Associations of albumin and nutritional index factors with delirium in patients admitted to the cardiac intensive care unit

Background Limited data are available on the association of malnutrition with the occurrence of delirium in the cardiac intensive care unit (CICU). Thus, we aimed to analyze whether nutritional indices and their components can predict the development of delirium in CICU. Methods We enrolled 2,783 patients admitted to the CICU of Samsung Medical Center for more than 24 h between September 2012 and December 2018. We assessed the nutritional status at admission using three indices, the Prognostic Nutrition Index (PNI), the Geriatric Nutritional Risk Index (GNRI), and the Controlling Nutritional Status (CONUT). Then, we compared predictive performances for the occurrence of delirium among nutritional indices using Delong’s test. Results Delirium developed in 678 patients (24.3%) assessed three times daily for 7 days of CICU stay. Nutritional indices had fair predictive performance for development of delirium in critically ill cardiac patients using the area under the receiver-operating characteristic curve (AUROC: 0.729 for the GNRI, 0.728 for PNI, and 0.762 for CONUT). Furthermore, the AUROC of albumin alone (0.77, 95% CI, 0.75–0.79) was significantly greater than that of either GNRI (p < 0.001) or PNI (p < 0.001). In a multivariable analysis including each component of nutritional indices, albumin was a significant predictor for delirium but not absolute lymphocyte count, bodyweight/ideal bodyweight, or total cholesterol level as a component of nutritional indices. Conclusion Predictive performances of nutritional indices for the occurrence of delirium were acceptable in patients admitted to CICU. Albumin alone might be a helpful and straightforward indicator for the occurrence of delirium.


Introduction
Delirium is a frequent complication in patients admitted to the intensive care unit (ICU). It has been associated with increased risk of mortality, development of post-ICU cognitive impairment, and longer ICU and hospital stays (1,2). In recent decades, the application of advanced medical devices for the treatment of cardiovascular disease has been significantly OPEN ACCESS EDITED BY Thomas A. Zelniker, Medical University of Vienna, Austria expanded, and the number of critically ill patients with underlying cardiac disease has increased (3). Accordingly, the incidence of delirium in the cardiac intensive care unit (CICU) has substantially increased and is comparable to that in other medical ICUs (4). Furthermore, recent studies have reported associations between delirium and clinical outcomes in critically ill cardiac patients (5,6).
Malnutrition is an essential risk factor for delirium in elderly patients (7). Several nutritional indices have been developed to assess the risk of malnutrition in critically ill patients. First, the Prognostic Nutritional Index (PNI) was created in 1980 to evaluate preoperative nutritional conditions and surgical risk for gastrointestinal surgery (8). The Geriatric Nutritional Risk Index (GNRI) was developed to estimate the risk of morbidity and mortality in elderly medical patients (9). The Controlling Nutritional Status (CONUT) is widely used for daily assessment of the nutritional status of inpatients (10).
Although several studies have reported an association of delirium with mortality in ICU patients, limited studies are available on the relationship between nutritional status and delirium in CICU patients. Therefore, the goal of the present study was (i) to assess the predictive performance of nutritional indices (PNI, GNRI, and CONUT) and their components in CICU patients and (ii) to investigate clinical factors associated with delirium in this population.

Study population
All consecutive patients who were admitted to the CICU between September 1, 2012, and December 31, 2018, were eligible for the study. A total of 4,261 patients who were 18 years old or older was admitted to the CICU at Samsung Medical Center in Seoul, Republic of Korea. We excluded 1,473 patients who stayed in the CICU for less than 24 h and three patients for whom we were not able to calculate nutritional indices due to limited data. The Institutional Review Board of Samsung Medical Center approved this study (IRB. No. 2020-10-102) and waived the requirement for informed consent because of the observational nature of the study. Patient information was anonymized and de-identified before analysis.

Cardiac intensive care unit management
Our unit is a 12-bed ICU with a 1:2 nurse-to-patient ratio offering level 1 care for critically ill cardiac patients. To provide comprehensive critical care to patients with various cardiovascular diseases and complex comorbidities, the CICU is equipped with invasive and noninvasive devices to monitor patients' hemodynamic status and provide advanced therapeutic technologies. With the high-intensity staffing model, patients are managed by a dedicated cardiac intensivist who is board certified in interventional cardiology and critical care medicine (11). Cardiac surgery support is promptly assessable. In addition, multidisciplinary care is provided via consultation with a dietitian, pharmacist, and respiratory care practitioner.
The clinical practice guidelines published by The Society of Critical Care Medicine were adopted for general intensive care (12). The Confusion Assessment Method for ICU (CAM-ICU) assessment was performed by nurses three times a day on patients with a Richmond Agitation-Sedation Scale of −3 (indicating movement or eye-opening to voice but no eye contact) or higher. The recorded CAM-ICU results were re-checked every day by a senior nurse. Nutrition support team consultation was available for nutrition therapy.

Nutritional indices
Nutritional status was assessed using the PNI, GNRI, and CONUT based on blood test results measured on the day of admission to CICU to identify risk scores of malnutrition in critically ill patients. The PNI is calculated as follows: 10 x serum albumin concentration (g/dL) + 0.005 x absolute lymphocyte count (ALC, number/mm 2 ) in peripheral blood (13). Patients with a PNI score > 38 were considered normal, patients with a score of 35-38 were at risk of moderate malnutrition, and patients with a score < 35 were classified as at risk of severe malnutrition. The GNRI is calculated as follows: 14.89 x serum albumin (g/dL) + 41.7 x [bodyweight/ideal bodyweight (IBW)] (9). Patients were categorized into four groups according to GNRI score: major risk (GNRI score < 82), moderate risk (82 ≤ GNRI score < 92), low risk (92 ≤ GNRI score < 98), and no risk (98 ≤ GNRI score). The CONUT was obtained based on serum albumin concentration, cholesterol level, and ALC (14). According to the CONUT score, patients were divided into four groups: normal (0-1 point), mild risk (2-4 points), moderate risk (5-8 points), and severe risk (9-12 points).

Data collection
After reviewing the medical records, a trained study coordinator retrospectively collected clinical, laboratory, and outcome data. To accurately evaluate the total number of days at risk of delirium, we did not include days when the patient was in a coma as defined by a Richmond Agitation-Sedation Scale assessment of −4 (unresponsive to voice but responded to physical stimulation) or − 5 (unresponsive to voice and physical stimulation) (15). Delirium was defined as CAM-ICU positivity at any time during the assessment three times daily for up to 7 days of CICU stay. Polypharmacy was defined as the routine use of five or more medications (16).

Statistical analysis
Descriptive statistics were performed to compare the clinical characteristics and outcomes between the delirium and non-delirium groups. Continuous variables were presented as median and interquartile range or mean ± standard deviation and were compared with a Mann-Whitney U test. Categorical variables were expressed as number and percentage and were analyzed using Chi-square tests or Fisher's exact tests, where applicable. The model's predictive ability to discriminate between patients who had and did not have delirium was estimated using the receiver operating characteristic curve. The receiver operating characteristic curve is presented to demonstrate the performance of delirium for the area under the curve (AUROC) and 95% confidence interval (CI). Discriminant functions were compared by a 2-tailed paired comparison of the ROC curve by DeLong's method (17). Variables with a p value less than 0.05 on univariate analyzes and clinically relevant a priori variables were entered into the multiple logistic regression model. Results were reported as each variable's odds ratio (OR) with 95% CI. A two-tailed p value less than 0.05 was considered statistically significant for all analyzes. Data were analyzed using R Statistical Software (Version 3.2.5; R Foundation for Statistical Computing, Vienna, Austria).

Clinical characteristics
During the study period, 2,783 patients were finally enrolled in this analysis; 678 patients (24.3%) have occurred delirium during CICU admission. The characteristics of the patients are shown in Table 1. Patients in the delirium group were older and higher-risk subjects with lower body mass index and more comorbidities than those in the non-delirium group. There was no difference between the two groups in the CICU admission diagnosis and polypharmacy. Cardiac arrest before CICU admission (20.9% vs. 6.3%; p < 0.001) and Sequential Organ Failure Assessment score (6.97 ± 3.47 vs. 3.39 ± 2.71; p < 0.001) were higher in the delirium group than in the non-delirium group. In the laboratory findings at CICU admission, ALC

Cardiac intensive care unit management and clinical outcomes
Patients in the delirium group were more likely to receive vasopressor, mechanical ventilator, intraaortic balloon pump, and extracorporeal membrane oxygenator interventions than were those in the non-delirium group during the CICU stay (Table 2). Compared with the non-delirium group, CICU length of stay (8.15 ± 11.35 days vs. 2.88 ± 3.52 days; p < 0.001), CICU mortality (11.4% vs. 1.8%; p < 0.001), and in-hospital mortality (18.3% vs. 3.1%; p < 0.001) were higher in the delirium group.

Predictive performance for delirium
The distribution of delirium according to the value of each component in the three nutritional indices is shown in cubic spline curves ( Figure 1). According to the increasing level of albumin, the occurrence of delirium tended to decrease linearly ( Figure 1A). However, U-shaped correlation was observed for increased level of  Figures 1B-D). The performances of albumin, PNI, GNRI, and CONUT for delirium prediction were compared and are shown in Figure 2. Among three nutritional indices, the AUROC of the CONUT was higher than those of the PNI and GNRI. Algorithm discrimination performance of albumin alone (AUC = 0.770; 95% CI, 0.750-0.790) and CONUT (AUC = 0.762; 95% CI, 0.750-0.790; p = 0.275) for delirium were not significantly different; however, discrimination was superior to that of the GNRI (AUC = 0.729; 95% CI, 0.706-0.753; p < 0.001) and PNI (AUC = 0.728; 95% CI, 0.706-0.751; p < 0.001).

Factors associated with delirium
Univariable logistic regression analysis indicated that 15 characteristics of CICU patients were associated with delirium (Table 3). In multivariable analysis, age, sex, smoking, stroke, dementia, Sequential Organ Failure Assessment score, mechanical ventilation, and white blood cell were associated with increased risk of delirium, while albumin level at CICU admission (adjusted OR 0.555; 95% CI 0.402-0.767; p < 0.001) was associated with decreased risk of delirium.

Discussion
We investigated the associations of nutritional indices and their components with delirium in CICU patients from a single-center registry. The major findings were as follows: (i) nutritional indices (PNI, GNRI, and CONUT) had fair predictive performances for delirium in CICU patients; (ii) predictive performance of albumin alone for delirium was superior to those of PNI and GNRI and similar to that of CONUT; and (iii) among the components of nutritional indices, albumin was an independent predictor of delirium, whereas ALC, bodyweight/IBW, and total cholesterol were not.
Delirium is an emerging complication of CICU stay, with a reported incidence range from 5.7 to 40% (18)(19)(20). There are several possible causes for the increased incidence of delirium in CICU. First of all, acute respiratory failure caused by cardiac origin has been a leading indication for CICU admission, and recent studies have described temporal increases in mechanical ventilation support in CICU populations (21). Although the clinical practice guidelines for adult patients in the ICU recommend light sedation, some mechanically ventilated patients require deep sedation to control pain and agitation during the early phase of respiratory failure (12). For deep sedation, intravenous administration of opioids and benzodiazepines is often used, and their use is associated with delirium (22). Furthermore, the CICU has become considerably more complex with advanced medical devices, increasing CICU patient complications (3). Therefore, delirium has recently been recognized as a common comorbidity for CICU patients.  Values are the mean ± standard deviation, median (interquartile range), or n (%). *Chronic kidney disease is defined as end-stage renal disease on hemodialysis or peritoneal dialysis. **Polypharmacy is defined as the routine use of five or more medications. ALC, absolute lymphocyte count; CICU, cardiac intensive care unit; HDL, high density lipoprotein; LDL, low density lipoprotein; NT pro-BNP, N-terminal pro-brain natriuretic peptide; SOFA score, Sequential Organ Failure Assessment score. Several prospective cohort studies have evaluated the incidence of postoperative delirium in malnourished patients undergoing coronary artery bypass grafting and have found that malnutrition is associated with postoperative delirium (7,23). Meta-analysis to determine the risk factors of delirium after orthopedic surgery in elderly patients has also shown malnutrition as an important factor for delirium after surgery (24). In CICU patients, Sugita et al. conducted a prospective cohort study to evaluate the association between malnutrition and delirium in critically ill cardiac patients (25). Of 653 CICU patients, 58 (8.9%) developed delirium, for which PNI and CONUT were both reported to be independent predictors of delirium, but GNRI was not. They suggested that ALC, included in PNI and CONUT but not in GNRI, might be an important prognostic factor associated with inflammation in delirium patients. However, in the present study, U-shaped correlation was observed between increased level of ALC and delirium. Furthermore, the discrimination performance of PNI for delirium was similar to those of CONUT and GNRI.
Albumin is the most abundant protein in the plasma and is a major transporter of many compounds in blood (26). A recent study of surgical ICU patients by Yeh DD, et al. found that serum albumin levels and pre-albumin levels may reflect baseline nutrition status. However, changes in serum albumin and pre-albumin levels during ICU stays may not indicate nutritional adequacy in the ICU (27). This is because albumin synthesis can be affected by numerous acute and chronic disease processes, and serum albumin level can provide predictive insights into the effects of disease on protein metabolism. Hypoalbuminemia can be observed in acute illness due to increased capillary leakage or decreased synthesis due to inflammation, underlying malnutrition, or hepatic synthetic dysfunction (28). Therefore, albumin is a well-known predictor of mortality in critically ill patients (29). Recent studies have reported the  The receiver-operating characteristic curve comparing the predictive performances for delirium. The predictive performances of nutritional indices were calculated using Delong's test and compared to those of albumin. AUC, area under the curve; CONUT, Controlling Nutritional Status; GNRI, Geriatric Nutritional Risk Index; PNI, Prognostic Nutritional Index.
Frontiers in Cardiovascular Medicine 06 frontiersin.org association between albumin level and delirium, and meta-analysis of risk factors of delirium occurrence among elderly patients in medical ICUs has found that low albumin level is associated with delirium (30). Zhang et al. analyzed 700 elderly patients admitted to the ICU after non-cardiac surgery and found that preoperative hypoalbuminemia was associated with an increased risk of postoperative delirium (31). Similarly, we found the decrease in albumin level is a risk factor of developing delirium in CICU patients. Furthermore, albumin alone revealed an inverse linear correlation with delirium, and its predictive performance was superior to those of PNI and GNRI and similar to that of CONUT. Compared with other variables included in nutritional indices, only albumin showed a linear correlation with delirium. Therefore, albumin might be a helpful and simple indicator for predicting delirium in CICU patients. This finding should be validated in future studies, and the mechanism of the relationship between albumin and delirium should be investigated further.
Although this study provides additional information on the associations of albumin and nutritional indices with delirium in CICU patients, several limitations should be noted. First, our study used nonrandomized registry data. Therefore, selection bias and confounding factors might have affected our results. Second, we evaluated delirium only using CAM-ICU. However, the CAM-ICU  Frontiers in Cardiovascular Medicine 07 frontiersin.org is generally used for delirium assessment in general ICUs, and we used this tool for routine practice in ICU management including CICU. Finally, we did not include individual information on nutritional support during CICU management collect pre albumin level for comparing albumin for nutritional status. Further studies including detailed knowledge of nutritional support would be informative.

Conclusion
Predictive performances of nutritional indices for the occurrence of delirium were acceptable, and albumin alone as a component of nutritional indices might be a simple and helpful indicator of delirium in patients admitted to a CICU.

Data availability statement
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement
The studies involving human participants were reviewed and approved by the Institutional Review Board of Samsung Medical Center approved this study (IRB. No. 2020-10-102) and waived the requirement for informed consent because of the observational nature of the study.

Author contributions
TK, RK, SN, CC, KC, C-MP, and JY contributed to conception and design of the study. TK, SN, CC, and KC organized the database. TK, RK, and JY performed the data analysis and interpretation. TK wrote the first draft of the manuscript. TK, RK, C-MP, and JY revised the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.