Article Text

Original research
Association between haemoglobin-to-red blood cell distribution width ratio at admission and all-cause mortality in adult patients with sepsis in an intensive care unit: a secondary analysis of the MIMIC-IV database
  1. Liping Zhong,
  2. Yuting Zhong,
  3. Weiming Chen,
  4. Fei Liang,
  5. Yilin Liao,
  6. Yuanjun Zhou
  1. Meizhou People's Hospital, Meizhou, Guangdong, China
  1. Correspondence to Dr Yuanjun Zhou; dr_chou{at}163.com

Abstract

Objective The association between haemoglobin-to-red blood cell distribution width ratio (HRR) and all-cause mortality remains poorly understood. This study aimed to examine the influence of HRR at the time of admission mortality over 1 year and 30 days in patients with sepsis.

Design This was a secondary analysis.

Setting This study was conducted in intensive care units (ICUs).

Participants Adult patients with sepsis were identified and included from an intensive care database based on eligibility criteria.

Primary outcome and measure The primary outcome was the rate of death within 1 year. The secondary outcome was the death rate within 30 days.

Results A total of 4233 patients with sepsis who met the inclusion criteria were analysed, excluding those ineligible. These participants were divided into quartiles based on their HRR at admission. The overall mortality rates at 1 year and 30 days were 42.9% and 25.5%, respectively. A significant inverse association was observed between HRR quartiles and all-cause mortality (p<0.001). Pairwise comparisons using Kaplan-Meier analysis showed significant differences in 1-year mortality rates across the quartiles. However, no significant difference was detected in 30-day mortality between the Q3 and Q4 groups (p=0.222). Multivariate Cox regression analysis demonstrated that a higher HRR at ICU admission was independently associated with reduced mortality at 1 year (HR, 0.935; 95% CI 0.913 to 0.958; p<0.001) and 30 days (HR, 0.969; 95% CI 0.939 to 0.999; p=0.043). Furthermore, restricted cubic spline models indicated a non-linear relationship between HRR and mortality at both 1 year and 30 days (p<0.001 for both).

Conclusions This retrospective analysis demonstrated that the HRR at the time of admission was a significant prognostic marker for long-term mortality in patients with sepsis.

  • mortality
  • adult intensive & critical care
  • infectious diseases

Data availability statement

Data are available upon reasonable request.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

STRENGTHS AND LIMITATIONS OF THE STUDY

  • Our study leveraged a reputable, widely accessible international database to examine the impact of the haemoglobin-to-red blood cell distribution width ratio (HRR) on the short-term and long-term prognosis of patients with sepsis.

  • The non-linearity between HRR and adverse outcomes was explored using restricted cubic splines.

  • The exclusion of patients with incomplete data could have introduced selection bias, potentially affecting the study’s results.

  • While the findings illustrate a correlation between HRR and long-term outcomes in patients with sepsis, they do not confirm causality.

Introduction

Sepsis1 represents a severe and potentially lethal condition that poses significant health challenges globally.2 Characterised by a dysregulated immune response to infection leading to organ dysfunction and heightened mortality, sepsis prompts significant concern.3 4 Despite advancements in managing sepsis, affected individuals continue to face a heightened risk of adverse outcomes. The early and practical assessment of illness severity and prognosis using simple indicators is essential for life-saving interventions and alleviating the healthcare burden imposed by sepsis. The identification of rapidly measurable indicators with high prognostic value is critical for delivering targeted interventions and enhancing patient survival rates.

Extensive research has identified several indices predictive of sepsis outcomes, including the neutrophil-to-lymphocyte ratio, albumin levels, platelet count and the ratio of red blood cell distribution width (RDW) to platelet count.5–8 RDW has emerged as a significant and reliable marker for acute and chronic systemic inflammation.9 It measures the variability in size of erythrocytes circulating in the blood, with increased RDW levels during sepsis being associated with elevated mortality risk. Furthermore, it is frequently observed that haemoglobin (Hb) concentrations decrease in the acute phase of an infection. This reduction can result from severe infections leading to lower serum iron levels, damage to red blood cells (RBC) and insufficient RBC production, ultimately leading to anaemia and decreased survival rates.10 11 Lower Hb levels are correlated with an increased necessity for transfusions in patients who are critical.

The haemoglobin-to-red blood cell distribution width ratio (HRR), defined as the ratio of Hb concentration (g/L) to RDW (%), is a novel composite indicator. The inherent association between RDW and anaemia is well documented, with RDW serving as a measure of anisocytosis and a diagnostic tool for various anaemia types. Furthermore, recent studies have demonstrated a significant association between reduced HRR levels and adverse outcomes in malignant diseases.12–14 Baseline HRR has been proposed as an effective prognostic indicator for patients with sepsis and atrial fibrillation (AF).15

However, the predictive value of low intensive care unit (ICU) admission HRR for outcomes in patients with sepsis remains unclear. This study aimed to investigate the association between low admission HRR and adverse outcomes in patients with sepsis, aiming to establish an effective and convenient predictor for identifying patients at high risk.

Methods and materials

Data source and permission for use

The data used in this research were exclusively sourced from the publicly available database, Medical Information Mart for Intensive Care-IV (MIMIC-IV, version 2.2).16 This database was developed by researchers at the Laboratory of Massachusetts Institute of Technology, including the Computational Physiology team, in collaboration with various research groups. The Beth Israel Deaconess Medical Center contributes high-quality clinical data from thousands of individuals, secured under stringent privacy guidelines, spanning from 2008 to 2019. Dr. Yuanjun Zhou, the principal investigator, received authorisation to conduct the authors’ research using this database.

Study population

This study focused on adult patients diagnosed with sepsis (≥18 years), who had a documented HRR within 24 hours of admission to the ICU during a hospital stay. Sepsis identification adhered to the criteria established by the Third International Consensus Definitions for Sepsis and Septic Shock, specifically characterised by the occurrence of infection accompanied by Sequential Organ Failure Assessment (SOFA) scores of two or higher.1

Exclusion criteria: patients hospitalised for ≤24 hours or lacking predefined clinical variable (table 1) were excluded from the analysis.

Table 1

Baseline characteristics and outcomes for patients with sepsis grouped by the HRR level

Data acquisition

Data were meticulously extracted from the MIMIC-IV (V.2.2) using PostgreSQL (V.13.0). All extracted variables and the details were presented in table 1, which were not described repeatedly. ICU admission information encompassed demographic details, the primary source of sepsis, the recorded comorbid conditions, treatment interventions, laboratory findings and the physiological taken at ICU admission. Evaluation scores such as SOFA, Simplified Acute Physiology Score II, Acute Physiology and Chronic Health Evaluation III, Glasgow Coma Score, Oxford Acute Severity of Illness Score and Charlson Comorbidity Index (CCI) were also collected. These data points represent the initial measurements recorded on admission to the ICU.

Outcomes

The primary endpoint was to assess the all-cause mortality rate 1 year post admission, with the secondary objective of examining the mortality rate within 30 days. Mortality data were ascertained from the ICU admission records and documented dates of death.

Statistical analysis

The HRR was categorised into four groups through quartile-based segmentation. A collinearity assessment of all variables to eliminate duplicates before proceeding with the analysis. Categorical variables were presented as proportions (%), and the χ2 test was applied to examine unordered categorical variables. Due to the non-normal distribution of continuous variables, as indicated by the Kolmogorov-Smirnov test, they were presented as medians and IQR. The comparison of baseline information disparities between deceased patients and survivors was also conducted. The differences across quartiles were analysed using the Kruskal-Wallis test for independent samples. Survival probabilities at 1 year and 30 days for patients with sepsis with varying HRR levels were examined using Kaplan-Meier curves. The influence of HRR on survival duration was analysed through univariate and multivariate Cox regression models. The non-linear relationship between continuous HRR and mortality at 1 year and 30 days was examined using restricted cubic splines (RCS) with 5 knots. Subgroup analyses were conducted to evaluate the association between baseline HRR and mortality at 1 year and 30 days, considering various factors such as age, sex, SOFA scores and CCI categories, using multivariate Cox regression models. Statistical analyses were performed using R language (V.4.3.1), setting a significance threshold of p<0.05.

Reporting guideline

The presentation of our research findings adhered to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.17

Patient and public involvement

No patients or members of the public were involved in the execution of this study.

Results

Cohort composition

Figure 1 details the inclusion and exclusion procedure employed in this research. Our study commenced with an initial screening of 23 073 patients with sepsis for potential inclusion. Following a thorough assessment against our study’s stringent inclusion criteria, 4233 individuals were deemed eligible and subsequently incorporated into our analysis, leading to the exclusion of the remainder.

Figure 1

Flow chart of the inclusion and exclusion procedure and quartile binning. The flow chart illustrating the inclusion and exclusion criteria. The range for Q1 was 1.03–5.62, for Q2 was 5.62–7.22, for Q3 was 7.22–9.06 and for Q4 was 9.06–15.4. ICUs, intensive care units; LOS, length of stay; HRR, haemoglobin-to-red blood cell distribution width ratio.

HRR computation and quartile segmentation

The HRR was determined by dividing the Hb (g/L) concentration by the RDW (%). RDW is presented as the SD (mean corpuscular volume (MCV))/MCV×100. The cohort was stratified into four quartiles based on HRR values to ensure an approximately equal distribution of patients across quartiles. The quartile ranges were defined as follows: Q1 (1.03–5.62), Q2 (5.62–7.22), Q3 (7.22–9.06) and Q4 (9.06–15.4) (table 1).

Baseline characteristics overview

The analysis categorised patients with sepsis into quartiles derived from their HRR values, with detailed baseline demographics and clinical characteristics outlined in table 1. Our collinearity assessment revealed no significant overlaps among the studied variables (online supplemental table 1). Significant variations were observed across the quartiles concerning several demographics and clinical variables, including age, sex, ethnicity, comorbidities and therapies on ICU admission. However, the source of sepsis did not significantly vary across groups (online supplemental table 2). Furthermore, significant disparities were noted in disease scoring, select laboratory parameters and vital signs across groups as per the Kruskal-Wallis test findings (table 1). The baseline characteristics comparison between survivors and non-survivors were described in online supplemental table 3.

Study outcomes

Kaplan-Meier survival analysis demonstrated a significant difference in 1 year (p<0.001) and 30-day mortality (p<0.001) rates among the quartiles (figure 2). Patients categorised in the lower HRR quartiles exhibited an increased mortality risk relative to those in the higher HRR quartiles. Detailed pairwise comparisons highlighted significant differences in 1-year mortality rates across all quartiles (figure 2A). However, the mortality rate difference between Q3 and Q4 was insignificant (p=0.222) (figure 2B). The multivariate Cox regression analysis identified a higher HRR as an independent protective factor against 1-year (p<0.001) and 30-day mortalities (p=0.043) in patients with sepsis (table 2). A decline in HRR levels was associated with an increased risk of 1-year mortality across the groups, with the exception of the comparison between Q3 and Q4 (p=0.787), using Q4 as the benchmark. However, no considerable disparities were detected in the 30-day mortality rates among the groups with Q4 serving as the reference (table 2). The RCS analysis revealed a non-linear association between HRR and the 1-year and 30-day mortality rates (p<0.001, p<0.001, respectively) (figure 3).

Figure 2

Kaplan-Meier analysis for 1-year/30-day mortality. HRR, haemoglobin-to-red blood cell distribution width ratio. The range for Q1 was 1.03–5.62, for Q2 was 5.62–7.22, for Q3 was 7.22–9.06 and for Q4 was 9.06–15.4. (A) The 1-year mortality was compared between groups. All p for pairwise comparison were <0.001. (B) The 30-day mortality was compared between groups. P for Q1 versus Q2 was 0.005; p for Q1 versus Q3 was <0.001; p for Q1 versus Q4 was <0.001; p for Q2 versus Q3 was <0.001; p for Q2 versus Q4 was <0.001; p for Q3 versus Q4 was 0.222.

Table 2

Results of Cox proportional hazards models for 1-year mortality and 30-day mortality

Figure 3

Restricted cubic spline (RCS) plot for 1-year/30-day mortality. HRR, hemoglobin-to-red blood cell distribution width ratio. (A) RCS of 1-year mortality and HRR. P for non-linearity <0.001 and p for overall <0.001; HRR reference (red line) = 7.22; 1-year mortality=42.9% (1814/4233). (B) RCS of 30-day mortality and HRR. P for non-linearity <0.001 and p for overall <0.001; HRR reference (red line) = 7.22; 30-day mortality = 25.5% (1081/4233).

Subgroup analysis

Further investigation into the association between HRR at ICU admission and mortality rates over 1-year and 30-day periods among patients with sepsis was conducted, taking into account variables such as age, sex, SOFA score and CCI (online supplemental figure 1). For patients with a SOFA score <3, no significant difference in all-cause 1-year mortality was observed across the quartiles. The 30-day mortality analysis yielded no significant findings across all subgroups, except in women and those with a SOFA score ≥3. In patients with CCI >2, a lower HRR consistently emerged as an independent risk factor for both 1-year and 30-day mortality, a trend not observed in patients with CCI ≤2. No significant interaction effects were detected among the subgroups, underscoring the reliability of HRR at admission as a reliable predictor of 1-year mortality in patients with sepsis.

External validation

The external validation, using independent datasets, corroborated the primary study’s findings, reinforcing the reliability of the results (online supplemental table 4).

Discussion

The present study explored the potential relationship between HRR at ICU admission and the risk of all-cause mortality among patients with sepsis, indicating that individuals with lower HRR levels face higher mortality risks at 1-year and 30-day intervals. These findings are significant, suggesting HRR’s utility as a valuable marker for overall survival in patients with sepsis. The novelty of our findings lies in the limited existing research on HRR’s predictive value for mortality in patients with sepsis, thereby providing a valuable addition to the clinical assessment tools available for managing sepsis.

Wang et al’s research identified a correlation between the HRR and reduced survival rates in patients with sepsis and AF.15 Similarly, another investigation associated low HRR to increased mortality in patients suffering from sepsis-associated encephalopathy (SAE).18 Our study expands this scope to include a broader demographic of patients with sepsis, diverging from the narrower focus on those with AF or SAE. Prior research has associated low HRR levels with decreased long-term survival in individuals with cancer or heart disease,13 14 19–21 suggesting that HRR at ICU admission might serve as a novel prognostic indicator for patients who are critically ill. Furthermore, our RCS analysis demonstrates a non-linear association between HRR and all-cause mortality, indicating that mortality rates decline with increasing HRR up to a threshold of 7.22, beyond which HRR does not influence mortality rates. This non-linear relationship suggests that HRR’s impact on mortality is modulated by various factors, warranting additional investigation. Our examination of subgroup effects found no significant differences in all-cause mortality among patients with mild sepsis or those with a low CCI. This outcome may result from the lesser extent of organ damage in patients with low SOFA scores and CCI, underscoring the need for careful consideration when applying HRR in evaluating patients’ organ functional statuses. Although our adjusted Cox regression analysis suggested a potential association between higher HRR and reduced 30-day mortality in patients with sepsis, most subgroup analyses did not confirm this, indicating a potentially tenuous connection between HRR and 30-day mortality that could be influenced by the subgroup characteristics. These findings advocate for a cautious interpretation and highlight the necessity for further research to explore the application of HRR across different population groups to improve personalised medical assessments and interventions.

The significant correlation between HRR alterations and severity of sepsis hints at complex underlying mechanisms, which could be elucidated by considering the impact of sepsis-associated increase in RDW and decrease in Hb. Systemic inflammation and oxidative stress can disrupt erythropoiesis by reducing erythropoietin (EPO) levels and iron availability, promoting RBC apoptosis, and leading to an increase in RDW and reduction in tissue oxygen utilisation. Moreover, inflammation may interfere with maturation of RBCs within the bone marrow, resulting in the release of immature RBCs into the circulation, further complicating the understanding of HRR’s role in sepsis.22–25 Inflammation can hinder the production and maturation of RBCs within the bone marrow, resulting in a surge of immature RBCs in the bloodstream.26 This disruption can lead to haemolytic anaemia, characterised by the destruction of these fragile RBCs and subsequent release of reticulocytes. In the context of sepsis-induced anaemia, EPO production is typically upregulated in response to anaemia, thereby promoting the proliferation and differentiation of erythroid progenitors. This process culminates in the elevation of mature RBC and Hb levels. However, in severe infections, the anticipated rise in EPO levels may be inhibited, consequently failing to produce a significant increase in Hb levels.27–31 Anaemia itself can contribute to an increased RDW.32

Monitoring HRR levels is crucial for managing sepsis and predicting long-term survival in patients who are critically ill. While previous studies have explored the associations of Hb and RDW with disease progression in such patients,33 34 neither of them may adequately capture the full scope of the systemic inflammation.21 Factors such as inflammation35 36 and oxidative stress37 38 can affect Hb and RDW levels. HRR, by integrating these markers, offers a more accurate reflection of systemic inflammatory status, anaemia and oxidative stress. In conditions such as oesophageal carcinoma or contrast-induced nephropathy, admission baseline HRR has been demonstrated to be a superior prognostic indicator than marker compared with Hb or RDW individually.39 40 This observation underscores the need for further investigation into its clinical utility.

Limitations

This retrospective study is subject to limitations, such as selection bias and data incompleteness. The exclusion of approximately 80% of the target population due to missing data constrains the result interpretation. Post-hoc analyses of the HRR and all-cause mortality in the excluded population showed consistent results, increasing the result reliability (online supplemental table 5). This study’s generalisability may be limited by its reliance on a single data source, despite a sizeable sample size. Moreover, while this research establishes an association between HRR and long-term adverse outcomes, it does not establish causality. Limited external validation, due to technical constraints in releasing more detailed population data, also constrains the study.

Conclusion

Our findings suggest a strong correlation between low HRR and increased overall mortality risk in patients with sepsis, positioning HRR as a potentially independent and significant indicator for assessing long-term mortality risk in this population. This underscores the necessity for further research to elucidate the underlying mechanisms and broader implications of HRR in the management and prognosis of sepsis.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

This study has received approval from the Institutional Review Boards (IRBs) of both the Massachusetts Institute of Technology (MIT) and Beth Israel Deaconess Medical Center (BIDMC) (No. 39149215).

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors The study design, data collection and examination, data analysis and manuscript drafting were performed by LZ. YZhong contributed to the study design, data examination, data analysis, manuscript drafting and supervision of the study process. WC and FL were responsible for data examination and analysis. YL oversaw the study process. YZhou was involved in all aspects of the study, including study design, data collection and examination, data analysis, manuscript drafting, supervision of the study process, responsible for the overall content as the guarantor.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.