Construction of a risk prediction model for sepsis-associated delirium in the intensive unit patients: A retrospective cohort study

doi:10.21203/rs.3.rs-4882108/v1

Download PDF

Article

Construction of a risk prediction model for sepsis-associated delirium in the intensive unit patients: A retrospective cohort study

https://doi.org/10.21203/rs.3.rs-4882108/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background: To identify the risk factors associated with sepsis-associated delirium and to develop a nomogram for predicting the risk of delirium in patients with sepsis on the basis of the Medical Information Mart for Intensive Care IV database.

Methods: A total of 12029 patients from the Medical Information Mart for Intensive Care-IV database (MIMIC-IV 2.0) were divided into a training set and a testing set at a ratio of 7:3. The primary outcome was the occurrence of sepsis-associated delirium (SAD) in ICU patients with sepsis. Hierarchical logistic regression analyses were conducted in the training set, sequentially analyzing basic information and clinical features (Model 1) followed by additional nursing management factors (Model 2), to identify factors associated with the risk of SAD. These identified factors were utilized to construct a nomogram for predicting the risk of SAD. The predictive performance of the nomogram was evaluated via receiver operating characteristic curves and calibration curves.

Results: A total of 12,029 patients were enrolled, of whom 1,666 patients (13.3%) died in the hospital and 4,807 patients (39.96%) had SAD. In-hospital mortality in the SAD group was greater than that in the non-SAD group (23.4% vs. 7.5%, p < 0.001). By hierarchical logistic regression, two models were compared, and a nomogram integrating basic information, clinical features and easily overlooked nursing management factors was constructed. The risk factors for sepsis-associated delirium include age ≥ 65 years, tobacco use, SOFA score, intensive care unit (ICU) type at admission, hypernatremia, hypocalcemia and diet type on the first day. CRRT during sepsis and receiving skin care decrease the risk of sepsis-associated delirium. The AUCs of the developed model were 0.832 (95% CI 0.823–0.841) in the training set and 0.832 (95% CI 0.818–0.846) in the testing set. The calibration curve showed good concordance between the predicted and observed delirium rates among sepsis patients.

Conclusions: We developed a nomogram incorporating ten features of clinical work,especially adding factors related to nursing management. The constructed nomogram has good performance and clinical applicability and can help identify SAD patients in a timely manner for early intervention and improve neurological outcomes.

Health sciences/Diseases

Health sciences/Medical research

Health sciences/Risk factors

Sepsis

delirium

nomogram

intensive care units

MIMIC-IV database

Sepsis is a life-threatening syndrome characterized by organ dysfunction, stemming from physiological, pathological, and biochemical abnormalities triggered by the host's dysregulated response to infection, which represents a major public health concern¹. The neurological manifestations of multiple organ dysfunction are prevalent yet often underestimated symptoms, with approximately 70% of sepsis patients experiencing neurological dysfunction. This diffuse neurological dysfunction and cognitive impairment, not caused by direct central nervous system infection or structural brain injury, is called sepsis-associated encephalopathy (SAE)². Sepsis-associated delirium (SAD) is one of the most common manifestations of SAE. It is clinically characterized by alterations in consciousness, impaired attention, disorientation, hallucinations and delusions³. Delirium is a frequent neurological complication observed in patients with sepsis in the intensive care unit, with an incidence ranging from 17.7–48%. The severity of delirium is strongly associated with the prognosis of patients^4,5, demonstrating a strong correlation with long-term cognitive impairments such as dementia. Additionally, it is associated with an elevated mortality rate among patients following discharge, leading to significant distress and a poor prognosis for individuals. This condition further imposes a substantial burden on their families^6,7.

Despite its profound clinical impact, the additional pathophysiological mechanisms underlying delirium remain incompletely understood. Central cholinergic deficiency, dopamine dysregulation, inflammation, chronic stress, hypoxemia, compromised blood flow and tissue perfusion, and metabolic disturbances (such as hyponatremia, hypernatremia, and hypoglycemia) are considered the key pathophysiological factors contributing to delirium⁸.

At present, there are no established diagnostic criteria for SAD and no specific treatment protocols. In accordance with the 2018 PADIS guidelines on pain, agitation, and delirium, the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) score is deemed the most effective tool for diagnosing and assessing delirium in adult ICU patients^9, and the need for early identification and elimination of risk factors for SAD is emphasized.

Therefore, it is imperative to deepen our comprehension of the factors pertaining to delirium in sepsis patients, promptly recognize risk factors, and implement suitable interventions. These efforts are pivotal in the early prevention and management of sepsis-associated delirium (SAD) ¹⁰. The risk factors for SAD have been delved into several studies^4,5,11–13, but there is still a lack of early predictive tools for SAD, particularly in terms of environmental factors, nutritional factors and nursing management in the ICU. Delirium is a sensitive and dependent outcome¹⁴, highlighting the crucial role of nursing management. In this study, our goal was to fill this gap by developing a novel SAD prediction model that integrates both clinical features and nursing management factors. By comparing the hierarchical nature of the clinical characteristics of sepsis patients with commonly overlooked nursing management factors in the ICU, our model aims to facilitate the early identification of patients at high risk and enable the provision of timely preventive and interventional measures.

2.1 Source of data

This study was a retrospective cohort study based on the Medical Information Mart for Intensive Care-IV database (MIMIC-IV 2.0), which is a single-center and freely available database containing comprehensive and elaborate data from more than 50,000 patients in the ICU at the Beth Israel Deaconess Medical Center (BIDMC) between 2008 and 2019¹⁵. The author responsible for data extraction has finished the National Institutes of Health’s web-based course and obtained access to the database with certification number 57042543.

2.2 Study design and participants

In this study, 28245 patients diagnosed with sepsis were selected from the MIMIC-IV. The inclusion criteria were as follows: (1) patients older than 18 years of age; (2) first ICU admission; and (3) ICU stay ≥ 24 h. A suspected infection combined with a SOFA score ≥ 2 was defined as sepsis according to the Sepsis-3 criteria¹⁶. The exclusion criteria were as follows: (1) cerebrovascular disease (4080); (2) dementia (895); and (3) use of sedatives or anesthetics before sepsis (5725). Finally, 12029 patients were enrolled, with another 5516 patients unable to undergo the assessment (UTA) excluded. Among these patients, 4807 patients were positive for delirium, and 7222 patients were negative. The enrollment flow chart is shown in Fig. 1.

Delirium: Patients were initially assessed via the Richmond Agitation-Distress Scale (RASS). Those with a RASS score of ≥ -3 were further evaluated via the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU)¹⁷. The RASS categorizes patients into unarousable, deep sedation, moderate sedation, light sedation, drowsy, alert and calm, restless, agitated, and combative, with scores ranging from − 5 to + 4. According to the CAM-ICU procedure, delirium was ascertained under the following conditions: (i) the RASS score is ≥-3; (ii) there is an acute change or fluctuating course of mental status; (iii) there is attention disorder (> 2 errors in the attention test); and (iv) either the RASS score is not 0 or the combined number of errors in the cognitive test is > 1. Additionally, delirium was assessed after the onset of sepsis, thereby providing SAD with a temporal definition.

2.3 Variable extraction and data collection

A total of 88 variables related to delirium were selected from the MIMIC-IV database (Table S1) on the basis of the literature and clinical knowledge. Through the four steps of the data processing (described in detail in section 2.5), 43 factors, including patient factors, clinical features, and nursing management, were selected and grouped into three sets to conduct the following hierarchical logistic regression. The patient factors included age, sex, body mass index (BMI, kg/m2), marital status, living situation, insurance, tobacco condition, care unit type and patient comorbidities. The clinical features included vital signs within the ICU, clinical laboratory results, therapeutic measures and severity scores. Nursing management included pressure ulcers present, pain present, diet type, urine source, skin integrity, skin care, sputum amount, head of bed, bed bath, and appetite. If a variable was recorded multiple times within a single ICU day, the value corresponding to the greatest severity of illness was utilized for further evaluation. Variables with more than 40% missing values were excluded from the analysis (Supplementary Materials Table S2). The specific selection and classification strategy for variables with multiple measurements is detailed in Supplementary Materials Table S3.

2.4 Baseline characteristics and outcomes

The baseline data, including age, sex, BMI, marital status, insurance, care unit, living situation, comorbidity, severity of illness, and treatment for patients, were collected on the initial day of admission to the intensive care unit. The primary outcome was defined as the occurrence of delirium after the onset of sepsis in ICU patients, and the secondary outcomes included the length of ICU stay (ICU LOS), length of hospital stay (Hospital LOS) and in-hospital mortality.

2.5 Statistical analysis

The data were analyzed via R version 4.3.2. The study population was categorized into CAM-ICU-positive (SAD) and CAM-ICU-negative (non-SAD) groups. The values are presented as the means (standard deviations) or medians [interquartile ranges (IQRs)] for normally distributed or nonnormally distributed continuous variables, respectively, and the number of cases (percentages) for categorical data.

For group comparisons, the χ² test or Fisher’s exact test was used for categorical variables, and either Student’s t test or the Mann–Whitney U test was used for continuous variables.

The four steps of data processing are as follows: 1) variables with missing data ≥ 40% were excluded; 2) univariate analyses (t tests, χ2) were performed with each variable (Supplementary Materials Table S5), and variables were selected when statistically significant (p value < 0.05); 3) variables highly correlated were only chosen from one of them; and 4) other factors that were considered of little relevance, 43 factors, were included in the following hierarchical logistic regression analysis, and missing values were imputed via the “MICE” package for multivariate imputation in R studio¹⁸.

The enrolled participants were divided into a training set and a testing set at a ratio of 7:3 via the “Sample” function in R. In the training set (n = 8420), patients were divided into a delirium group and a non-delirium group, and the delirium risk factors were classified into two hierarchies and were assessed in the following order: patient factors and clinical features (30 variables) and nursing management (10 variables). C statistics and Nagelkerke R² values were used to assess the fitness of the logistic model. In addition, we selected 9 variables according to the results of logistic regression analysis and clinical knowledge to construct a new risk prediction model.

Finally, receiver operating characteristic (ROC) curves and calibration curves were used to assess the predictive performance of the developed model. All the statistical tests were two-sided, and a p value of < 0.05 was considered statistically significant.

2.6 Ethics approval and consent to participate.

The MIMIC-IV database received approval from the Institutional Review Boards of Beth Israel Deaconess Medical Center and the Massachusetts Institute of Technology¹⁹. All protected health information in the database was deidentified, obviating the need for individual consent. All procedures were conducted in strict accordance with applicable guidelines and regulations.

3.1 Demographic and clinical characteristics

A total of 12,029 sepsis patients in the intensive care unit were included in the study, with 1,666 patients expiring in the hospital (mortality rate of 13.8%). Notably, 39.96% (4,807) of these patients exhibited delirium. Compared with patients in the non-delirium group, patients in the delirium group presented a more severe clinical picture, characterized by higher SAPS-II scores (43 [35, 53] vs. 35 [28, 43]), SOFA scores (8 [6, 11] vs. 5 [3, 7]), and lower GCS scores (12 [8, 14] vs. 14 [14, 15]). Additionally, there was a greater likelihood of receiving mechanical ventilation (2484 (51.7%) vs. 1653 (22.9%)), bronchoscopy (738 (15.4%) vs. 181 (2.5%)), vasoactive agents (3132 (65.2%) vs. 3286 (45.5%)), and invasive lines (2750 (57.2%) vs. 2832 (39.2%)) in the delirium group, with all p values < 0.001. Furthermore, delirium patients exhibited prolonged lengths of ICU stay (5.07 days [2.89, 9.64] vs. 2.21 days [1.52, 3.51], p<0.001) and hospital stay (12.11 days [7.13, 20.81] vs. 7.37 days [5.02, 11.80], p<0.001), as well as a significantly higher in-hospital mortality rate (1124 (23.4%) vs. 542 (7.5%), p<0.001) than non-delirium patients did (Table 1).

3.2 Hierarchical logistic regression model to explore risk factors for SAD

A total of 12029 enrolled patients were divided into a training set (n=8420) and a testing set (n=3609) at a ratio of 7:3. The baseline variables did not significantly differ between the two groups (p>0.05) (Supplementary Material Table S4). In the training set, 40 risk factors for SAD were adopted to perform hierarchical logistic regression analysis in sequence of basic information and clinical features (Model 1) and nursing management (Model 2) (Table 2). Model 2 included nursing management in the analysis, with a higher C statistic of 0.86 and Nagelkerke R² of 0.48, indicating model fitness. For Model 1, the risk of SAD was elevated in patients aged 65 years and older (OR = 1.19; 95% CI = 1.05–1.36, p = 0.007), those who had smoked within one year (OR = 1.65; 95% CI = 1.35–2.01, p < 0.001), and individuals who required mechanical ventilation (OR = 1.89; 95% CI = 1.66–2.16, p < 0.001). Additionally, the use of anticoagulant therapy with heparin within 48 hours (OR = 1.19; 95% CI = 1.07–1.34, p = 0.002), the performance of bronchoscopy (OR = 3.61; 95% CI = 2.89–4.50, p < 0.001), a higher SOFA score (OR = 1.27; 95% CI = 1.24–1.30, p < 0.001), and having a lower minimum calcium level on the first day of ICU admission (OR = 1.28; 95% CI = 1.19–1.37, p<0.001) were associated with an increased risk of SAD.

When considering clinical features and nursing management, we observed that patients with different diet types presented an increased risk of SAD compared with those with NPOs (OR_EN = 3.97, 95% CI = 3.33–4.74, p<0.001; OR_TPN = 2.50, 95% CI = 1.46–4.28, p<0.001; OR_Normal = 0.44, 95% CI = 0.31–0.64, p<0.001). Additionally, patients in special positions (such as the Reverse T-berg or Trendelenburg position) had a greater risk of SAD than did those in low-head positions (OR = 1.59; 95% CI = 1.19–1.37, p<0.001). Furthermore, the presence of a small amount of sputum also increased the risk of SAD (OR = 2.23; 95% CI = 1.77–2.82, p<0.001). However, the performance of CRRT (OR = 0.16; 95% CI = 0.12–0.21, p<0.001), the presence of pain (OR = 0.72; 95% CI = 0.64–0.81, p<0.001), and the provision of skin care (OR = 0.21; 95% CI = 0.07–0.57, p=0.002) were associated with a decreased risk of SAD (Table 2).

3.3 Construction of the prediction model

Using hierarchical logistic regression in the training set, a comprehensive assessment of the impact of various risk factors was conducted. The factors identified as significant included age, care unit type, tobacco condition, use of CRRT, bronchoscopy, SOFA score, Ca²⁺, Na⁺, diet type, and skin care. These variables were then integrated to establish a risk prediction model for the occurrence of sepsis-associated delirium (SAD). To facilitate the understanding and application of this model, a nomogram was constructed, which visually represents the relationship between these factors and the predicted probability of SAD occurrence (Figure 2). ROC analysis revealed high predictive capacity for the model in both the training and testing sets, with AUCs of 0.832 (95% CI 0.823–0.841) and 0.832 (95% CI 0.818–0.846), respectively, demonstrating its accuracy.

3.4 Validation of the prediction model

Calibration curves were applied to demonstrate the clinical accuracy of the prediction model. The prediction model was calibrated in both the training and testing sets via calibration curves, and a bias-corrected line was formed via a bootstrap method with 1000 repetitions (Figure 3). Decision curve analysis (DCA) was used to assess the net clinical benefit derived from using the nomogram. The results demonstrated that the nomogram provided a clinical net benefit in both the training set and testing set (Figure 3).

Sepsis remains a leading cause of death among critically ill patients, with SAD being a critical neurological complication. The occurrence of delirium significantly prolongs hospital stays, increases expenses, worsens long-term cognitive impairment, and elevates both short- and long-term mortality rates, drawing increasing attention from healthcare providers ^20,21. However, SAD continues to be underdiagnosed, and data concerning effective methods for the early detection and prevention of sepsis-associated delirium are scarce ²².

After excluding patients with preexisting cerebrovascular disease or dementia and patients who had used sedatives or anesthetics prior to the diagnosis of delirium¹¹, 12029 sepsis patients were included in the hierarchical logistic regression. By considering several often-neglected nursing management factors, we aimed to increase the prediction accuracy. Consequently, a novel predictive nomogram was developed, incorporating nine key factors: age, ICU type, tobacco condition, SOFA score, bronchoscopy, CRRT, Na⁺, Ca²⁺, diet type in the first 24 h and skin care. This nomogram was validated in both training and testing sets, potentially aiding clinicians in swiftly and precisely identifying high-risk patients for more precise medical decision-making.

Age was an independent risk factor for SAD²³. A recent systematic review identified advanced age and cognitive impairment as the top two risk factors²⁴. With the acceleration of population aging, delirium has emerged as a serious and prevalent issue among elderly individuals. Our data also revealed that those aged ≥ 65 years are more likely to suffer from delirium, so more attention should be given to the prevention of delirium in older patients, especially nonpharmacologic prevention and treatment²⁵. Another patient factor that needs attention is smoking. Nicotine exposure is an important and modifiable risk factor for ICU delirium²⁶^, and a meta-analysis independently confirmed the association between nicotine exposure and delirium and revealed a dose‒response relationship²⁷. This is also revealed in our model, and patients who smoke within one year are at greater risk. Interestingly, our study revealed that quitting smoking for more than 1 year can affect those who have never smoked.

In our study, care unit type was highly correlated with the risk of delirium, which may be closely related to the primary diseases that occur in patients. Patients admitted to the NSICU (neurosurgical intensive care unit) are likely to undergo neurosurgery. Delirium is caused by neuronal activity disorders, and intracranial surgery may cause a substantial inflammatory response, leading to oxidative stress, which is then exacerbated by the formation of edema and impaired oxygenation of the surrounding tissue²⁸. In addition, patients in the CCU (coronary care unit and cardiac intensive care unit) have a relatively lower risk of developing delirium. Although cardiac surgery involving cardiopulmonary bypass may lead to cerebral hypoperfusion and significantly increase the risk of delirium²⁹, the degree of trauma and inflammatory conditions in the CCU may be relatively mild compared with those in the MICU or SICU.

As a classical tool for assessing disease severity, the SOFA score is used for the identification of septic patients and has also been shown to be positively associated with the occurrence of delirium^30,31. In the present study, patients with higher SOFA scores tended to suffer from delirium (OR = 1.36, p<0.001) (Table 2). Electrolyte imbalances are highly prevalent among sepsis patients. The analysis of laboratory data from the first day of intensive care unit (ICU) admission revealed that the occurrence of delirium was significantly related to hypernatremia and hypocalcemia according to multivariate logistic regression (Table 2). Hypernatremia has also been previously described as an independent risk factor for agitated delirium and mortality in ICU patients³². A previous study demonstrated a significant association between hypernatremia and SAD by logistic regression analysis (hypernatremia >145 mmol/l, OR = 2.30, 95% CI=1.48–3.57, p<0.001)³³. It is common that circulating Ca²⁺ levels are decreased in critically ill patients, which is associated with an increased risk of death^34,35. It has been proposed that irregular levels of circulating Ca²⁺ in critically ill patients could damage the cell membrane, leading to increased permeability of the cell membrane to Ca²⁺³⁶.

Critically ill patients may undergo many treatments, including invasive lines, early anticoagulation heparin, CRRT, bronchoscopy, and mechanical ventilation. Our data revealed that bronchoscopy was associated with an increase in delirium risk (OR = 1.75, 95% CI=1.38–2.23, p<0.001) because patients receiving bronchoscopes may suffer misery and tend to have more severe lung infections and increased sputum volume. Patient undergoing CRRT before access for delirium, however, was considered to be a protective factor (OR = 0.23, 95% CI=0.17--0.29, p<0.001), as CRRT helps reduce fluid load, removes inflammatory factors and relieves sepsis complicated with acute kidney injury, although several studies have reported an elevated risk of delirium in dialysis patients^5,37. Further clinical studies on the association between CRRT and delirium are urgently needed.

Uniquely, several nursing management factors, which are crucial in the intensive care unit, were incorporated into our research¹⁴. Ten nursing management factors were included, and skin care and diet type in the first 24 h were significantly related to the occurrence of delirium. Elderly critically ill patients are at high nutritional risk³⁸, and the timing and path of nutritional support therapy are very important. Our results indicate that in sepsis procedures, early initiation of both EN (OR=6.47, 95% CI=5.51--7.59, p<0.001) and TPN (OR=6.47, 95% CI=5.51--7.59, p<0.001) increases the risk of delirium, whereas a House-Regular diet reduces the risk of delirium compared with NPO. This is partly because motor dysfunction and barrier dysfunction often coexist with hemodynamic instability and gastrointestinal stress in the early stage of sepsis³⁹. Skin care is also a pivotal aspect of intensive care. Since critically ill patients are often confined to bed for extended periods, they face an elevated risk of developing conditions such as incontinence dermatitis and pressure ulcers, which can significantly worsen their discomfort. Our study underscores the vital role of adequate skin care in mitigating these risks, serving as a crucial protective measure in such scenarios.

The present study has several notable strengths. First, the relatively large sample size lends credibility to the research findings. Second, we developed an intuitive and user-friendly model based on certain clinical indicators and nursing management factors to predict the risk of SAD in patients with sepsis. Additionally, the internal validation results demonstrated that this prediction model exhibited excellent discrimination and accuracy in predicting the risk of delirium among sepsis patients. However, there are several limitations to this study. First, the use of data from the MIMIC-IV limits it to a single-center retrospective cohort study, introducing potential bias compared with other populations. Second, future external validation is crucial. Third, the lack of clear SAD diagnostic criteria and varying delirium assessment rates among ICU staff can lead to misdiagnosis and introduce bias, necessitating further precise prospective studies.

In this study, two hierarchies of clinical factors were compared, and a more comprehensive prediction model was constructed. The results of this study may help improve the identification of SAD and assist medical staff in adjusting medical decisions in a timely manner.

Acknowledgments

Not applicable

Authors' contributions

ZP-J, YS-W and T-X contributed equally to this work. ZP-J collected the data and drafted the manuscript. YS-W and T-X conceived the idea and analyzed the data. L-Z and Y-L conceptualized the idea and provided funding for the acquisition. Y-L also undertook revisions and editing of the manuscript. JC, JL, MJL, XD, and DXW contributed to data collection and provided valuable feedback on the manuscript draft. All the authors have thoroughly reviewed and approved the final manuscript.

Funding

This study is supported by the Key Specialty of Changning District of Shanghai (No. 20231005) and the Foundation for Young Scientists of Changning District of Shanghai (No. 2023QN23).

Availability of data and materials

Data are available upon reasonable request.

Ethics approval and consent to participate

Given that this study is based on an anonymous public database provided by a third party and has been approved by the institutional review board, ethical review is not needed.

Consent for publication

All the authors have approved this submission.

Competing interests

The authors declare that they have no competing interests.

Ely EW, Shintani A, Truman B, et al. Delirium as a predictor of mortality in mechanically ventilated patients in the intensive care unit. JAMA 2004; 291: 1753–62.
Golzari SE, Mahmoodpoor A. Sepsis-associated encephalopathy versus sepsis-induced encephalopathy. Lancet Neurol 2014; 13: 967–8.
European Delirium Association, American Delirium Society. The DSM-5 criteria, level of arousal and delirium diagnosis: inclusiveness is safer. BMC Med 2014; 12: 141.
Kim Y, Jin Y, Jin T, Lee S-M. Risk factors and outcomes of sepsis-associated delirium in intensive care unit patients: A secondary data analysis. Intensive Crit Care Nurs 2020; 59: 102844.
Yamamoto T, Mizobata Y, Kawazoe Y, et al. Incidence, risk factors, and outcomes for sepsis-associated delirium in patients with mechanical ventilation: A subanalysis of a multicenter randomized controlled trial. J Crit Care 2020; 56: 140–4.
Girard TD, Thompson JL, Pandharipande PP, et al. Clinical phenotypes of delirium during critical illness and severity of subsequent long-term cognitive impairment: a prospective cohort study. Lancet Respir Med 2018; 6: 213–22.
Vasilevskis EE, Chandrasekhar R, Holtze CH, et al. The Cost of ICU Delirium and Coma in the Intensive Care Unit Patient. Med Care 2018; 56: 890–7.
Wilson JE, Mart MF, Cunningham C, et al. Delirium. Nat Rev Dis Primers 2020; 6: 90.
Barr J, Fraser GL, Puntillo K, et al. Clinical practice guidelines for the management of pain, agitation, and delirium in adult patients in the intensive care unit. Crit Care Med 2013; 41: 263–306.
Atterton B, Paulino MC, Povoa P, Martin-Loeches I. Sepsis Associated Delirium. Medicina (Kaunas) 2020; 56: 240.
Zhao Q, Xiao J, Liu X, Liu H. The nomogram to predict the occurrence of sepsis-associated encephalopathy in elderly patients in the intensive care units: A retrospective cohort study. Front Neurol 2023; 14: 1084868.
Zhao L, Wang Y, Ge Z, Zhu H, Li Y. Mechanical Learning for Prediction of Sepsis-Associated Encephalopathy. Front Comput Neurosci 2021; 15: 739265.
Sonneville R, de Montmollin E, Poujade J, et al. Potentially modifiable factors contributing to sepsis-associated encephalopathy. Intensive Care Med 2017; 43: 1075–84.
Stalpers D, de Brouwer BJM, Kaljouw MJ, Schuurmans MJ. Associations between characteristics of the nurse work environment and five nurse-sensitive patient outcomes in hospitals: a systematic review of literature. Int J Nurs Stud 2015; 52: 817–35.
Johnson AEW, Bulgarelli L, Shen L, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data 2023; 10: 1.
Singer M, Deutschman CS, Seymour CW, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA 2016; 315: 801–10.
Ely EW, Margolin R, Francis J, et al. Evaluation of delirium in critically ill patients: validation of the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU). Crit Care Med 2001; 29: 1370–9.
Zhang Z. Multiple imputation with multivariate imputation by chained equation (MICE) package. Ann Transl Med 2016; 4: 30.
Jhou H-J, Chen P-H, Yang L-Y, Chang S-H, Lee C-H. Plasma Anion Gap and Risk of In-Hospital Mortality in Patients with Acute Ischemic Stroke: Analysis from the MIMIC-IV Database. J Pers Med 2021; 11: 1004.
Widmann CN, Heneka MT. Long-term cerebral consequences of sepsis. Lancet Neurol 2014; 13: 630–6.
Vasilevskis EE, Han JH, Hughes CG, Ely EW. Epidemiology and risk factors for delirium across hospital settings. Best Pract Res Clin Anesthesiol 2012; 26: 277–87.
Tauber SC, Djukic M, Gossner J, Eiffert H, Brück W, Nau R. Sepsis-associated encephalopathy and septic encephalitis: an update. Expert Rev Anti Infect Ther 2021; 19: 215–31.
Bugiani O. Why is delirium more frequent in the elderly? Neurol Sci 2021; 42: 3491–503.
Ormseth CH, LaHue SC, Oldham MA, Josephson SA, Whitaker E, Douglas VC. Predisposing and Precipitating Factors Associated With Delirium: A Systematic Review. JAMA Netw Open 2023; 6: e2249950.
Inouye SK, Westendorp RGJ, Saczynski JS. Delirium in elderly people. Lancet 2014; 383: 911–22.
Jayaswal AK, Sampath H, Soohinda G, Dutta S. Delirium in medical intensive care units: Incidence, subtypes, risk factors, and outcome. Indian J Psychiatry 2019; 61: 352–8.
Hsieh SJ, Shum M, Lee AN, Hasselmark F, Gong MN. Cigarette smoking as a risk factor for delirium in hospitalized and intensive care unit patients. A systematic review. Ann Am Thorac Soc 2013; 10: 496–503.
Maldonado JR. Neuropathogenesis of delirium: review of current etiologic theories and common pathways. Am J Geriatr Psychiatry 2013; 21: 1190–222.
Ibrahim K, McCarthy CP, McCarthy KJ, et al. Delirium in the Cardiac Intensive Care Unit. J Am Heart Assoc 2018; 7: e008568.
Rahimi-Bashar F, Abolhasani G, Manouchehrian N, Jiryaee N, Vahedian-Azimi A, Sahebkar A. Incidence and Risk Factors of Delirium in the Intensive Care Unit: A Prospective Cohort. Biomed Res Int 2021; 2021: 6219678.
Liu H, Zhao Q, Liu X, Hu X, Wang L, Zhou F. Incidence and interaction factors of delirium as an independent risk of mortality in elderly patients in the intensive units: a retrospective analysis from MIMIC-IV database. Aging Clin Exp Res 2022; 34: 2865–72.
Jaber S, Chanques G, Altairac C, et al. A prospective study of agitation in a medical-surgical ICU: incidence, risk factors, and outcomes. Chest 2005; 128: 2749–57.
Yang J, Li Y, Liu Q, et al. Brief introduction of medical database and data mining technology in big data era. J Evid Based Med 2020; 13: 57–69.
Shahbaz AU, Zhao T, Zhao W, et al. Calcium and zinc dyshomeostasis during isoproterenol-induced acute stressor state. Am J Physiol Heart Circ Physiol 2011; 300: H636-644.
Zhang Z, Xu X, Ni H, Deng H. Predictive value of ionized calcium in critically ill patients: an analysis of a large clinical database MIMIC II. PLoS One 2014; 9: e95204.
Hansen B-A, Bruserud Ø. Hypomagnesemia in critically ill patients. J Intensive Care 2018; 6: 21.
Yoon JS, Kim YR, Choi JW, Ko JS, Gwak MS, Kim GS. Risk factors of postoperative delirium following liver transplantation. Korean J Anesthesiol 2009; 57: 584–9.
Sheean PM, Peterson SJ, Chen Y, Liu D, Lateef O, Braunschweig CA. Utilizing multiple methods to classify malnutrition among elderly patients admitted to the medical and surgical intensive care units (ICU). Clin Nutr 2013; 32: 752–7.
Hill TL. Gastrointestinal Tract Dysfunction With Critical Illness: Clinical Assessment and Management. Top Companion Anim Med 2019; 35: 47–52.

Table 1

Baseline characteristics and outcome data in Delirium group and Non-delirium group.
Variables	Total patients (n = 12029)	Non-delirium group (n = 7222)	Delirium group (n = 4807)	P value
Age (years, M [IQR])	68.36 [57.96, 78.22]	68.24 [57.97, 78.06]	68.51 [57.94, 78.62]	0.116
Gender [Male, n(%)]	7010 (58.3)	4214 (58.3)	2796 (58.2)	0.856
BMI (M [IQR])	27.30 [23.59, 32.28]	27.14 [23.59, 31.97]	27.59 [23.58, 32.82]	0.002
Marital status [n(%)]				< 0.001
Single	3380 (28.1)	1987 (27.5)	1393 (29.0)
Married	5422 (45.1)	3433 (47.5)	1989 (41.4)
Others	3227 (26.8)	1802 (25.0)	1425 (29.6)
Insurance [n(%)]				< 0.001
Medicare	765 (6.4)	445 (6.2)	320 (6.7)
Medicaid	6047 (50.3)	3536 (49.0)	2511 (52.2)
Other	5217 (43.4)	3241 (44.9)	1976 (41.1)
Careunit [n(%)]				< 0.001
CCU	3359 (27.9)	2405 (33.3)	954 (19.8)
MICU	5965 (49.6)	3377 (46.8)	2588 (53.8)
NSICU	268 (2.2)	124 (1.7)	144 (3.0)
SICU	2437 (20.3)	1316 (18.2)	1121 (23.3)
Living situation [n(%)]				< 0.001
Lives alone	1592 (13.2)	1055 (14.6)	537 (11.2)
With spouse	2582 (21.5)	1714 (23.7)	868 (18.1)
With family	2302 (19.1)	1530 (21.2)	772 (16.1)
Others	5553 (46.2)	2923 (40.5)	2630 (54.7)
Comorbidity [n(%)]
Myocardial infarct	2255 (18.7)	1350 (18.7)	905 (18.8)	0.873
Congestive heart failure	4370 (36.3)	2533 (35.1)	1837 (38.2)	< 0.001
Peripheral vascular disease	1471 (12.2)	833 (11.5)	638 (13.3)	0.005
Chronic pulmonary disease	3570 (29.7)	2096 (29.0)	1474 (30.7)	0.056
Rheumatic disease	505 (4.2)	297 (4.1)	208 (4.3)	0.597
Diabetes with cc	1647 (13.7)	958 (13.3)	689 (14.3)	0.101
Malignant cancer	1990 (16.5)	1239 (17.2)	751 (15.6)	0.028
Renal disease	3531 (29.4)	2070 (28.7)	1461 (30.4)	0.043
Severe liver disease	1185 (9.9)	523 (7.2)	662 (13.8)	< 0.001
Metastatic solid tumor	907 (7.5)	571 (7.9)	336 (7.0)	0.067
AIDS	119 (1.0)	69 (1.0)	50 (1.0)	0.714
Severity of illness (M [IQR])
SAPSⅡ	38.00 [31.00, 47.00]	35.00 [28.00, 43.00]	43.00 [35.00, 53.00]	< 0.001
SOFA	6.00 [4.00, 9.00]	5.00 [3.00, 7.00]	8.00 [6.00, 11.00]	< 0.001
GCS	14.00 [11.00, 15.00]	14.00 [14.00, 15.00]	12.00 [8.00, 14.00]	< 0.001
Treatment [n(%)]
Mechanical ventilation	4137 (34.4)	1653 (22.9)	2484 (51.7)	< 0.001
Vasoactive	6418 (53.4)	3286 (45.5)	3132 (65.2)	< 0.001
CRRT	692 (5.8)	499 (6.9)	193 (4.0)	< 0.001
Bronchoscopy	919 (7.6)	181 (2.5)	738 (15.4)	< 0.001
Invasive line	5582 (46.4)	2832 (39.2)	2750 (57.2)	< 0.001
Heparin within 48 h	5098 (42.4)	2897 (40.1)	2201 (45.8)	< 0.001
Elective surgery	487 (4.0)	387 (5.4)	100 (2.1)	< 0.001
ICU LOS (M [IQR])	2.95 [1.82, 5.43]	2.21 [1.52, 3.51]	5.07 [2.89, 9.64]	< 0.001
Hospital LOS (M [IQR])	8.78 [5.53, 15.15]	7.37 [5.02, 11.80]	12.11 [7.13, 20.81]	< 0.001
In-hospital mortality [n(%)]	1666 (13.8)	542 (7.5)	1124 (23.4)	< 0.001
BMI: Body mass index; CCU: Cardiac care unit; MICU: Medical Intensive Care Unit; NSICU: Neurosurgical intensive care unit; SICU: Surgical intensive care unit; ARDS: Acute respiratory distress syndrome; SAPS II Simplified acute physiology score II; SOFA Sequential organ failure assessment; GCS: Glasgow coma scale; CRRT: Continuous renal replacement therapy

Table 2

Hierarchical logistic regression model examining risk of Sepsis-associated delirium (training set).
Hierarchy	Variables	Value	N (%) or mean (SD)		OR (95%CI, p value)
Hierarchy	Variables	Value	SAD	Non-SAD	Model 1	Model 2
Patient factors and clinical features	Age	>=65	2947 (58.3%)	2002 (59.5%)	1.19 (1.05–1.36, p = 0.007)	1.16 (1.01–1.33, p = 0.037)
	BMI	Mean ± SD	28.4 ± 7.5	29.2 ± 8.2	1.00 (0.99–1.01, p = 0.624)	1.00 (0.99–1.01, p = 0.602)
	Marital status	MARRIED/SINGLE	2393 (47.3%)	1387 (41.2%)	0.93 (0.82–1.07, p = 0.309)	0.93 (0.80–1.07, p = 0.279)
	Care unit	CCU	1693 (33.5%)	671 (19.9%)
		MICU/CCU	2356 (46.6%)	1797 (53.4%)	1.79 (1.55–2.08, p < 0.001)	1.72 (1.46–2.03, p < 0.001)
		NSICU/CCU	98 (1.9%)	107 (3.2%)	5.31 (3.77–7.47, p < 0.001)	4.74 (3.24–6.94, p < 0.001)
		SICU/CCU	909 (18%)	789 (23.5%)	1.96 (1.66–2.32, p < 0.001)	1.56 (1.30–1.87, p < 0.001)
	Tobacco condition	Never used	1232 (24.4%)	495 (14.7%)
		Stop tobacco > 1 year	1084 (21.4%)	454 (13.5%)	1.02 (0.86–1.22, p = 0.787)	1.01 (0.84–1.22, p = 0.923)
		Stop tobacco ≤ 1 year	574 (11.4%)	382 (11.4%)	1.65 (1.35–2.01, p < 0.001)	1.55 (1.25–1.91, p < 0.001)
	Temperature	Mean ± SD	36.8 ± 1.1	36.9 ± 1.3	1.06 (1.01–1.12, p = 0.016)	1.01 (0.96–1.07, p = 0.681)
	Urine output	Mean ± SD	1883.5 ± 1294.6	1556.6 ± 1251.8	1.00 (1.00–1.00, p < 0.001)	1.00 (1.00–1.00, p = 0.024)
	Respiratory rate	Mean ± SD	19.6 ± 3.9	20.4 ± 4.2	1.00 (0.99–1.02, p = 0.619)	0.98 (0.97-1.00, p = 0.060)
	Heart rate	Mean ± SD	86.4 ± 15.8	89.8 ± 17.0	1.01 (1.00-1.01, p < 0.001)	1.01 (1.00-1.01, p = 0.001)
	Mechanical ventilation	Yes	1175 (23.2%)	1743 (51.8%)	1.89 (1.66–2.16, p < 0.001)	0.85 (0.73-1.00, p = 0.055)
	Elective surgery	Yes	269 (5.3%)	69 (2.1%)	0.67 (0.49–0.91, p = 0.011)	0.82 (0.59–1.14, p = 0.242)
	Invasive line	Yes	1988 (39.3%)	1940 (57.7%)	1.10 (0.97–1.25, p = 0.131)	1.04 (0.91–1.19, p = 0.542)
	Heparin within 48 h	Yes	2021 (40%)	1533 (45.6%)	1.19 (1.07–1.34, p = 0.002)	1.00 (0.89–1.13, p = 1.000)
	Vasoactive	Yes	2301 (45.5%)	2199 (65.4%)	1.08 (0.94–1.23, p = 0.298)	0.99 (0.86–1.15, p = 0.930)
	CRRT	Yes	363 (7.2%)	139 (4.1%)	0.19 (0.14–0.25, p < 0.001)	0.16 (0.12–0.21, p < 0.001)
	Bronchoscopy	Yes	136 (2.7%)	527 (15.7%)	3.61 (2.89–4.50, p < 0.001)	1.42 (1.10–1.82, p = 0.006)
	SOFA	Mean ± SD	5.4 ± 2.9	8.6 ± 4.0	1.27 (1.24–1.30, p < 0.001)	1.20 (1.16–1.23, p < 0.001)
	SAPS II	Mean ± SD	36.3 ± 11.8	44.6 ± 13.9	1.01 (1.00-1.01, p = 0.003)	1.01 (1.01–1.02, p < 0.001)
	Glucose	Mean ± SD	7.8 ± 2.6	8.2 ± 2.8	1.00 (0.98–1.02, p = 0.801)	0.99 (0.97–1.01, p = 0.288)
	Hemoglobin	Mean ± SD	9.5 ± 2.1	9.4 ± 2.2	1.01 (0.98–1.04, p = 0.515)	1.02 (0.99–1.05, p = 0.291)
	Platelets	Mean ± SD	173.1 ± 104.0	170.2 ± 109.0	1.00 (1.00–1.00, p < 0.001)	1.00 (1.00–1.00, p = 0.005)
	Albumin	Mean ± SD	3.1 ± 0.7	3.0 ± 0.7	0.87 (0.80–0.95, p = 0.002)	0.88 (0.80–0.97, p = 0.008)
	Bicarbonate	Mean ± SD	21.3 ± 4.8	20.2 ± 5.7	0.98 (0.97–0.99, p = 0.004)	0.97 (0.96–0.98, p < 0.001)
	Calcium	Mean ± SD	8.0 ± 0.8	8.0 ± 0.9	1.28 (1.19–1.37, p < 0.001)	1.26 (1.17–1.36, p < 0.001)
	Chloride	Mean ± SD	100.7 ± 6.7	100.3 ± 7.5	0.96 (0.95–0.97, p < 0.001)	0.97 (0.95–0.98, p < 0.001)
	Creatinine M	Mean ± SD	1.8 ± 2.0	2.2 ± 2.1	1.02 (0.98–1.05, p = 0.318)	1.04 (1.01–1.08, p = 0.020)
	Sodium	Mean ± SD	138.7 ± 4.7	139.8 ± 5.9	1.07 (1.05–1.08, p < 0.001)	1.06 (1.04–1.07, p < 0.001)
	Bilirubin total	Mean ± SD	1.7 ± 3.6	2.4 ± 5.5	0.98 (0.97-1.00, p = 0.012)	0.98 (0.97-1.00, p = 0.016)
	Lactate	Mean ± SD	2.5 ± 1.9	3.1 ± 2.8	0.95 (0.93–0.98, p < 0.001)	0.96 (0.93–0.98, p = 0.002)
	NLR	Mean ± SD	10.7 ± 15.1	14.0 ± 16.9	1.00 (1.00-1.01, p = 0.019)	1.00 (1.00-1.01, p = 0.102)
Nursing managements	Diet type	NPO	3147 (62.2%)	1310 (38.9%)
		EN	311 (6.2%)	1480 (44%)		3.97 (3.33–4.74, p < 0.001)
		TPN	31 (0.6%)	50 (1.5%)		2.50 (1.46–4.28, p < 0.001)
		Normal	309 (6.1%)	37 (1.1%)		0.44 (0.31–0.64, p < 0.001)
	Appetite type	Good	2129 (42.1%)	770 (22.9%)
		Middle	62 (1.2%)	50 (1.5%)		0.93 (0.57–1.51, p = 0.765)
		Poor	588 (11.6%)	310 (9.2%)		1.03 (0.84–1.26, p = 0.780)
	Pain	Yes	2378 (47%)	1466 (40.7%)		0.96 (0.84–1.21, p = 0.187)
	Pressure ulcer	Yes	246 (4.9%)	258 (7.7%)		1.09 (0.85–1.39, p = 0.495)
	Urine type	Voids	2432 (48.1%)	1295 (38.5%)
		Catheter	2417 (47.8%)	1911 (56.8%)		1.10 (0.97–1.25, p = 0.126)
	Skin integrity	Impaired	437 (8.6%)	372 (11.1%)
		Intact	4619 (91.4%)	2992 (88.9%)		0.96 (0.79–1.17, p = 0.681)
	Skin care	Yes	386 (7.6%)	137 (4.1%)		0.21 (0.07–0.57, p = 0.002)
	Bed bath	Yes	329 (6.5%)	133 (4%)		1.81 (0.64–5.12, p = 0.266)
	Sputum amount	None	443 (8.8%)	160 (4.8%)
		Small	1157 (22.9%)	2242 (66.6%)		2.23 (1.77–2.82, p < 0.001)
		Moderate	129 (2.6%)	71 (2.1%)		1.34 (0.91–1.98, p = 0.134)
		Copious	14 (0.3%)	2 (0.1%)		0.17 (0.03–0.91, p = 0.038)
		Unknown	3313 (65.5%)	889 (26.4%)		0.84 (0.67–1.04, p = 0.114)
	Head of bed	Low head<45°	2479 (49%)	1387 (41.2%)
		High head>45°	2052 (40.6%)	1198 (35.6%)		1.07 (0.94–1.21, p = 0.309)
		Special position	525 (10.4%)	779 (23.2%)		1.59 (1.33–1.90, p < 0.001)
C-statistics					0.82	0.86
NagelkerkeR²					0.39	0.48
BMI: Body mass index; CCU: Cardiac care unit; MICU: Medical Intensive Care Unit; NSICU: Neurosurgical intensive care unit; SICU: Surgical intensive care unit; SAPS II: Simplified acute physiology score II; SOFA: Sequential organ failure assessment; CRRT: Continuous renal replacement therapy; NLR: Neutrophil to lymphocyte ratio

No competing interests reported.

supplymentarymaterials.docx

Download PDF

Editor assigned by journal
14 Oct, 2024
Editor invited by journal
14 Aug, 2024
Submission checks completed at journal
12 Aug, 2024
First submitted to journal
08 Aug, 2024

You are reading this latest preprint version

Construction of a risk prediction model for sepsis-associated delirium in the intensive unit patients: A retrospective cohort study

Status:

Version 1

Abstract

Figures

Introduction

Patients and methods

2.1 Source of data

2.2 Study design and participants

2.3 Variable extraction and data collection

2.4 Baseline characteristics and outcomes

2.5 Statistical analysis

Results

Discussion

Conclusion

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Version 1