prediction in HFpEF patients. Besides, the continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were carried out to evaluate the reclassification [19].
The R Studio (Version 4.2.2) and the Free Statistics (Version 2.3 beta, Beijing, China) were utilized for all statistical analyses. Cox regression results are reported as hazards ratios (HRs) with 95% confidence intervals (CIs). Statistical significance was defined by a two-tailed P-value <0.05.
3. Results
3.1. Baseline characteristics
This study analyzed 1962 critically ill HFpEF patients categorized by RDW tertiles (T1: <14.2%; T2: 14.2–15.8%; T3: >15.8%). The baseline characteristics across different RDW tertiles was shown in Table 1. Patients in higher RDW tertiles were older and exhibited lower systolic, diastolic, and mean blood pressures. Coagulation profiles worsened with increasing RDW (higher INR, PT, APTT). Renal dysfunction (BUN, creatinine) and illness severity scores (SAPS II, APS III, SOFA) progressively increased across RDW tertiles. Comorbidities including hypertension, diabetes, chronic pulmonary disease, and CKD were more prevalent in higher RDW groups. Medication use varied, with lower aspirin and ACEI/ARB but higher digitalis and amiodarone use in T3. Mortality rates demonstrated a pronounced gradient across all time points (28-day T1 9.4% vs T3 22.9%; P < 0.001), with longer hospital and ICU stays in the highest tertile.
Table 1.
Baseline characteristics of patients grouped according to RDW tertiles.
| Variables | Total (n = 1962) | T1 (n = 615) | T2 (n = 670) | T3 (n = 677) | P value |
|---|---|---|---|---|---|
| Age, year | 74.3 ± 12.7 | 72.9 ± 13.2 | 75.2 ± 12.6 | 74.6 ± 12.2 | 0.002 |
| Sex, n (%) | 0.050 | ||||
| Female | 1005 (51.2) | 293 (47.6) | 365 (54.5) | 347 (51.3) | |
| Male | 957 (48.8) | 322 (52.4) | 305 (45.5) | 330 (48.7) | |
| Race, n (%) | 0.072 | ||||
| White | 1369 (69.8) | 443 (72) | 470 (70.1) | 456 (67.4) | |
| Non-white | 343 (17.5) | 87 (14.1) | 124 (18.5) | 132 (19.5) | |
| Unknown | 250 (12.7) | 85 (13.8) | 76 (11.3) | 89 (13.1) | |
| BMI, kg/m2 | 31.3 ± 9.3 | 30.9 ± 8.5 | 31.2 ± 9.3 | 31.6 ± 9.8 | 0.439 |
| Vital signs | |||||
| SBP, mmHg | 117.3 ± 15.2 | 118.6 ± 16.0 | 117.8 ± 14.8 | 115.8 ± 14.7 | 0.003 |
| DBP, mmHg | 58.2 ± 10.2 | 59.4 ± 10.5 | 57.9 ± 10.0 | 57.4 ± 10.1 | 0.001 |
| MBP, mmHg | 75.0 ± 9.7 | 76.4 ± 9.8 | 75.0 ± 9.4 | 73.8 ± 9.7 | <0.001 |
| HR, beats/min | 82.7 ± 15.2 | 81.7 ± 14.6 | 82.0 ± 14.7 | 84.4 ± 16.0 | 0.002 |
| RR, times/min | 19.7 ± 3.7 | 19.4 ± 3.6 | 19.7 ± 3.6 | 19.9 ± 3.7 | 0.036 |
| SpO2, % | 96.8 ± 2.1 | 96.9 ± 2.0 | 96.7 ± 2.1 | 96.8 ± 2.2 | 0.350 |
| Temperature, °C | 36.8 ± 0.5 | 36.8 ± 0.5 | 36.8 ± 0.5 | 36.7 ± 0.5 | 0.041 |
| Laboratory data | |||||
| RBC, K/μL | 3.4 ± 0.7 | 3.5 ± 0.7 | 3.4 ± 0.7 | 3.3 ± 0.8 | <0.001 |
| WBC, K/μL | 12.9 ± 8.2 | 12.7 ± 5.7 | 13.0 ± 7.6 | 13.0 ± 10.4 | 0.797 |
| Hb, g/dL | 10.1 ± 2.1 | 10.7 ± 2.0 | 10.1 ± 2.0 | 9.5 ± 2.1 | <0.001 |
| Platelet, K/μL | 197.5 ± 92.6 | 191.1 ± 80.0 | 201.7 ± 91.1 | 199.2 ± 103.8 | 0.100 |
| RDW, % | 15.4 ± 2.1 | 13.4 ± 0.6 | 14.9 ± 0.4 | 17.7 ± 1.8 | <0.001 |
| BUN, mg/dL | 26.0 (17.0, 41.0) | 20.0 (15.0, 31.0) | 26.0 (18.0, 38.0) | 32.0 (20.0, 51.0) | <0.001 |
| Creatinine, mg/dL | 1.2 (0.9, 1.8) | 1.0 (0.8, 1.3) | 1.2 (0.9, 1.7) | 1.4 (1.0, 2.3) | <0.001 |
| Sodium, mEq/L | 138.6 ± 4.9 | 138.4 ± 4.4 | 138.5 ± 4.9 | 138.8 ± 5.4 | 0.355 |
| Potassium, mEq/L | 4.3 ± 0.7 | 4.3 ± 0.6 | 4.3 ± 0.7 | 4.4 ± 0.7 | 0.013 |
| Chloride, mEq/L | 103.6 ± 6.8 | 104.5 ± 6.4 | 103.6 ± 6.7 | 102.7 ± 7.2 | <0.001 |
| Glucose, g/dL | 145.4 ± 48.3 | 146.0 ± 48.5 | 145.7 ± 48.9 | 144.7 ± 47.6 | 0.888 |
| Lactate, mmol/L | 1.7 (1.2, 2.5) | 1.7 (1.2, 2.5) | 1.8 (1.2, 2.5) | 1.7 (1.3, 2.5) | 0.942 |
| INR | 1.6 ± 1.0 | 1.5 ± 1.0 | 1.6 ± 1.0 | 1.7 ± 1.1 | <0.001 |
| PT, s | 17.4 ± 10.2 | 15.8 ± 8.5 | 17.4 ± 10.1 | 18.9 ± 11.4 | <0.001 |
| APTT, s | 40.7 ± 26.3 | 38.7 ± 25.1 | 40.6 ± 25.6 | 42.6 ± 27.8 | 0.030 |
| SIRS | 2.6 ± 0.9 | 2.6 ± 0.9 | 2.6 ± 0.9 | 2.6 ± 0.9 | 0.197 |
| SAPS II | 41.9 ± 13.1 | 38.9 ± 12.7 | 42.4 ± 12.9 | 44.0 ± 13.1 | <0.001 |
| APS III | 48.9 ± 19.2 | 44.1 ± 18.3 | 48.8 ± 19.2 | 53.4 ± 18.9 | <0.001 |
| SOFA | 5.7 ± 3.4 | 5.1 ± 3.1 | 5.8 ± 3.4 | 6.3 ± 3.4 | <0.001 |
| GCS | 14.3 ± 2.2 | 14.4 ± 2.3 | 14.3 ± 2.3 | 14.4 ± 2.0 | 0.887 |
| Past history, n (%) | |||||
| Myocardial infarct | 473 (24.1) | 158 (25.7) | 172 (25.7) | 143 (21.1) | 0.081 |
| Hypertension | 583 (29.7) | 198 (32.2) | 223 (33.3) | 162 (23.9) | <0.001 |
| Diabetes | 838 (42.7) | 218 (35.4) | 288 (43.0) | 332 (49.0) | <0.001 |
| Cerebrovascular disease | 247 (12.6) | 90 (14.6) | 75 (11.2) | 82 (12.1) | 0.160 |
| Peripheral vascular disease | 299 (15.2) | 89 (14.5) | 100 (14.9) | 110 (16.2) | 0.649 |
| Chronic pulmonary disease | 858 (43.7) | 226 (36.7) | 303 (45.2) | 329 (48.6) | <0.001 |
| Chronic kidney disease | 786 (40.1) | 172 (28) | 273 (40.7) | 341 (50.4) | <0.001 |
| Medication, n (%) | |||||
| Aspirin | 1392 (70.9) | 490 (79.7) | 482 (71.9) | 420 (62) | <0.001 |
| Clopidogrel | 332 (16.9) | 120 (19.5) | 116 (17.3) | 96 (14.2) | 0.036 |
| Beta blocker | 1583 (80.7) | 507 (82.4) | 549 (81.9) | 527 (77.8) | 0.067 |
| Diuretic | 1765 (90.0) | 555 (90.2) | 613 (91.5) | 597 (88.2) | 0.125 |
| ACEI/ARB | 718 (36.6) | 267 (43.4) | 253 (37.8) | 198 (29.2) | <0.001 |
| Heparin | 1711 (87.2) | 523 (85) | 587 (87.6) | 601 (88.8) | 0.124 |
| Warfarin | 690 (35.2) | 199 (32.4) | 246 (36.7) | 245 (36.2) | 0.208 |
| Amiodarone | 442 (22.5) | 156 (25.4) | 153 (22.8) | 133 (19.6) | 0.047 |
| Digitalis | 186 (9.5) | 47 (7.6) | 59 (8.8) | 80 (11.8) | 0.029 |
| CCB | 741 (37.8) | 230 (37.4) | 265 (39.6) | 246 (36.3) | 0.465 |
| Statin | 1273 (64.9) | 391 (63.6) | 464 (69.3) | 418 (61.7) | 0.011 |
| Mortality, n (%) | |||||
| 28-day mortality | 301 (15.3) | 58 (9.4) | 88 (13.1) | 155 (22.9) | <0.001 |
| 90-day mortality | 441 (22.5) | 86 (14) | 138 (20.6) | 217 (32.1) | <0.001 |
| 180-day mortality | 533 (27.2) | 103 (16.7) | 176 (26.3) | 254 (37.5) | <0.001 |
| 365-day mortality | 638 (32.5) | 126 (20.5) | 217 (32.4) | 295 (43.6) | <0.001 |
| LOS in hospital, days | 9.9 (6.4, 15.1) | 8.7 (5.9, 13.4) | 9.9 (6.3, 14.8) | 11.2 (7.0, 18.0) | <0.001 |
| LOS in ICU, days | 3.4 (2.0, 6.2) | 3.0 (1.7, 5.3) | 3.3 (2.0, 6.8) | 3.9 (2.2, 7.0) | <0.001 |
Continuous variables are presented as mean ± SD if normally distributed, and median (interquartile range) if not normally distributed. Categorical variables are presented as counts and percentage (%). BMI, body mass index; MI, myocardial infarction; CPD, chronic pulmonary disease; CKD, chronic kidney disease; SBP, systolic blood pressure; DBP, diastolic blood pressure; MBP, mean blood pressure; HR, heart rate; RR, respiratory rate; SpO2, oxygen saturation; BUN, urea nitrogen; PT, prothrombin time; APTT, activated partial prothrombin time; WBC, white blood cell; RBC, red blood cell; RDW, red cell distribution width; GCS, Glasgow coma scale; SOFA, sequential organ failure assessment; SAPS II, simplified acute physiology score; LOS, length of stay.
3.2. Survival analysis and Cox proportional-hazards regression model
The Kaplan-Meier survival curves indicated that patients in the higher RDW tertiles had significantly higher 28-day mortality than those in the lower RDW group (log rank P < 0.001, Fig. 2). In unadjusted analyses, each unit increase in RDW (continuous) was associated with a 22% higher mortality risk (HR 1.22, 95% CI 1.17–1.27, P < 0.001). Patients in the highest RDW tertile (T3) had a substantially elevated mortality risk compared to the lowest tertile (T1) (HR 2.62, 95% CI 1.94–3.54, P < 0.001). To identify confounders for inclusion in the multivariable regression model, we performed univariable Cox regression analysis and generated a heatmap illustrating the correlation between RDW and other variables, as detailed in Supplementary Table 1 and Supplementary Fig. 1. This association persisted robustly for RDW as a continuous variable across sequentially adjusted models (Model 4 HR 1.13, 95% CI 1.07–1.19, P < 0.001). When analyzed categorically, the significantly higher risk associated with the highest RDW tertile (T3) remained evident even after extensive multivariable adjustment for demographics, vital signs, laboratory parameters, disease severity scores, comorbidities, and medications (Model 4 HR 1.66, 95% CI 1.20–2.29, P = 0.002). In contrast, the association for middle tertile (T2) became non-significant after adjustment (Model 4 HR 1.12, 95% CI 0.80–1.58, P = 0.510), as detailed in Table 2.
Fig. 2.
Kaplan–Meier survival curve analysis for all-cause mortality.
Table 2.
Cox regression results showing HR for mortality across different RDW tertiles.
| Model | RDW (continuous) |
RDW T1 (n = 615) |
RDW T2 (n = 670) |
RDW T3 (n = 677) |
|||
|---|---|---|---|---|---|---|---|
| HR (95% CI) | P value | Reference | HR (95% CI) | P value | HR (95% CI) | P value | |
| Unadjusted model | 1.22 (1.17–1.27) | <0.001 | Reference | 1.41 (1.01–1.97) | 0.042 | 2.62 (1.94–3.54) | <0.001 |
| Model 1 | 1.22 (1.17–1.28) | <0.001 | Reference | 1.35 (0.97–1.88) | 0.079 | 2.55 (1.88–3.45) | <0.001 |
| Model 2 | 1.16 (1.10–1.21) | <0.001 | Reference | 1.22 (0.87–1.70) | 0.252 | 1.87 (1.37–2.57) | <0.001 |
| Model 3 | 1.15 (1.10–1.21) | <0.001 | Reference | 1.11 (0.79–1.56) | 0.544 | 1.80 (1.31–2.47) | <0.001 |
| Model 4 | 1.13 (1.07–1.19) | <0.001 | Reference | 1.12 (0.80–1.58) | 0.510 | 1.66 (1.20–2.29) | 0.002 |
Unadjusted model: includes no covariates.
Model 1: adjusted for age, sex, race, and BMI.
Model 2: further included HR, Spo2, Temperaturef, WBC, BUN, creatinine, glucose, lactate, potassium, chloride, INR.
Model 3: further adjusted for SIRI, SOFA, APSIII, and comorbidities including hypertension, diabetes, CKD, CPD.
Model 4: further adjusted for medication during ICU admission, including aspirin, clopidogrel, beta blocker, ACEI/ARB, amiodarone, statin.
3.3. Restricted cubic spline (RCS) analysis
RCS analysis revealed a significant linear association between RDW and 28-day mortality risk (P < 0.001 for overall association; P = 0.222 for non-linearity), with higher RDW conferring a monotonically increased risk of death (Fig. 3).
Fig. 3.
RCS curve analysis showing the non-linear relationship between RDW levels and 28-day mortality risk.
3.4. Predictive performance of the RDW
A basic prognostic model was constructed using Lasso-Cox regression, incorporating age, race, BUN, lactate, potassium, chloride, SOFA score, APSIII score, and use of aspirin, beta-blockers, and ACEI/ARBs (Supplementary Fig. 3). The nomogram, C-statistics, calibration curve, and clinical decision curve for the prediction model was presented in Supplementary Figs. 4–5. The apparent C-statistic of the basic model was 0.814, with an optimism-corrected estimate of 0.809. After adding RDW, it yielded a modest but statistically significant improvement in predictive accuracy for 28-day mortality (AUC = 0.819). Furthermore, the RDW-inclusive model demonstrated significantly improved reclassification compared with the baseline model, as evidenced by an IDI of 0.017 (95% CI: 0.006–0.036; P = 0.008) and an NRI of 0.140 (95% CI: 0.041–0.203; P = 0.012) (Fig. 4; Table 3).
Fig. 4.
Comparison of the ROC curves between the RDW and RDW plus the basic model in predicting all-cause mortality. The basic model was constructed using Lasso-Cox regression incorporating age, race, BUN, lactate, potassium, chloride, SOFA, APSIII, aspirin, beta blocker, ACEI/ARB. RDW was incorporated as a continuous linear term.
Table 3.
Discrimination and reclassification of the adding the RDW to the basic model in predicting all-cause mortality.
| Basic model | IDI |
P value | NRI (95% CI) |
P value |
|---|---|---|---|---|
| Reference | Reference | |||
| +RDW | 0.017 (0.006–0.036) | 0.008 | 0.140 (0.041–0.203) | 0.012 |
NRI, net reclassification improvement; IDI, integrated discrimination improvement. Basic model included age, race, BUN, lactate, potassium, chloride, SOFA, APSIII, aspirin, beta blocker, ACEI/ARB.
4. Discussion
Utilizing the MIMIC-IV database, this study enrolled critically ill patients with HFpEF admitted to the ICU for the first time and identified a significant linear association between RDW and 28-day all-cause mortality. Elevated RDW levels at admission were significantly associated with an increased risk of all-cause mortality. A predictive model for mortality in HFpEF was developed using Lasso-Cox regression, demonstrating that incorporating RDW slightly enhanced the model's predictive performance.
Heart failure has emerged as the final frontier in cardiovascular disease management this century, with HFpEF accounting for over half of all HF cases. Distinct from heart failure with reduced ejection fraction (HFrEF) in both pathophysiology and therapeutic outcomes, HFpEF predominantly manifests in patients with hypertension, diabetes, obesity or CKD, often accompanied by pulmonary hypertension and reduced exercise tolerance [20]. By contrast, patients with HFrEF primarily results from ischemic or dilated cardiomyopathy. Regarding treatment, conventional HFrEF therapies - such as beta-blockers and renin-angiotensin system inhibitors, remain controversial in HFpEF. Conversely, sodium-glucose cotransporter 2 (SGLT2) inhibitors, novel nonsteroidal mineralocorticoid receptor antagonists (Finerenone), and glucagon-like peptide 1 (GLP-1) receptor agonists have demonstrated therapeutic benefits in HFpEF patients [21], [22], [23]. Given the complex etiology and population heterogeneity of HFpEF, coupled with the current lack of standardized effective treatments, early and precise prediction of adverse outcomes and identification of high-risk patients are of paramount importance. B-Type Natriuretic Peptide (BNP) and NT-proBNP serve as reference biomarkers for diagnosing and prognosticating heart failure. However, NT-proBNP concentrations can be influenced by various comorbidities, such as renal dysfunction, advanced age, and cardiac arrhythmias [24], [25]. Furthermore, NT-proBNP demonstrates reduced reliability in HFpEF owing to confounding clinical features including obesity, atrial fibrillation, and renal impairment [26].
RDW is an accessible, cost-effective hematological parameter quantifying variations in circulating erythrocyte size. Its prognostic significance has been validated across diverse HF populations [24], [27], [28]. Multiple investigations have established that elevated RDW levels correlate strongly with adverse outcomes in HF, including heightened all-cause and cardiovascular mortality, as well as increased HF-related rehospitalization risk [8], [29], [30]. Prior research confirms RDW as an independent predictor of short-, intermediate-, and long-term mortality in HF cohorts [24], [31]. Combining RDW with NT-proBNP enhances prognostic accuracy [24]. Additionally, a linear correlation exists between baseline hemoglobin-to-RDW ratio (HRR) and in-hospital mortality among chronic HF patients with or without acute kidney injury [32], [33]. Subgroup analyses revealed significant interactions between ejection fraction strata and HRR categories, indicating enhanced predictive capacity in patients with EF <40% but diminished utility in those with EF ≥40% [34]. Contemporary studies propose that RDW not only acts as a static biomarker but also provides dynamic monitoring of disease trajectory through serial measurements. Greater in-hospital ΔRDW increments significantly predict higher 30-day readmission risks and short-term mortality [31]. Novel composite indices, including the RDW-to-albumin ratio (RAR) [35] and RDW-to-lymphocyte ratio (RLR) [36], show promise for long-term outcome assessment in acute HF, expanding the clinical applications of RDW-based risk models.
In the HFpEF population, one study demonstrated that anisocytosis (RDW >14.5%) was observed in 49.5% of patients and correlated with greater baseline comorbidity burden, reduced peak VO2, and more frequent HF hospitalizations [28]. RDW levels further correlated with established fibrosis and inflammation biomarkers, including carboxy-terminal collagen telopeptide, procollagen III N-terminal peptide, and galectin-3 [28]. Cardiac MRI analyses confirmed that RDW remained independently associated with extracellular volume and myocardial fibrosis after adjusting for age, sex, LVEF and hematocrit. Moreover, patients exhibiting both anisocytosis and myocardial fibrosis constituted a subgroup at high risk for 2-year mortality (HR 16.28, 95% CI: 4.30–61.66; P < 0.001) [28]. RDW levels at admission independently predicted non-cardiac adverse events in acute decompensated HFpEF [37]. This study establishes a significant association between admission RDW levels and 28-day all-cause mortality in critically ill HFpEF patients, demonstrating a distinct linear relationship. Furthermore, a prognostic model developed by Lasso-Cox regression confirmed that RDW integration enhances discriminatory performance for short-term adverse outcomes. Implementation of such models may enable earlier identification of high-risk individuals in clinical pathways.
The potential mechanisms underlying the association between elevated RDW and poor prognosis in HF involve multiple interrelated pathophysiological processes, including chronic inflammation, oxidative stress, malnutrition, iron deficiency, and bone marrow dysfunction [6], [7]. First, chronic inflammation plays a central role in the development and progression of HF, and increased RDW may serve as a marker reflecting the persistent inflammatory state in these patients [38]. Second, oxidative stress contributes to erythrocyte membrane damage, leading to reduced red blood cell stability, increased hemolysis, and consequently elevated RDW [39]. Third, iron deficiency is highly prevalent among individuals with HF and can impair normal erythropoiesis, thereby increasing RDW [40]. More broadly, metabolic disturbances such as malnutrition or micronutrient deficiencies may disrupt RBC production and further widen RDW [41]. In critically ill patients with HFpEF, chronic hypoxia and impaired microcirculatory perfusion are commonly observed. Hypoxic conditions stimulate accelerated RBC production in the bone marrow, resulting in greater heterogeneity in RBC volume and an increased RDW. Additionally, systemic inflammation may compromise bone marrow function, promoting the premature release of immature erythrocytes into circulation, which also contributes to RDW rise [42]. Elevated anisocytosis is associated with decreased erythrocyte deformability and diminished oxygen-carrying capacity, potentially leading to reduced oxygen delivery to peripheral and myocardial tissues. This hypoxia may exacerbate cardiac dysfunction, thereby creating a self-perpetuating cycle that worsens the clinical course of HFpEF.
Several limitations warrant consideration. First, as the analysis utilized the US-based MIMIC-IV database, the cohort predominantly comprised critically ill White patients with substantially elevated all-cause mortality. Second, the final analytic cohort was derived after a high exclusion rate, mainly because of missing BMI. This substantial exclusion may have introduced selection bias and may limit generalizability to broader, unselected critically ill HFpEF populations. Third, the diagnosis of HFpEF relied on ICD-9/10 codes, which may be subject to coding inaccuracies and lack independent echocardiographic confirmation of LVEF ≥50%. Although these codes have been validated and used in prior MIMIC-IV studies, misclassification bias cannot be excluded. Fourth, we did not perform formal correction for multiple comparisons across the exploratory analyses and multiple adjusted models reported in this study. Therefore, some nominally significant findings may represent false-positive results, and the associations identified should be interpreted as hypothesis-generating rather than definitive. Fifth, only baseline RDW measurements at admission were assessed, precluding assessment of serial RDW changes and their value for predicting prognosis in HFpEF patients. Future studies are warranted to determine whether serial RDW assessment could enhance long-term prognostic risk stratification.
5. Conclusion
In conclusion, this study establishes a significantly linear association between RDW and short-term poor prognosis in critically ill HFpEF patients. Admission RDW levels enhance the predictive accuracy for 28-day all-cause mortality in this population. As an accessible, cost-effective hematologic index, RDW demonstrates significant prognostic utility in heart failure. Future integration of mechanistic investigations and combined biomarker analyses may further advance its clinical utility for individualized risk stratification in HFpEF, particularly accounting for its pathophysiological heterogeneity.
Abbreviations
- HFpEF
Heart failure with preserved ejection fraction
- BMI
Body mass index
- SBP
Systolic blood pressure
- DBP
Diastolic blood pressure
- MBP
Mean blood pressure
- HR
Heart rate
- RR
Respiratory rate
- SpO2,
Oxygen saturation
- WBC
White blood cell
- RBC
Red blood cell
- RDW
Red cell distribution width
- BUN
Urea nitrogen
- INR
International normalized ratio
- PT
Prothrombin time
- APTT
Partial prothrombin time
- MI
Myocardial infarction
- CPD
Chronic pulmonary disease
- CKD
Chronic kidney disease
- GCS
Glasgow coma scale
- SOFA
Sequential organ failure assessment
- SAPS II
Simplified acute physiology score
- BNP
B-type natriuretic peptide
Ethics declaration
Informed consent and patient details
The authors have NOT obtained informed consent from participants or their legal representatives. Maintained by Beth Israel Deaconess Medical Center (BIDMC) in Boston, Massachusetts, this openly available, de-identified repository provides extensive patient demographics, laboratory results, medication histories, and comprehensive clinical documentation, serving as a vital asset for detailed healthcare investigations.
Studies in human
This study was performed in compliance with relevant laws, regulatory frameworks and guidelines where the research took place.
The BIDMC Institutional Review Board granted an exemption. The authors provided the following explanation: Maintained by Beth Israel Deaconess Medical Center (BIDMC) in Boston, Massachusetts, this openly available, de-identified repository provides extensive patient demographics, laboratory results, medication histories, and comprehensive clinical documentation, serving as a vital asset for detailed healthcare investigations.
CRediT authorship contribution statement
Zhihui Yao: Writing – original draft, Methodology, Investigation, Formal analysis, Data curation. Jing Xiao: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation. Ying Li: Writing – review & editing, Software, Formal analysis, Data curation. Xiyu Gao: Writing – review & editing, Software, Formal analysis, Data curation. Chunyan Zhang: Writing – review & editing, Methodology, Funding acquisition, Conceptualization. Yan Zhang: Writing – review & editing, Methodology, Funding acquisition, Conceptualization. Tuo Han: Writing – review & editing, Supervision, Project administration, Funding acquisition, Conceptualization.
Clinical trial number
Not applicable.
Ethics approval and consent to participate
The review boards of the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center have approved the use of the MIMIC-IV database. Since the participants in the study were anonymized and deidentified, this study was exempt from the requirements for ethical approval and informed consent.
Funding
This work was supported by National Natural Science Foundation of China (82100359), Natural Science Basic Research Program of Shaanxi Province (2023JC-YB-709) and Xi'an Science and Technology Plan Project (24YXYJ0148).
Declaration of competing interest
The authors declared that there are no conflicts of interests.
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.ahjo.2026.100905.
Appendix A. Supplementary data
Supplementary figures
Supplementary tables
Data availability
The publicly available datasets used in this study can be accessed from online repositories. The data can be obtained from https://physionet.org/content/mimiciv/3.1/.
References
- 1.Campbell P., Rutten F.H., Lee M.M., et al. Heart failure with preserved ejection fraction: everything the clinician needs to know. Lancet. 2024;403:1083–1092. doi: 10.1016/S0140-6736(23)02756-3. [DOI] [PubMed] [Google Scholar]
- 2.Omote K., Verbrugge F.H., Borlaug B.A. Heart failure with preserved ejection fraction: mechanisms and treatment strategies. Annu. Rev. Med. 2022;73:321–337. doi: 10.1146/annurev-med-042220-022745. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Desai A.S., Lam C.S.P., McMurray J.J.V., et al. How to manage heart failure with preserved ejection fraction: practical guidance for clinicians. JACC Heart Fail. 2023;11:619–636. doi: 10.1016/j.jchf.2023.03.011. [DOI] [PubMed] [Google Scholar]
- 4.Mishra S., Kass D.A. Cellular and molecular pathobiology of heart failure with preserved ejection fraction. Nat. Rev. Cardiol. 2021;18:400–423. doi: 10.1038/s41569-020-00480-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Abudureyimu M., Luo X., Wang X., et al. Heart failure with preserved ejection fraction (HFpEF) in type 2 diabetes mellitus: from pathophysiology to therapeutics. J. Mol. Cell Biol. 2022;14 doi: 10.1093/jmcb/mjac028. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Salvagno G.L., Sanchis-Gomar F., Picanza A., et al. Red blood cell distribution width: a simple parameter with multiple clinical applications. Crit. Rev. Clin. Lab. Sci. 2015;52:86–105. doi: 10.3109/10408363.2014.992064. [DOI] [PubMed] [Google Scholar]
- 7.Xanthopoulos A., Giamouzis G., Dimos A., et al. Red blood cell distribution width in heart failure: pathophysiology, prognostic role, controversies and dilemmas. J. Clin. Med. 2022;11 doi: 10.3390/jcm11071951. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Hou H., Sun T., Li C., et al. An overall and dose-response meta-analysis of red blood cell distribution width and CVD outcomes. Sci. Rep. 2017;7:43420. doi: 10.1038/srep43420. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Chen W., Yu P., Chen C., et al. Association between the red blood cell distribution width and 30-day mortality in intensive care patients undergoing cardiac surgery: a retrospective observational study based on the medical information mart for intensive care-IV database. Ann. Lab. Med. 2024;44:401–409. doi: 10.3343/alm.2023.0345. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lai H., Wu G., Zhong Y., et al. Red blood cell distribution width improves the prediction of 28-day mortality for patients with sepsis-induced acute kidney injury: a retrospective analysis from MIMIC-IV database using propensity score matching. J. Intensive Med. 2023;3:275–282. doi: 10.1016/j.jointm.2023.02.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hua R., Liu X., Yuan E. Red blood cell distribution width at admission predicts outcome in critically ill patients with kidney failure: a retrospective cohort study based on the MIMIC-IV database. Ren. Fail. 2022;44:1182–1191. doi: 10.1080/0886022X.2022.2098766. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Butt J.H., McDowell K., Kondo T., et al. Heart failure with preserved ejection fraction, red cell distribution width, and sacubitril/valsartan. ESC Heart Fail. 2024;11:65–77. doi: 10.1002/ehf2.14558. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Kim M., Lee C.J., Kang H.J., et al. Red cell distribution width as a prognosticator in patients with heart failure. ESC Heart Fail. 2023;10:834–845. doi: 10.1002/ehf2.14231. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Johnson A.E.W., Bulgarelli L., Shen L., et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci. Data. 2023;10:1. doi: 10.1038/s41597-022-01899-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Vandenbroucke J.P., von Elm E., Altman D.G., et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. PLoS Med. 2007;4 doi: 10.1371/journal.pmed.0040297. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Song L., Ying J., Li M., et al. Propensity score matched cohort study on magnesium supplementation and mortality in critically ill patients with HFpEF. Sci. Rep. 2025;15 doi: 10.1038/s41598-025-85931-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Little R.J. Missing data analysis. Annu. Rev. Clin. Psychol. 2024;20:149–173. doi: 10.1146/annurev-clinpsy-080822-051727. [DOI] [PubMed] [Google Scholar]
- 18.Azur M.J., Stuart E.A., Frangakis C., et al. Multiple imputation by chained equations: what is it and how does it work? Int. J. Methods Psychiatr. Res. 2011;20:40–49. doi: 10.1002/mpr.329. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Feng J., Huang L., Zhao X., et al. Construction of a metabolism-malnutrition-inflammation prognostic risk score in patients with heart failure with preserved ejection fraction: a machine learning based Lasso-Cox model. Nutr. Metab. (Lond.) 2024;21:77. doi: 10.1186/s12986-024-00856-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Li R., Liu Y., Zhao Z., et al. Machine learning-based phenotyping and assessment of treatment responses in heart failure with preserved ejection fraction. EClinicalMedicine. 2025;88 doi: 10.1016/j.eclinm.2025.103462. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Solomon S.D., McMurray J.J.V., Claggett B., et al. Dapagliflozin in heart failure with mildly reduced or preserved ejection fraction. N. Engl. J. Med. 2022;387:1089–1098. doi: 10.1056/NEJMoa2206286. [DOI] [PubMed] [Google Scholar]
- 22.Solomon S.D., McMurray J.J.V., Vaduganathan M., et al. Finerenone in heart failure with mildly reduced or preserved ejection fraction. N. Engl. J. Med. 2024;391:1475–1485. doi: 10.1056/NEJMoa2407107. [DOI] [PubMed] [Google Scholar]
- 23.Patel R., Wadid M., Makwana B., et al. GLP-1 receptor agonists among patients with overweight or obesity, diabetes, and HFpEF on SGLT2 inhibitors. JACC Heart Fail. 2024;12:1814–1826. doi: 10.1016/j.jchf.2024.07.006. [DOI] [PubMed] [Google Scholar]
- 24.Liang L., Huang L., Zhao X., et al. Prognostic value of RDW alone and in combination with NT-proBNP in patients with heart failure. Clin. Cardiol. 2022;45:802–813. doi: 10.1002/clc.23850. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Vrabie A.M., Totolici S., Delcea C., et al. Biomarkers in heart failure with preserved ejection fraction: a perpetually evolving frontier. J. Clin. Med. 2024;13 doi: 10.3390/jcm13164627. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Nishikimi T., Nakagawa Y. B-type natriuretic peptide (BNP) revisited-is BNP still a biomarker for heart failure in the angiotensin receptor/neprilysin inhibitor era? Biology. 2022;11 doi: 10.3390/biology11071034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Varga A., Cristescu L., Marusteri M.S., et al. Prognostic value of the red cell distribution width-to-eGFR ratio (RGR) across chronic heart failure phenotypes: a retrospective observational pilot study. J. Clin. Med. 2025;14 doi: 10.3390/jcm14082852. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Bevan G.H., Rana M., Al-Furaih N., et al. Anisocytosis is associated with myocardial fibrosis and exercise capacity in heart failure with preserved ejection fraction. Heart Lung. 2022;54:68–73. doi: 10.1016/j.hrtlng.2022.03.013. [DOI] [PubMed] [Google Scholar]
- 29.Shao Q., Li L., Li G., et al. Prognostic value of red blood cell distribution width in heart failure patients: a meta-analysis. Int. J. Cardiol. 2015;179:495–499. doi: 10.1016/j.ijcard.2014.11.042. [DOI] [PubMed] [Google Scholar]
- 30.Arba I.F., Multazam C.E.C.Z., Widiarti W., et al. A systematic review and dose-response meta-analysis of red blood cell distribution width to albumin ratio as mortality predictor in cardiovascular disease. Sci. Rep. 2025;15 doi: 10.1038/s41598-024-81876-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Zhang Q., Zhou B., Li X., et al. In-hospital changes in the red blood cell distribution width and mortality in critically ill patients with heart failure. ESC Heart Fail. 2023;10:3287–3298. doi: 10.1002/ehf2.14513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Li Y., Xu C., Qin Z., et al. Relationship between the hemoglobin-to-red cell distribution width ratio and in-hospital mortality in patients with chronic heart failure. Vasc. Health Risk Manag. 2024;20:553–565. doi: 10.2147/VHRM.S486075. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Xu X., Yang R., Yin Y., et al. Association of hemoglobin-to-red blood cell distribution width ratio with mortality in critically ill patients with heart failure and acute kidney injury: insights from the MIMIC-IV database. BMC Cardiovasc. Disord. 2025;25:214. doi: 10.1186/s12872-025-04632-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Zhou J., Ma W., Wan Y., et al. Predictive value of the Hb/RDW ratio for the risk of all-cause death in patients with heart failure with different ejection fractions. Cardiology. 2024;149:237–247. doi: 10.1159/000536440. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.He Q., Hu S., Xie J., et al. The red blood cell distribution width to albumin ratio was a potential prognostic biomarker for acute respiratory failure: a retrospective study. BMC Med. Inform. Decis. Mak. 2024;24 doi: 10.1186/s12911-024-02639-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Michalak A., Guz M., Kozicka J., et al. Red blood cell distribution width derivatives in alcohol-related liver cirrhosis and metabolic-associated fatty liver disease. World J. Gastroenterol. 2022;28:5636–5647. doi: 10.3748/wjg.v28.i38.5636. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Imai R., Uemura Y., Okumura T., et al. Impact of red blood cell distribution width on non-cardiac mortality in patients with acute decompensated heart failure with preserved ejection fraction. J. Cardiol. 2017;70:591–597. doi: 10.1016/j.jjcc.2017.03.010. [DOI] [PubMed] [Google Scholar]
- 38.Galindo-Martín C.A., Chong-Aviña P.A., Godinez-Breacher V., et al. Malnutrition: muscle wasting, inflammation, RDW, and their relation with adverse outcomes. Cir. Cir. 2024;92:150–158. doi: 10.24875/CIRU.22000634. [DOI] [PubMed] [Google Scholar]
- 39.Joosse H.J., van Oirschot B.A., Kooijmans S.A.A., et al. In-vitro and in-silico evidence for oxidative stress as drivers for RDW. Sci. Rep. 2023;13:9223. doi: 10.1038/s41598-023-36514-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Tkaczyszyn M., Comín-Colet J., Voors A.A., et al. Iron deficiency and red cell indices in patients with heart failure. Eur. J. Heart Fail. 2018;20:114–122. doi: 10.1002/ejhf.820. [DOI] [PubMed] [Google Scholar]
- 41.Haenggi E., Kaegi-Braun N., Wunderle C., et al. Red blood cell distribution width (RDW) - a new nutritional biomarker to assess nutritional risk and response to nutritional therapy? Clin. Nutr. 2024;43:575–585. doi: 10.1016/j.clnu.2024.01.001. [DOI] [PubMed] [Google Scholar]
- 42.Xie X., Su M., Ren K., et al. Clonal hematopoiesis and bone marrow inflammation. Transl. Res. 2023;255:159–170. doi: 10.1016/j.trsl.2022.11.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary figures
Supplementary tables
Data Availability Statement
The publicly available datasets used in this study can be accessed from online repositories. The data can be obtained from https://physionet.org/content/mimiciv/3.1/.