Abstract
Objectives
Red blood cell distribution width (RDW) with prognosis in various infectious diseases. For fractured patients admitted to the intensive care units (ICU), an accurate and fast appraisal is essential. To investigate the association between RDW and prognosis in fractured patients admitted to the ICU utilizing the MIMIC‐III database.
Methods
A retrospective cohort from the MIMIC III database from 2001 and 2012 was constructed. RDW and other information were collected with in‐hospital mortality as the primary outcome and 90‐day mortality and hospital and intensive care unit (ICU) length of stay (LOS) as secondary outcomes. Univariate and multivariate logistic regression models with propensity score inverse probability of treatment weighting (IPTW) were used to investigate the prognostic value of RDW. A nomogram was built with significant prognostic factors to predict in‐hospital mortality, and the performance of the nomogram was evaluated and compared with other severity assessment scores. Subgroup analysis was also conducted.
Results
A total of 2721 fracture patients admitted to the ICU were identified. After IPTW, the group with higher RDW was significantly associated with elevated in‐hospital mortality (odds ratio [OR]: 1.68, 95% confidence interval [CI]: 1.19–2.37), 90‐day mortality (OR: 1.39, 95% CI: 1.04–1.86), prolonged hospital LOS (OR: 1.25, 95% CI: 1.03–1.50), and ICU LOS significantly (OR: 1.26, 95% CI: 1.05–1.53) in the multivariate logistics model. The nomogram showed optimal discriminative ability and predictive accuracy with an area under the receiver operating characteristic curve of 0.77.
Conclusion
RDW independently predicted in‐hospital mortality, 90‐day mortality, and hospital and ICU LOS in fractured patients admitted to ICU. The nomogram including RDW could also be a promising tool with potential clinical benefits.
Keywords: Fracture Patients, Mortality, Prognosis, Red Blood Cell Distribution Width
Elevated RDW was linked to the poor prognosis for critically ill fracture patients. The higher RDW was associated with the increased risk of in‐hospital mortality and 90‐day mortality, as well as prolonged ICU and hospital LOS. RDW: Red blood cell distribution width; ICU: Intensive care unit; LOS: length of stay.
Introduction
Bone fracture is a common physical injury resulting from many factors, such as trauma, osteoporosis, and cancer. It could be a great burden for patients both economically and mentally. 1 , 2 , 3 The incidences of osteoporosis fracture and hip fracture are estimated to reach almost 175,000 and 2.6 million by 2050, respectively. 4 , 5 Patients with hip fractures are more likely to experience surgical complications, disability, and high 1‐year mortality estimated at 30%. 6 Severe fracture patients usually require surgery to help with bone healing and body recovery, while the healing process is complex, both biologically 7 and biomechanically. 8 , 9
Accurate and timely judgment of patients' conditions is crucial in fractured patients admitted to ICU for better clinical treatment decisions. C‐creative protein, white blood cells, and hematocrit are widely used indicators at present. 10 In one retrospective single‐center cohort analysis, we found that the serum anion gap (AG) can be used as a risk stratification tool for hip fracture. 11 However, they generally lack accuracy and specificity. 12 , 13 For individuals with isolated hip fractures, the severity of illness (SOI) score may be a better indicator of outcomes. 14 Other scores, the Nottingham Hip Fracture Score and the orthopaedic version of the Physiologic and Operative Severity Score for the Enumeration of Mortality and Morbidity, and previously published risk prediction models could be time‐consuming and inconvenient. 15 , 16 , 17
The red blood cell distribution width (RDW), which is the indication of size diversity among circulating red blood cells, has been as a biomarker reflecting systemic inflammation and malnutrition among elderly people. 18 Nevertheless, its association with prognosis in various infectious diseases 19 , 20 and cancers 21 , 22 , 23 , 24 , 25 , 26 has been recently noted. Recent studies have investigated the potential relationship between vertebral body fracture, hip fracture, and RDW. 27 , 28 However, to our knowledge, few studies have investigated the significance of RDW and RDW in fractured patients admitted to ICU.
Hence, the goal of this study was to determine whether RDW had predictive value in fractured patients admitted to ICU and develop a nomogram to predict the probability of in‐hospital mortality with performance evaluation.
Methods
Study Design and Data Source
The study used data from the Medical Information Mart for Intensive Care‐III (MIMIC‐III) database and is a retrospective cohort study. 29 The MIMIC‐III is a publicly available critical care database that includes 50,000 hospital admissions comprising 38,645 adults as well as 7875 neonates admitted to surgical, trauma surgery, coronary, and cardiac surgery recovery intensive care units (ICUs) of Beth Israel Deaconess Medical Center in Boston from 2001 to 2012. The institutional review boards of both Beth Israel Deaconess Medical Center and Massachusetts Institute of Technology Affiliates allowed access to the database (authorization code: 40043439). We acquired anonymized data from a database; thus, informed permission was not needed. Ethical approval and consent were not required for the present study. This research is reported in compliance with the Strengthening the Reporting of Observational Studies in Epidemiology statement.
Study Population
The patients with fractures were identified by the International Classification of Diseases‐9 term associated with fracture. Patients from 18 to 89 years old were enrolled in the study. If patients were hospitalized many times, only the first hospital admission with ICU stay was examined. Patients who spent fewer than 24 h in the ICU were also excluded considering that the patient's condition would be either too mild or too severe.
Data Collection and Definitions
The data were extracted from the database using structure query language (SQL) with PostgreSQL (version 9.4.6, www.postgresql.org). The variables in this study included: (1) demographics; (2) hospitalization and prognosis: in‐hospital mortality, 90‐day mortality, ICU and hospital length of stay (LOS); (3) mean value of severity scores containing simplified acute physiology score (SAPS II), sequential organ failure assessment (SOFA), Glasgow coma scale (GCS), and Elixhauser scores in the first 24 h after ICU admission; (4) comorbidities; (5) mean laboratory results in the first 24 h after ICU admission; and (6) mean vital signs value in the first 24 h after ICU admission. The RDW was examined both as a continuous variable and as quartiles. The ICU and hospital length of stay were dichotomized into two groups for the following analysis.
To avoid potential bias, variables with more than 30% missing values were omitted from the following analysis. Using the multiple imputation method, we completed variables with fewer than 30% missing data. 30
In‐hospital mortality was chosen as the primary outcome because we were interested in the prognosis of fracture patients. Secondary outcomes included 90‐day mortality and hospital and ICU length of stay (LOS).
Statistical Analysis
The median and standard deviation (SD) for continuous variables and proportions for categorical variables were used to report demographics and clinical features. To determine normality, the Kolmogorov–Smirnov test was performed on each continuous variable. T‐tests or the Mann–Whitney U test were used for continuous variables, while for categorical variables, chi‐square or Fisher's exact tests were used.
A logistic linear regression model was implemented to identify the associations between the covariates and prognosis. First, we assessed the covariates with significant associations with outcomes using univariate logistic linear regression. The statistically significant covariates (p value <0.05) and a change in the effect estimate exceeding 10% in the univariate logistic analysis regarding the four outcomes were identified. We excluded the severity scores from the covariates for the multivariate analysis to avoid potential interference with the results. The multivariate analysis was conducted with the remaining covariates. An inverse probability of treatment weighting (IPTW) analysis was applied in the logistic models after adjusting the following covariates: age, fracture position, gender, ethnicity, admission type, congestive heart failure, hypertension, chronic pulmonary, renal failure, liver disease, rheumatoid arthritis, obesity, diabetes, and anemia. The IPTW analysis was derived to reduce selection bias by statistically adjusting for background factors using propensity scores on all observations before matching. 31 Based on the significant covariates in the multivariate analysis, a dynamic nomogram for in‐hospital mortality was constructed. The performance of the nomogram was assessed by discrimination and accuracy by the area under the receiver operating characteristic curve (AUC), calibration plot, and the Hosmer‐Lemeshow test (H‐L test). For the calibration plot, the nomogram was subjected to 1000 bootstrap resamples for internal validation. Additionally, we performed subgroup analyses for the unmatched cohort using the nomogram to further evaluate the prognostic value of RDW regarding age, anemia, diabetes, and ICU length of stay. All data cleaning, statistical analyses, and part of the illustrations were performed in R software (version 4.0.3) with “tableone,” 32 “ggplot2,” 33 “tidyverse,” 34 “lubridate,” 35 “pROC,” 36 “surve,” 37 “DynNom,” 38 “rsconnect,” 39 “rms,” 40 and “ResourceSelection.” 41 A p value <0.05 was considered statistically significant.
Results
Baseline Characteristics
A total of 2721 fracture patients were eventually enrolled in this study, as shown in Figure 1. Patients were stratified by the median RDW value: 13.85, and the basic characteristics are shown in Table 1. The majority of fracture patients were identified as having a skull fracture (67.4%) and lower limb fracture (16.2%). The other types of fracture were the upper limb fracture (8.4%) and pathologic or stress fracture (8.0%). The proportions of anemia among the lower and higher RDW groups were 7.3% and 16.1%, respectively.
FIGURE 1.
Flowchart of included patients. ICU: intensive care unit; RDW: red blood cell distribution width, N: number of patients
TABLE 1.
Baseline patient characteristics
| Lower RDW group | Higher RDW group | p value | |
|---|---|---|---|
| Number of patients | 1345 | 1376 | |
| Age, mean (SD) | 47.5 (21.3) | 61.0 (19.2) | <0.01 |
| Gender: male, n (%) | 933 (69.4) | 795 (57.8) | <0.01 |
| Admission type, n (%) | <0.01 | ||
| Elective | 11 (0.8) | 44 (3.2) | |
| Emergency | 1326 (98.6) | 1319 (95.9) | |
| Urgent | 8 (0.6) | 13 (0.9) | |
| Ethnicity, n (%) | 0.14 | ||
| White | 980 (72.9) | 1016 (73.8) | |
| Not specified | 249 (18.5) | 231 (16.8) | |
| Black | 42 (3.1) | 63 (4.6) | |
| Hispanic | 57 (4.2) | 45 (3.3) | |
| Asian | 17 (1.3) | 21 (1.5) | |
| Severity score, mean (SD) | |||
| GCS | 13.9 (2.3) | 13.6 (2.5) | 0.01 |
| SOFA | 2.5 (2.1) | 4.1 (2.9) | <0.01 |
| SPAS II | 25.6 (12.3) | 35.1 (13.7) | <0.01 |
| Elixhauser scores | 0.1 (1.1) | 0.4 (2.1) | <0.01 |
| Fracture position, n (%) | <0.01 | ||
| Lower limb fracture | 152 (11.3) | 288 (20.9) | |
| Pathologic or stress fracture | 38 (2.8) | 181 (13.2) | |
| Skull fracture | 1037 (77.1) | 796 (57.8) | |
| Upper limb fracture | 118 (8.8) | 111 (8.1) | |
| Comorbidities, n (%) | |||
| Congestive heart failure | 69 (5.1) | 212 (15.4) | <0.01 |
| Hypertension | 17 (1.3) | 100 (7.3) | <0.01 |
| Chronic pulmonary disease | 105 (7.8) | 211 (15.3) | <0.01 |
| Renal failure | 15 (1.1) | 123 (8.9) | <0.01 |
| Liver disease | 24 (1.8) | 87 (6.3) | <0.01 |
| Rheumatoid arthritis | 15 (1.1) | 55 (4.0) | <0.01 |
| Obesity | 20 (1.5) | 68 (4.9) | <0.01 |
| Diabetes | 109 (8.1) | 268 (19.5) | <0.01 |
| Anemia | 98 (7.3) | 221 (16.1) | <0.01 |
| Vital signs, mean (SD) | |||
| Heart rate | 86.2 (15.6) | 89.5 (15.9) | <0.01 |
| SBP | 124.7 (14.5) | 121.6 (16.0) | <0.01 |
| DBP | 64.1 (9.9) | 62.0 (10.8) | <0.01 |
| MBP | 82.1 (10.1) | 80.0 (10.9) | <0.01 |
| RR | 17.7 (3.4) | 18.6 (3.8) | <0.01 |
| T | 37.1 (0.6) | 37.0 (0.6) | <0.01 |
| SpO2 | 97.9 (1.7) | 97.6 (2.0) | <0.01 |
| Laboratory results, mean (SD) | |||
| Hematocrit, % | 33.8 (4.8) | 30.8 (4.9) | <0.01 |
| Hemoglobin, g/dL | 11.7 (1.7) | 10.5 (1.7) | <0.01 |
| Platelet count, 109/L | 212.9 (68.5) | 195.6 (101.8) | <0.01 |
| WBC, 109/L | 12.1 (4.2) | 11.6 (6.1) | 0.01 |
| RBC, m/μL | 3.8 (0.6) | 3.5 (0.6) | <0.01 |
| MCV, fL | 89.7 (4.9) | 88.9 (6.4) | <0.01 |
| RDW, % | 13.1 (0.5) | 15.4 (1.6) | <0.01 |
| Glucose, mg/dL | 134.6 (35.8) | 142.6 (37.8) | <0.01 |
| Anion gap, mEq/L | 13.4 (2.7) | 13.5 (3.0) | 0.29 |
| Bicarbonate, mEq/L | 24.1 (3.0) | 23.5 (4.0) | <0.01 |
| Creatinine, mg/dL | 0.8 (0.4) | 1.1 (1.0) | <0.01 |
| Chloride, mEq/L | 105.7 (4.6) | 106.6 (5.5) | <0.01 |
| Potassium, mEq/L | 4.0 (0.4) | 4.1 (0.5) | <0.01 |
| PTT, s | 29.0 (11.9) | 33.1 (15.3) | <0.01 |
| INR | 1.2 (0.3) | 1.3 (0.5) | <0.01 |
| PT, s | 13.8 (3.2) | 14.8 (4.1) | <0.01 |
| Sodium, mEq/L | 139.0 (3.6) | 139.2 (4.2) | 0.17 |
| BUN, mg/dL | 14.5 (8.4) | 20.8 (15.5) | <0.01 |
| Calcium, mg/dL | 8.3 (0.6) | 8.1 (0.7) | <0.01 |
| Hospitalization, mean (SD) | |||
| ICU interval, day | 5.3 (6.8) | 6.3 (7.6) | <0.01 |
| Hospital interval, day | 11.5 (11.9) | 13.9 (12.4) | <0.01 |
| Prognosis, n (%) | |||
| In‐hospital mortality | 70 (5.2) | 193 (14.0) | <0.01 |
| 90‐day mortality | 112 (8.3) | 331 (24.1) | <0.01 |
Abbreviations: BUN, Blood urea nitrogen; DBP, Diastolic blood pressure; GCS, Glasgow coma scale; ICU, Intensive care unit; INR, International normalized ratio; MBP, Mean blood pressure; MCV, Mean corpuscular volume; PT, Prothrombin time; PTT, Partial thromboplastin time; RBC, Red blood cell; RDW, Red blood cell distribution width; RDW, Red blood cell distribution width; RR, Respire rate; SBP, Systolic blood pressure; SD, Standard deviation; SOFA, Sequential organ failure assessment score; SPAS II, Simplified acute physiology score II; SpO2, Oxygen saturation; T, Temperature; WBC, White blood cell.
Primary and Secondary Outcomes
As shown in Figure 2, in the univariable logistic regression model, the higher RDW group was significantly associated with elevated in‐hospital mortality (odds ratio [OR]: 2.97, 95% confidence interval [CI]: 2.24–3.95), 90‐day mortality (OR: 3.49, 95% CI: 2.77–4.39), hospital LOS (OR: 1.57, 95% CI: 1.35–1.83), and ICU LOS (OR: 1.61, 95% CI: 1.39–1.88). The balance pre‐ and post‐IPTW were shown in Table S1. After IPTW matching, the higher RDW group remained significantly associated with in‐hospital mortality (OR: 2.08, 95% CI: 1.44–3.00), 90‐day mortality (OR: 1.64, 95% CI: 1.22–2.22), hospital LOS (OR: 1.43, 95% CI: 1.20–1.71), and ICU LOS (OR: 1.56, 95% CI: 1.30–1.86). The mean RDW showed a similar significant association with these four outcomes (Table S2). In the IPTW matched cohort of multivariable logistic regression results (Figure 2), the higher RDW group was significantly associated with elevated in‐hospital mortality (OR: 1.68, 95% CI: 1.19–2.37), 90‐day mortality (OR: 1.39, 95% CI: 1.04–1.86), prolonged hospital LOS (OR: 1.25, 95% CI: 1.03–1.50), and ICU LOS (OR: 1.26, 95% CI: 1.05–1.53). The detailed results of univariate and multivariate results are provided in Tables S2 and S3.
FIGURE 2.
The univariate and multivariate results of the RDW groups and clinical outcomes in fractured patients admitted to ICU before and post IPTW matching. The forest plot shows the odds ratio (black circle), lower and upper levels (two ends of the line) of the 95% odds ratio. RDW: red blood cell distribution width; ICU intensive care unit; LOS: length of stay; N: number of patients; OR: odds ratio; CI: confidence interval; IPTW: inverse probability of treatment weighting. All p values <0.05 are bolded
With the aforementioned method, the prognostic nomogram was established based on fracture position, congestive heart failure, bicarbonate, anion gap, sodium, and RDW group. As the respiration rate and temperature could be extreme in certain circumstances, we excluded these significant covariates for the generalization of the model and nomogram. This final multivariable regression model demonstrated optimal predictive discrimination for in‐hospital mortality with an AUC of 0.77 (95% CI: 0.74–0.79) (Figure 3A). To validate the accuracy, the model was performed using bootstrap analyses with 1000 resamples before plotting the calibration plot (Figure 3B), which indicates good agreement between the predicted and observed values. Moreover, the H‐L test result showed a p value = 0.48, suggesting the goodness of the model fitting.
FIGURE 3.
(A) The receiver operating characteristic plot demonstrated optimal predictive discrimination (AUC: 0.77, 95% CI: 0.74–0.79) for the multivariate logistic model. (B) Model accuracy is visualized by comparing predicted vs. actual probabilities, demonstrating apparent and bias‐corrected predictive ability. The vertical lines at the top of the plot showed the relative prevalence of probability levels
The probabilities of in‐hospital mortality can be estimated for fractured patients admitted to ICU. The dynamic nomogram was created allowing automated estimation of probabilities with 95% confidence intervals based on the inputs (https://ly-scu-wch.shinyapps.io/DynNomapp/, Figure 4). With the aid of a nomogram, it was possible to effectively predict prognosis according to patient information. The discrimination ability of the nomogram was compared with the other severity scores, as illustrated in Table 2. Interestingly, SOFA, Elixhauser scores, and GCS showed a significant decrease in AUC compared with the nomogram, while SPAS II showed an insignificant AUC increase.
FIGURE 4.
Dynamic nomogram (https://ly‐scu‐wch.shinyapps.io/DynNomapp/) for in‐hospital mortality risk estimation of fractured patients admitted to ICU
TABLE 2.
Odds ratio, discrimination ability, and H‐L test for the in‐hospital mortality prognostic model comprised of different severity scores and compared with the dynamic nomogram containing red blood cell distribution width
| Severity scores | OR (95% CI)a | p valueb | AUC | Compared with dynamic nomogramc | p valued | H‐L test |
|---|---|---|---|---|---|---|
| SOFA | 1.34 (1.28–1.41) | <0.01 | 0.73 (0.69, 0.76) | −1.97 (−0.08, 0) | 0.05 | 0.36 |
| SPAS II | 1.08 (1.07–1.09) | <0.01 | 0.79 (0.76, 0.82) | 1.22 (−0.01, 0.06) | 0.22 | 0.55 |
| Elixhauser scores | 1.1 (1.03–1.17) | <0.01 | 0.52 (0.48, 0.56) | −10.78 (−0.29, −0.2) | <0.01 | 0.06 |
| GCS | 0.88 (0.83–0.93) | <0.01 | 0.53 (0.51, 0.54) | −14.78 (−0.27, −0.21) | <0.01 | 1 |
Note: a,b, the OR, 95% CI, and the p value of univariates logistic analysis for in‐hospital mortality.
Note: c,d, the Z value, 95% CI, and p value for the comparison of AUROC value between the severity scores with the dynamic nomogram containing red blood cell distribution width.
Abbreviations: AUC, Area under the receiver operating characteristic curve; CI, confidential interval; GCS, Glasgow coma scale; H‐L, Hosmer–Lemeshow; OR, odds ratio; SOFA, sequential organ failure assessment score; SPAS II, simplified acute physiology score II.
Subgroup Analysis
According to the results of the subgroup analysis (Table 3), the higher RDW group was linked to higher in‐hospital death rates in elderly and younger patients (OR: 1.91, 95% CI: 1.31–2.77; and OR: 2.12, 95% CI: 1.20–3.73, respectively), patients without anemia (OR: 2.85, 95% CI: 2.05–3.94), patients with diabetes (OR: 3.40, 95% CI: 1.13–10.2), patients without diabetes (OR: 2.47, 95% CI: 1.74–3.41), patients with a longer ICU LOS (OR: 2.13, 95% CI: 1.46–3.09), and patients with a shorter ICU LOS (OR: 3.30, 95% CI: 1.92–5.68). Notably, in the nonanemic patients, the higher RDW group did not show a significant association with in‐hospital mortality (OR: 0.73, 95% CI: 0.29–1.82).
TABLE 3.
Subgroup analysis results
| Subgroups | RDW group | p value | |
|---|---|---|---|
| N | OR (95% CI) | ||
| Age ≥ 55 | 894 | 1.91 (1.31–2.77) | <0.01 |
| Age < 55 | 482 | 2.12 (1.2–3.73) | 0.01 |
| With anemia | 221 | 0.73 (0.29–1.82) | 0.5 |
| Without anemia | 1155 | 2.85 (2.05–3.94) | <0.01 |
| With diabetes | 268 | 3.4 (1.13–10.2) | 0.03 |
| Without diabetes | 1108 | 2.47 (1.79–3.41) | <0.01 |
| ICU LOS longer | 771 | 2.13 (1.46–3.09) | <0.01 |
| ICU LOS shorter | 605 | 3.3 (1.92–5.68) | <0.01 |
Note: All p values <0.05 are bolded.
Abbreviations: CI, confidential interval; ICU, intensive care unit; LOS, length of stay; N, number of patients; OR, odds ratio; RDW, red blood cell distribution width.
Discussion
Summarize Results and Previous Studies
In our study, elevated RDW was independently associated with an increased risk of in‐hospital mortality and 90‐day mortality and prolonged ICU and hospital LOS in fracture patients. Additionally, a multivariate logistic model including RDW together with a dynamic nomogram was established to predict in‐hospital mortality, and its discrimination and calibration ability was proven to be optimal.
RDW has been studied as a valid short‐term and long‐term prognostic factor in hip fracture and osteoporotic vertebral fracture patients. 28 , 42 , 43 In osteoporotic vertebral fracture patients, Sakai et al. demonstrated that elevated RDW (>15.0) was an independent factor associated with abasia (OR: 1.227, 95% CI: 1.003–1.500) for a 1‐year follow‐up period with 460 patients. 27 In another study, 28 with 203 patients divided by RDW less or greater than 13.35, elevated RDW was the key predictor of 30‐day mortality in older patients undergoing hip fracture surgery (hazard ratio (HR): 2.73, 95% CI: 2.06–3.62). Lv et al. 43 investigated the prognostic value of RDW over a 2‐year follow‐up period with 1479 patients. RDW showed a significant association with both 2‐year mortality (HR: 1.183, 95% CI: 1.017 to 1.376) and 4‐year mortality independently (HR: 1.244, 95% CI, 1.052 to 1.471). 43 To our knowledge, this is the first study to use a publicly available database to investigate the prognostic value of RDW in critically fractured patients admitted to ICU. Systemic inflammation might be a pivotal mediator in the association between RDW and the mortality of fractured patients admitted to ICU. The potential mechanism may be that the inflammatory responses could suppress renal erythropoietin (EPO) production, impair red cell survival, and cause the release of premature red cells into the circulation, resulting in an elevation of RDW. 44 , 45 This is more likely the case in fracture patients admitted to ICU whose injury might be complicated by prevalent inflammation.
New Insights of Study
Notably, in this study, the authors found a significant association between elevated RDW values and increased odds of all‐cause mortality in nonanemic patients but not in anemic patients, 43 which was persistent in the present study. Although the validity of this result in our study can be undermined by the small number of patients, special attention should be paid, especially to anemic patients, as RDW might be affected by hematogenesis leading to a decrease in prognostic value. 42 , 43 , 46 Therefore, for anemic fracture patients, the value of RDW in the differential diagnosis of anemia may outweigh the prognostic value of mortality. 43 , 47
On the other hand, the predictive value of elevated RDW for prognosis has been investigated in various studies and groups of patients, e.g., cardiovascular disorders, 46 pulmonary hypertension, 48 sepsis, 24 , 49 and COVID‐19. 50 For heart failure patients, RDW was predictive of mortality 46 (hazard ratio (HR): 1.18; 95% CI: 1.12, 1.24). Hampole et al. demonstrated that the highest of the three tertiles of RDW was associated with the mortality of pulmonary patients in a multivariate model (HR: 2.4, 95% CI: 1.02, 5.84) with a mean follow‐up duration of 2.1 years. 48 In a recently published article, the authors found that RDW was independently associated with mortality and that 24‐h RDW and admission RDW, when added to the severity scores, could improve the discrimination ability of SOFA, Logistic Organ Dysfunction System, Acute Physiology and Chronic Health Evaluation‐II, and SAPS‐II. 49 However, the underlying mechanism linking RDW with poor prognosis remains unclear. Researchers have revealed that the association of RDW and RPR with oxidative stress and inflammatory cytokines such as tumor necrosis factor (TNF)‐alpha, interleukin (IL)‐1, and IL‐6 could contribute to its predictive potential. 51 , 52 , 53
Strengths and Limitations
There are some strengths to our study. First, the included patients were admitted to ICU of multiple institutions and the number of included patients was more than that of previous retrospective observational studies. Also, we performed the subgroup analysis according to the causes of bone fracture and the conditions of patients to investigate the association between the RDW and the prognosis in different groups of patients.
This study also has several limitations. First, the establishment of the nomogram and prognostic model was based on a single retrospective cohort from the MIMIC dataset, which might influence the results by possible selection bias and its inherent retrospective nature. Second, although we addressed the association between RDW and the prognosis of fracture patients, the causation remains unknown. Third, some clinical information was excluded during the patient selection and data cleaning process. Moreover, the results of our study still need more external validation.
Conclusion
In summary, in fracture patients admitted to ICU, RDW appears to be a simple‐to‐use independent predictive indicator. For the first time, we constructed a prognostic nomogram for fractured patients admitted to ICU, which could be an easily accessible clinical tool facilitating counseling. These findings, however, need further verification and external validation.
Competing Interest
The authors declare no competing interests.
Author's Contributions
Kaibo Sun and Bin Shen contributed to the conception and design. Yuangang Wu , Yi Zeng, and Jiawen Xu contributed to the acquisition, analysis, and visualization of the data. Kaibo Sun and Yannan Zhou wrote the main manuscript text. Bin Shen contributed to the supervision and review. All authors have approved the submission version.
Ethical Statement
This study was in accordance with ethical approvals and all methods were in compliance with relevant guidelines and regulations.
Availability of Data and Materials
The datasets used or analyzed during the current study are available from the corresponding author on reasonable request.
Authorship Declaration
All authors listed meet the authorship criteria according to the latest guidelines of the International Committee of Medical Journal Editors. All authors are in agreement with the manuscript.
Disclosure Statement
All authors declared no financial support or relationships that may pose a conflict of interest.
Supporting information
Table S1. The covariates balancing in pre‐IPTW and post‐IPTW matching
Table S2. The univariate logistic regression models with pre‐IPTW and post‐IPTW matching in each outcome
Table S3. The multivariate logistic regression models with pre‐IPTW and post‐IPTW matching in each outcome
Acknowledgments
This study was funded by the National Natural Science Foundation of China (Program No. 81974347), China Postdoctoral Science Foundation (No.2021M702351), Medical Science and Technology Project of the Health Commission of Sichuan Provincial (No.21PJ040).
Kaibo Sun and Yannan Zhou Contributed equally to this work.
References
- 1. Mohd‐Tahir NA, Li SC. Economic burden of osteoporosis‐related hip fracture in Asia: a systematic review. Osteoporos Int. 2017;28(7):2035–44. [DOI] [PubMed] [Google Scholar]
- 2. Beall DP, Chambers MR, Thomas S, Amburgy J, Webb JR Jr, Goodman BS, et al. Prospective and multicenter evaluation of outcomes for quality of life and activities of daily living for balloon kyphoplasty in the treatment of vertebral compression fractures: the EVOLVE trial. Neurosurgery. 2019;84(1):169–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Gold DT. The nonskeletal consequences of osteoporotic fractures: psychologic and social outcomes. Rheum Dis Clin North Am. 2001;27(1):255–62. [DOI] [PubMed] [Google Scholar]
- 4. Abrahamsen B, Van Staa T, Ariely R, Olson M, Cooper C. Excess mortality following hip fracture: a systematic epidemiological review. Osteoporos Int. 2009;20(10):1633–50. [DOI] [PubMed] [Google Scholar]
- 5. Mithal A, Kaur P. Osteoporosis in Asia: a call to action. Curr Osteoporos Rep. 2012;10(4):245–7. [DOI] [PubMed] [Google Scholar]
- 6. Aguado HJ, Castillón‐Bernal P, Ventura‐Wichner PS, Cervera‐Díaz MC, Abarca‐Vegas J, García‐Flórez L, et al. Impact of subtrochanteric fractures in the geriatric population: better pre‐fracture condition but poorer outcome than pertrochanteric fractures: evidence from the Spanish hip fracture registry. J Orthop Traumatol. 2022;23(1):17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Marsell R, Einhorn TA. The biology of fracture healing. Injury. 2011;42(6):551–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Glatt V, Evans CH, Tetsworth K. A concert between biology and biomechanics: the influence of the mechanical environment on bone healing. Front Physiol. 2017;7:678. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. White AA, Panjabi MM, Southwick WO. The four biomechanical stages of fracture repair. J Bone Jt Surg. 1977;59(2):188–92. [PubMed] [Google Scholar]
- 10. Scherer MA, Neumaier M, von Gumppenberg S. C‐reactive protein in patients who had operative fracture treatment. Clin Orthop Relat Res. 2001;393:287–293. [DOI] [PubMed] [Google Scholar]
- 11. Zhang XB, Shu WB, Li AB, Lan GH. The anion gap and mortality in critically ill patients with hip fractures. Contrast Media Mol Imaging. 2022;2022:1591507–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Simon L, Gauvin F, Amre DK, Saint‐Louis P, Lacroix J. Serum procalcitonin and C‐reactive protein levels as markers of bacterial infection: a systematic review and meta‐analysis. Clin Infect Dis. 2004;39(2):206–17. [DOI] [PubMed] [Google Scholar]
- 13. Berbari E, Mabry T, Tsaras G, Spangehl M, Erwin PJ, Murad MH, et al. Inflammatory blood laboratory levels as markers of prosthetic joint infection: a systematic review and meta‐analysis. J Bone Jt Surg. 2010;92(11):2102–9. [DOI] [PubMed] [Google Scholar]
- 14. Bauman ZM, Phillips P, Cantrell E, Raposo‐Hadley A, Patel M, Evans CH, et al. Severity of illness impacts outcomes more than admitting service for isolated hip fracture patients. World J Surg. 2022;46:2344–9. [DOI] [PubMed] [Google Scholar]
- 15. Rushton PRP, Reed MR, Pratt RK. Independent validation of the Nottingham hip fracture score and identification of regional variation in patient risk within England. Bone Jt J. 2015;97(1):100–3. [DOI] [PubMed] [Google Scholar]
- 16. Copeland GP, Jones D, Walters M. POSSUM: a scoring system for surgical audit. Br J Surg. 1991;78(3):355–60. [DOI] [PubMed] [Google Scholar]
- 17. van Zeeland MLP, Genovesi IPO, Mulder J‐WR, Strating PR, Glas AS, Engel AF. POSSUM predicts hospital mortality and long‐term survival in patients with hip fractures. J Trauma Acute Care Surg. 2011;70(4):E67–72. [DOI] [PubMed] [Google Scholar]
- 18. Hong N, Kim CO, Youm Y, Choi JY, Kim HC, Rhee Y. Elevated red blood cell distribution width is associated with morphometric vertebral fracture in community‐dwelling older adults, independent of anemia, inflammation, and nutritional status: the Korean urban rural elderly (KURE) study. Calcif Tissue Int. 2019;104(1):26–33. [DOI] [PubMed] [Google Scholar]
- 19. Chen P‐C, Sung F‐C, Chien K‐L, Hsu H‐C, Su T‐C, Lee Y‐T. Red blood cell distribution width and risk of cardiovascular events and mortality in a community cohort in Taiwan. Am J Epidemiol. 2010;171(2):214–20. [DOI] [PubMed] [Google Scholar]
- 20. Bojakowski K, Dzabic M, Kurzejamska E, Styczynski G, Andziak P, Gaciong Z, et al. A high red blood cell distribution width predicts failure of arteriovenous fistula. PLoS One. 2012;7(5):e36482. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Duarte‐Gamas L, Pereira‐Neves A, Jácome F, Fragão‐Marques M, Vaz RP, Andrade JP, et al. Red blood cell distribution width as a 5‐year prognostic marker in patients submitted to carotid endarterectomy. Cerebrovasc Dis Extra. 2020;10(3):181–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Kim J, Kim K, Lee JH, Jo YH, Rhee JE, Kim TY, et al. Red blood cell distribution width as an independent predictor of all‐cause mortality in out of hospital cardiac arrest. Resuscitation. 2012;83(10):1248–52. [DOI] [PubMed] [Google Scholar]
- 23. 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]
- 24. Jo YH, Kim K, Lee JH, Kang C, Kim T, Park H‐M, et al. Red cell distribution width is a prognostic factor in severe sepsis and septic shock. Am J Emerg Med. 2013;31(3):545–8. [DOI] [PubMed] [Google Scholar]
- 25. Lee H, Kong S‐Y, Sohn JY, Shim H, Youn HS, Lee S, et al. Elevated red blood cell distribution width as a simple prognostic factor in patients with symptomatic multiple myeloma. Biomed Res Int. 2014;2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Hu L, Li M, Ding Y, Pu L, Liu J, Xie J, et al. Prognostic value of RDW in cancers: a systematic review and meta‐analysis. Oncotarget. 2017;8(9):16027–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Sakai Y, Wakao N, Matsui H, Watanabe T, Iida H, Katsumi A. Elevated red blood cell distribution width is associated with poor outcome in osteoporotic vertebral fracture. J Bone Miner Metab. 2021;39(6):1048–57. [DOI] [PubMed] [Google Scholar]
- 28. Wei‐Hsiang H, Zhu Y, Zhang J, Zhang Y. Pretreatment red blood cell distribution width as an efficient predictor of survival in older patients undergoing hip fracture surgery. Int J Clin Pract. 2021;75(12):e14791. [DOI] [PubMed] [Google Scholar]
- 29. Johnson AEW, Pollard TJ, Shen L, L‐wH L, Feng M, Ghassemi M, et al. MIMIC‐III, a freely accessible critical care database. Sci Data. 2016;3(1):1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Van Buuren S, Groothuis‐Oudshoorn K. Mice: multivariate imputation by chained equations in R. J Stat Softw. 2011;45:1–67. [Google Scholar]
- 31. Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med. 2015;34(28):3661–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Yoshida K, Bohn J, Yoshida MK. Package ‘tableone’. Vienna, Austria: R Foundation for Statistical Computing; 2020. [Google Scholar]
- 33. Wickham H. ggplot2: Elegant Graphics for Data Analysis. Springer; 2016. [Google Scholar]
- 34. Wickham H, Averick M, Bryan J, Chang W, McGowan LDA, François R, et al. Welcome to the Tidyverse. J Open Source Softw. 2019;4(43):1686. [Google Scholar]
- 35. Grolemund G, Wickham H. Dates and times made easy with lubridate. J Stat Softw. 2011;40:1–25. [Google Scholar]
- 36. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez J‐C, et al. pROC: an open‐source package for R and S+ to analyze and compare ROC curves. BMC Bioinf. 2011;12(1):1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Lumley T. Analysis of complex survey samples. J Stat Softw. 2004;9:1–19. [Google Scholar]
- 38. Jalali A, Roshan D, Alvarez‐Iglesias A, Newell J. DynNom: Visualising Statistical Models using Dynamic Nomograms; 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Allaire J. rsconnect: Deployment interface for R markdown documents and Shiny applications. R package version 08. 2019;16.
- 40. Harrell FE. Regression modeling strategies. Bios. 2018;2017(330):14. [Google Scholar]
- 41. Lele SR, Keim JL, Solymos P, Solymos MP, Imports M. Package ‘ResourceSelection’. Google Scholar; 2019. [Google Scholar]
- 42. Garbharran U, Chinthapalli S, Hopper I, George M, Back DL, Dockery F. Red cell distribution width is an independent predictor of mortality in hip fracture. Age Ageing. 2013;42(2):258–61. [DOI] [PubMed] [Google Scholar]
- 43. Lv H, Zhang L, Long A, Mao Z, Shen J, Yin P, et al. Red cell distribution width as an independent predictor of long‐term mortality in hip fracture patients: a prospective cohort study. J Bone Miner Res. 2016;31(1):223–33. [DOI] [PubMed] [Google Scholar]
- 44. Macdougall IC, Cooper A. The inflammatory response and epoetin sensitivity. Nephrol Dial Transplant. 2002;17(Suppl 1):48–52. [DOI] [PubMed] [Google Scholar]
- 45. Felker GM, Allen LA, Pocock SJ, Shaw LK, McMurray JJ, Pfeffer MA, et al. Red cell distribution width as a novel prognostic marker in heart failure: data from the CHARM program and the Duke databank. J Am Coll Cardiol. 2007;50(1):40–7. [DOI] [PubMed] [Google Scholar]
- 46. Núñez J, Núñez E, Rizopoulos D, Miñana G, Bodí V, Bondanza L, et al. Red blood cell distribution width is longitudinally associated with mortality and anemia in heart failure patients. Circ J. 2014;78(2):410–8. [DOI] [PubMed] [Google Scholar]
- 47. van Kimmenade RRJ, Mohammed AA, Uthamalingam S, van der Meer P, Felker GM, Januzzi JL Jr. Red blood cell distribution width and 1‐year mortality in acute heart failure. Eur J Heart Fail. 2010;12(2):129–36. [DOI] [PubMed] [Google Scholar]
- 48. Hampole CV, Mehrotra AK, Thenappan T, Gomberg‐Maitland M, Shah SJ. Usefulness of red cell distribution width as a prognostic marker in pulmonary hypertension. Am J Cardiol. 2009;104(6):868–72. [DOI] [PubMed] [Google Scholar]
- 49. Moreno‐Torres V, Royuela A, Múñez‐Rubio E, Gutierrez‐Rojas Á, Mills‐Sánchez P, Ortega A, et al. Red blood cell distribution width as prognostic factor in sepsis: a new use for a classical parameter. J Crit Care. 2022;71:154069. [DOI] [PubMed] [Google Scholar]
- 50. Lorente L, Martín MM, Argueso M, Solé‐Violán J, Perez A, Ramos JAMY, et al. Association between red blood cell distribution width and mortality of COVID‐19 patients. Anaesth Crit Care Pain Med. 2021;40(1):100777. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51. Pierce CN, Larson DF. Inflammatory cytokine inhibition of erythropoiesis in patients implanted with a mechanical circulatory assist device. Perfusion. 2005;20(2):83–90. [DOI] [PubMed] [Google Scholar]
- 52. Scharte M, Fink MP. Red blood cell physiology in critical illness. Crit Care Med. 2003;31(12):S651–S7. [DOI] [PubMed] [Google Scholar]
- 53. Ghaffari S. Oxidative stress in the regulation of normal and neoplastic hematopoiesis. Antioxid Redox Signal. 2008;10(11):1923–40. [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
Table S1. The covariates balancing in pre‐IPTW and post‐IPTW matching
Table S2. The univariate logistic regression models with pre‐IPTW and post‐IPTW matching in each outcome
Table S3. The multivariate logistic regression models with pre‐IPTW and post‐IPTW matching in each outcome
Data Availability Statement
The datasets used or analyzed during the current study are available from the corresponding author on reasonable request.