This study is an observational and analytic study of a retrospective cohort. Patients who were followed up in Akdeniz University (Türkiye) Pediatric Intensive Care Unit between January 2019 and September 2021 were included in the study. All children older than 1 month and younger than 18 years of age, who needed intensive care and survived at the end of the first 24 h for surgical and medical purposes were included in the study. The exclusion criteria were patients whose length of stay in the PICU less than 24 h,20,21 patients with deficiencies in their data in the registry system, chronic kidney failure patients and those who were admitted to the intensive care unit due to kidney transplantation. The informed consent form was obtained from both the children themselves and their parents if adequate communication could be established, and if communication could not be established, it was obtained from the parents of the children. The study was approved by the Akdeniz University clinical studies ethics committee, and was conducted in accordance with the terms specified in the Declaration of Helsinki.

Daily fluid data was obtained for each patient during the study period. Clinical, demographic, diagnosis-related and laboratory data also collected through the first 24 h after ICU admission from electronic medical records. The total follow-up period of the study patients was 28 days or upon discharge from the PICU.

Total of fluid inputs and outputs were measured daily to quantify fluid overload. Total daily input was calculated as the sum of all oral and intravenous fluids administered to the patient including maintenance fluids, blood products, medications and nutritional support. Total daily output was calculated as the sum of all output volumes including urine, stool, drain output, gastrointestinal aspirates and fluid removal by renal replacement therapies. Urine output was calculated by measuring the urine volume if the patient had a urethral catheter and by weighing disposable diapers if the patient did not. For standardisation, net fluid calculations were done based on 8 am every day. Insensible losses were not taken into calculations. Daily fluid accumulation = [net daily input (L)—net daily output (L)]/PICU admission weight (Kg) × 100.22 Bhaskar et al. found that in their cohort, both survivors and non-survivors, fluid accumulation peaked on the second day, therefore we recorded the data of our patients for 3 days.19 The term 'early fluid overload' was defined for the fluid accumulation occuring during the first 24 h of admission to PICU.23 The term 'peak fluid overload' was defined for the highest daily fluid accumulation on any day of the first 3 days.15 The term 'cumulative fluid overload' was defined for the total net fluid accumulation of the first 3 days. Daily maintenance fluid requirement was determined according to the body surface area of ​​the patients. The decision to restrict the amount of daily fluid intake or replace ongoing losses was made by the medical team.

The illness severity was evaluated by the PRISM IV,24 PELOD II score,25 which were previously defined in the original studies. Physiologic variables were measured only in the first 4 h of PICU care and laboratory variables were measured in the time period from 2 h before PICU admission through the first 4 h.

International pediatric sepsis consensus conference: definitions for sepsis and organ dysfunction in pediatrics 2005 criteria were used to define shock and organ failure.26 'The Pediatric Acute Lung Injury Consensus Conference Group' criteria were used for the diagnosis and classification of ALI.27 The presence of acute kidney injury (AKI) was defined according to the KDIGO criteria.28 While performing KDIGO staging, we used the AKI score resulting from the cumulative application of creatinine and urine output, which was applied in the study of Sutherland et al.29 PICU admission creatinine values were accepted as basal creatinine. If the patient's hospitalization creatinine value was high, the values in the last 2 weeks were considered as 'basal creatinine'.30 For all patients older than 1 year who had no 'baseline creatinine' value in the system in the last 2 weeks, the normal creatinine level was calculated with an age-dependent equation.31 For children younger than 1 year, the curve of estimated creatinine values was used.32 Information on the use of furosemide during the first three day after PICU admission was recorded, frequencies and the cumulative doses per kilogram of body weight were calculated.33

Our primary outcome was PICU mortality. Secondary outcomes were intensive care unit-free days (IFD) and ventilator-free days (VFD) up to 28 days. The term 'VFD' was used for the days when patients did not need a mechanical ventilator in the first 28 days. If the patient remained on mechanical ventilator for more than 28 days or died within 28 days without weaning from the mechanical ventilator, the day without ventilator was recorded as zero. The term 'IFD' was used to evaluate the PICU hospitalization of the patients. For patients who were followed up in the PICU for more than 28 days or died within 28 days, the day without intensive care was recorded as zero.

Statistical tests were performed using the 'Statistical Package for Social Science' (SPSS) 23 program. Descriptive results were expressed as mean ± standard deviation (SD) for continuous variables. All definitions related with FO were analyzed separately as a continuous and categorical variable. Fluid overload was categorized according to ‘5%’ and ‘10%’ thresholds and analyzed for demographic characteristics, mortality and secondary outcomes. Results were compared simultaneously as early, peak and cumulative fluid overload. These values were determined by reference to previous studies.8,17,19,23,34–37 Depending on the distribution of the data, the student’s t test for unpaired samples and the Mann–Whitney U test were used to determine the differences between two groups. To compare the differences of categorical variables between groups, we used Fisher's exact test or the chi-square test. To examine correlations, we used Spearman's analysis. We performed univariate regression analysis to calculate the 95% confidence interval (CI) and odds ratio (OR) to determine the associations of fluid accumulation with mortality. Predictors of mortality were determined using multivariable logistic regression models. All potential confounders were included in the initial model. Stepwise regression procedures were used to produce the final models. In order to determine whether there was a significant multicollinearity between the variables, collinearity diagnostics were performed using variance inflation factor (VIF) and tolerance values and variables with a VIF value greater than 5 were not included in the model. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was calculated to assess the predictive ability. The Youden index was calculated to assess the best alignment between sensitivity and specificity. All statistical tests were 2-sided and p values <0.05 were considered statistically significant.

During the study period, a total of 1064 children were admitted to the PICU. 28 patients with chronic renal failure and 25 patients whose cardiopulmonary arrest could not be stabilized 2 h after admission to the intensive care unit were excluded from the study. 430 patients with less than 24 h of intensive care stay and missing information in the registry system were excluded from the study. A total of 581 patients were included in the study (Fig. 1). Respiratory diseases were the most frequent reason for PICU admission, accounting for 27.2% of cases. This was followed by postoperative diseases at 26% and neurologic diseases at 12.2% (Suplemental Table 1). The PICU mortality rate of the whole cohort was 11.7%. 42% of patients were admitted to the PICU directly from the emergency service. The overall median age was 53 (1–215) months. 335 (57.7%) of the patients participating in the study were male.

Figure 1.

Study population.

(0.2MB).

There was no significant difference between these two groups in terms of age, weight and gender. Non-survivors had higher admission lactate levels, PRISM IV and PELOD II scores compared to survivors (p < 0.001). A higher rate of AKI, shock, and ALI observed in non-survivors during PICU stay (p < 0.001). Also compared to survivors, non-survivors had fewer VFD and IFD (p < 0.001). It was observed that renal support treatments such as RRT and diuretic administration were used more frequently in non-survivors (p < 0.001) (Table 1).

When fluid accumulation was analyzed in terms of mortality, the percentage of fluid overload was 1.86 ± 2.58 and 3.35 ± 4.06 on the first day, peak fluid overload was 2.87 ± 2.75 and 5.54 ± 4.03, cumulative fluid overload was 3.40 ± 4.7 and 8.16 ± 7.26, in survivors and non-survivors respectively (Table 2, Fig. 2). During the initial 3 days in the ICU, the mean early, peak and cumulative fluid accumulation was higher for non-survivors than for survivors. ≥5% early fluid overload occurred in 63 patients, ≥5% peak fluid overload occurred in 111 patients and ≥10% cumulative fluid overload occurred in 55 patients.

Figure 2.

Relationship of fluid overload to survival.

(0.14MB).

The threshold value was determined as ‘5%’ for early and peak fluid overload and ‘10%’ for cumulative fluid overload. Patients with ≥5% early fluid overload had fewer VFD and IFD, higher rates of mortality, need for RRT, AKI and shock development were observed compared to <5% fluid overload. Similarly patients with ≥5% peak fluid overload had fewer VFD and IFD, higher rates of mortality, need for RRT, AKI, shock and ALI development were observed compared to <5% fluid overload. Also patients with ≥10% cumulative fluid overload had fewer VFD and IFD, higher rates of mortality, need for RRT, AKI, shock and ALI development were observed compared to <10% fluid overload (Table 3).

To determine whether fluid overload was independently associated with mortality in the general PICU population, multiple regression models were constructed using variables with p < 0.05 in univariate statistics. Variables with high correlations and high VIF values ​​were not included in the same model. Finally, two different models were generated.

In the first model for fluid overload as a continuous variable analysis, the final model retained ALI (aOR 5.74; 95% CI 2.4–13.7; p < 0.001), shock (aOR 2.37; 95% CI 1.05–5.33; p = 0.036), PRISM IV (aOR 1.04; 95% CI 1.02–1.1; p < 0.001), AKI (aOR 14.4; 95% CI 4.67–44.3; p < 0.001), mechanical ventilator (aOR 13.45; 95% CI 1.42–126.7; p = 0.023) and peak fluid overload (aOR 1.15; 95% CI 1.05–1.26; p = 0.002) associated with mortality (Table 4).

In second model for fluid overload as a continuous variable analysis, the final model retained ALI (aOR 5.39; 95% CI 2.2–13; p < 0.001), shock (aOR 2.41; 95% CI 1.1–5.4; p = 0.033), PRISM IV (aOR 1.04; 95% CI 1.02–1.1; p < 0.001), AKI (aOR 15.2; 95% CI 4.9–46.9; p < 0.001), mechanical ventilator (aOR 12.59; 95% CI 1.5–108.1; p = 0.021) and cumulative fluid overload (aOR 1.10; 95% CI 1.04–1.2; p < 0.001) associated with mortality (Table 5).

Although early fluid overload associated with mortality in univariate analysis, the association did not remain significant after controlling for ALI, shock, PRISM IV, AKI, MV, peak FO and cumulative FO.

The predictive ability of fluid overloads for PICU mortality was showed in Fig. 3. The peak fluid overload achieved AUC of 0.731 (95% CI, 0.66−0.80; p < 0.001) for predicting mortality. For the 5.05% cut-off value, detected sensitivity was 51% and the specificity was 85%. The cumulative fluid overload achieved AUC of 0.725 (95% CI, 0.65−0.80; p < 0.001) for predicting mortality. For the 7.58 % cut-off value, detected sensitivity was 51% and the specificity was 89% (Table 6).

Figure 3.

ROC curves for ability of peak and cumulative fluid overload to predict mortality.

(0.09MB).

Spearman correlation analysis was used to show the relationship between fluid accumulation and IFD-VFD. There was a significant negative weak correlation between the peak fluid overload and IFD (ρ, −0.359; p < 0.001). Furthermore, there was a negative weak correlation between the peak fluid overload and the VFD (ρ, −0.288; p < 0.001). The cumulative fluid overload had significant negative weak correlation with IFD (ρ, −0.315; p < 0.001) and VFD (ρ, −0.228; p < 0.001).