Abstract
Background
Chronic kidney disease (CKD) complicated with hypertension constitutes a major global public health burden. Early identification of reliable prognostic biomarkers is essential to optimize clinical management and reduce mortality among these patients.
Methods
A total of 6,731 intensive care unit (ICU) patients with hypertensive CKD were enrolled from the MIMIC-IV database. Three hematological markers were analyzed: red blood cell distribution width (RDW), hemoglobin to RDW ratio (HRR), and RDW to platelet ratio (RPR). Cox regression, Kaplan-Meier, and restricted cubic spline analyses were used to evaluate associations with prognosis. ROC curves, C-index, NRI, and IDI assessed predictive performance and incremental value, and decision curve analysis (DCA) evaluated clinical net benefit. Subgroup and sensitivity analyses were also performed.
Results
Among 6,731 patients with hypertensive CKD, 2,505 (37.2%) died within one year. After multivariate adjustment, elevated RDW was associated with an increased risk of one-year all-cause mortality (HR = 1.19, 95% CI: 1.16–1.21, P < 0.001), while a higher HRR correlated with reduced mortality risk (HR = 0.89, 95% CI: 0.87–0.91, P < 0.001). The RPR showed a U-shaped correlation with one-year all-cause mortality. The areas under the ROCs for RDW, HRR and RPR were 0.668, 0.600 and 0.515, respectively. RDW and HRR can improve the predictive performance of conventional risk models, while the incremental value of RPR was heterogeneous and limited to specific scoring systems. DCA demonstrated that RDW, HRR, and RPR all provided favorable clinical net benefits across a wide range of high-risk thresholds, suggesting their potential clinical applicability. Subgroup analyses demonstrated significant interactions of RDW with CRRT use and CKD stage (P < 0.001), as well as a notable interaction between HRR and CRRT (P < 0.001). Sensitivity analysis verified the robustness of the present findings.
Conclusions
RDW, HRR and RPR are associated with one-year mortality in ICU patients with hypertensive CKD. RDW exhibited modest predictive discrimination among the three markers and provides supplementary information for risk stratification when combined with established clinical risk scores. Prospective studies with external validation are warranted to confirm these findings and elucidate underlying mechanisms.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s12872-026-06423-2.
Keywords: Hypertensive chronic kidney disease, Red Blood Cell Distribution Width, Hemoglobin to RDW ratio, RDW to platelet ratio, Prognostic biomarkers, Intensive care unit
Introduction
Chronic Kidney Disease (CKD) is a prevalent global public health problem, and the annual incidence of new CKD cases rose from 7.8 million in 1990 to 18.99 million, severely impairing the quality of life of countless patients worldwide [1]. The causes of CKD are numerous, including hypertension, diabetes, genetic factors, medications, and other causes [2]. Among these, hypertension is considered one of the most important causes [3]. According to statistics, CKD caused by hypertension accounted for 66% of newly diagnosed cases attributed to known causes in 2019 [1]. These data emphasize the close relationship between hypertension and CKD, highlighting the need for special attention to hypertensive CKD.
Hypertensive CKD is generally a progressive condition. The impact of hypertension alters the hemodynamics of the renal microvascular system, and sustained elevated pressure results in ischemia and hypoxia in the renal tubules and glomeruli. This ongoing damage diminishes the kidney’s filtration capacity, hindering its ability to excrete metabolic waste and maintain electrolyte balance [4]. Over time, significant pathological alterations may occur in renal tissue, including local inflammation, cell apoptosis, fibrosis, and glomerulosclerosis, ultimately contributing to the onset and progression of CKD [5–7]. The relationship between CKD and hypertension is complex, involving a bidirectional interaction. As CKD worsens, the kidneys lose their ability to manage fluid, sodium, and electrolyte balance, making patients more vulnerable to hypertension, especially with increased fluid intake or high salt consumption [8, 9]. This creates a harmful cycle: hypertension damages the kidneys further, and kidney damage worsens hypertension. This vicious cycle elevates the risks of cardiovascular disease, heart failure and premature mortality, imposing a substantial burden on individuals and society [10]. Hence, simple and practical biomarkers are urgently needed to identify patients at high risk and guide early clinical interventions.
Red blood cell distribution width (RDW) has been correlated with adverse clinical outcomes in multiple conditions, including cardiovascular disease, sepsis and liver disease [11–14]. Elevated RDW levels are closely linked to inflammation, tissue hypoxia and malnutrition [15]. In addition, several indicators have been derived based on RDW, such as the hemoglobin to RDW ratio (HRR) and the RDW to platelet ratio (RPR). However, there is currently a lack of research on whether these indicators can be used to predict the prognosis of hypertensive CKD patients, especially in critically ill patients.
The aim of this study was to investigate whether RDW and its derived indicators (HRR and RPR) could act as prognostic biomarkers for patients with hypertensive CKD in the ICU setting. We extracted data on hypertensive CKD patients from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database and then followed these patients for one year to evaluate the predictive value of RDW and its derived indicators for all-cause mortality. By elucidating the value of RDW and its derived indicators in this ICU cohort, this research may inform future risk assessment strategies and hypothesis generation for larger validation studies.
Methods
Study population
This study utilized data from the MIMIC-IV database (version 3.1), a well-known publicly available resource managed by the Computational Physiology Laboratory at the Massachusetts Institute of Technology. The database includes patient information collected from 2008 to 2022 through hourly physiological measurements recorded by bedside monitors, which were validated by ICU nursing staff. It features a diverse cohort of ICU patients and provides extensive details on demographic factors, medical histories, laboratory test results, and prescribed medications. Access to the database was granted upon the successful completion of the National Institutes of Health online course titled “Protecting Human Research Participants”, which ensures ethical standards in research involving human subjects. Specifically, access was awarded to individuals who completed the Collaborative Institutional Training Initiative examination. We thank Degang Mo for assistance with data acquisition (Certification No. 65748833). In accordance with the International Classification of Diseases (ICD), 9th and 10th Revisions (ICD: 40390, I129, 40391, 40310, 40311, 40301, 40300), this study enrolled patients diagnosed with hypertensive CKD. The exclusion criteria were: (1) patients not diagnosed with hypertensive CKD; (2) patients younger than 18 years; (3) patients with prior ICU admissions; and (4) patients with an ICU stay of less than 24 h. As a result, a total of 6,731 patients were included in this study (Fig. 1).
Fig. 1.
The flowchart of the study
Data collection
The software PostgreSQL (version 13.7.2) was used to extract information. In this study, demographic data extracted from the database included age, gender, race, and weight. The vital signs of patients encompassed temperature, heart rate, respiratory rate, systolic blood pressure (SBP), diastolic blood pressure (DBP), mean blood pressure (MBP), and peripheral capillary oxygen saturation (SpO2). Comorbidities recorded included type 2 diabetes mellitus (T2DM), heart failure, acute myocardial infarction (AMI), cancer, cirrhosis, pneumonia, stroke, hyperlipidemia, chronic obstructive pulmonary disease (COPD), respiratory failure (RF), obesity, sepsis and acute kidney injury (AKI). The scoring system data incorporated the Sequential Organ Failure Assessment (SOFA), Simplified Acute Physiology Score II (SAPS II), Oxford Acute Severity of Illness Score (OASIS), and Glasgow Coma Scale (GCS). Laboratory test results were defined as the first measurements obtained after ICU admission, including white blood cells (WBC), red blood cells (RBC), platelets (PLT), hemoglobin (Hb), RDW, HRR, RPR, sodium, potassium, calcium, chloride, glucose, international normalized ratio (INR), blood urea nitrogen (BUN), and creatinine. Baseline estimated glomerular filtration rate (eGFR) was calculated using the 2021 CKD-EPI creatinine equation [16]. For females, eGFR = 142 × (Scr/0.7)^α × 0.9938^Age× 1.012, with α = -0.241 if Scr ≤ 0.7 mg/dL and α = -1.200 if Scr > 0.7 mg/dL. For males, eGFR = 142 × (Scr/0.9)^α × 0.9938^Age, with α = -0.302 if Scr ≤ 0.9 mg/dL and α = -1.200 if Scr > 0.9 mg/dL. Treatment measures included continuous renal replacement therapy (CRRT) and ventilation.
Outcomes and definitions
The primary outcome of this study was all-cause mortality within one year of ICU admission. The secondary outcome was 30-day all-cause mortality, which was examined in sensitivity analyses. We used the MIMIC-IV database (version 3.1) in this study. In addition to in-hospital electronic health records, this database is linked to the Massachusetts State Registry of Vital Records and Statistics to enable long-term mortality follow-up. Survival time was calculated from the date of ICU admission to the date of death. Patients who were still alive after 365 days of follow-up were defined as right-censored cases. A total of 6,731 patients were enrolled in the cohort, among whom 2,505 died within one year and 4,226 were right-censored. The loss to follow-up rate was 0%.
Statistical analysis
This study employed the R 4.3.0 statistical analysis software. The Shapiro-Wilk test was conducted to assess the normality of the data distribution, revealing that all continuous variables deviated from a normal distribution. Skewed continuous data were represented as median (Q₁, Q₃), and the Mann-Whitney U test was utilized for independent samples. Categorical data were expressed as n (%), with the chi-square test applied as appropriate. P < 0.05 was deemed indicative of statistical significance. Bonferroni correction was applied for multiple comparisons. Cox regression models were utilized to assess the hazard ratio (HR) between RDW and its derived indicators and the one-year mortality rate of patients. RPR had inherently low raw values. Accordingly, all RPR values were multiplied by 10 prior to Cox regression analyses for better presentation and interpretation of regression results. RDW and HRR were analyzed using their original raw values without any transformation. Model 1 was unadjusted. Model 2 adjusted for gender, age, weight, SpO2, and temperature. Model 3 further adjusted for cancer, cirrhosis, pneumonia, stroke, hyperlipidemia, COPD, obesity, calcium, chloride, INR, and eGFR. The proportional hazards (PH) assumption for Cox regression models was tested using Schoenfeld residuals. Multicollinearity among the covariates was assessed using the variance inflation factor (VIF) and tolerance (1/VIF). Kaplan-Meier survival analysis was employed to assess mortality rates associated with various levels of RDW and its derived indicators. The log-rank test was subsequently applied to analyze differences in these mortality rates. To further analyze the dose-response relationships between patient prognosis and RDW with its derivatives, we established restricted cubic spline (RCS) models. All RCS models were fitted with 4 knots located at the 5th, 35th, 65th and 95th percentiles of each continuous variable, following the classic knot configuration proposed by Harrell [17]. The median value of each biomarker was defined as the reference point (HR = 1), which is a common practice in Cox regression-based RCS analyses for clinical cohorts. Two RCS models were established: an unadjusted model with no covariates, and a fully adjusted model incorporating all covariates from Cox regression model 3. We constructed receiver operating characteristic (ROC) curves to compare the predictive capabilities for one-year mortality rates among RDW, HRR, and RPR. Pairwise comparisons of the AUCs were performed using the DeLong test. To evaluate the incremental prognostic value of RDW, HRR and RPR, we calculated the C-index, net reclassification improvement (NRI) and integrated discrimination improvement (IDI) before and after adding these three biomarkers to each baseline model. The baseline models were constructed based on SOFA, SAPS II, OASIS, GCS, age and eGFR. Subgroup analyses were further performed to explore the consistency of the associations between RDW, HRR, RPR and one-year all-cause mortality across different clinical subgroups. Decision curve analysis (DCA) was further performed to assess the clinical net benefit. Stratification factors included CRRT administration (yes vs. no) and CKD stage (stage 1–2, 3a, 3b, 4, and 5). Stratified Cox regression models were constructed for each subgroup. The interaction term was incorporated into the overall Cox model to test the between-group heterogeneity of effect sizes. The proportion of missing values for each variable was summarized (Supplementary Table 1). The missing data proportions were generally low for most variables (ranging from 0.01% to 0.86%), with the highest missing rates observed for INR (6.49%) and calcium (1.95%). Given the low overall missing proportion and the assumption that data were missing at random, multiple imputation was performed to handle missing values [18]. To evaluate the influence of missing data and verify the robustness of our findings, we performed sensitivity analysis via complete case analysis by excluding all participants with missing variables. The original three Cox regression models were reconstructed in the remaining 6,001 patients, and the outcomes were compared with the primary results to examine whether missing data altered the main conclusions. Second, we reanalyzed all Cox regression models using 30-day all-cause mortality as an alternative clinical endpoint to further explore the associations between the three hematological markers and short-term prognosis.
Results
Patient characteristics
A total of 6,731 patients diagnosed with Hypertensive CKD were included in the study. The average age of the cohort was 76 years, and males accounted for 61.66% (4,150 individuals) of the total population. Among the participants, there were 2,505 reported deaths (37.22%) within one year. There was a significant difference in RDW between the survival group and the death group, with values of 14.60 (13.70, 15.90) and 15.90 (14.60, 17.50), respectively (P < 0.001). There was a significant difference in HRR between the survival group and the death group, with values of 6.65 (5.55, 7.92) for the survival group and 6.00 (4.97, 7.27) for the death group (P < 0.001). A significant difference in the RPR was observed between the survival and death groups, with the survival group having a value of 0.08 (0.06, 0.11) and the death group a value of 0.08 (0.06, 0.12) (P = 0.035). In the analysis of the survival and mortality groups, significant differences were observed in demographic characteristics such as age, gender, and weight (P < 0.05). Additionally, vital signs, including temperature, heart rate, respiratory rate, SBP, MBP, and SpO2, also exhibited significant differences between the two groups. Comorbidities recorded included heart failure, cancer, cirrhosis, pneumonia, stroke, hyperlipidemia, COPD, RF, obesity, sepsis and AKI. The scoring systems assessed, including SOFA, SAPS II, OASIS, and GCS, all exhibited significant differences between the groups. The laboratory test results, which included WBC, RBC, PLT, Hb, chloride, INR, BUN, creatinine and eGFR, demonstrated significant differences between the two groups. The use of CRRT and mechanical ventilation also differed significantly between the two groups (both P < 0.05). For specific details, please refer to Table 1.
Table 1.
Baseline characteristics of patients with hypertensive chronic kidney disease
| Characteristics | Total (n = 6731) |
Survivor (n = 4226) |
Non-Survivor (n = 2505) |
P-value |
|---|---|---|---|---|
| Demographic characteristics | ||||
| Age, year | 76.00 (66.00, 84.00) | 74.00 (64.00, 82.00) | 79.00 (70.00, 86.00) | < 0.001 |
| Gender, n (%) | 0.028 | |||
| Female | 2581 (38.34) | 1578 (37.34) | 1003 (40.04) | |
| Male | 4150 (61.66) | 2648 (62.66) | 1502 (59.96) | |
| Race, n (%) | 0.326 | |||
| Asian | 184 (2.73) | 119 (2.82) | 65 (2.59) | |
| White | 4475 (66.48) | 2779 (65.76) | 1696 (67.70) | |
| Black | 1021 (15.17) | 664 (15.71) | 357 (14.25) | |
| Others | 1051 (15.61) | 664 (15.71) | 387 (15.45) | |
| Weight, kg | 79.60 (67.25, 93.40) | 81.90 (69.50, 96.00) | 75.40 (64.10, 88.60) | < 0.001 |
| Vital signs | ||||
| Temperature, ℃ | 36.67 (36.44, 37.00) | 36.72 (36.44, 37.00) | 36.67 (36.39, 36.96) | < 0.001 |
| Heart rate, bpm | 82.00 (72.00, 95.00) | 81.00 (71.00, 93.00) | 85.00 (73.00, 99.00) | < 0.001 |
| Respiratory rate, bpm | 18.00 (15.00, 22.00) | 17.00 (14.00, 21.00) | 19.00 (16.00, 23.00) | < 0.001 |
| SBP, mmHg | 123.00 (106.00, 142.00) | 125.00 (108.00, 144.00) | 121.00 (103.00, 139.00) | < 0.001 |
| DBP, mmHg | 63.00 (53.00, 76.00) | 64.00 (54.00, 76.00) | 63.00 (52.00, 76.00) | 0.056 |
| MBP, mmHg | 78.00 (68.00, 91.00) | 79.00 (68.00, 92.00) | 77.00 (66.00, 91.00) | < 0.001 |
| SpO2, % | 98.00 (96.00, 100.00) | 98.00 (96.00, 100.00) | 98.00 (95.00, 100.00) | < 0.001 |
| Comorbidity | ||||
| T2DM, n (%) | 3094 (45.97) | 1961 (46.40) | 1133 (45.23) | 0.350 |
| Heart failure, n (%) | 2587 (38.43) | 1428 (33.79) | 1159 (46.27) | < 0.001 |
| AMI, n (%) | 465 (6.91) | 282 (6.67) | 183 (7.31) | 0.323 |
| Cancer, n (%) | 1308 (19.43) | 706 (16.71) | 602 (24.03) | < 0.001 |
| Cirrhosis, n (%) | 363 (5.39) | 162 (3.83) | 201 (8.02) | < 0.001 |
| Pneumonia, n (%) | 1980 (29.42) | 1002 (23.71) | 978 (39.04) | < 0.001 |
| Stroke, n (%) | 815 (12.11) | 480 (11.36) | 335 (13.37) | 0.014 |
| Hyperlipidemia, n (%) | 3417 (50.77) | 2314 (54.76) | 1103 (44.03) | < 0.001 |
| COPD, n (%) | 471 (7.00) | 247 (5.84) | 224 (8.94) | < 0.001 |
| RF, n (%) | 1812 (26.92) | 808 (19.12) | 1004 (40.08) | < 0.001 |
| Obesity, n (%) | 920 (13.67) | 662 (15.66) | 258 (10.30) | < 0.001 |
| Sepsis, n (%) | 3971 (59.00) | 2231 (52.79) | 1740 (69.46) | < 0.001 |
| AKI, n (%) | 5453 (81.01) | 3276 (77.52) | 2177 (86.91) | < 0.001 |
| Scoring System | ||||
| SOFA | 5.00 (3.00, 7.00) | 5.00 (3.00, 7.00) | 6.00 (4.00, 9.00) | < 0.001 |
| SAPS II | 41.00 (34.00, 49.50) | 38.00 (31.00, 46.00) | 46.00 (37.00, 55.00) | < 0.001 |
| OASIS | 32.00 (26.00, 38.00) | 31.00 (25.00, 36.00) | 35.00 (29.00, 41.00) | < 0.001 |
| GCS | 15.00 (14.00, 15.00) | 15.00 (14.00, 15.00) | 15.00 (13.00, 15.00) | < 0.001 |
| Laboratory tests | ||||
| WBC, K/uL | 10.00 (7.30, 13.70) | 10.00 (7.30, 13.30) | 10.20 (7.40, 14.50) | 0.004 |
| RBC, m/uL | 3.29 (2.85, 3.80) | 3.31 (2.86, 3.83) | 3.25 (2.83, 3.75) | 0.001 |
| PLT, K/uL | 181.00 (132.00, 243.00) | 178.00 (132.00, 236.00) | 187.00 (132.00, 257.00) | 0.001 |
| Hb, g/dL | 9.80 (8.50, 11.20) | 9.80 (8.50, 11.40) | 9.60 (8.50, 11.00) | < 0.001 |
| RDW, % | 15.00 (13.90, 16.60) | 14.60 (13.70, 15.90) | 15.90 (14.60, 17.50) | < 0.001 |
| HRR | 6.41 (5.34, 7.71) | 6.65 (5.55, 7.92) | 6.00 (4.97, 7.27) | < 0.001 |
| RPR | 0.08 (0.06, 0.12) | 0.08 (0.06, 0.11) | 0.08 (0.06, 0.12) | 0.035 |
| Sodium, mEq/L | 139.00 (136.00, 141.00) | 139.00 (136.00, 141.00) | 138.86 (135.00, 141.00) | 0.202 |
| Potassium, mEq/L | 4.30 (3.90, 4.80) | 4.30 (3.90, 4.80) | 4.30 (3.90, 4.80) | 0.403 |
| Calcium, mEq/L | 8.40 (8.00, 8.90) | 8.40 (8.00, 8.90) | 8.40 (7.90, 8.90) | 0.162 |
| Chloride, mEq/L | 105.00 (100.00, 109.00) | 105.00 (101.00, 109.00) | 104.00 (99.00, 108.00) | < 0.001 |
| Glucose, mg/dL | 127.00 (103.00, 160.00) | 126.00 (104.00, 157.00) | 129.00 (102.00, 166.00) | 0.115 |
| INR | 1.30 (1.10, 1.50) | 1.26 (1.10, 1.42) | 1.30 (1.12, 1.50) | < 0.001 |
| BUN, mg/dL | 33.00 (23.00, 51.00) | 30.00 (21.00, 45.00) | 40.00 (27.00, 58.00) | < 0.001 |
| Creatinine, mg/dL | 1.70 (1.30, 2.80) | 1.60 (1.20, 2.40) | 2.00 (1.40, 3.20) | < 0.001 |
| eGFR, (mL/min/1.73 m²) | 36.90 (20.70, 53.45) | 41.30 (24.30, 57.18) | 28.90 (17.30, 45.10) | < 0.001 |
| CKD stage | < 0.001 | |||
| 1–2 | 1195 (17.75) | 917 (21.70) | 278 (11.10) | |
| 3a | 1263 (18.76) | 913 (21.60) | 350 (13.97) | |
| 3b | 1617 (24.02) | 1035 (24.49) | 582 (23.23) | |
| 4 | 1554 (23.09) | 761 (18.01) | 793 (31.66) | |
| 5 | 1102 (16.37) | 600 (14.20) | 502 (20.04) | |
| Treatment | ||||
| CRRT, n (%) | 502 (7.46) | 219 (5.18) | 283 (11.30) | < 0.001 |
| Ventilation, n (%) | 5357 (79.59) | 3327 (78.73) | 2030 (81.04) | 0.023 |
| ICU stay, days | 2.53 (1.64, 4.57) | 2.26 (1.51, 3.99) | 3.04 (1.84, 5.90) | < 0.001 |
Abbreviation: SBP Systolic Blood Pressure, DBP Diastolic Blood Pressure, MBP Mean Blood Pressure, SpO2 Peripheral Capillary Oxygen Saturation, T2DM Type 2 Diabetes Mellitus, AMI Acute Myocardial Infarction, COPD Chronic Obstructive Pulmonary Disease, RF Respiratory Failure, AKI Acute Kidney Injury, SOFA Sequential Organ Failure Assessment, SAPS II Simplified Acute Physiology Score II, OASIS Oxford Acute Severity of Illness Score, GCS Glasgow Coma Scale, WBC White Blood Cells, RBC Red Blood Cells, PLT Platelets, Hb Hemoglobin, RDW Red Cell Distribution Width, HRR Hemoglobin / RDW, RPR RDW / Platelets, INR International Normalized Ratio, BUN Blood Urea Nitrogen, eGFR Estimated Glomerular Filtration Rate, CRRT Continuous Renal Replacement Therapy, ICU Intensive Care Unit
The effects of RDW, HRR, and RPR on the survival time of patients
To analyze the effects of RDW, HRR, and RPR on the one-year survival time of patients, we developed multiple Cox regression survival models. In Model 1, which was not adjusted for any additional variables, we found that RDW [HR: 1.25, 95% confidence interval (CI): 1.22–1.27; P < 0.001] and RPR (HR: 1.25, 95% CI: 1.13–1.38; P < 0.001) were associated with an increased risk of one-year all-cause mortality. Conversely, HRR (HR: 0.85, 95% CI: 0.83–0.87; P < 0.001) was associated with a decreased mortality risk. In order to further validate the association between RDW, HRR, and RPR with one-year mortality in patients, we categorized the continuous variables into four groups based on quartiles and subsequently incorporated them into the Cox regression model for analysis. For RDW, in Model 1, the Q1 was used as the reference group. The HRs for the remaining three quartiles were as follows: Q2 had an HR of 1.55 (95% CI: 1.36–1.78; P < 0.001), Q3 had an HR of 2.14 (95% CI: 1.89–2.44; P < 0.001), and Q4 had an HR of 3.55 (95% CI: 3.15–4.01; P < 0.001). These results indicated a significant increase in the risk of mortality associated with higher RDW quartiles. For HRR, in Model 1, the Q1 was used as the reference group. The HRs for the remaining three quartiles were as follows: Q2 had an HR of 0.77 (95% CI: 0.70–0.86; P < 0.001), Q3 had an HR of 0.59 (95% CI: 0.53–0.65; P < 0.001), and Q4 had an HR of 0.51 (95% CI: 0.46–0.57; P < 0.001). These findings suggested a significant decrease in the risk of mortality associated with increasing HRR quartiles. For RPR, in Model 1, the Q1 was used as the reference group. The HRs for the remaining three quartiles were as follows: Q2 had an HR of 0.84 (95% CI: 0.75–0.94; P = 0.003), Q3 had an HR of 0.79 (95% CI: 0.70–0.88; P < 0.001), and Q4 had an HR of 1.15 (95% CI: 1.03–1.27; P = 0.011). These results indicated a significant reduction in the risk of mortality for Q2 and Q3, while Q4 demonstrated a significant increase in risk compared to the reference group. After adjusting for other influencing factors, models 2 and 3 also showed the same trend of effects (Table 2). The proportional hazards (PH) assumption of the Cox regression models was verified using Schoenfeld residuals. The global test results showed no significant deviation from the PH assumption for the RDW, HRR and RPR models (all P > 0.05), indicating that the PH assumption was well satisfied (Supplementary Table 2). Multicollinearity among the included covariates was assessed by variance inflation factor (VIF) and tolerance. All variables had a VIF lower than 2 and tolerance above 0.7, suggesting no obvious multicollinearity between covariates (Supplementary Table 3).
Table 2.
Cox regression models for all-cause mortality within one year of RDW, HRR and RPR
| Categories | Model 1 | Model 2 | Model 3 | ||||||
|---|---|---|---|---|---|---|---|---|---|
| HR | 95% CI | P | HR | 95% CI | P | HR | 95% CI | P | |
| RDW Continuous | 1.25 | 1.22 ~ 1.27 | < 0.001 | 1.25 | 1.23 ~ 1.27 | < 0.001 | 1.19 | 1.16 ~ 1.21 | < 0.001 |
| RDW Quartile 1 | Reference | Reference | Reference | ||||||
| RDW Quartile 2 | 1.55 | 1.36 ~ 1.78 | < 0.001 | 1.55 | 1.35 ~ 1.77 | < 0.001 | 1.34 | 1.17 ~ 1.54 | < 0.001 |
| RDW Quartile 3 | 2.14 | 1.89 ~ 2.44 | < 0.001 | 2.12 | 1.87 ~ 2.42 | < 0.001 | 1.69 | 1.48 ~ 1.93 | < 0.001 |
| RDW Quartile 4 | 3.55 | 3.15 ~ 4.01 | < 0.001 | 3.57 | 3.16 ~ 4.03 | < 0.001 | 2.58 | 2.27 ~ 2.94 | < 0.001 |
| HRR Continuous | 0.85 | 0.83 ~ 0.87 | < 0.001 | 0.84 | 0.82 ~ 0.86 | < 0.001 | 0.89 | 0.87 ~ 0.91 | < 0.001 |
| HRR Quartile 1 | Reference | Reference | Reference | ||||||
| HRR Quartile 2 | 0.77 | 0.70 ~ 0.86 | < 0.001 | 0.75 | 0.68 ~ 0.83 | < 0.001 | 0.81 | 0.73 ~ 0.90 | < 0.001 |
| HRR Quartile 3 | 0.59 | 0.53 ~ 0.65 | < 0.001 | 0.57 | 0.51 ~ 0.64 | < 0.001 | 0.66 | 0.59 ~ 0.73 | < 0.001 |
| HRR Quartile 4 | 0.51 | 0.46 ~ 0.57 | < 0.001 | 0.49 | 0.44 ~ 0.55 | < 0.001 | 0.64 | 0.57 ~ 0.72 | < 0.001 |
| RPR Continuous | 1.25 | 1.13 ~ 1.38 | < 0.001 | 1.27 | 1.15 ~ 1.40 | < 0.001 | 1.12 | 1.01 ~ 1.25 | 0.003 |
| RPR Quartile 1 | Reference | Reference | Reference | ||||||
| RPR Quartile 2 | 0.84 | 0.75 ~ 0.94 | 0.003 | 0.85 | 0.76 ~ 0.96 | 0.006 | 0.90 | 0.80 ~ 1.01 | 0.067 |
| RPR Quartile 3 | 0.79 | 0.70 ~ 0.88 | < 0.001 | 0.78 | 0.69 ~ 0.87 | < 0.001 | 0.81 | 0.72 ~ 0.91 | < 0.001 |
| RPR Quartile 4 | 1.15 | 1.03 ~ 1.27 | 0.011 | 1.16 | 1.05 ~ 1.29 | 0.005 | 1.06 | 0.95 ~ 1.19 | 0.273 |
Model1: Crude
Model2: Adjust: gender, age, weight, SpO2, and temperature
Model3: Adjust: gender, age, weight, SpO2, temperature, Cancer, Cirrhosis, Pneumonia, Stroke, Hyperlipidemia, COPD, Obesity, Calcium, Chloride, INR, and eGFR
RDW Red Cell Distribution Width, HRR Hemoglobin / RDW, RPR RDW / Platelets, HR Hazard ratio, CI Confidence interval
Bonferroni correction (α = 0.05/3 = 0.0167) applied to p-values for quartile comparisons
To further investigate the relationship between RDW, HRR, RPR, and the one-year mortality rate of patients, we performed separate Kaplan-Meier survival analyses. Each variable was analyzed using quartiles. The findings revealed that patients with elevated RDW and RPR experienced lower survival rates, while the trend was reversed for HRR (Fig. 2). Log-rank test results indicated statistically significant differences among the four quartile groups for each variable (P < 0.001).
Fig. 2.
Kaplan-Meier survival curves for one-year all-cause mortality stratified by quartiles of RDW (A), HRR (B) and RPR (C) in patients with hypertensive CKD
The association between RDW, HRR, RPR and one-year mortality
To further investigate the association between RDW, HRR, RPR, and one-year mortality, we performed both unadjusted and adjusted RCS analyses for each variable. In the analysis of RDW, both the unadjusted RCS model (Fig. 3A; P for nonlinear < 0.001) and the adjusted RCS model (Fig. 3D; P for nonlinear = 0.002) revealed a positive non-linear relationship between RDW and mortality. The investigation of HRR indicated a negative non-linear relationship between HRR and mortality, as evidenced by both the unadjusted RCS model (Fig. 3B, P for nonlinear = 0.001) and the adjusted RCS model (Fig. 3E, P for nonlinear = 0.001). The assessment of RPR demonstrated a U-shaped relationship with mortality, evidenced by both the unadjusted RCS model (Fig. 3C, P for nonlinear < 0.001) and the adjusted RCS model (Fig. 3F, P for nonlinear < 0.001).
Fig. 3.
Restricted cubic spline (RCS) analyses of the associations between RDW, HRR, RPR and one-year all-cause mortality.The median value of each biomarker (RDW = 15, HRR = 6.4, RPR = 0.08) was defined as the reference (HR = 1) for all RCS curves. Panels A–C display unadjusted models, while panels D–F display fully adjusted models that incorporated all covariates in Cox regression Model 3
Comparison of predictive abilities for one-year mortality among RDW, HRR, and RPR
To compare the predictive abilities of RDW, HRR, and RPR for one-year mortality in patients with hypertensive CKD, we conducted a ROC analysis. The results indicated that the areas under the curve (AUC) for RDW, HRR, and RPR were modest: RDW (0.668, 95% CI: 0.657–0.679), HRR (0.600, 95% CI: 0.588–0.612) and RPR (0.515, 95% CI: 0.503–0.527) (Fig. 4). The optimal cut-off values derived from ROC curve analysis were 15.95 for RDW, 6.266 for HRR, and 0.123 for RPR. For RDW, the corresponding sensitivity was 0.759 (95% CI: 0.746–0.772) and specificity was 0.485 (95% CI: 0.466–0.505). For HRR, the sensitivity was 0.410 (95% CI: 0.395–0.425) and specificity was 0.436 (95% CI: 0.416–0.455). For RPR, the sensitivity reached 0.809 (95% CI: 0.797–0.821), with a specificity of 0.261 (95% CI: 0.244–0.279). Pairwise comparisons among the three measures indicated statistically significant differences as determined by the DeLong test ( all P < 0.001).
Fig. 4.
Receiver operating characteristic (ROC) curves of RDW, HRR and RPR for predicting one-year all-cause mortality
Incremental prognostic value and clinical net benefit of RDW, HRR and RPR
We further assessed the incremental prognostic performance of RDW, HRR and RPR by calculating C-index, NRI and IDI after incorporating each biomarker into six baseline models including SOFA, SAPS II, OASIS, GCS, age and eGFR.
RDW significantly improved the predictive performance of all baseline models in terms of C-index, NRI and IDI (all P < 0.001), with the most remarkable improvement seen in the age model. HRR also brought consistent and significant incremental benefits to all baseline models (all P < 0.001), though its effect was weaker than RDW. The incremental value of RPR was heterogeneous. It failed to significantly enhance the predictive ability of SOFA and SAPS II models (all P > 0.05), but produced statistically significant improvements in the OASIS, GCS, age and eGFR models (all P < 0.05) (Table 3). DCA demonstrated that RDW, HRR, and RPR all provided favorable clinical net benefits across a wide range of high-risk thresholds, suggesting their potential clinical applicability (Fig. 5).
Table 3.
Incremental prognostic value of RDW, HRR and RPR for one-year all-cause mortality
| Variables | C-index (95% CI) | NRI (%) (95% Cl) | P-value | IDI (%)(95% Cl) | P-value |
|---|---|---|---|---|---|
| SOFA | 0.62 (0.61–0.63) | Ref | Ref | Ref | Ref |
| SOFA + RDW | 0.66 (0.65–0.67) | 43.38 (38.58–48.18) | < 0.001 | 5.77(5.19–6.34) | < 0.001 |
| SOFA + HRR | 0.63 (0.62–0.64) | 24.31(19.42–29.20) | < 0.001 | 1.55(1.26–1.85) | < 0.001 |
| SOFA + RPR | 0.62 (0.61–0.63) | 0.98(-3.93-5.89) | 0.6954 | 0.07(-0.01-0.14) | 0.7785 |
| SAPS II | 0.67 (0.65–0.68) | Ref | Ref | Ref | Ref |
| SAPS II + RDW | 0.70 (0.69–0.71) | 45.92(41.14–50.71) | < 0.001 | 5.79(5.21–6.37) | < 0.001 |
| SAPS II + HRR | 0.67 (0.66–0.68) | 26.83(21.95–31.71) | < 0.001 | 1.57(1.27–1.87) | < 0.001 |
| SAPS II + RPR | 0.66 (0.65–0.68) | 3.67(-1.21-8.55) | 0.1407 | 0.05(0.00-0.11) | 0.0681 |
| OASIS | 0.63 (0.62–0.64) | Ref | Ref | Ref | Ref |
| OASIS + RDW | 0.68 (0.67–0.69) | 47.34(42.56–52.13) | < 0.001 | 6.53(5.91–7.15) | < 0.001 |
| OASIS + HRR | 0.65 (0.64–0.66) | 28.64(23.77–33.51) | < 0.001 | 2.17(1.81–2.54) | < 0.001 |
| OASIS + RPR | 0.63 (0.62–0.65) | 5.36(0.48–10.24) | 0.0312 | 0.18(0.07–0.30) | 0.002 |
| GCS | 0.55 (0.54–0.56) | Ref | Ref | Ref | Ref |
| GCS + RDW | 0.65 (0.64–0.66) | 47.74(42.96–52.53) | < 0.001 | 7.63(6.97–8.29) | < 0.001 |
| GCS + HRR | 0.60 (0.59–0.61) | 30.22(25.35–35.09) | < 0.001 | 2.83(2.43–3.24) | < 0.001 |
| GCS + RPR | 0.56 (0.55–0.57) | 7.67(1.59–13.75) | 0.0135 | 0.24(0.14–0.35) | < 0.001 |
| Age | 0.59 (0.58–0.60) | Ref | Ref | Ref | Ref |
| Age + RDW | 0.66 (0.65–0.68) | 49.93(45.16–54.70) | < 0.001 | 7.53(6.87–8.19) | < 0.001 |
| Age + HRR | 0.62 (0.61–0.63) | 31.42(26.56–36.29) | < 0.001 | 2.87(2.45–3.28) | < 0.001 |
| Age + RPR | 0.59 (0.58–0.60) | 5.26(0.39–10.14) | 0.0344 | 0.22(0.09–0.35) | 0.001 |
| eGFR | 0.60 (0.58–0.61) | Ref | Ref | Ref | Ref |
| eGFR + RDW | 0.65 (0.64–0.66) | 41.38(36.57–46.19) | < 0.001 | 5.29(4.72–5.86) | < 0.001 |
| eGFR + HRR | 0.61 (0.60–0.62) | 22.64(17.74–27.54) | < 0.001 | 1.64(1.33–1.95) | < 0.001 |
| eGFR + RPR | 0.60 (0.59–0.61) | 5.28(0.4-10.16) | 0.0338 | 0.27(0.14–0.40) | < 0.001 |
C-index Concordance index, NRI Net reclassification improvement, IDI Integrated discrimination improvement;
CI Confidence interval, Ref Reference group. Baseline models included SOFA, SAPS II, OASIS, GCS, age and eGFR
Fig. 5.
Decision curve analysis (DCA) for RDW, HRR and RPR in predicting one-year all-cause mortality
Subgroup analyses
We conducted subgroup analyses stratified by CRRT status and CKD stage to explore potential effect modification between the three biomarkers and one-year all-cause mortality (Table 4).
Table 4.
Subgroup analyses of the associations between RDW, HRR, RPR and one-year all-cause mortality
| Variables | n (%) | HR (95%CI) | P | P for interaction |
|---|---|---|---|---|
| RDW | ||||
| All patients | 6731 (100.00) | 1.25 (1.22 ~ 1.27) | < 0.001 | |
| CRRT | < 0.001 | |||
| No | 6229 (92.54) | 1.26 (1.23 ~ 1.28) | < 0.001 | |
| Yes | 502 (7.46) | 1.09 (1.03 ~ 1.15) | 0.004 | |
| CKD stage | < 0.001 | |||
| 1–2 | 1195 (17.75) | 1.31 (1.24 ~ 1.37) | < 0.001 | |
| 3a | 1263 (18.76) | 1.30 (1.23 ~ 1.36) | < 0.001 | |
| 3b | 1617 (24.02) | 1.22 (1.18 ~ 1.27) | < 0.001 | |
| 4 | 1554 (23.09) | 1.15 (1.12 ~ 1.19) | < 0.001 | |
| 5 | 1102 (16.37) | 1.19 (1.14 ~ 1.25) | < 0.001 | |
| HRR | ||||
| All patients | 6731 (100.00) | 0.85 (0.83 ~ 0.87) | < 0.001 | |
| CRRT | < 0.001 | |||
| No | 6229 (92.54) | 0.84 (0.82 ~ 0.86) | < 0.001 | |
| Yes | 502 (7.46) | 1.05 (0.98 ~ 1.13) | 0.175 | |
| CKD stage | 0.365 | |||
| 1–2 | 1195 (17.75) | 0.84 (0.79 ~ 0.91) | < 0.001 | |
| 3a | 1263 (18.76) | 0.87 (0.81 ~ 0.92) | < 0.001 | |
| 3b | 1617 (24.02) | 0.86 (0.82 ~ 0.90) | < 0.001 | |
| 4 | 1554 (23.09) | 0.91 (0.87 ~ 0.95) | < 0.001 | |
| 5 | 1102 (16.37) | 0.89 (0.83 ~ 0.95) | < 0.001 | |
| RPR | ||||
| All patients | 6731 (100.00) | 1.25 (1.13 ~ 1.38) | < 0.001 | |
| CRRT | 0.130 | |||
| No | 6229 (92.54) | 1.24 (1.12 ~ 1.39) | < 0.001 | |
| Yes | 502 (7.46) | 1.00 (0.76 ~ 1.31) | 0.985 | |
| CKD stage | 0.070 | |||
| 1–2 | 1195 (17.75) | 0.92 (0.67 ~ 1.27) | 0.624 | |
| 3a | 1263 (18.76) | 1.07 (0.81 ~ 1.42) | 0.640 | |
| 3b | 1617 (24.02) | 1.15 (0.93 ~ 1.42) | 0.210 | |
| 4 | 1554 (23.09) | 1.35 (1.14 ~ 1.61) | < 0.001 | |
| 5 | 1102 (16.37) | 1.47 (1.20 ~ 1.82) | < 0.001 |
For RDW, significant interactions were detected with both CRRT status (P for interaction < 0.001) and CKD stage (P for interaction < 0.001). In the overall cohort, RDW was associated with an increased risk of one-year all-cause mortality (HR = 1.25, 95% CI: 1.22–1.27, P < 0.001). Stratified analysis showed that this positive association existed in all subgroups. The magnitude of the association was weaker in patients receiving CRRT (HR = 1.09, 95% CI: 1.03–1.15, P = 0.004) compared with those not receiving CRRT (HR = 1.26, 95% CI: 1.23–1.28, P < 0.001). Across different CKD stages, RDW remained positively correlated with mortality risk in all subgroups.
For HRR, a significant interaction was found between HRR and CRRT status (P for interaction < 0.001), while no significant interaction was observed across CKD stages (P for interaction = 0.365). In the overall population, higher HRR was related to lower mortality risk (HR = 0.85, 95% CI: 0.83–0.87, P < 0.001). This inverse association remained stable across all CKD subgroups.
For RPR, there was no statistically significant interaction with CRRT status (P for interaction = 0.130) or CKD stage (P for interaction = 0.070). RPR was associated with elevated mortality risk in the total cohort (HR = 1.25, 95% CI: 1.13–1.38, P < 0.001). When stratified by CRRT, the association was significant in patients without CRRT (HR = 1.24, 95% CI: 1.12–1.39, P < 0.001), but not in those with CRRT. In subgroup analyses by CKD stage, RPR was significantly linked to higher mortality risk in patients with stage 4 and 5 CKD, whereas no obvious association was seen in patients at stages 1–3.
Sensitivity analysis
Sensitivity analysis was performed using the complete case dataset by excluding all participants with missing variables, and all three Cox regression models were reconstructed to verify the robustness of the primary findings (Supplementary Table 4). In the complete case cohort, the association patterns of RDW, HRR and RPR with one-year all-cause mortality were highly consistent with the main analysis. For continuous variables, the HR values and corresponding P values of RDW, HRR and RPR showed no obvious changes across the unadjusted and multivariable adjusted models. We further evaluated 30-day all-cause mortality as an alternative endpoint (Supplementary Table 5). In the fully adjusted model, the associations remained largely consistent with the primary findings for RDW (HR = 1.17, 95% CI: 1.14–1.20, P < 0.001), HRR (HR = 0.93, 95% CI: 0.90–0.97, P < 0.001), and RPR (HR = 1.21, 95% CI: 1.05–1.40, P = 0.009). Quartile analyses showed similar trends.
Discussion
Our findings show moderate prognostic associations of RDW and its derived indicators, which deliver supplementary predictive value when combined with traditional risk models. We analyzed data from 6,731 patients with hypertensive CKD, and the findings demonstrated that elevated RDW was associated with higher one-year all-cause mortality, while higher HRR was correlated with lower mortality risk. Additionally, RPR displayed a complex U-shaped relationship with mortality. ROC analyses in this cohort revealed that the three markers had only moderate standalone discriminatory ability. Importantly, RDW and HRR could deliver significant incremental prognostic value when incorporated into existing risk prediction models in this ICU population. Consistent with prior CKD-related studies [19–22], RDW was associated with adverse clinical outcomes in our cohort. Prior small-sample and cross-sectional studies have indicated the prognostic role of RDW in hypertensive patients with early renal damage [23]. Our study focused on critically ill ICU patients with hypertensive CKD and adopted a large sample. Previous studies demonstrated that HRR was associated with adverse cardiovascular and cerebrovascular prognoses [24, 25]. To our knowledge, this is the first report addressing the association between HRR and one-year mortality in hypertensive CKD patients in the ICU setting. RPR has also been identified as a prognostic marker in several acute illnesses including acute pancreatitis, liver diseases, and acute myocardial infarction [26–28]. A study [29] indicated that elevated RPR at admission was significantly linked to a higher risk of mortality. Notably, the U-shaped correlation between RPR and one-year mortality observed in our cohort has not been described in prior relevant research.
This section explores potential biological correlates to interpret the observed associations. Elevated RDW is correlated with systemic inflammation, altered erythrocyte morphology, and endothelial dysfunction, all of which are linked to poor prognosis in patients with hypertensive CKD [30–32]. Chronic inflammatory cytokines induce oxidative stress and inhibit bone marrow erythropoiesis [33]. In patients with hypertensive chronic kidney disease, reduced renal clearance causes hepcidin accumulation, disrupting iron transport and triggering CKD-associated anemia, which eventually increases erythrocyte volume heterogeneity [34]. In parallel, hypertension together with uremic toxins impairs vascular endothelial function, lowers nitric oxide bioavailability and aggravates renal microcirculatory hypoxia [35]. The synergistic effects of inflammatory activation, oxidative damage, hepcidin dysregulation, suppressed erythropoiesis and endothelial dysfunction collectively explain the rising trend of RDW across advancing CKD stages [36]. HRR reflects the combined levels of hemoglobin and RDW. Higher HRR values correspond to sufficient oxygen transport and stable erythrocyte profiles [37], which are correlated with favorable clinical outcomes in CKD patients [38]. Furthermore, the association between HRR and enhanced survival may reflect an adequate nutritional status, which is linked to a reduced prevalence of associated complications, given that malnutrition and inflammation are prevalent among patients with CKD [39]. RPR has emerged as a novel prognostic biomarker, demonstrating a complex U-shaped relationship in patients with hypertensive CKD, warranting further exploration. The U-shaped relationship suggests that both very high and very low levels of RPR are linked to negative health outcomes. Elevated RPR may indicate heightened cardiovascular risk and increased inflammation [40], whereas reduced RPR may reflect insufficient platelet response and a potential risk of bleeding. In summary, RDW, HRR, and RPR may serve as indicators of inflammation, nutritional status, and coagulation function, and in this ICU cohort they showed meaningful associations with long-term mortality among critically ill patients with hypertensive CKD.
Notably, the cohort of this study comprised critically ill ICU patients with substantial clinical heterogeneity. Given that RDW is highly sensitive to acute inflammatory stress and critical illness severity, its predictive signal within our cohort represents a combined manifestation of both acute organ injury and chronic renal damage. To isolate the prognostic effect of RDW from the confounding influence of acute disease burden, we adjusted for critical comorbidities in multivariate Cox regression models and conducted stratified subgroup analyses based on CRRT administration and CKD stages. The consistently significant statistical associations observed across all subgroups further verified that RDW still delivers supplementary prognostic information after accounting for acute critical illness factors.
Nevertheless, the biological plausibility of these associations does not automatically translate into strong clinical discriminatory power. Although RDW, HRR and RPR showed statistically significant associations with one‑year mortality, their modest AUC values (0.668, 0.600 and 0.515, respectively) indicate limited discriminatory ability when used in isolation. Rather, their clinical utility lies in serving as supplementary, readily available indicators that may refine risk perception when integrated into established prognostic scores — as supported by our NRI and IDI analyses. NRI and IDI analyses demonstrated that incorporating RDW and HRR into conventional scoring systems improved risk reclassification. Although these three biomarkers exhibited statistically significant incremental prognostic value, their low standalone AUC values limit their direct clinical application. Clinicians may only regard RDW > 15.95 or HRR < 6.27 as auxiliary warning indicators to intensify patient monitoring, nutritional support and inflammatory control, and clinical intervention strategies should never be formulated solely based on these hematological ratios. Further prospective studies with clinical endpoints as primary outcomes are warranted to verify whether integrating RDW and HRR into risk models can yield tangible clinical benefits for patients.
Stratified subgroup analyses further revealed effect modification across different clinical conditions. A significant interaction was observed between RDW and both CRRT administration and CKD stage. The positive association between elevated RDW and mortality risk persisted in all subgroups, while the correlation strength was weaker among patients receiving CRRT. For HRR, significant interaction was only found with CRRT status, and its protective correlation with mortality remained stable across different CKD stages. No obvious interaction was detected for RPR in either stratification factor. These subgroup discrepancies may be related to differences in systemic inflammation, hemodynamic status and organ function among patients receiving CRRT or at varying CKD stages. CRRT can regulate fluid balance and alleviate systemic inflammatory burden, which may weaken the prognostic implication of RDW. Meanwhile, the stable performance of HRR across CKD stages suggests its correlation with mortality is less dependent on renal impairment severity. Overall, the heterogeneous effects remind clinicians to interpret these biomarkers individually according to patients’ treatment and disease status in clinical practice.
This study has several strengths. Firstly, it is a large-scale retrospective cohort study including 6,731 hypertensive CKD patients. The large sample size enhances the reliability of the observed associations and reduces random error. Secondly, we specifically focused on critically ill hypertensive CKD patients admitted to ICU, a subgroup less frequently investigated in current literature, which fills relevant research gaps. Third, we comprehensively evaluated and compared the prognostic performance of RDW, HRR and RPR simultaneously, providing evidence for their potential application in risk stratification. These preliminary findings provide tentative evidence that these routine laboratory indicators may bring incremental predictive value when integrated into existing risk assessment models, and their clinical practicability needs further external verification.
This study has several notable limitations. First, its retrospective design and exclusive focus on ICU patients with hypertensive CKD limit generalizability to non-ICU or outpatient populations, and the lack of external validation precludes broad extrapolation of our results to all clinical scenarios. Second, residual confounding stemming from missing medication history cannot be ruled out, as MIMIC-IV contains no standardized records of long-term pre-admission medications. We could not extract or adjust for chronic use of RAAS inhibitors, other antihypertensive agents, statins and erythropoiesis-stimulating agents, all of which modify renal outcomes, anemia levels and mortality risk, leading to unmeasured bias within our prognostic models. Third, we only included subjects with pre-coordinated causal codes of ICD-9 category 403 and ICD-10 category I12. According to the official ICD coding guidelines, combination codes I12/403 can only be assigned when clinicians identify hypertension as the primary cause of renal injury. Even with standardized ICD grouping, misclassification and overdiagnosis of hypertensive nephropathy are common in epidemiological studies based on registry databases. To address this limitation, future studies should adopt renal biopsy pathological findings as a crucial diagnostic tool to exclude misclassified patients complicated with other primary renal diseases such as glomerulopathy, tubulointerstitial nephropathy and polycystic kidney disease, given that the histological features of hypertensive nephropathy itself are non-specific. Fourth, the lack of external validation in an independent cohort limits the generalizability of our findings. Although MIMIC-IV includes multiple ICU types with considerable clinical heterogeneity, it remains a single-center dataset. Future studies should validate our results in external multicenter cohorts before these biomarkers can be recommended for broader clinical application. Fifth, our primary Cox models omitted key ICU severity metrics (SOFA, SAPS II, OASIS) and critical comorbidities (sepsis, acute kidney injury, heart failure), leading to residual confounding that may bias the true associations of RDW, HRR and RPR with 1-year mortality. We partially addressed this via subgroup analyses by CRRT and CKD stage, and evaluated the incremental predictive value of the three biomarkers against baseline models incorporating these ICU scores, age and eGFR. Future studies should fully adjust for all critical illness indices to reduce residual confounding and validate our results. Finally, RDW lacks disease specificity, and its derivative indices HRR and RPR are subject to confounding by a variety of clinical conditions, including anemia, systemic inflammation, malnutrition (as reflected by low albumin levels), malignancy, history of blood transfusion, and acute illness severity. In addition, the absence of key laboratory parameters in the MIMIC-IV database—including iron metabolism markers (ferritin and transferrin saturation), inflammatory markers (C-reactive protein), and urinary albumin—together with the lack of standardized documentation regarding pre-admission pharmacotherapy (e.g., RAAS inhibitors, other antihypertensives, statins, erythropoiesis-stimulating agents, and iron supplementation), introduces residual confounding that further limits the prognostic utility of these biomarkers.
Conclusion
In this cohort of ICU patients with hypertensive CKD, RDW, HRR and RPR showed significant associations with one-year all-cause mortality. In this population, higher RDW was associated with increased mortality risk, while higher HRR was associated with reduced mortality risk, and RPR presented a U-shaped relationship with mortality. This study identifies independent associations between three hematological indices and one-year all-cause mortality. However, their standalone predictive ability is limited, and they should be used in conjunction with traditional prognostic scores to provide incremental information. Further large prospective investigations, independent external cohort validation and in-depth mechanistic research are urgently warranted to verify whether these indices can bring meaningful clinical benefit for risk stratification.
Supplementary Information
Acknowledgements
We express our gratitude to all the participants in the MIMIC-IV study and the project team. We would also like to extend our sincere thanks to Degang Mo for his valuable assistance with data acquisition and database access.
Abbreviations
- CKD
Chronic kidney disease
- RDW
Red blood cell distribution width
- HRR
Hemoglobin to RDW ratio
- RPR
RDW to platelet ratio
- ICU
Intensive care unit
- ICD
International Classification of Diseases
- MIMIC-IV
Medical Information Mart for Intensive Care-IV
- SBP
Systolic blood pressure
- DBP
Diastolic blood pressure
- MBP
Mean blood pressure
- SpO2
Peripheral capillary oxygen saturation
- T2DM
Type 2 diabetes mellitus
- COPD
Chronic obstructive pulmonary disease
- AMI
Acute myocardial infarction
- RF
Respiratory failure
- SOFA
Sequential organ failure assessment
- SAPS II
Simplified acute physiology score II
- OASIS
Oxford acute severity of illness score
- GCS
Glasgow Coma Scale
- WBC
White blood cells
- RBC
Red blood cells
- PLT
Platelets
- Hb
Hemoglobin
- INR
International normalized ratio
- BUN
Blood urea nitrogen
- eGFR
Estimated glomerular filtration rate
- CRRT
Continuous renal replacement therapy
- HR
Hazard ratio
- PH
Proportional hazards
- VIF
Variance inflation factor
- RCS
Restricted cubic spline
- ROC
Receiver operating characteristic
- NRI
Net reclassification improvement
- IDI
Integrated discrimination improvement
- CI
Confidence interval
- AUC
Area under the curve
Authors’ contributions
L.L. and Y.L. (Yanyu Lin) were responsible for the design and conceptualization of the study, in addition to drafting and revising the manuscript. S.Z. and Y.L. (Yuzhi Liu) contributed to data collecting, statistical analysis, and result interpretation. Y.L. (Yuzhi Liu) was responsible for the data results visualization, read and revised the manuscript. All authors read and approved the final manuscript.
Funding
The authors declare that no financial support was received for the research, authorship, and/or publication of this article.
Data availability
The data that support the findings of this study are publicly available in the MIMIC-IV database (https://doi.org/10.13026/kpb9-mt58).
Declarations
Ethics approval and consent to participate
The study was performed according to the guidelines of the Helsinki Declaration. The use of the MIMIC-IV database was approved by the review committee of Massachusetts Institute of Technology and Beth Israel Deaconess Medical Center. The data is publicly available (in the MIMIC-IV database), therefore, the ethical approval statement and the requirement for informed consent were waived for this study.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Ying M, Shao X, Qin H, Yin P, Lin Y, Wu J, et al. Disease Burden and Epidemiological Trends of Chronic Kidney Disease at the Global, Regional, National Levels from 1990 to 2019. Nephron. 2024;148(2):113–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Webster AC, Nagler EV, Morton RL, Masson P. Chronic Kidney Disease Lancet. 2017;389(10075):1238–52. [DOI] [PubMed] [Google Scholar]
- 3.Burnier M, Damianaki A. Hypertension as Cardiovascular Risk Factor in Chronic Kidney Disease. Circ Res. 2023;132(8):1050–63. [DOI] [PubMed] [Google Scholar]
- 4.Delrue C, Speeckaert MM. Beyond Blood Pressure: Emerging Pathways and Precision Approaches in Hypertension-Induced Kidney Damage. Int J Mol Sci. 2025;26(15):7606. [DOI] [PMC free article] [PubMed]
- 5.Wei J, Zhang J, Wang L, Cha BJ, Jiang S, Liu R. A new low-nephron CKD model with hypertension, progressive decline of renal function, and enhanced inflammation in C57BL/6 mice. Am J Physiol Ren Physiol. 2018;314(5):F1008–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Naas S, Schiffer M, Schodel J. Hypoxia and renal fibrosis. Am J Physiol Cell Physiol. 2023;325(4):C999–1016. [DOI] [PubMed] [Google Scholar]
- 7.Ono H, Ono Y. Nephrosclerosis and hypertension. Med Clin North Am. 1997;81(6):1273–88. [DOI] [PubMed] [Google Scholar]
- 8.Gargiulo R, Suhail F, Lerma EV. Hypertension and chronic kidney disease. Dis Mon. 2015;61(9):387–95. [DOI] [PubMed] [Google Scholar]
- 9.Lawson JS, Jepson RE. Feline comorbidities: The intermingled relationship between chronic kidney disease and hypertension. J Feline Med Surg. 2021;23(9):812–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Provenzano M, Coppolino G, Faga T, Garofalo C, Serra R, Andreucci M. Epidemiology of cardiovascular risk in chronic kidney disease patients: the real silent killer. Rev Cardiovasc Med. 2019;20(4):209–20. [DOI] [PubMed] [Google Scholar]
- 11.Ni Q, Wang X, Wang J, Chen P. The red blood cell distribution width-albumin ratio: A promising predictor of mortality in heart failure patients - A cohort study. Clin Chim Acta. 2022;527:38–46. [DOI] [PubMed] [Google Scholar]
- 12.Zhao N, Hu W, Wu Z, Wu X, Li W, Wang Y, et al. The Red Blood Cell Distribution Width-Albumin Ratio: A Promising Predictor of Mortality in Stroke Patients. Int J Gen Med. 2021;14:3737–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Tan M, Liu B, You R, Huang Q, Lin L, Cai D, et al. Red Blood Cell Distribution Width as a Potential Valuable Survival Predictor in Hepatitis B Virus-related Hepatocellular Carcinoma. Int J Med Sci. 2023;20(7):976–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Zhou Y, Zhong L, Chen W, Liang F, Liao Y, Zhong Y. Enhanced red blood cell distribution width to platelet ratio is a predictor of mortality in patients with sepsis: a propensity score matching analysis based on the MIMIC-IV database. BMJ Open. 2022;12(9):e062245. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Valenti AC, Vitolo M, Imberti JF, Malavasi VL, Boriani G. Red Cell Distribution Width: A Routinely Available Biomarker with Important Clinical Implications in Patients with Atrial Fibrillation. Curr Pharm Des. 2021;27(37):3901–12. [DOI] [PubMed] [Google Scholar]
- 16.Inker LA, Eneanya ND, Coresh J, Tighiouart H, Wang D, Sang Y, et al. New Creatinine- and Cystatin C-Based Equations to Estimate GFR without Race. N Engl J Med. 2021;385(19):1737–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Harrell FE Jr. Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis. 2nd ed. Springer; 2015.
- 18.Sterne JA, White IR, Carlin JB, Spratt M, Royston P, Kenward MG, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ. 2009;338:b2393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lu YA, Fan PC, Lee CC, Wu VC, Tian YC, Yang CW, et al. Red cell distribution width associated with adverse cardiovascular outcomes in patients with chronic kidney disease. BMC Nephrol. 2017;18(1):361. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Yilmaz F, Sozel H. Red blood cell distribution width is a predictor of chronic kidney disease progression and all-cause mortality. Bratisl Lek Listy. 2021;122(1):49–55. [DOI] [PubMed] [Google Scholar]
- 21.Hsieh YP, Chang CC, Kor CT, Yang Y, Wen YK, Chiu PF. The Predictive Role of Red Cell Distribution Width in Mortality among Chronic Kidney Disease Patients. PLoS ONE. 2016;11(12):e0162025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Deng X, Gao B, Wang F, Zhao MH, Wang J, Zhang L. Red Blood Cell Distribution Width Is Associated With Adverse Kidney Outcomes in Patients With Chronic Kidney Disease. Front Med (Lausanne). 2022;9:877220. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Li ZZ, Chen L, Yuan H, Zhou T, Kuang ZM. Relationship between red blood cell distribution width and early-stage renal function damage in patients with essential hypertension. J Hypertens. 2014;32(12):2450–5. discussion 6. [DOI] [PubMed] [Google Scholar]
- 24.Xiu WJ, Zheng YY, Wu TT, Hou XG, Yang Y, Ma YT, et al. Hemoglobin-to-Red-Cell Distribution Width Ratio Is a Novel Predictor of Long-Term Patient Outcomes After Percutaneous Coronary Intervention: A Retrospective Cohort Study. Front Cardiovasc Med. 2022;9:726025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Rahamim E, Zwas DR, Keren A, Elbaz-Greener G, Ibrahimli M, Amir O et al. The Ratio of Hemoglobin to Red Cell Distribution Width: A Strong Predictor of Clinical Outcome in Patients with Heart Failure. J Clin Med. 2022;11(3):886. [DOI] [PMC free article] [PubMed]
- 26.Cetinkaya E, Senol K, Saylam B, Tez M. Red cell distribution width to platelet ratio: new and promising prognostic marker in acute pancreatitis. World J Gastroenterol. 2014;20(39):14450–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Chen B, Ye B, Zhang J, Ying L, Chen Y. RDW to platelet ratio: a novel noninvasive index for predicting hepatic fibrosis and cirrhosis in chronic hepatitis B. PLoS ONE. 2013;8(7):e68780. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Pusuroglu H, Cakmak HA, Akgul O, Erturk M, Surgit O, Akkaya E, et al. The prognostic value of admission red cell distribution width-to-platelet ratio in patients with ST-segment elevation myocardial infarction undergoing primary percutaneous coronary intervention. Rev Port Cardiol. 2015;34(10):597–606. [DOI] [PubMed] [Google Scholar]
- 29.Wu J, Huang L, He H, Zhao Y, Niu D, Lyu J. Red Cell Distribution Width to Platelet Ratio Is Associated with Increasing In-Hospital Mortality in Critically Ill Patients with Acute Kidney Injury. Dis Markers. 2022;2022:4802702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Salvagno GL, Sanchis-Gomar F, Picanza A, Lippi G. Red blood cell distribution width: A simple parameter with multiple clinical applications. Crit Rev Clin Lab Sci. 2015;52(2):86–105. [DOI] [PubMed] [Google Scholar]
- 31.Montagnana M, Cervellin G, Meschi T, Lippi G. The role of red blood cell distribution width in cardiovascular and thrombotic disorders. Clin Chem Lab Med. 2011;50(4):635–41. [DOI] [PubMed] [Google Scholar]
- 32.Solak Y, Yilmaz MI, Saglam M, Caglar K, Verim S, Unal HU, et al. Red cell distribution width is independently related to endothelial dysfunction in patients with chronic kidney disease. Am J Med Sci. 2014;347(2):118–24. [DOI] [PubMed] [Google Scholar]
- 33.Yan Z, Xu G. A Novel Choice to Correct Inflammation-Induced Anemia in CKD: Oral Hypoxia-Inducible Factor Prolyl Hydroxylase Inhibitor Roxadustat. Front Med (Lausanne). 2020;7:393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Mansour GK, Hajjar AW, Sajid MR. Therapeutic targeting of the hepcidin-ferroportin axis and erythropoietic modulators: a narrative review. Front Med (Lausanne). 2025;12:1726337. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Baaten C, Vondenhoff S, Noels H. Endothelial Cell Dysfunction and Increased Cardiovascular Risk in Patients With Chronic Kidney Disease. Circ Res. 2023;132(8):970–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Yonemoto S, Hamano T, Fujii N, Shimada K, Yamaguchi S, Matsumoto A, et al. Red cell distribution width and renal outcome in patients with non-dialysis-dependent chronic kidney disease. PLoS ONE. 2018;13(6):e0198825. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Xu R, Wu Z, Liu Z. Hemoglobin-to-red blood cell distribution width ratio: a new insight into cognitive protection for obese individuals. Front Med (Lausanne). 2025;12:1625542. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Babitt JL, Lin HY. Mechanisms of anemia in CKD. J Am Soc Nephrol. 2012;23(10):1631–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Piccoli GB, Cederholm T, Avesani CM, Bakker SJL, Bellizzi V, Cuerda C, et al. Nutritional status and the risk of malnutrition in older adults with chronic kidney disease - implications for low protein intake and nutritional care: A critical review endorsed by ERN-ERA and ESPEN. Clin Nutr. 2023;42(4):443–57. [DOI] [PubMed] [Google Scholar]
- 40.Liu J, Huang X, Yue S, Wang J, Ye E, Huang J, et al. Association of Red Cell Distribution Width-to-Platelet Ratio and Mortality in Patients with Sepsis. Mediators Inflamm. 2022;2022:4915887. [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
Data Availability Statement
The data that support the findings of this study are publicly available in the MIMIC-IV database (https://doi.org/10.13026/kpb9-mt58).