The Hemoglobin-to-Red Cell Distribution Width Ratio to Predict All-Cause Mortality in Patients with Sepsis-Associated Encephalopathy in the MIMIC-IV Database

Objective The hemoglobin-to-red cell distribution width ratio (HRR) is associated with the prognosis of sepsis-associated encephalopathy (SAE). This study aimed to determine the relationship between HRR and SAE and to clarify the possible mechanism of HRR as a prognostic factor for SAE. Methods A multivariate Cox proportional-hazards regression model was used to assess the correlation between HRR and all-cause mortality. Piecewise linear regression and smooth-curve Cox proportional-hazards regression models were used to observe whether there was a nonlinear relationship between HRR and all-cause mortality in SAE. Results This study included 8853 patients with SAE. A nonlinear relationship between HRR and SAE was observed through a two-segment regression model. The left inflection point for the HRR threshold was calculated to be 15.54, which was negatively correlated with all-cause mortality (HR = 0.83, 95% CI = 0.76–0.91, p < 0.001). Subgroup analyses revealed significant interactions between white blood cell count, glucose, and patients who received dialysis and HRR. The inverse correlation between HRR and SAE was more pronounced in patients who did not receive vasopressin (HR = 0.91, 95% CI = 0.87–0.96, p < 0.001) than in those who did receive vasopressin (HR = 0.94, 95% CI = 0.88–1.02, p=0.152) and was significantly more pronounced in patients without myocardial infarction (HR = 0.91, 95% CI = 0.88–0.96, p < 0.001) than in those with myocardial infarction (HR = 0.94, 95% CI = 0.87–1.02, p < 0.114). Conclusion This large retrospective study found a nonlinear relationship between all-cause mortality and HRR in patients with SAE in intensive care units, with low HRR being inversely associated with increased all-cause mortality in patients with SAE.


Introduction
Sepsis-associated encephalopathy (SAE), which is a syndrome of difuse neurological dysfunction, with no associated infectious foci in the central nervous system (CNS). Since host dysregulation does not occur in CNS infections, clear diagnostic criteria are still lacking for clinical practice [1,2]. SAE occurrence not only afects the autonomic nervous system and neuroendocrine function of patients with general sepsis but also leads to diferent degrees of cognitive decline, and even accelerates the occurrence and progression of multiple organ failure in sepsis, thus somewhat prolonging the length of stay in intensive care units (ICU). Statistics suggest that the mortality rate caused by SAE in patients with sepsis in ICU has been well over 60% [3]. However, for SAE diagnosis, in addition to the basic screening of the nervous system, EEG, CT, and MRI examinations can only be used to exclude organic CNS lesions. Identifying more risk factors is therefore important for early assessments of survival and prognosis of patients with SAE [4].
Various clinical biomarkers are currently in the diagnosis and prognosis of SAE. Hemoglobin (Hb) is widely applied to the overall anemic condition of patients with sepsis, including nutrition, infammation, and general condition. Tere is a very strong correlation between sepsis and encephalopathy caused by shock and prognosis [5]. Te red-blood-cell distribution width (RDW) was used as a biochemical index to measure red blood cell (RBC) variation, and increased RDW is not only indicative of malnutrition caused by oversized or undersized RBC, stress response infammation, dyslipidemia, hypertension, and other abnormal reactions but can also predict early coronary heart disease, cerebral infarction, pulmonary embolism, other cardiovascular and cerebrovascular diseases, and cancer. Previous studies have investigated the efect of RDW on patients with sepsis [6][7][8]. Hb and RDW are two important clinical parameters. Te Hb/RDW ratio (HRR) can especially refect the early abnormal manifestations of SAE. Meanwhile, no specifc SAE biomarker has been found for use in clinical practice [9,10].
Tis study therefore utilized data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database, which is a generalized and publicly available database that includes baseline data from many patients with critical diseases, to analyze the connection between the HRR and inhospital deaths due to SAE. Te aim was to determine whether HRR can be used as an important specifc indicator to distinguish SAE from other infammatory diseases and to provide new insights for early and efective SAE interventions.

Data Source.
Tis retrospective study was based on the MIMIC-IV database [11,12]. Te MIMIC database is a multiparameter, free-access, open-source, critical care database established by the Massachusetts Institute of Technology. Tis database is determined to be under the Safe Harbor Provision of the Health Insurance Portability and Accountability Act [13,14]. It contains detailed clinical information on more than 380,000 patients in ICU during 2008-2019. We completed the collaborating agency training program (certifcation number: 45351934). Te requirement to obtain individual patient consent was waived since the program does not afect clinical care, and all protected health information was anonymized.

Study Population and Data Extraction.
In this study, we used structured query language (SQL) via Navicat Premium (version 15.0) to extract meaningful clinical baseline data based on the subject_id of each ICU patient. According to the third internationally recognized defnition of sepsis (pyosis-3), sepsis was diagnosed using the latest sepsis 3 criteria, defned as life-threatening infection combined with sequential organ failure assessment (SOFA) score ≥ 2 [15]. SOFA was included according to the abovementioned criteria. Patients were screened according to the following criteria [16]: (1) patients dying or leaving within 24 h since ICU admission; (2) patients with primary brain injury (traumatic brain injury, ischemic stroke, hemorrhagic stroke, epilepsy, or intracranial infection); (3) patients with severe burn and trauma; (4) patients with chronic alcohol or drug abuse; (5) patients with severe liver disease or encephalopathy caused by kidney disease; and (6) patients without a laboratory indicator of HRR.

Sepsis-Associated Encephalopathy.
As far as we know, Glasgow Coma Scale (GCS) has been proved to be an important tool to assess the degree of consciousness disturbance in SAE patients. Sepsis patients with GCS ≤ 14 and excluding other unrelated factors can be initially diagnosed as sepsis-related encephalopathy. Terefore, SAE was further defned as sepsis with GCS score ≤14 or delirium diagnosed according to ICD9 code on the frst day of admission, and GCS scores assessed before surgery or before sedation and anesthesia will be included as far as possible, and excluded psychiatric disorders, neurological disorders, alcoholism, or trauma from the initial diagnosis of admission [17].

Statistical Analysis
All baseline values of continuous variables are expressed as mean ± standard deviation or median values, and the values of categorical variables are expressed as counts and proportions. All continuous variables in this study were compared using t-tests or Wilcoxon rank-sum tests, and the resulting proportions were compared using chi-square or Fisher's exact tests. Diferences among HRR tertiles were determined using one-way ANOVA or Kruskal-Wallis H tests. We used a Cox proportional-hazards model to determine the association between HRR and in-hospital mortality in patients with SAE, and developed three diferent models based on this: an unadjusted model, with no adjustment for covariates; model 1, partially adjusted model (for age and sex only); and model 2, fully adjusted for potential confounders. Te results are expressed as hazard ratios (HRs) with 95% confdence intervals (CIs). Results are provided for unadjusted, microadjusted, and fully adjusted analyses.
Furthermore, a generalized additive model was applied to verify the nonlinear relationship between HRR and SAE after it was confrmed using a logarithmic likelihood ratio test (p < 0.05). We used cubic-spline function ftting and the smooth curves of two linear regression models of all-cause mortality in the smooth fgure calculation for the HRR threshold. Moreover, a recursive method was used to automatically calculate the infection points in the ftted graph and infer the likelihood ratio of the optimal model. We analyzed the relationship between HRR and all-cause mortality in patients with SAE between diferent subgroups to determine whether there were diferences in the efect of HRR. Te statistical software package R (the R Foundation; https://www.r-project.org) was used to perform data analyses in this study.

Baseline Results.
After the above screening criteria, a total of 8853 SAE patients were included in the fnal cohort, and the inclusion process is shown in Figure 1. In all included population, the median age was 67 years (IQR=55-78 years) and comprised 4007 males (45.30%) and 4845 females (54.70%).. Final outcomes of in-hospital death and survival occurred in 693 (7.80%) and 8159 (92.2%) patients, respectively, with a signifcant diference (p < 0.001). Quartiles 1 (≥3.98 to <14.00), 2 (≥14.00 to <17.10), and 3 (≥17.10 to ≤28.66) of HRR values comprised 2937, 2967, and 2948 cases, respectively. Diferences between groups were assessed using Pearson's chi-square test for categorical variables and t-test for continuous variables. All the variables are listed in Table 1.

Analysis of the Nonlinear Relationship between the HRR and All-Cause Mortality.
Tis study found a signifcant nonlinear relationship between HRR and patients with SAE. We included the following confounding variables that were meaningful to the model: sex, age, sodium, diabetes, calcium, LODS score, SOFA score, GCS score, chloride, myocardial infarction, respiratory rate, temperature, SpO 2 , PT, WBC count, peripheral vascular disease, MBP, paraplegia, norepinephrine, vasopressin, anion gap, bicarbonate, blood urea nitrogen, creatinine, INR, Hb, OASIS, SAPSII, norepinephrine, chronic pulmonary disease, and congestive heart failure. Figure 3 and Table 3 display the nonlinearity of the relationship between HRR and SAE all-cause mortality. Notably, there was an inverse association between the left infection-point threshold and all-cause mortality (HR � 0.83, 95% CI 0.76-0.91, p < 0.001), while there was no signifcant correlation at the right infection-point threshold (HR � 0.95, 95% CI � 0.91-1.01, p � 0.085).

Discussion
No previous study that we were aware of investigated the connection between HRR and SAE mortality, and this was the frst time that a certain relationship was assumed between the ratio of patients with SAE and HRR based on the large MIMIC-IV database. Tis retrospective study demonstrated a signifcant nonlinear association between HRR and all-cause mortality in patients with SAE. In addition, even in the presence of certain traditional factors for lifethreatening cerebrovascular disease, patients with low HRR were still at a higher risk of death. Tis study therefore found that lower HRR remained an important predictor of mortality in patients with SAE. SAE is most common in patients with sepsis in ICU and one of its most serious complications. Its pathogenesis has not yet been fully elucidated, but it may be related to autoimmune injury of the central body infammatory reaction [20,21], a blood fow regulating mechanism for the damage caused by cerebral perfusion abnormalities and microvascular dysfunction [22,23], or a large release of cytokines in sepsis that leads to neurotransmitter imbalance [24]. Polyneuropathy is often more severe in the later stage of SAE International Journal of Clinical Practice development, which not only afects peripheral nerves but also rapidly causes cognitive defcits, personality changes, generalized epilepsy, and other symptoms. Many patients who survive SAE often have a depressive tendency, posttraumatic stress disorder, and other consciousness disorders. Te clinical diagnosis of SAE is still performed as an exclusion diagnosis, and laboratory tests have not identifed specifc SAE biomarkers. It is therefore highly signifcant to identify specifc early warning and diagnostic indicators among the risk factors for SAE in critically ill patients with sepsis for facilitating its early intervention and treatment [25,26]. HRR has been previously explored as a novel infammatory marker in many studies on the relationship between HRR and the clinical outcomes of various diseases. Tere is increasing evidence that HRR is a new prognostic indicator in critically ill patients with malignant diseases. For example, Qin et al. explored the nonlinear relationship between all-cause mortality and HRR value in patients with ischemic stroke [27], and Rahamim et al. found that HRR was an important prognostic tool for predicting mortality in heart failure and the hospitalization of patients with cardiovascular disease [9]. Chi et al. found that low HRR were associated with a dual risk of disease progression or cancer recurrence in patients and used them to predict poor prognosis in patients with cancer [28]. Te present singlecenter, retrospective study confrmed the hypothesis that HRR afects the prognosis of patients with SAE. We found that SAE had an obvious nonlinear relationship with HRR, with a low HRR greatly increasing the risk of a poor prognosis for SAE. A piecewise linear regression model was used in an attempt to calculate the critical infection point for HRR. It was found that HRR ≤ 15.57 was negatively correlated with all-cause mortality.
Te underlying mechanism of HRR predicting the risk of mortality in SAE remains unclear. We attempted to explore the link from a pathophysiological point of view. First, SAE caused by prolonged septic infection often manifests as the nervous system involvement, cerebral hemorrhage, cerebral ischemia, signifcant hypotension, organ dysfunction, and shock, leading to massive red blood cell loss and failure to timely replantation of bone marrow hematopoiesis, resulting in anemia [29,30]. Hb somewhat refects the severity and prognosis of patients   International Journal of Clinical Practice  International Journal of Clinical Practice with SAE [31]. On the other hand, as another parameter of HRR, RDW is a variation coefcient for measuring the heterogeneity of RBC sizes [32]. Generally, a larger RDW means greater heterogeneity of RBC sizes, or larger diferences in shape and size [33,34]. It can be better explained by the decreased RBC deformation ability under the infuence of infammatory cytokines in patients with sepsis aggravating the abnormal microcirculation in blood perfusion, or the CNS infammation induced by the oxidative stress response, which is an important indicator of the potential accelerated development of infammation in SAE [35][36][37]. Te present study was the frst to use the HRR as an independent predictive biochemical index, which was found to be signifcantly more stable than RDW alone, and excluded the infuence of traditional factors such as cerebrovascular disease, malignancy, infectious liver disease, severe liver and kidney dysfunction, and trauma. A clear nonlinear relationship was observed between low HRR and poor prognosis in patients with SAE. Previous studies have confrmed the importance of GCS score as a neuroscientifc assessment tool to identify patients with SAE at a high risk of short-term mortality early [38,39]. Terefore, in order to strengthen early interventions for patients with SAE, HRR, and GCS assessments can be used as further prognostic indicators for patients with SAE. We look forward to providing supplementary prognostic information for patients with SAE at earlier stages. Te study had some limitations. First, this large retrospective study had a single-center retrospective design, and the selection of population and covariate factors was somewhat biased, which needs to be addressed by designing prospective multicenter studies in the future. Secondly, the laboratory indicators extracted in this study are relatively few, and the dynamic changes of biochemical indicators are not analyzed in detail. Other undiscovered confounding factors may exist, and it is difcult to further prove the causal association between HRR and SAE all-cause mortality. Terefore, further longitudinal data collection and more advanced statistical methods are needed to conduct causal inference. Tirdly, regarding the cutof range of HRR values, there has been heterogeneity in previous studies on diferent critical diseases, and there is no evident standard cutof value for this ratio in clinical practice. Finally, long-term follow-up events were not analyzed, instead the focus being on inhospital all-cause mortality among patients with SAE. We look forward to further research with more rigor and depth, especially to address these issues through studies with a multicenter, follow-up, and prospective design.

Conclusion
In summary, this large retrospective study has identifed a nonlinear association between all-cause mortality and HRR in patients with SAE, with a low HRR being negatively associated with increased ICU mortality in patients with SAE.

Data Availability
Publicly available datasets are analyzed in this study. Te data are available on the MIMIC-IV website at https:// mimic-iv.mit.edu/.

Ethical Approval
All procedures performed in studies involving human participants were in accordance with the Ethical Standards of the Institutional and National Research Committee and with the 1964 Helsinki Declaration and its later amendments or comparable Ethical Standards.

Consent
Consent was obtained for the original data collection. Te need for informed consent was waived for studies using this database.

Disclosure
Xiaxuan Huang and Shiqi Yuan are the co-frst authors.

Conflicts of Interest
Te authors declare that they have no conficts of interest to disclose.