Prognostic Value of the Red Blood Cell Distribution Width-to-Albumin Ratio in Critically Ill Older Patients with Acute Kidney Injury: A Retrospective Database Study

Background There is no evidence suggesting that red blood cell distribution width-to-albumin ratio (RA) predicts outcomes in severely ill older individuals with acute kidney injury (AKI). We hypothesized that RA is associated with all-cause mortality in critically ill older patients with AKI. Methods We recorded demographics, laboratory tests, comorbidities, vital signs, and other clinical information from the MIMIC-III V1.4 dataset. The primary endpoint was 90-day all-cause mortality, and the secondary endpoints were 30-day mortality, one-year mortality, renal replacement treatment (RRT), duration of stay in the intensive care unit (ICU), sepsis, and septic shock. We generated Cox proportional hazards and logistic regression models to determine RA's prognostic values and subgroup analyses to determine the subgroups' mortality. We conducted a Pearson correlation analysis on RA and C-reactive protein (CRP) in the cohort of patients from the Second Affiliated Hospital of Wenzhou Medical University. Results A total of 6,361 patients were extracted from MIMIC-III based on the inclusion and exclusion criteria. RA levels directly and linearly correlated with 90-day all-cause mortality. After controlling for ethnicity, gender, age, and other confounding variables in multivariate analysis, higher RA was significantly associated with an increased risk of 30-day, 90-day, and one-year all-cause mortality as opposed to the reduced levels of RA (tertile 3 vs. tertile 1: hazard ratios (HRs), 95% confidence intervals (CIs): 1.70, 1.43–2.01; 1.90, 1.64–2.19; and 1.95, 1.72–2.20, respectively). These results suggested that elevated levels of RA were linked to an elevated risk of 30-day, 90-day, and one-year all-cause death. There was a similar trend between RA and the use of RRT, length of stay in ICUs, sepsis, and septic shock. The subgroup analysis did not reveal any considerable interplay among strata. When areas under the curve were compared, RA was a weaker predictor than the SAPS II score but a stronger predictor than red blood cell distribution width (RDW) or albumin alone (P < 0.001); RA combined with SAPS II has better predictive power than SAPS II alone (P < 0.001). The Second Affiliated Hospital of Wenzhou Medical University cohort showed that CRP positively correlated with RA, with a coefficient of 0.2607 (P < 0.001). Conclusions RA was an independent prognostic predictor in critically ill older patients with AKI, and greater RA was linked to a higher probability of death. The risk of AKI is complicated when RRT occurs; sepsis and septic shock increase with RA levels.


Introduction
Acute kidney injury (AKI) is diagnosed using serum creatinine and urine output criteria and is linked to morbidity and mortality, particularly in critical diseases [1,2]. AKI is more likely to occur in critically ill older patients due to age-associated physiological changes, reduced renal reserves, and numerous comorbidities that increase vulnerability to acute renal impairment. Older adults usually use more medications and undergo procedures that might jeopardize renal function [3]. Older illness survivors are frequently unable to regain renal function and rely on long-term dialysis [4], which imposes high fnancial costs [5]. Given the increasing prevalence of AKI and the dismal outcomes in critical illness, researchers have looked for risk factors for death from AKI [6,7].
Te red blood cell distribution width (RDW) represents the diferences in the sizes of circulating red blood cells and is computed in automated complete blood counts [8]. In individuals with cardiovascular diseases [9], multiple myeloma [10], systemic sclerosis [11], and acute ischemic stroke [12], elevated RDW was signifcantly related to poor outcomes. Our previous study showed that RDW appeared to be an independent prognostic indicator in ill patients with AKI and that elevated RDW was linked to an elevated risk of death in this group [13]; other studies of the MIMIC database have shown that RDW was an independent prognostic factor of long-term outcomes in critically ill patients with AKI [14]. Albumin is an essential protein that regulates osmotic pressure and has anti-infammatory and antioxidant properties [15,16]; it has also been linked to AKI [17]. However, whether combining RDW with albumin could predict outcomes in critically sick older patients with AKI is unknown. Terefore, we hypothesized that the red blood cell distribution width-to-albumin ratio (RA) is associated with all-cause mortality in critically ill older patients with AKI.

Data Source.
We used the methodology of Jia et al. [14,15]. Multiparameter Intelligent Monitoring in Intensive Care III version 1.4 (MIMIC-III v1.4) is a widely recognized publicly accessible repository containing health information from 40,000 critical care patients between 2001 and 2012 [18]. We completed the Protecting Human Research Participants test and received a certifcate (No. 6182750) before applying for database access. Approval of the study was obtained from the Institutional Review Boards of the Massachusetts Institute of Technology and the Beth Israel Deaconess Medical Center with informed consent waivers. In addition, the other part of the data was also obtained from the Second Afliated Hospital of Wenzhou Medical University from January 2018 to December 2020. Our use of these data was also approved by the Institute of Institutional Research and Ethics of the Second Afliated Hospital and Yuying Children's Hospital of Wenzhou Medical University (No. LCKY2019-04). Because the data were anonymous, the requirement for informed consent was waived.

Selection Criteria.
We limited our search to AKI patients aged 65 years or older. Te incidence of AKI was verifed using the Kidney Disease: Improving Global Outcomes criteria [19], and additional material included the structured query language for retrieving AKI. Patients had to be included in the ICU for over two days at the time of their initial admission. Exclusion criteria were no RDW and albumin within the frst 24 h of admission and over 10% of individual information missing. Subjects in the Second Afliated Hospital of Wenzhou Medical University cohort also had to meet these criteria. . Te information retried from MIM-IC-III included demographics, scoring systems, comorbidities, laboratory tests, vital signs, and other factors recorded.  Comorbidities included peripheral vascular disease, sepsis,  septic shock, arrhythmia, heart valve disease, diabetes, hypertension, renal failure, hemorrhagic anemia, alcohol abuse, metastatic cancer, solid tumor, and congestive heart failure. Laboratory data included RDW, albumin, prothrombin time (PT), glucose, chloride, potassium, blood urea nitrogen (BUN), sodium, white blood cell count (WBC), platelet, hemoglobin, hematocrit, activated partial thromboplastin time (APTT), international normalized ratio (INR), creatinine, bilirubin, bicarbonate, lactate, and anion gap. RDW was expressed as %, albumin was expressed as g/dL, and RA was calculated as the ratio of RDW to albumin. Other information, including gender, age, systolic blood pressure (SBP), diastolic blood pressure (DBP), renal replacement therapy (RRT), sequential organ failure assessment (SOFA), and simplifed acute physiology score II (SAPS II), was also collected.

Data Extraction
Our endpoints were 30-day, 90-day, and one-year allcause mortality beginning from when the patients were admitted to the ICU. Te Social Security Death Index records were utilized to acquire vital status survival statistics and calculate mortality rates at diferent times. Te primary endpoint was 90-day all-cause mortality, and the secondary endpoints were 30-day mortality, one-year mortality, RRT, duration of stay in ICUs, sepsis, and septic shock. Te data from the Second Afliated Hospital and Yuying Children's Hospital of Wenzhou Medical University were similarly collected from the medical record system. Te collected indicators include clinical parameters, vital signs, laboratory parameters, comorbidities, scoring systems, and mortality (Table S1). Since the number of included cases is not up to the standard, the outcome cannot be directly analyzed, so this portion of the data was placed temporarily into Supplementary Materials.

Statistical Analysis.
Continuous variables were expressed as the mean ± SD or interquartile ranges and medians, and categorical variables were expressed as percentages or frequencies. Kruskal-Wallis H tests, one-way analysis of variance, and chi-square tests were performed to determine any signifcant diferences between the cohorts. A generalized additive model was used to determine whether the relationship between RA and 90-day all-cause mortality was linear. Cox proportional hazards models were used to calculate relationships between RA levels and 30-day, 90day, and one-year all-cause mortality, and the fndings were expressed as HRs with 95% CIs. Logistic regression models were used to assess the link among RA and RRT, duration of stay in ICUs, sepsis, and septic shock, showing odds ratios with 95% CIs. As possible confounding factors, variables premised on the biological and epidemiological background were included, and those confounding factors with a change in the efect estimate of more than 10% were utilized to construct an adjusted model [20]. Tree multivariate models were created for each endpoint. Stratifed linear regression All probability estimates were two-sided, with values <0.05 deemed statistically signifcant. All statistical data were analyzed by using R software (Version 3.6.1).

Subject Characteristics.
A total of 6,361 patients were extracted from MIMIC-III based on the inclusion and exclusion criteria. Table 1 summarizes the baseline variables of these subjects, split into tertiles according to RA. Tere were 3,401 men and 2,960 women, and patients with higher RA (RA ≥ 5.68 ml/g) had a higher likelihood of having a history of sepsis, septic shock, peripheral vascular disease, hemorrhagic anemia, coagulopathy, electrolyte disorder, hypothyroidism, congestive heart failure, metastatic cancer, and a solid tumor. Patients in the high-RA cohort (RA ≥ 5.68 ml/ g) exhibited high WBC, BUN, chloride, PT, use of RRT, mortality, SOFA score, SAPS II score, lactate, bilirubin, and duration of stay in the ICU. Table S1 in Supplementary Materials shows the baseline variables of patients in the Second Afliated Hospital of Wenzhou Medical University cohort. Because the number of cases included is not very large, we can only observe the general trend.

Te Linear Association between Levels of RA and 90-Day All-Cause Mortality.
Tere was a linear association between levels of RA and 90-day all-cause mortality, with mortality increasing as RA levels increased (Figure 1).

Te Relationship between RA and Other Endpoints.
Various models were developed to evaluate the relationship between RA and the use of RRT, duration of stay in ICUs, sepsis, and septic shock. Tere were similar trends between RA levels and these partial secondary outcomes (Table 3).

Emergency Medicine International
Te risk of AKI complicated with RRT, sepsis, and septic shock increased with RA levels.

Subgroup Analyses.
We undertook subgroup analyses to ascertain whether the connection between RA and the risk of 90-day all-cause mortality was consistent (Table 4). Te subgroup analysis revealed no signifcant interactions, suggesting that the results of this study were stable in these patients.

Mortality Prediction.
According to the ROC curves, the AUCs for RA, RDW, albumin, SAPS II, and SOFA scores were 0.656, 0.612, 0.635, 0.692, and 0.758, respectively (Figure 2(a)). RA was a weaker predictor than the SAPS II score when AUCs were compared, although it was more potent than RDW or albumin alone (P < 0.001). SAPS II scores had an AUC of 0.692, contrasted with 0.718 for RA plus SAPS II values (P < 0.001).
A comparison of the AUCs showed that RA combined with SAPS II has better predictive power than SAPS II alone (Figure 2(b)).

Pearson Correlation Analysis.
We generated CRP and RA scatter plots for patient data from the Second Afliated    Hospital of Wenzhou Medical University. CRP was positively correlated with RA ( Figure 3). Te correlation coefcient between CRP and RA was 0.2607 (P < 0.001).

Discussion
To the best of our knowledge, there have been no epidemiological studies of the prognostic signifcance of RA in critically ill older adults with AKI. We found a linear connection between RA levels and 90-day all-cause mortality, with mortality increasing as RA levels increased. Elevated levels of RA were linked to an elevated risk of 30-day, 90-day, and one-year all-cause death in the fully adjusted model. Te risk of AKI complicated with RRT, sepsis, and septic shock increased with RA levels. Te subgroup analysis indicated that there was no considerable interplay among strata. Te ROC curve revealed that RA was lower than the SAPS II value, although it was a stronger predictor than RDW or albumin alone. A comparison of the AUCs showed that RA combined with SAPS II has better predictive power than SAPS II alone. Te data obtained from the Second Afliated Hospital of Wenzhou Medical University cohort showed that RA was positively correlated with CRP in critically ill older individuals with AKI, suggesting that RA may be related to the infammatory response in these patients.
AKI is a frequent, life-threatening illness with an elevated fatality rate [21]. Te etiologies can be classifed as prerenal, renal, or postrenal [22]. Prerenal AKI in older adults is caused by volume depletion and reduced efective arterial blood volume, resulting in renal hypoperfusion. Because of the age-associated reduced glomerular fltration rate, reduced renal reserves, and poor autoregulation, vomiting, diarrhea, and excessive diuretic usage induce AKI more commonly and rapidly in older people [23]. In a retrospective investigation of 381 individuals over the age of 80,  [26]. Other intrinsic factors that cause AKI in older people include acute interstitial nephritis, which is frequently caused by hypersensitivity to drugs (especially nonsteroidal anti-infammatories and antibiotics), glomerulonephritis, and renal vascular disorders [3]. AKI involves a complex physiological process induced by many variables, and its etiology is unknown [27]. A study suggested several theories, one of which is that excessive amounts of infammatory mediators in the bloodstream signifcantly contribute to AKI [28]. Te infammatory mediators linked to AKI and its outcome include RDW, albumin concentrations, CRP, TNF-R-II, tumor necrosis factor receptor I, interleukin-(IL-) 6, IL-10, platelets, lymphocytes, and neutrophils [29,30]. Impairment of the autoregulation of renal blood fow is another potential cause of AKI [31]. Reduced renal blood fow exhausts intracellular adenosine triphosphate, compromises the cytoskeleton's integrity, triggers infammatory pathways, produces free radicals, and disrupts intracellular calcium homeostasis [32,33]. Tese lesions cause hypoxic damage to tubular cells, resulting in casts that block renal tubules.
Our previous study illustrated that RDW was among these biomarkers for the outcome of AKI [13]. RDW is a readily accessible biomarker to predict the development of various illnesses and organ dysfunctions [34,35]. In several observational studies, investigators have shown a link between elevated RDW and alterations in infammatory biomarkers [36,37]. Tese fndings suggest that the systemic infammatory response is instrumental in explaining the possible connection between RDW and mortality in severely ill AKI patients. Moreover, according to several research reports, hypoalbuminemia has also been linked to the progression of AKI and poor outcomes in patients with severe illnesses [38,39]. Albumin protects the kidneys from toxins and keeps glial pressure at optimal levels for adequate renal perfusion [40]. Te present study results suggest that RA is a stronger independent predictor of all-cause mortality in critically ill older patients with AKI than albumin or RDW alone, and we have grounds to assume that RA is clinically signifcant.

Emergency Medicine International
Tere were some limitations in our research. First, biases were unavoidable because this was a retrospective study based on only one center. Second, we computed RA only after the patient was admitted to the ICU, and only one measure of RA might alter the results' accuracy. Tird, despite our best eforts to minimize bias using a multivariate model, countless additional known and unknown variables exist. Finally, retrospective databases entail numerous faws; hence, multicenter prospective studies must validate our fndings.

Conclusions
RA was an independent prognostic predictor in critically ill older patients with AKI, and elevated RA was linked to a greater risk of death among these patients. Te risk of AKI complicated with RRT, sepsis, and septic shock increased with RA levels. Extensive prospective multicenter investigations are needed to validate these results.

Data Availability
Te clinical data for this study were provided by MIMIC-III v.1.4. Researchers must complete the National Institutes of Health's online course Protecting Human Research Participants to apply for permission to access the database. Te other part of the data was also obtained from the Second Afliated Hospital of Wenzhou Medical University; due to the small number of participants and specifc nature of the service, the data would not be shared for the time being.

Ethical Approval
Te study was approved by the Institute of Institutional Research and Ethics of the Second Afliated Hospital and Yuying Children's Hospital of Wenzhou Medical University (No. LCKY2019-04).

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