Death in patients with immune-mediated inflammatory diseases in the intensive care unit: First week data

Abstract

BACKGROUND

Systemic immune-mediated diseases can be quite severe in both course and complications, causing multiple organ failure and death. These patients are often required to be hospitalized in an intensive care unit (ICU).

AIM

To find early predictors of death in patients with immune-inflammatory diseases hospitalized in the ICU.

METHODS

The study included 51 patients (23 boys, 28 girls) with immune-inflammatory diseases, including multisystem inflammatory syndrome associated with coronavirus disease 2019 (n = 18), systemic rheumatic diseases (n = 24), and generalized infections (n = 9) aged from 7 months up to 17 years old, admitted to the ICU of the clinic of Saint Petersburg State Pediatric Medical University in the period from 2007 to 2023. All patients were divided into those with a fatal outcome (n = 13) and those who recovered (n = 38). Macrophage activation syndrome (MAS) was diagnosed by the 2016 European League Against Rheumatism/American College of Rheumatology/Paediatric Rheumatology International Trials Organisation criteria.

RESULTS

First-day predictors were white blood cell ≤ 3.1 × 10⁹/L, platelets ≤ 168 × 10⁹/L, diuresis ≤ 1.5 mL/kg/hour, low saturation, K⁺ > 4.7 mmol/L, albumin ≤ 30 g/L, creatinine > 74 μmol/L, pH ≤ 7.36, HCO3^- ≤ 22.2, Glasgow Coma Scale score ≤ 13, Sequential Organ Failure Assessment (SOFA) score > 2, oxygen therapy, mechanical ventilation (MV), fresh frozen plasma transfusions and biological treatment. The third-day predictors were: White blood cell ≤ 4.0 × 10⁹/L, platelets ≤ 63 × 10⁹/L, hemoglobin ≤ 87 g/L, C-reactive protein (CRP) > 129 mg/L, triglycerides > 2.45 mmol/L, albumin ≤ 28 g/L, creatinine > 83.5 μmol/L, pH ≤ 7.38, Glasgow Coma Scale score ≤ 10, SOFA score > 2 and need in MV, intravenous immunoglobulin, and blood transfusion requirements. On the fifth day, the main predictors were CRP > 28 mg/L, triglycerides > 2.3 mmol/L, creatinine > 58 μmol/L, fibrinogen > 3.3 g/L, compliance with the MAS criteria, Glasgow Coma Scale score ≤ 14, SOFA score > 2, and need for MV, vasopressors, and anticoagulant therapy, as well as blood and fresh frozen plasma transfusions. The seventh-day predictors were CRP > 19.1 mg/L, albumin ≤ 35 g/L, total protein ≤ 55 g/L, compliance with the MAS criteria, Glasgow Coma Scale score ≤ 12, SOFA score > 3, and need for MV and biological and anticoagulant therapy.

CONCLUSION

Hemaphagocytosis (leukopenia, thrombocytopenia, hyperferritinemia, increased histochemistry score), progressive decline in Glasgow Coma Scale, increasing SOFA scores, and persistent high CRP levels were markers of an unfavorable outcome in patients with immune-mediated inflammatory diseases.

Key Words: Intensive care unit; Systemic inflammation; Systemic rheumatic diseases; Sepsis; Multisystem inflammatory syndrome associated with COVID-19 in children; Children

Core Tip: Systemic immune-mediated inflammatory diseases can be severe in both course and complications, quite often requiring intensive care unit admission and, in some cases, leading to death. For such patients, careful observation and dynamic monitoring of their condition, along with risk assessment, is critical. The risk factors for a lethal outcome during the first week of intensive care unit stay included signs of hemophagocytosis, progressive decline in the Glasgow Coma Scale, increasing Sequential Organ Failure Assessment scores, and persistent high C-reactive protein levels. Monitoring of these parameters may help to timely identify the patients at high risk of lethal outcome, allowing for the prompt prescription of appropriate treatment.

INTRODUCTION

Systemic immune-mediated inflammatory diseases can be pretty severe in both course and complications. They lead to multiple organs and systems involvement, often requiring urgent intensive care unit (ICU) admission. Around half of children with multisystem inflammatory syndrome associated with coronavirus disease 2019 (MIS-C) required ICU admissions[1,2], with mortality rates of around 1.5%-2.5%[3-6]. Systemic rheumatic diseases, such as systemic lupus erythematosus (SLE), systemic vasculitis, juvenile dermatomyositis, and juvenile idiopathic arthritis (JIA) with systemic onset, also have a high probability of ICU hospitalizations[7]. SLE in this scroll is the main reason for ICU admissions and takes the first place (62.5%) among other systemic rheumatic diseases[8].

Immunosuppressive drugs are the basis for the treatment of systemic rheumatic diseases patients. However, they can be complicated by various infectious diseases, e.g., systemic mycosis, gram-negative bacterial supremacy, especially Staphylococcus aureus and Enterococcus spp., as well as other microorganisms such as Klebsiella spp., Escherichia coli, Pseudomonas aeruginosa, and Acinetobacter baumannii, which is also associated with ICU hospitalization[9-13].

There are many reasons for fatal outcomes in children hospitalized in the ICU. Unfortunately, in the literature, there are few publications on early predictors of lethal outcomes in children with immune-mediated inflammatory diseases hospitalized in the ICU. Some authors have described associations between leukocyte and platelet counts, C-reactive protein (CRP) and ferritin levels, and hypoalbuminemia in children who subsequently died in the ICU; however, these studies were not stratified by the duration of ICU observation and treatment[14-17].

MATERIALS AND METHODS

Study design, data sources, and measurements

The retrospective continuous cohort study included the data about 51 children (23 boys and 28 girls) aged 7 months to 17 years with systemic immune-mediated diseases, in the research: 18 patients with MIS-C, five patients with JIA, six patients with SLE, 13 patients with systemic vasculitis, and nine patients with systemic infections. All of them were admitted to the ICU of Saint Petersburg State Pediatric Medical University from 2007 to 2023. Children with non-inflammatory diseases, accident trauma, and malignant diseases were not included in the study. All patients were divided into two groups. The condition and therapy were assessed in the first week after admission to the ICU. Death and time to death were the primary outcomes of the study.

Data collection

Examination of all patients included the collection of the following data: (1) Demography: Date of birth, date of disease onset, date of ICU admission, and date of death for dead patients, gender, diagnosis. The delayed fatal outcome in our study refers to any fatal cases that occurred during the seven-day observation period; (2) Clinical features: Maximal fever, rash, respiratory rate, heart rate, blood pressure, hepatomegaly and splenomegaly, minute diuresis; (3) Laboratory tests: Hemoglobin level, complete blood cell count, platelets count, electrolytes (Na⁺, K⁺, Phosphate), CRP, erythrocyte sedimentation rate, lactate dehydrogenase, alanine aminotransferase, aspartate aminotransferase, ferritin, procalcitonin, triglycerides, albumin, total protein, creatinine, bilirubin, coagulation (prothrombin index, D-dimer, fibrinogen, blood acid-base state (pH, pO₂, pCO₂, base excess, HCO₃^-); (4) Assessment with different scales: To assess the functional state of the body’s organs and systems, the Sequential Organ Failure Assessment (SOFA) score was used[18]. The Glasgow Coma Scale was used to assess patients’ consciousness[19]. Assessment of hemophagocytosis was performed using various previously published tools, such as the current hemophagocytic lymphohistiocytosis 2004 criteria[20] and the criteria for macrophage activation syndrome (MAS) in systemic JIA, as provided by the Ravelli et al[21], criteria of European League Against Rheumatism/American College of Rheumatology/Paediatric Rheumatology International Trials Organisation (EULAR/ACR/PRINTO)[22] and hemophagocytic syndrome diagnostic score (HScore) calculation[23]; (5) Treatment: Corticosteroid and biological therapy, antibiotic treatment, intravenous immunoglobulin treatment, use of anticoagulants, diuretics, transfusion of erythrocyte suspension, and fresh frozen plasma (FFP); and (6) Intensive treatment: Infusion therapy, oxygen therapy, mechanical ventilation, vasopressors, and inotropes required.

Statistical analysis

The STATISTICA software package, version 10.0 (StatSoft Inc., Tulsa, OK, United States), was used to analyze the obtained data. We used the Kolmogorov-Smirnov test to assess the normality of the distribution and employed non-parametric statistics due to the absence of a normal distribution. The median was used to describe quantitative indicators (i.e., the 25^th and 75^th percentiles). The Pearson criterion χ² was used to compare categorical indicators. The Mann–Whitney t-test was used to compare quantitative indicators. Sensitivity and specificity were determined to show the ability of each trait to differentiate patients with binary traits. Using area under the curve-receiver operating characteristic analysis with a 95% confidence interval determination and calculation of odds ratio, without considering the time of event development using 2 × 2 tables, cut-off values were calculated for quantitative variables. For each studied parameter, sensitivity and specificity were evaluated. Using binary logistic regression, which included quantitative and qualitative indicators associated with the dependent variable, independent predictors were identified. Differences or connections were considered statistically significant at P < 0.05.

RESULTS

General characteristics of patients with systemic immune-inflammatory diseases

The studied cohort consisted of 51 patients hospitalized in the ICU with various immune-mediated inflammatory diseases. There were slightly more girls (n = 28, 54.9%) than boys (n = 23, 45.1%). The median age of patients was 10 years. The patients were divided into three groups based on their diagnosis: MIS-C in children, systemic rheumatic diseases, and generalized infections. The patient groups were as follows: MIS-C, 18 (35.3%) patients; systemic rheumatic diseases, 24 (47.1%) patients; and generalized infections, 9 (17.6%) patients. The group of systemic rheumatic diseases included patients with SLE (n = 8), systemic vasculitis (n = 12), and systemic JIA (n = 4). The median time between disease onset and ICU admission was 10.0 (6.0, 30.0) days, with a range of 2 days to 720 days. Among all children hospitalized in the ICU, 18 (35.3%) received intensive immunosuppressive therapy before admission.

A fatal outcome was recorded in 13 (25.5%) children hospitalized in the ICU in 39 (1762) days after admission. The mortality rates in the nosologic groups were as follows: MIS-C, 1/18 (5.6%); systemic rheumatic diseases, 10/24 (41.7%); and sepsis, 2/9 (22.2%) (P = 0.028). Some patients hospitalized in the ICU developed hemophagocytic syndrome, with a frequency ranging from 8.3% according to the current hemophagocytic lymphohistiocytosis 2004 criteria to 39.6% according to the PRES/EULAR/ACR criteria, and 72.9% according to the MAS criteria for systemic JIA (2005). The median hemophagocytosis index score was 97 (72, 140) points. The data are presented in Table 1.

Table 1 General characteristics of patients hospitalized in the intensive care unit and their main outcomes, n (%)/median (interquartile range).

General characteristics of patients	n = 51
Patients’ demography
Gender, male	23 (45.1)
Age, months	121 (60, 182)
Disease groups
MIS-C	18 (35.3)
Systemic rheumatic diseases	24 (47.1)
Infections	9 (17.6)
Preceding immunosuppression	18 (35.3)
Assessment of MAS
Hscore, points	97 (72, 140)
Correspondence to criteria
HLH 2004	4/48 (8.3)
MAS 2005	35/48 (72.9)
EULAR/ACR/PRINTO	19/48 (39.6)
ICU admission and outcomes
Time to ICU admission from the onset of disease, days	10.0 (6.0, 30.0)
Duration of stay in ICU, days	6.0 (4.0, 15.0)
Lethal outcome	13 (25.5)
Days before deaths	39 (17, 62)

MIS-C: Multisystem inflammatory syndrome associated with coronavirus disease 2019 in children; MAS: Macrophage activation syndrome; HLH: Hemophagocytic lymphohistiocytosis; EULAR: European League Against Rheumatism; ACR: American College of Rheumatology; PRINTO: Paediatric Rheumatology International Trials Organisation; ICU: Intensive care unit.

Comparative analysis between patients from the three studied groups revealed the following differences: Younger patients [41 (19, 60) months] were more likely to have sepsis, older children more often had MIS-C [106 (64, 137) months], and systemic rheumatic diseases [175 (115, 93) months] P = 0.001. The longest time between hospitalization in the ICU and the onset of disease was in patients with systemic rheumatic diseases 30 (3, 8) days, shorter in patients with MIS-C 6 (4, 10) days, and with sepsis 7 (7, 8) days, P = 0.00001.

Treatment

Children with generalized infections stayed in the ICU longer than patients in the other groups. However, there were more patients with systemic rheumatic diseases. Intensive immunosuppressive therapy was more often provided before admission to the ICU in children with systemic rheumatic diseases (62.5%) than in patients with MIS-C (5.6%) or generalized infections (22.2%) (P = 0.0005). Corticosteroids and anakinra were used in children with generalized infection who developed MAS. When calculating the hstochemistry score (hemophagocytosis index), its highest levels were observed in children with systemic rheumatic diseases and generalized infections, and a low level was found in children with MIS-C. Using Raveli-2005 criteria[21], a high prevalence of MAS was established in patients with MIS-C (17.7%). The lowest ICU mortality rate was observed in patients with MIS-C, and in patients with systemic rheumatic diseases and generalized infections, this indicator was higher.

Dynamic analysis of predictors of delayed mortality based on outcomes during the first week of stay in the ICU

Analysis of fatal outcomes in patients hospitalized in the ICU: Among 51 patients with immune-mediated inflammatory diseases hospitalized in the ICU, 13 (25.5%) children died. Fatal outcomes were more common in older children (175 months vs 102 months, P = 0.064). Deaths were documented more frequently in children with systemic rheumatic diseases (41.7%) compared to other diagnoses (MIS-C 7.7%, sepsis 15.4%, P = 0.028). Patients with lethal outcomes were treated with preceding immunosuppression before ICU admission more frequently (53.9% vs 18.4%, P = 0.013). Lethality was associated with a longer time to hospitalization in the ICU [28 (13, 145) days] compared to survivors [7 (6, 16) days, (P = 0.006)]. At the same time, the length of stay in the ICU was higher in deceased patients [39 (9, 62) days], as they were unable to be discharged, compared to the survivors’ group [5 (4, 7) days, P = 0.0006]. To establish predictors of fatal outcomes, an analysis of the dynamics of the main parameters was conducted in patients during the first week of stay in the ICU. Data are presented in Supplementary Table 1.

Analysis of delayed mortality predictors based on first-day outcomes in the ICU: Two patients died during the first day of stay in the ICU. Six patients were discharged from the ICU during the first three days: Two with JIA, two with systemic anti-neutrophil cytoplasmic antibodies-associated vasculitis, one with SLE, and one with MIS-C. In patients who subsequently died, the following changes were noted on the first day of their stay in the ICU compared to survived: Lower diuresis (1.1 mL/kg/hour vs 1.9 mL/kg/hour, P = 0.015), low fluid balance (50% vs 76%, P = 0.026) and saturation on atmospheric air (87 vs 95, P = 0.018). Among laboratory parameters, a tendency to cytopenia was revealed: A lower level of leukocytes [5.1 (1.1, 16.3) × 10⁹/L vs 16.3 (9.9, 20.8) × 10⁹/L, P = 0.015], and platelets [87 (44, 134) × 10⁹/L vs 213 (124, 414) × 10⁹/L, P = 0.001], hypoalbuminemia was also noted [24 (19, 30) g/L vs 30.7 (27, 35.3) g/L, P = 0.009]. According to the acid-base balance assessment, the following differences were noted: Base excess [-4.4 (-7.2, -3.9) mmol/L vs -1.2 (-4.5, 1.3) mmol/L, P = 0.017] and HCO₃^- [20.2 (19, 21.3) mmol/L vs 23.4 (20.7, 26) mmol/L, P = 0.031]. Also, in the group of patients with fatal outcomes, higher levels of potassium [4.9 (3.9, 5.6) mmol/L vs 4.0 (3.8, 4.4) mmol/L, P = 0.025] and creatinine [131 (78, 189) μmol/L vs 50.8 (38, 122) μmol/L, P = 0.021] were observed. Patients who died had relatively lower Glasgow Coma Scale scores [13 (8, 14) points vs 14 (14, 15) points, P = 0.019] and higher SOFA scores [7 (5, 11) points vs 2 (1, 5) points, P = 0.0006] on the first day of ICU stay. Subsequently, patients with fatal outcomes required respiratory support with more stringent parameters of mechanical ventilation [the median respiratory rate was 27 (22, 30) vs 16 (13, 20), P = 0.031]; these patients more often needed mask oxygen therapy (81.8% vs 42.1%, P = 0.020) and mechanical ventilation (41.7% vs 13.2%, P = 0.032). Patients who died were more often treated with plasma transfusions (50% vs 16.7%, P = 0.021) and biological drugs (15.7% vs 0%, P = 0.015). They also required blood transfusions twice as often and were treated with anticoagulants twice as often, but the data were insignificant. It is worth noting that despite the inflammatory nature of the diseases, no differences were found in the levels of CRP, procalcitonin, ferritin, and other indicators of the systemic inflammatory response and cytokine storm on the first day. The analysis identified predictors of delayed mortality based on the results of the first day of patients' stay in the ICU (Table 2).

Table 2 Predictors of lethal outcome (data on the first day of stay in the intensive care unit), median (interquartile range).

Predictor	Se	Sp	OR (95%CI)	RR (95%CI)	P value
Diuresis ≤ 1.5 mL/kg/hour	91.7	61.8	17.8 (2.1-154.2)	2.1 (1.5-3.8)	0.002
Water balance ≤ 74%	100.0	55.9	-	-	0.001
Saturation without oxygen support ≤ 92%	90.9	58.1	15.8 (13.8-139)	2.4 (1.5-3.8)	0.003
Leukocytes ≤ 3.1 × 109/L	50.0	100.0	-	-	0.000004
Platelets ≤ 168 × 109/L	100.0	59.5	-	-	0.0003
Potassium > 4.7 mmol/L	66.7	91.4	15.5 (3.2-75.9)	5.8 (2.1-15.9)	0.0002
Albumin ≤ 30 g/L	90.9	54.3	12.5 (1.4-108.2)	2.1 (1.4-3.1)	0.007
Creatinine > 74 μmol/L	75.0	65.7	5.1 (1.2-22.2)	2.0 (1.2-3.5)	0.023
pH ≤ 7.36	58.3	85.3	8.2 (1.8-36.0)	4.0 (1.6-10.2)	0.003
HCO3- ≤ 22.2	90.0	61.5	14.4 (10.6-131.5)	2.3 (1.4-4.0)	0.006
Glasgow Coma Scale ≤ 13 points	58.3	77.8	4.9 (1.2-19.7)	2.6 (1.5-5.7)	0.019
SOFA > 2 points	100.0	51.5	-	-	0.002
MV-RR > 20 per minute	80.0	83.3	20 (0.9-429.9)	4.8 (0.8-30.3)	0.036
Oxygen therapy via face mask	81.8	78.6	1.9 (1.2-3.1)	6.2 (1.2-32.6)	0.020
Need for mechanical ventilation	41.7	86.8	4.7 (1.1-20.8)	3.2 (1.1-9.1)	0.032
Treatment with bDMARD	15.4	100.0	-	-	0.015
Transfusion of FFP	50.0	83.3	5.0 (1.2-20.9)	3.0 (1.2-7.6)	0.021

Se: Sensitivity; Sp: Specificity; CI: Confidence interval; OR: Odds ratio; RR: Respiratory rate; SOFA: Sequential Organ Failure Assessment; MV: Mechanical ventilation; bDMAD: Biological disease-modifying antirheumatic drugs; FFP: Fresh frozen plasma.

Predictors with the highest sensitivity were diuresis ≤ 1.5 mL/kg/hour, hydro-balance < 74%, saturation in atmospheric air ≤ 92%, platelets ≤ 168 × 10⁹/L, albumin ≤ 30 g/L, and pediatric SOFA scale > 2 points. Predictors with the highest specificity were leukocytes ≤ 3.1 × 10⁹/L, potassium level > 4.7 mmol/L, pH ≤ 7.36, use of biologic drugs, need for mechanical ventilation, and transfusion of FFP. The highest odds ratio was associated with the following predictors: Mechanical ventilation, increased potassium levels, decreased diuresis, and decreased oxygen saturation.

Analysis of delayed mortality predictors based on third-day outcomes in the ICU: Nine patients were discharged from the ICU on the third to fifth day of stay: Five were with MIS-C, two with a septic infection, and two patients with systemic rheumatic diseases. No fatal outcomes were recorded on days 2-5 of patients’ stay in the ICU. The analysis of the third day of ICU stay included 43 patients (11 fatal, 32 survivors), since by this period of data analysis, two patients had died and six had been discharged. When comparing the studied parameters on the third day of stay in the ICU, the following features were noted in deceased patients compared to survivors: In laboratory data - severe leukopenia [2.8 (1.4, 11.2) × 10⁹/L vs 14.4 (9.7, 20.9) × 10⁹/L, P = 0.004] and thrombocytopenia [59 (30, 113) × 10⁹/L vs 211 (101, 359) × 10⁹/L, P = 0.001], more severe anemia [85 (80, 87) g/L vs 98 (90, 114) g/L, P = 0.014], higher levels of CRP [151 (68, 194) mg/L vs 49.7 (12, 123) mg/L, P = 0.025] and triglycerides [4.4 (2.6, 6.6) mmol/L vs 1.7 (1.4, 2.5) mmol/L, P = 0.050], higher creatinine level [91 (66, 278) μmol/L vs 45.3 (33, 74) μmol/L, P = 0.028], more pronounced hypoalbuminemia [27.8 (25, 28) g/L vs 32.9 (31, 35) g/L, P = 0.028], and a lower pH level was revealed in the group of deceased patients [7.38 (7.33, 7.44) vs 7.45 (7.43, 7.48), P = 0.031]. The deceased had a lower score on the Glasgow Coma Scale [7 (3, 14) points vs 15 (14, 15) points, P = 0.0001] and a higher score on the pediatric multiple organ failure scale SOFA [10 (6, 13) points vs 1 (1, 4) points, P = 0.0001]. The deceased patients more often needed mechanical ventilation (70% vs 6.9%, P = 0.00004), plasma transfusion (55.6% vs 20%, P = 0.033), and less anticoagulant therapy (30% vs 71.4%, P = 0.017). A tendency towards a higher hemophagocytosis index HScore was revealed [98 (57, 134) vs 72 (38, 96), P = 0.074]. Detailed data are presented in the Supplementary Table 1.

The third-day predictors of death were established (Table 3). The SOFA scale with more than 2 points had the highest sensitivity. Treatment with intravenous immunoglobulin (IVIG) and blood transfusion on day 3 significantly decreased the risk of the delayed fatal outcome. Leukopenia (≤ 4.0 × 10⁹/L), creatinine level > 83.5 μmol/L, pH ≤ 7.38, Glasgow Coma Scale ≤ 10 points, and mechanical ventilation had the highest specificity. Thrombocytopenia (platelet count ≤ 63 × 10⁹/L), hypertriglyceridemia (triglyceride level > 2.45 mmol/L), hypoalbuminemia (albumin level ≤ 28 g/L), anemia (hemoglobin level ≤ 87 g/L), and increased CRP (CRP level > 129 mg/L) had similar sensitivity and specificity (Table 3).

Table 3 Predictors of lethal outcome (data on the third day of staying in the intensive care unit), median (interquartile range).

Predictor	Se	Sp	OR (95%CI)	RR (95%CI)	Р value
Leukocytes ≤ 4.0 × 109/L	60.0	97.1	46.5 (4.4-492.1)	7.9 (3.0-21.0)	0.00001
Platelets ≤ 63 × 109/L	70.0	88.2	17.5 (3.2-96.6)	7.0 (2.2-22.5)	0.0002
Hemoglobin ≤ 87 g/L	80.0	82.4	18.7 (3.1-111.0)	8.6 (2.1-35.2)	0.0002
CRP > 129 mg/L	66.7	78.8	8.0 (1.6-40.2)	4.5 (1.3-15.1)	0.009
Triglycerides > 2.45 mmol/L	75.0	81.8	13.5 (0.9-207.6)	3.0 (0.8-44.0)	0.039
Albumin ≤ 28 g/L	77.8	86.2	21.9 (3.3-145.2)	8.6 (2.1-35.1)	0.0002
Creatinine > 83.5 μmol/L	66.7	83.5	11.6 (2.2-62.2)	5.1 (1.5-16.9)	0.002
pH ≤ 7.38	55.6	89.7	12.1 (2.1-71.1)	4.7 (1.6-13.5)	0.004
Glasgow Coma Scale score ≤ 10 points	70.0	91.2	24.1 (4.0-145.6)	7.9 (2.5-25.2)	0.00005
SOFA > 2 points	100.0	71.0	-	-	0.00009
Mechanical ventilation	70.0	93.1	31.5 (4.4-226.5)	7.8 (2.5-24.1)	0.00004
No IVIG treatment	96.9	27.3	11.6 (1.1-127.3)	3.2 (0.6-17.5)	0.018
Blood transfusion	27.3	21.9	0.11 (0.02-0.5)	0.2 (1.1-3.3)	0.003

Se: Sensitivity; Sp: Specificity; CI: Confidence interval; OR: Odds ratio; RR: Respiratory rate; CRP: С-reactive protein; SOFA: Sequential Organ Failure Assessment; IVIG: Intravenous immunoglobulin.

Analysis of delayed mortality predictors based on third-day outcomes in the ICU: The analysis of the fifth day included 34 patients: 11 patients who subsequently died and 23 patients who survived. On the fifth day of stay in ICU the following differences were observed: Patients from the lethal group had tachycardia [118 (110, 130) bpm vs 90 (80, 110) bpm, P = 0.016], higher levels of CRP [69.3 (58.8, 196.5) mg/L vs 16.5 (14.3, 27.3) mg/L, P = 0.001], triglycerides [4.5 (2.9, 6.1) mmol/L vs 2.2 (1.1, 2.8) mmol/L, P = 0.043], creatinine [80 (77, 208) μmol/L vs 48.6 (38, 65) μmol/L, P = 0.023], and fibrinogen [4.5 (3.6, 6.1) g/L vs 1.8 (1.2, 2.6) g/L, P = 0.001] were revealed. When evaluating diagnostic scales, deceased patients had a lower Glasgow Coma Scale score [8 (3, 14) vs 15 (14, 15), P = 0.001] and higher SOFA pediatric scores [7.5 (6, 12) points vs 1 (1, 6) points, P = 0.001]. Deceased patients more often required blood transfusions (40% vs 4.4%, P = 0.009) and FFP transfusions (40% vs 13%, P = 0.021), but they were less likely to use anticoagulant therapy (30% vs 82.6%, P = 0.003). They also required vasopressor support more often (30% vs 4.4%, P = 0.038). Among the deceased patients, nearly half met the 2016 EULAR/ACR/PRINTO criteria for MAS, whereas among the survivors, the frequency of MAS was four times lower (55.6% vs 13.6%, P = 0.015). Detailed data are presented in the Supplementary Table 1.

The predictors of fatal outcomes, based on the results of the fifth day of stay in the ICU, are presented in Table 4. Among the parameters with the highest sensitivity, it is necessary to note elevated inflammatory markers associated with hemophagocytosis, including CRP and fibrinogen, as well as hypertriglyceridemia and a more frequent presence of MAS, according to the EULAR/ACR/PRINTO 2016 criteria. Additionally, deceased patients with further complications had a higher pediatric SOFA scale, Glasgow Coma Scale score ≤ 14, and the presence of tachycardia (heart rate > 97 per minute). They require blood transfusion and FFP transfusion, as well as vasopressors and mechanical ventilation. Treatment with anticoagulants decreased the risk of the following lethal outcome. No fatal outcomes were recorded during the 6- to 7-day stay in the ICU; during this period, eight patients were discharged. Among the discharged patients, half (n = 4) had systemic rheumatic diseases, specifically JIA (n = 2), systemic vasculitis (n = 1), and SLE (n = 1). The other half of the patients (n = 4) had MIS-C. On the seventh day of the patient's stay in the ICU, the following differences were noted. In the lethal outcome group, tachycardia persisted [124 (110, 130) beats per minute vs 95 (80, 113) beats per minute, P = 0.008], higher levels of CRP [47 (26.4, 162.3) mg/L vs 12 (3, 15.4) mg/L, P = 0.001], bilirubin [19.9 (8.7, 29.6) μmol/L vs 9.1 (5.4, 13.2) μmol/L, P = 0.032], and fibrinogen [3.4 (2, 4.4) g/L vs 1.3 (1.3, 2.6) g/L, P = 0.015] were observed. Also, patients from this group had a greater tendency to hypoalbuminemia [albumin - 30 g/L (27, 35) vs 36 (34, 36.9) g/L, P = 0.013] and hypoproteinemia [total protein - 50 (46, 54) g/L vs 64 (61.1, 69) g/L, P = 0.013]. According to diagnostic scales, patients from the first group had a lower score on the Glasgow Coma Scale [8.5 (5, 14) vs 15 (14, 15), P = 0.001] and a higher score on the pediatric SOFA multiple organ failure scale [8 (5, 13) points vs 1 (0, 3) points, P = 0.001]. Patients who subsequently died were more likely to meet the ACR/EULAR/PRINTO MAS 2016 criteria (44.4% vs 15.4%, P = 0.015). They were more often prescribed biological therapy (30% vs 0%, P = 0.034) and less often anticoagulant therapy (30% vs 84.6%, P = 0.008). Detailed data are presented in Supplementary Table 1.

Table 4 Predictors of lethal outcome (data on the fifth day of staying in the intensive care unit), median (interquartile range).

Predictor	Se	Sp	OR (95%CI)	RR (95%CI)	Р value
Maximum HR > 97 per minute	91.7	18.5	2.5 (0.3-24.1)	2.0 (0.3-12.8)	0.039
CRP > 28 mg/L	88.9	85.7	48.0 (4.3-535.3)	13.8 (2.0-96.3)	0.0001
Triglycerides > 2.3 mmol/L	100.0	63.6	-	-	0.029
Creatinine > 58 μmol/L	80.0	69.6	9.1 (1.5-54.5)	4.8 (1.2-19.3)	0.009
Fibrinogen > 3.3 g/L	80.0	95.5	84.0 (6.7-1060)	10.2 (2.7-39.2)	0.00001
Glasgow coma scale score ≤ 14 points	90.0	69.6	20.6 (2.2-195.0)	9.6 (1.4-67.2)	0.002
SOFA > 2 points	100.0	70.0	-	-	0.0003
Need in MV	54.5	91.3	12.6 (1.9-82.1)	3.9 (1.6-9.4)	0.002
Need in vasopressors	27.3	95.7	7.9 (0.7-91.3)	2.8 (1.2-6.4)	0.038
Compliance with the MAS criteria EULAR/ACR/PRINTO 2016	55.6	86.4	7.9 (1.3-47.5)	3.6 (1.3-10.2)	0.015
No anticoagulant therapy	82.6	72.7	12.7 (2.3-70.0)	2.6 (1.1-5.9)	0.002
Blood transfusion	36.4	95.7	12.6 (1.2-131.9)	2.9 (1.4-6.4)	0.009
Transfusion of FFP	36.4	87.0	3.8 (0.7-21.4)	2.2 (0.9-5.5)	0.021

Se: Sensitivity; Sp: Specificity; CI: Confidence interval; OR: Odds ratio; RR: Respiratory rate; HR: Heart rate; CRP: C-reactive protein; SOFA: Sequential Organ Failure Assessment; MV: Mechanical ventilation; MAS: Macrophage activation syndrome; EULAR: European League Against Rheumatism; ACR: American College of Rheumatology; PRINTO: Paediatric Rheumatology International Trials Organisation; FFP: Fresh frozen plasma.

Based on the seventh day of stay in the ICU, the following predictors of a fatal outcome were identified: Tachycardia, elevated inflammatory markers - CRP and fibrinogen, elevated bilirubin levels, higher SOFA scores (pediatric), lower Glasgow Coma Scale values, hypoproteinemia and hypoalbuminemia, compliance with MAS EULAR/ACR/PRINTO 2016 criteria, more frequent use of biological therapy. Treatment with anticoagulants is associated with a lower probability of a lethal outcome. The data are presented in Table 5.

Table 5 Predictors of lethal outcome (data on the seventh day of staying in the intensive care unit), median (interquartile range).

Predictor	Se	Sp	OR (95%CI)	RR (95%CI)	Р value
Maximum HR > 115 per minute	70.0	91.7	25.7 (2.2-298.5)	4.1 (1.5-11.5)	0.003
CRP > 19.1 mg/L	100.0	92.3	-	-	0.00003
Albumin ≤ 35 g/L	88.9	61.5	12.8 (1.2-135.6)	5.5 (0.8-36.9)	0.018
Total protein ≤ 55 g/L	88.9	84.6	44.0 (3.4-573.4)	9.6 (1.4-64.3)	0.0007
Glasgow Coma Scale score ≤ 12 points	70.0	92.3	28.0 (2.4-323.7)	4.4 (1.5-12.5)	0.002
SOFA > 3 points	100.0	83.3	-	-	0.00009
Need in MV	70.0	91.7	25.7 (2.2-298.5)	4.1 (1.5-11.5)	0.003
Meeting the EULAR/ACR/PRINTO 2016 criteria	44.4	84.6	4.4 (0.6-32.5)	2.1 (0.9-5.4)	0.015
Treatment with bDMARD	30.0	100.0	-	-	0.034
Anticoagulant therapy	30.0	15.4	0.08 (0.01-0.59)	0.28 (0.1-0.8)	0.008

Se: Sensitivity; Sp: Specificity; CI: Confidence interval; OR: Odds ratio; RR: Respiratory rate; HR: Heart rate; CRP: C-reactive protein; SOFA: Sequential Organ Failure Assessment; MV: Mechanical ventilation; EULAR: European League Against Rheumatism; ACR: American College of Rheumatology; PRINTO: Paediatric Rheumatology International Trials Organisation; bDMARD: Biological disease-modifying antirheumatic drugs.

DISCUSSION

We identified the dynamics of the main clinical and laboratory parameters (on days 1, 3, 5, and 7) associated with the fatal outcome in patients with immune-mediated inflammatory diseases. Analysis of the dynamics of parameters of the first week of intensive care showed that persistent leukopenia, persistent thrombocytopenia, hyperferritinemia, hypercreatininemia, progressive acidosis, progressive decrease in the Glasgow Coma Scale, an increase in the SOFA index, a slight decrease in CRP, as well as a progressive increase in the hemophagocytosis index Hscore, were associated with the risk of death. Moreover, by the end of the first week of intensive care, 44.4% of patients met the EULAR/ACR/PRINTO 2016 criteria for MAS. Delayed blood transfusion, absence of IVIG treatment, and absence of anticoagulant therapy were associated with increased probability of the lethal outcome.

So, in general, the age over 162 months, time before admission to the ICU from the onset of a disease of more than 11 days, immunosuppressive therapy, invasive mycosis, and stay in the ICU for more than 1 week were associated with death in children of such a group. The majority of children hospitalized in the ICU had systemic rheumatic diseases. This group of patients, especially those who had SLE, more often had lethal outcomes. Also, Australian colleagues published data that, in patients with rheumatic diseases, age and severity of the disease were associated with 28-day mortality[24]. Treatment in the ICU for more than 15 days was associated with fatal outcomes and thus reflected the severity of the disease and the risk of developing nosocomial infections. At the same time, another study describes that a stay of 4 days or longer in the ICU (P = 0.001) is associated with unfavorable outcomes[25].

Additionally, we have identified specific dynamics in laboratory parameters associated with a lethal outcome, characterized by the most pronounced changes until the fifth day, with subsequent values approaching normal levels. Such dynamics demonstrate a positive relationship between the severity of the condition when the patient was just admitted to the ICU and further fatal outcomes.

Similar data have been reported in the literature: The lymphocyte/Leukocyte ratio (14.4%), CRP value (12.5%), comorbidities (12.5%), and leucocytopenia (8.9%) were associated with a fatal outcome in children with severe acute respiratory syndrome Coronavirus 2 (SARS-Cov-2) who were hospitalized in the ICU[14,15]. Among children with SLE who did not survive, the level of hemoglobin sharply decreased[16]. In adults, a positive relationship was observed between serum ferritin levels, thrombocytopenia, and fatal outcomes. Also, hypoalbuminemia was associated with a need for life support treatment[17].

Furthermore, in our study, mechanical ventilation, a respiratory rate of more than 20 breaths per minute, oxygen therapy through a face mask, biological therapy, and transfusion of FFP were identified as early predictors of a fatal outcome. As laboratory markers, such dynamics explain the significance of the initial condition of a patient when the child is just admitted to the ICU. If severe conditions occur when a patient is admitted to the ICU, there is a much higher risk of a fatal outcome. In the international literature, secondary diseases, the need for mechanical ventilation, and vasopressor support on admission are identified as factors that may lead to unfavorable outcomes[8,25]. Solórzano-Santos et al[26] show that the presence of more than one other disease can increase the odds of death by up to 10-fold in children with SARS-Cov-2 hospitalized in the ICU.

It is clear that children with rheumatic diseases who end up in the ICU are often in severe condition and receive biological therapy and immunosuppressive therapy due to the activity of the primary disease. In such a cohort of children, the risk of developing nosocomial infections and mycoses is very high. Invasive fungal infections were the leading cause of death (44.4%), and pulmonary infections were the predominant site (55.5%) in patients with lupus erythematosus[16]. For example, patients with lupus erythematosus and invasive mycosis were more likely to have fatal outcomes[27-29]. Limited access to IVIG and anakinra occurred during MIS-C, leading to the use of high-dose intravenous corticosteroids as first-line therapy and tocilizumab in severe cases[30]. An association was found between the severity of disease in patients with MIS-C and a higher incidence of hypotension/shock, neurologic and respiratory myocardial injury, and the incidence of MAS[3]. Borgia RE et al[31] studied SLE in children and found that the risk of death was higher in those who developed MAS. According to data from Spanish colleagues, the mortality rate in patients with JIA and MAS is about 6.5%[32].

It has been observed that reducing corticosteroid administration results in a decrease in the incidence of generalized infections among immunocompromised patients[33]. At the moment, factors predisposing to the development of invasive mycosis in patients with rheumatic diseases have been identified: Neutropenia < 500 cells/μL (< 0.5 × 10⁹/L), persisting for 10 days or more; long-term use of corticosteroids ≥ 0.3 mg/kg/day (prednisolone) for at least 3 weeks during the previous 60 days; treatment with T-cell suppressing drugs during the last 90 days; treatment with B-cell suppressing drugs[34,35].

By paying attention to prognostic features daily, it becomes possible to understand the disease’s further course most clearly. Avoiding a prolonged stay in the ICU is one of the factors associated with a favorable outcome in patients with rheumatic diseases. There should be alertness about invasive mycosis in patients with risk factors (immunosuppressive therapy and other features). Limitations of our study include its retrospective design, single-center setting, relatively small sample size, incomplete data, and varying study time intervals. Also, we did not investigate the pathophysiological mechanisms linking identified predictors to mortality. Therefore, only predictors with statistical significance were considered.

CONCLUSION

The signs associated with hemophagocytosis, such as leukopenia, thrombocytopenia, hyperferritinemia, hypercreatininemia, as well as a progressive increase in the hemophagocytosis index Hscore, tendency to acidosis, progressive decrease in the Glasgow Coma Scale, and an increase in the SOFA index, persistent high CRP, can be used as markers of an unfavorable outcome among patients with immune-mediated inflammatory diseases. The role of hemophagocytosis in patients for whom it is not suspected is underestimated. Careful monitoring of these parameters may enable the timely identification of patients at high risk of an unfavorable outcome, allowing for the prompt prescription of appropriate treatment. Future trials are required.