Abstract
Background:
The consensus definition of acute kidney injury (AKI) has evolved since developing the original multiple organ dysfunction syndrome (MODS) definitions. Whether or not risk for adverse short- and long-term outcomes can be identified using the refined AKI criteria in the setting of MODS has not been studied. We hypothesize that incorporation of kidney disease: improving global outcome (KDIGO) AKI criteria into existing MODS definitions will have a higher association with major adverse kidney events at 30 days (MAKE30) and will increase the number of patients with MODS.
Methods:
Post-hoc analysis of 410 children admitted to a tertiary care pediatric intensive care unit (PICU). MODS was defined using two existing criteria (Goldstein and Proulx) during the first 7 days following ICU admission and then modified by replacement of the kidney injury criteria using the KDIGO AKI definitions (G’ and P’).
Results:
MAKE30 occurred in 65 of 410 (16%) children. After substituting KDIGO kidney injury criteria, identification of MAKE30 increased from 46 children (71%) to 53 (82%) and 29 children (45%) to 43 (66%) for the Goldstein and Proulx criteria, respectively. Additionally, identification of MODS increased from 194 (47%) by Goldstein to 224 (55%) by G’ and 95 children (23%) by Proulx to 132 (32%) by P’.
Conclusion:
Substituting KDIGO AKI criteria into existing MODS criteria increases the sensitivity for major adverse kidney events as well as the identification of MODS, improving the detection of children at risk for long-term adverse renal outcomes.
Keywords: acute kidney injury, multiple organ dysfunction syndrome, Kidney Disease: Improving Global Outcomes, major adverse kidney events
Introduction:
Nearly 1 in 5 children in the pediatric intensive care unit (PICU) experience multiple organ dysfunction syndrome (MODS), a life-threatening condition defined as two or more organ systems with dysfunction in an acutely ill child [1–4]. Multiple studies have demonstrated higher mortality and worse functional outcomes with increasing number of failing organ systems in various pediatric populations with MODS [1, 2, 5, 6]. The diagnostic criteria for MODS were not scientifically validated nor weighted for importance, leading to the identification of the sickest patients but with a varied identification of MODS-associated outcomes [7].
The consensus definition of acute kidney injury (AKI) has evolved since the foundational MODS descriptions, while MODS definitions have not [8–10]. The Kidney Disease: Improving Global Outcomes (KDIGO) Work Group introduced an improved definition of AKI in 2012 [10]. With this definition, over 1 in 4 critically ill neonates and children are identified to have AKI in the first seven days of ICU stay, and more severe AKI is associated with higher mortality [11]. Contemporary MODS definitions of kidney injury utilize blood urea nitrogen (BUN) levels and absolute cut offs or relative age-based changes for serum creatinine, and do not stage or weight the severity of injury. Alternatively, KDIGO criteria incorporate both urine output as well as validated stratifications of creatinine measurements in order to define and stage kidney injury, which are not part of current MODS definitions [8–10]. In fact, urine output alone will identify nearly 20% of children with AKI, with similar outcomes as definitions based solely on creatinine [12]. As such, current MODS definitions do not reflect current definitions of AKI but instead only account for severe kidney injury.
Finally, major adverse kidney events within 30 days (MAKE30) has been identified as an important patient-centered outcome and has recently been utilized in evaluations of outcomes in pediatrics [13, 14]. In a database cohort of critically ill children, MAKE30 was common, occurring in nearly 10% [13]. However, this study included children based on ICD-9 coding that does not reflect current AKI definitions and therefore may not fully represent a population at risk of poor outcomes [13].
In order to evaluate the impact on long-term outcomes of utilizing updated AKI definitions within MODS criteria, we tested if the rate of identification of MAKE30 would change when the current AKI criteria in MODS were substituted with KDIGO AKI criteria in a cohort of children at risk for MODS. Secondly, in order to evaluate the impact of changing definitions on overall detection of MODS, we measured the rate of detection of MODS after inclusion of KDIGO AKI criteria. We hypothesize that incorporation of KDIGO AKI criteria into existing MODS definitions will have a higher association with MAKE30 and will increase the identification of the number of patients with MODS.
Materials and Methods:
Study Population
A prospective, observational convenience cohort of patients admitted to a PICU at a tertiary care children’s hospital from March 2017 through August 2018 was evaluated, as part of a larger study of clinician identification and prediction of MODS [15]. All patients admitted during the study period with an expected length of stay greater than 2 days at admission, as determined by a member of the care team, and had complete renal injury information were included. Complete renal injury information consisted of at least 1 serum creatinine measurement or urine output measurements from ICU admission. Patients were excluded if there was incomplete renal injury information (lacking serum creatinine and urine output information) or who were admitted during the study period but not part of the parent study. The Institutional Review Board (IRB) at the University of Michigan approved this study.
Data Collection
The incidence of organ dysfunction meeting MODS criteria and AKI were measured over the first 7 days after ICU admission, per the parent study protocol [15]. Hourly urine output was measured and recorded by PICU nursing staff per unit protocol. In an effort to minimize false positive identification of oliguria, in patients without urethral catheters, the total output of mixed stool and urine was used to define urine output. Additionally, if the patient had an unmeasurable output, it was defined as a normal amount for age by default. To further protect against false positive identification of oliguria in patients without urethral catheters, total urine output was divided over the period of time between instances of output, allowing for normal sleep periods of >6 hours with no urine output. This protected false identification of stage 1 AKI by KDIGO.
Baseline creatinine was measured based on the lowest measured creatinine in the previous three months prior to hospitalization [16–18]. If the patient was hospitalized for the last three months prior to PICU admission, the lowest creatinine measurement more than one month prior to PICU admission was used. If this measurement was not available, baseline creatinine was imputed using age dependent calculations with the assumption of a glomerular filtration rate (GFR) of 120 ml/min/1.73m2, as previously published [18]. Admission creatinine was defined as the highest creatinine in the 24 hours prior to admission or the 12 hours after admission.
Definitions
MODS is defined as the concurrent dysfunction of two or more organ systems including: respiratory, cardiovascular, hematological, neurological, gastrointestinal, hepatic, and renal. We utilized two unique MODS diagnostic criteria put forth by Proulx and Goldstein [8, 9].We defined AKI using the kidney injury criteria from Proulx, Goldstein, and KDIGO definitions at PICU admission as well as the development at any point over 7 days from PICU admission [10]. Figure 1 compares the definition and staging definitions of kidney injury by Proulx, Goldstein, and KDIGO. KDIGO AKI stage was determined based on the worst stage achieved by serum creatinine or urine output at any point over the first 7 days after PICU admission.
Figure 1:
Proulx, Goldstein, and KDIGO Renal Dysfunction Definitions
At least one of the criteria have to be met to be included
SCr = Serum creatinine; UOP = Urine Output; RRT = Renal Replacement Therapy; BUN = Blood Urea Nitrogen
*In the absence of known renal dysfunction
^ Patients are included in this criteria automatically after initiation of RRT
In order to assess the impact of KDIGO AKI definitions in the context of MODS, the KDIGO AKI definitions replaced the existing AKI definitions within the Proulx MODS criteria (leading to: P’) and within the Goldstein MODS criteria (G’). The number of extra-renal organ dysfunctions were calculated each day based on Goldstein and Proulx criteria individually and reported as the maximum value achieved. Vasoactive-inotropic score (VIS) was calculated each day and reported as the maximum VIS [19].
Outcomes
The primary outcome was the identification of the development of major adverse kidney events at 30 days (MAKE30) in patients meeting MODS criteria by Goldstein, Proulx, G’, and P’. Major adverse kidney events (MAKE) is a composite outcome of death, new renal replacement therapy (RRT), or persistent renal dysfunction [14]. Persistent renal dysfunction was defined as a creatinine at least 150% of baseline or an absolute increase of at least 0.3 mg/dL at 30 days after ICU admission, otherwise meeting KDIGO Stage 1 criteria. Thirty days was chosen as an endpoint because it has been established as a realistic kidney injury timepoint in clinical trials [13, 20].
Secondary evaluations included 1) incidence of AKI by Goldstein, Proulx, and KDIGO kidney injury criteria, 2) the overall number of patients meeting MODS criteria within the first seven days of PICU admission using Goldstein, Proulx, G’, and P’ definitions, and 3) detection rate of MAKE30 of each MODS criteria in a population of patients with measured baseline serum creatinine.
Statistical Analysis
Statistical analysis was performed using Stata 16 (StataCorp, LLC, College Station, TX, USA). Categorical data are presented as number and percentages and continuous data are presented as medians and interquartile ranges (IQR), as we anticipated non-normally distributed data. Repeated admissions were included and considered as a separate, discrete risk of developing AKI. For AKI related outcomes, medians were evaluated using Wilcoxon ranksum tests. Because patients may meet MODS or AKI criteria by multiple definitions, this creates a nested cohort of patients occurs and as such, outcomes cannot be directly compared. Therefore, we calculated Youden’s Index to signify the collective performance of each criteria. Youden’s Index is a measure of the combined sensitivity and specificity of a measure, with a higher value indicating increased precision [21]. Any relationships that were compared were done comparing the additional patients identified with new definitions vs. the standard definitions. Specifically, in order to address the concern that those patients meeting the new MODS criteria may have a lower severity of illness, the relative risk of developing MAKE30 was evaluated. Risk ratios were calculated and significance was established via the chi-square test. Significance was set at p < 0.05. Predictive characteristics of Proulx, Goldstein, P’, and G’ MODS criteria were measured for MAKE30.
Results:
Over the study period, 479 admissions occurred among 433 patients. Complete admission kidney injury data was available in 410 admissions and were included in our study. The cohort was 50% male (n=205), with a median age of 5 years and PICU length of stay of 4 days (Table 1). 146 (36%) patients had measured baseline serum creatinine prior to admission. Of the 410 patients, 65 (16%) met the primary endpoint of MAKE30, with 17 (4%) deaths, 16 (4%) patients with new RRT, and 46 (11%) patients with persistent renal dysfunction. (Table 1, Figure 2) Of the 146 patients who had a measured baseline serum creatinine, 40 (27%) met the primary endpoint of MAKE 30, with 8 (5%) deaths, 5 (3%) patients with new RRT, and 35 (24%) patients with persistent renal dysfunction.
Table 1:
Acute Kidney Injury Population Characteristics
| Characteristics | All Patients (n=410) | No AKI^ (n=219) | AKI | ||
|---|---|---|---|---|---|
| Proulx (n=34) | Goldstein (n=77) | KDIGO (n=191) | |||
| Female | 205 (50.0%) | 102 (46.6%) | 23 (67.6%) | 46 (59.7%) | 103 (53.9%) |
| Age (y) | 5 (1.3–13) | 4 (1–12) | 14 (4–17) | 8.5 (1.67–15) | 6 (1.6–14) |
| Severity of Illness | |||||
| PRISM-III | 3 (0–7) | 2 (0–4) | 11 (7–18)* | 8 (3.5–13.5)* | 5 (0–11)* |
| VIS | 0 (0–0) | 0 (0–0) | 4.5 (0–18) | 0 (0–5) | 0 (0–0) |
| Mechanical Ventilation (n) | 172 (42.0%) | 80 (36.5%) | 22 (64.7%) | 39 (50.6%) | 92 (48.2%) |
| Extrarenal Organ Dysfunction | |||||
| Proulx | 1 (0–1) | 1 (0–1) | 2 (0–4) | -- | 1 (0–2) |
| Goldstein | 1 (0–2) | 1 (0–2) | -- | 2 (1–3) | 2 (0–3) |
| Outcome | |||||
| MAKE30 (total) | 65 (15.9%) | 5 (2.3%) | 21 (61.8%) | 34 (44.2%) | 60 (31.4%) |
| Mortality | 17 (4.1%) | 3 (1.4%) | 5 (14.7%) | 10 (13.0%) | 14 (7.3%) |
| New RRT | 16 (3.9%) | 0 (0%) | 16 (47.1%) | 14 (18.2%) | 16 (8.4%) |
| Persistent Renal Dysfunction | 46 (11.2%) | 2 (0.9%) | 8 (23.5%) | 21 (27.3%) | 44 (23.0%) |
| ICU LOS (d) | 4 (2–8) | 3 (2–6) | 12 (3–19)* | 6 (2–13)* | 5 (2–11)* |
| Hospital LOS (d) | 8 (4–16) | 6 (3–11) | 17 (9–44)* | 15 (6–34)* | 11 (6–23)* |
Data are presented as median (IQR) for continuous measures, and n (%) for categorical measures.
PRISM-III = Pediatric Risk of Mortality III Score; VIS = Vasoactive-Inotropic Score (dopamine + dobutamine + (norepinephrine × 100) + (epinephrine × 100) + (milrinone × 10) + (vasopressin × 10); RRT = Renal Replacement Therapy; LOS = Length of Stay; (y) = year; (d) = day
No AKI by any criteria
p<0.001 by Wilcoxon rank-sum, compared to “No AKI” group
Figure 2:
MAKE30 Among Different Organ Dysfunction Definitions
Acute Kidney Injury Incidence
Kidney injury occurred in 77 (19%) and 34 (8%) patients by Goldstein and Proulx AKI definitions within the first 7 days of PICU admission, respectively. Using KDIGO definitions, 191 (47%) patients experienced AKI; 70 (37%) experiencing KDIGO Stage 1, 54 (28%) Stage 2, and 67 (35%) Stage 3. (Table 2) All patients who met kidney injury criteria by Goldstein or Proulx also met kidney injury criteria by KDIGO. In total, 32 patients met kidney injury criteria by all three definitions. Using Proulx or Goldstein kidney injury criteria, AKI was not identified in 58/70 (83%) of Stage 1, 37/54 (69%) of Stage 2, and 18/67 (27%) of Stage 3. Urine output alone was used to define AKI in 32 patients (17%).
Table 2:
Distribution of Renal Dysfunction Among AKI Stages
| KDIGO 0 | KDIGO 1 | KDIGO 2 | KDIGO 3 | |
|---|---|---|---|---|
| Proulx (n=34) | 0 (0%) | 2 (5.9%) | 1 (2.9%) | 31 (91.2%) |
| Goldstein (n=77) | 0 (0%) | 12 (15.6%) | 16 (20.8%) | 49 (63.6%) |
| All Patients (n=410) | 219 (53.4%) | 70 (17.1%) | 54 (13.2%) | 67 (16.3%) |
Data presented as number of patients who meet both criteria and percentage of total row (n(%))
MAKE30 Identification
Among the 65 patients with MAKE30, AKI within the first 7 days of ICU admission was detected in 21 (32%) by Proulx criteria, 34 (52%) by Goldstein criteria, and 60 (92%) by KDIGO criteria. Patients who never had AKI by any criteria had lower PRISM-III scores as well as shorter median ICU and hospital lengths of stays compared to those who did meet AKI criteria (p<0.001). (Table 1)
Among those who met MODS by Proulx definition, 29 of 95 (31%) had the MAKE30 outcome compared to 43 of 132 (33%) patients who met criteria by P’. Of those who met MODS criteria by Goldstein, 46 of 194 (24%) had the MAKE30 outcome compared to 53 of 224 (24%) of those who met MODS criteria by G’. (Figure 2). The 37 additional patients identified using the P’ criteria had a relative risk of 1.13 (95% CI 0.68–1.88, p = 0.64) of developing the MAKE30 outcome compared to standard Proulx criteria. The 30 additional patients identified by G’ criteria over the Goldstein criteria had a relative risk of 0.95 (95% CI 0.47 – 1.92, p = 0.25) of developing the MAKE30 outcome. While the relative risk of identifying MAKE30 was similar, a higher proportion of those patients who developed MAKE30 were detected by P’ (43/65) and G’ (53/65) compared to Proulx (29/65) and Goldstein (46/65) definitions.
Among patients with MODS, G’ and P’ criteria increased the sensitivity for the prediction of MAKE30 while increasing negative predictive values for MAKE30. (Table 4) The positive predictive value was improved with P’ criteria compared to Proulx alone, while it was unchanged with G’ criteria. Sensitivity for the prediction of MAKE30 increased from 45% by Proulx to 66% by P’, and 71% by Goldstein to 82% by G’. As a measure of overall performance of the prediction of MAKE30, Youden’s Index was calculated. G’ and P’ criteria increased Youden’s Index compared to traditional MODS criteria alone, signifying improved overall performance for the prediction of MAKE30. (Table 4)
Table 4:
Predictive Characteristics for MAKE30
| Sensitivity | Specificity | PPV | NPV | Yl | |
|---|---|---|---|---|---|
| AKI Criteria | |||||
| Proulx | 32.3% (21.2–45.1) | 96.2% (93.6–98.0) | 31.8% (43.6–77.8) | 88.3% (84.6–91.4) | 28.5 |
| Goldstein | 52.3% (39.5–24.9) | 87.5% (83.6–90.8) | 44.2% (32.8–55.9) | 90.7% (87.0–93.6) | 39.8 |
| KDIGO | 92.3% (83.0–97.5) | 62.0% (56.7–67.2) | 31.4% (24.9–38.5) | 97.7% (94.8–99.3) | 55.3 |
| MODS Criteria | |||||
| Proulx | 44.6% (32.3–57.5) | 80.9% (76.3–84.9) | 30.5% (21.5–40.8) | 88.6% (84.5–91.9) | 25.5 |
| P’ | 66.2% (53.4–77.4) | 74.2% (69.2–78.7) | 32.6% (24.7–41.3) | 92.1% (88.3–95.0) | 40.4 |
| Goldstein | 70.8% (58.2–81.4) | 57.1% (51.7–62.4) | 23.7% (17.9–30.3) | 91.2% (86.6–84.6) | 27.9 |
| G’ | 81.5% (70.0–90.1) | 50.4% (45.0–55.8) | 23.7% (18.3–29.8) | 93.5% (89.0–96.6) | 31.9 |
Data are presented as % (95% CI); PPV = Positive predictive value; NPV = negative predictive value; YI = Youden’s Index P’ = ≥2 organ dysfunctions with Proulx extra-renal criteria and KDIGO AKI criteria; G’ = ≥2 organ dysfunctions by Goldstien extra-renal criteria and KDIGO AKI criteria
Compared to patients with MODS based on traditional definitions, those patients who had MODS according to G’ and P’ criteria had similar rates of mortality and renal replacement therapy, but an increased proportion experiencing persistent renal dysfunction. (Figure 2) 46 patients experienced persistent renal dysfunction after ICU stay in our cohort. Of these, 26% (12/46) had MODS by Proulx criteria, 57% (26/46) by P’, 61% (28/46) by Goldstein, and 76% (35/46) by G’.
MODS Incidence
Among the 410 patients, MODS by Proulx criteria occurred in 95 (23%) patients and 194 (47%) patients by Goldstein in the first seven days after ICU admission. MODS was identified in 132 (32%) patients by P’ criteria and in 224 (54%) patients by G’ criteria. (Figure 3) This was an absolute increase of 37 patients (39%) over the Proulx criteria and 30 patients (15%) over the Goldstein criteria. The median number of extra-renal dysfunctional organs was not different among the four definitions, remaining at 2 in Proulx, Goldstein, P’, and G’. (Table 3)
Figure 3:
Organ Dysfunction Class Switching By Inclusion of KDIGO
Yes = meets criteria; No = does not meet critera; Proulx = Proulx MODS criteria; Goldstein = Goldstein MODS criteria; KDIGO = KDIGO AKI criteria; P’ = ≥2 organ dysfunctions with Proulx extra-renal criteria and KDIGO AKI criteria; G’ = ≥2 organ dysfunctions by Goldstien extra-renal criteria and KDIGO AKI criteria
Table 3:
MODS Population Characteristics
| Characteristics | All Patients (n=410) | ≥2 Organ Dysfunctions | |||
|---|---|---|---|---|---|
| Proulx (n=95) | P’ (n=132) | Goldstein (n=194) | G’ (n=224) | ||
| Female | 205 (50.0%) | 46 (48.4%) | 68 (51.5%) | 97 (50.0%) | 113 (50.4%) |
| Age (y) | 5 (1.3–13) | 6 (1–14) | 4 (.96–13) | 3.69 (.92–12) | 4.54 (1–12.15) |
| Severity of Illness | |||||
| PRISM-III | 3 (0–7) | 8 (3–14) | 6 (2–12) | 5 (2–11) | 4 (1–10) |
| VIS | 0 (0–0) | 1 (0–9) | 0 (0–5.5) | 0 (0–2) | 0 (0–0) |
| Mechanical Ventilation | 172 (42.0%) | 85 (89.5%) | 106 (80.3%) | 149 (76.8%) | 155 (69.2%) |
| Extrarenal Organ Dysfunction | |||||
| Proulx | 1 (0–1) | 2 (2–3) | 2 (1–3) | -- | -- |
| Goldstein | 1 (0–2) | -- | -- | 2 (2–3) | 2 (2–3) |
| Outcome | |||||
| MAKE30 (total) | 65 (15.9%) | 29 (30.5%) | 43 (32.6%) | 46 (23.7%) | 53 (23.7%) |
| Mortality | 17 (4.1%) | 13 (13.7%) | 14 (10.6%) | 17 (8.7%) | 17 (7.6%) |
| New RRT | 16 (3.9%) | 14 (14.7%) | 14 (10.6%) | 14 (7.2%) | 14 (6.3%) |
| Persistent Renal Dysfunction | 46 (11.2%) | 12 (12.6%) | 26 (19.7%) | 28 (14.4%) | 35 (15.6%) |
| ICU LOS (d) | 4 (2–8) | 10 (5–16) | 8 (4–14.5) | 6 (3–13) | 6 (3–12) |
| Hospital LOS (d) | 8 (4–16) | 18 (9–34) | 16 (8–35) | 13 (6–24.5) | 12 (6–23) |
Data are presented as median (IQR) for continuous measures, and n (%) for categorical measures.
PRISM-III = Pediatric Risk of Mortality III Score; VIS = Vasoactive-Inotropic Score (dopamine + dobutamine + (norepinephrine × 100) + (epinephrine × 100) + (milrinone × 10) + (vasopressin × 10); RRT = Renal Replacement Therapy; LOS = Length of Stay; (y) = year; (d) = day; P’ = ≥2 organ dysfunctions with Proulx extra-renal criteria and KDIGO AKI criteria; G’ = ≥2 organ dysfunctions by Goldstien extra-renal criteria and KDIGO AKI criteria
Discussion:
In this analysis of a population of children with critical illness, we demonstrate that major adverse kidney events are common among children with MODS, occurring in nearly 1 in 6. Additionally, the integration of updated AKI criteria improves the detection of children with persistent renal dysfunction after critical illness. This increase suggests that previous definitions underrepresent organ injury and may mis-estimate the patient-centered burden and long-term consequences of MODS.
The MAKE30 outcome occurred in 65 (16%) of PICU patients. While the KDIGO AKI definition alone identified over 90% of these patients, less than half were identified by Proulx kidney injury criteria in the first 7 days of PICU admission. The majority of these patients developed persistent renal dysfunction. Contrary to previous critiques that KDIGO criteria may identify “less ill” patients and suggesting this kidney injury is less important, we found that the relative risk of developing MAKE30 was similar before and after the inclusion of KDIGO criteria, suggesting that the newly identified cohort of patients carries similar risk of poor outcomes. In other words, the P’ and G’ criteria enhance the proportion of children found to have MODS and who developed major adverse events at 30 days.
Nearly 10% of our cohort had persistent renal dysfunction at 30 days after ICU admission, consistent with previous findings that 7.5% of patients in a PICU population were discharged without renal recovery [22]. Both G’ and P’ MODS criteria increased identification of those children with MAKE30, particularly among those with persistent renal dysfunction. Enhanced AKI and persistent kidney injury detection in critical illness is important, as it provides an opportunity for collaboration between pediatricians, nephrologists, and intensivists to provide a structured follow up for care in children at risk for negative long-term health consequences. Increasing evidence suggests an episode of acute kidney injury in critically ill children is associated with development of chronic kidney disease, future hypertension, and early cardiovascular events [23–27]. Persistent renal dysfunction is a risk factor for chronic kidney disease and is associated with increased mortality by 3 years after discharge [28]. Thus, improving the early and accurate recognition of patients with multiple pathways to clinically important outcomes is important as studies in MODS increasingly include predictive enrichment strategies to identify patients at risk for these same outcomes.
While the importance of urine output inclusion among diagnostic criteria for AKI has recently been demonstrated, it is not included in either MODS definition [8, 9, 11, 12]. Nearly one in five children with AKI do not have an increase in serum creatinine [12]. Children who had AKI based on urine output criteria alone had increased renal replacement therapy use and similar mortality rates to those who met creatinine-based criteria [12]. Those who met both creatinine and urine output criteria had the worst outcomes, including highest mortality, most renal replacement therapy, and the longest lengths of stay. Adults who met urine output criteria had consistently increased mortality compared to those who met a similar AKI stage based on creatinine alone [29]. Urine output alone diagnosed AKI in 17% of our patients. As mortality has been shown to increase with each additional dysfunctional organ system, inclusion of urine output represents an important area to refine current MODS definitions to capture the multifactorial etiologies and complex physiology of kidney injury [30].
Using KDIGO criteria we identified 191 children with AKI, over two and a half times more than would be detected by the Proulx or Goldstein kidney injury definitions. In both children and adults, AKI is independently associated with increased length of stay, mechanical ventilation, and mortality [8, 11, 12, 16, 17, 22, 31–39]. Similarly, we found children with AKI to have increased length of ICU and hospital stay. (Table 1) Our data suggest that the current MODS definitions skew toward the most severe kidney injury, missing over 80% episodes of mild kidney injury. Severe AKI (KIDGO stage 2/3) has been shown to double the odds of death or new moderate disability at hospital discharge compared to those with no or mild AKI [37]. This particularly vulnerable group was not well captured by traditional criteria. Indeed, nearly 75% of KDIGO stage 2 or 3 patients were not identified by Proulx. Thus, traditional MODS definitions underrepresent those at increased risk of death and disability.
We identified nearly 40% more instances of MODS when kidney injury definitions were based on KDIGO criteria. MODS has long been associated with poor functional outcomes and mortality among children [3, 6]. These adverse outcomes increase with each new organ system involved [2, 6, 8, 9, 40]. In a prior study of children with acute respiratory failure, only 5% had kidney injury by Goldstein criteria [40]. This level remained constant over 28 days, while the incidence of other extrapulmonary organ dysfunctions increased. In contrast, the AWARE study used KDIGO AKI criteria and found over 25% of patients had evidence of kidney injury [11]. Among patients with MODS in our cohort, kidney injury occurred in 26 and 31% of patients by Proulx and Goldstein based AKI definitions, respectively. In contrast, when KDIGO AKI definitions were applied, kidney injury occurred in 83% of children with MODS by Proulx and 63% of children with MODS by Goldstein criteria. These missed patients are underrepresented in studies of children with MODS, impacting conclusions drawn with MODS-based endpoint investigations.
There are several limitations in this study. First, we performed a single-center retrospective study, and therefore, our data may not be generalizable to other populations. However, we evaluated MODS and AKI in a population of more than 400 children over a 1.5 year period. Second, we were unable to confirm outside measurements of creatinine for assessment of baseline creatinine in every case, potentially under-representing cases of existing chronic kidney disease (CKD) and over-representing AKI in our cohort. Third, while we included patients who did not have a urinary catheter, we included all combined output as urine output by hour in these cases, potentially under-representing oliguria and under-classifying existing AKI. However, we were able to assess hour-by-hour urine output, which is often absent from retrospective AKI studies. Fourth, some patients met definitional agreement among all criteria, meaning the patient populations were inherently nested within one another, which did not allow for between group comparisons. Fifth, the definitions that are used in this study are not scientifically validated and a definition inherently will dictate the epidemiology of the disease. However, the KDIGO criteria are based on studies describing outcome stratification with changes in creatinine and urine output. Finally, given a composite outcome, there is a possibility that a single patient would meet each of the criteria for the composite. For this reason, p-values were not meaningful comparisons between population groups and conclusions drawn are limited to the relationship between a definition and the outcome, instead of primarily between definitions.
In this cohort of critically ill pediatric patients, inclusion of KDIGO kidney injury criteria into previously defined MODS criteria identified more children who experienced adverse kidney events at 30 days, particularly identifying those children with persistent renal dysfunction. This study suggests that increased granularity of definitions may be important to population stratification and subgroup identification for outcome evaluations and personalized therapeutic targets in future studies.
Footnotes
Conflicts of Interest and Source of Funding: None.
References
- 1.Farris RWD, Weiss NS, Zimmerman JJ (2013) Functional Outcomes in Pediatric Severe Sepsis. Pediatr Crit Care Med 14:835–842. 10.1097/pcc.0b013e3182a551c8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Lin JC, Spinella PC, Fitzgerald JC, et al. (2017) New or Progressive Multiple Organ Dysfunction Syndrome in Pediatric Severe Sepsis. Pediatr Crit Care Med 18:8–16. 10.1097/pcc.0000000000000978 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Typpo KV, Petersen NJ, Hallman DM, et al. (2009) Day 1 multiple organ dysfunction syndrome is associated with poor functional outcome and mortality in the pediatric intensive care unit. Pediatr Crit Care Med 10:562–570. 10.1097/pcc.0b013e3181a64be1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Leclerc F, Leteurtre S, Duhamel A, et al. (2005) Cumulative Influence of Organ Dysfunctions and Septic State on Mortality of Critically Ill Children. Am J Resp Crit Care 171:348–353. 10.1164/rccm.200405-630oc [DOI] [PubMed] [Google Scholar]
- 5.Leteurtre S, Martinot A, Duhamel A, et al. (2003) Validation of the paediatric logistic organ dysfunction (PELOD) score: prospective, observational, multicentre study. Lancet 362:192–197. 10.1016/s0140-6736(03)13908-6 [DOI] [PubMed] [Google Scholar]
- 6.Graciano AL, Balko JA, Rahn DS, et al. (2005) The Pediatric Multiple Organ Dysfunction Score (P-MODS): Development and validation of an objective scale to measure the severity of multiple organ dysfunction in critically ill children. Crit Care Med 33:1484–1491. 10.1097/01.ccm.0000170943.23633.47 [DOI] [PubMed] [Google Scholar]
- 7.Watson RS, Crow SS, Hartman ME, et al. (2017) Epidemiology and Outcomes of Pediatric Multiple Organ Dysfunction Syndrome. Pediatr Crit Care Med 18:S4–S16. 10.1097/pcc.0000000000001047 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Proulx F, Fayon M, Farrell CA, et al. (1996) Epidemiology of Sepsis and Multiple Organ Dysfunction Syndrome in Children. Chest 109:1033–1037. 10.1378/chest.109.4.1033 [DOI] [PubMed] [Google Scholar]
- 9.Goldstein B, Giroir B, Randolph A, Sepsis ICC on P (2005) International pediatric sepsis consensus conference: Definitions for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med 6:2–8. 10.1097/01.pcc.0000149131.72248.e6 [DOI] [PubMed] [Google Scholar]
- 10.Kidney Disease: Improving Global Outcomes (KDIGO) Acute Kidney Injury Work Group. (2012) KDIGO clinical practice guideline for acute kidney injury. Kidney Int Suppl 2:1–138. 10.1038/kisup.2012.2 [DOI] [Google Scholar]
- 11.Kaddourah A, Basu RK, Bagshaw SM, Goldstein SL (2017) Epidemiology of Acute Kidney Injury in Critically Ill Children and Young Adults. New Engl J Medicine 376:11–20. 10.1056/nejmoa1611391 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kaddourah A, Basu RK, Goldstein SL, et al. (2019) Oliguria and Acute Kidney Injury in Critically Ill Children. Pediatr Crit Care Med 20:332–339. 10.1097/pcc.0000000000001866 [DOI] [PubMed] [Google Scholar]
- 13.Weiss SL, Balamuth F, Thurm CW, et al. (2019) Major Adverse Kidney Events in Pediatric Sepsis. Clin J Am Soc Nephro 14:CJN.12201018. 10.2215/cjn.12201018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Billings FT, Shaw AD (2014) Clinical Trial Endpoints in Acute Kidney Injury. Nephron Clin Pract 127:89–93. 10.1159/000363725 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Carlton EF, Close J, Paice K, et al. (2020) Clinician Accuracy in Identifying and Predicting Organ Dysfunction in Critically Ill Children. Crit Care Med Publish Ahead of Print: 10.1097/ccm.0000000000004555 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Alkandari O, Eddington KA, Hyder A, et al. (2011) Acute kidney injury is an independent risk factor for pediatric intensive care unit mortality, longer length of stay and prolonged mechanical ventilation in critically ill children: a two-center retrospective cohort study. Crit Care 15:R146. 10.1186/cc10269 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Selewski DT, Cornell TT, Heung M, et al. (2014) Validation of the KDIGO acute kidney injury criteria in a pediatric critical care population. Intens Care Med 40:1481–1488. 10.1007/s00134-014-3391-8 [DOI] [PubMed] [Google Scholar]
- 18.Hessey E, Ali R, Dorais M, et al. (2017) Evaluation of height-dependent and heightindependent methods of estimating baseline serum creatinine in critically ill children. Pediatr Nephrol 32:1953–1962. 10.1007/s00467-017-3670-z [DOI] [PubMed] [Google Scholar]
- 19.Gaies MG, Gurney JG, Yen AH, et al. (2010) Vasoactive–inotropic score as a predictor of morbidity and mortality in infants after cardiopulmonary bypass* Pediatr Crit Care Med 11:234–238. 10.1097/pcc.0b013e3181b806fc [DOI] [PubMed] [Google Scholar]
- 20.Semler MW, Self WH, Wanderer JP, et al. (2018) Balanced Crystalloids versus Saline in Critically Ill Adults. New Engl J Medicine. 10.1056/nejmoa1711584 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Le CT (2006) A solution for the most basic optimization problem associated with an ROC curve. Stat Methods Med Res 15:571–584. 10.1177/0962280206070637 [DOI] [PubMed] [Google Scholar]
- 22.Hessey E, Ali R, Dorais M, et al. (2017) Renal Function Follow-Up and Renal Recovery After Acute Kidney Injury in Critically Ill Children* Pediatr Crit Care Med 18:733–740. 10.1097/pcc.0000000000001166 [DOI] [PubMed] [Google Scholar]
- 23.Chawla LS, Amdur RL, Shaw AD, et al. (2014) Association between AKI and Long-Term Renal and Cardiovascular Outcomes in United States Veterans. Clin J Am Soc Nephro 9:448–456. 10.2215/cjn.02440213 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Greenberg JH, Zappitelli M, Devarajan P, et al. (2016) Kidney Outcomes 5 Years After Pediatric Cardiac Surgery: The TRIBE-AKI Study. Jama Pediatr 170:1071. 10.1001/jamapediatrics.2016.1532 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Mammen C, Abbas AA, Skippen P, et al. (2012) Long-term Risk of CKD in Children Surviving Episodes of Acute Kidney Injury in the Intensive Care Unit: A Prospective Cohort Study. Am J Kidney Dis 59:523–530. 10.1053/j.ajkd.2011.10.048 [DOI] [PubMed] [Google Scholar]
- 26.Askenazi DJ, Feig DI, Graham NM, et al. (2006) 3–5 year longitudinal follow-up of pediatric patients after acute renal failure. Kidney Int 69:184–189. 10.1038/sj.ki.5000032 [DOI] [PubMed] [Google Scholar]
- 27.Chawla LS, Amdur RL, Faselis C, et al. (2017) Impact of Acute Kidney Injury in Patients Hospitalized With Pneumonia. Crit Care Med 45:600–606. 10.1097/ccm.0000000000002245 [DOI] [PubMed] [Google Scholar]
- 28.Fiorentino M, Tohme FA, Wang S, et al. (2018) Long-term survival in patients with septic acute kidney injury is strongly influenced by renal recovery. Plos One 13:e0198269. 10.1371/journal.pone.0198269 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Izawa J, Uchino S, Takinami M (2016) A detailed evaluation of the new acute kidney injury criteria by KDIGO in critically ill patients. J Anesth 30:215–222. 10.1007/s00540-015-2109-6 [DOI] [PubMed] [Google Scholar]
- 30.Villeneuve A, Joyal J-S, Proulx F, et al. (2016) Multiple organ dysfunction syndrome in critically ill children: clinical value of two lists of diagnostic criteria. Ann Intensive Care 6:40. 10.1186/s13613-016-0144-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Chawla LS, Bellomo R, Bihorac A, et al. (2017) Acute kidney disease and renal recovery: consensus report of the Acute Disease Quality Initiative (ADQI) 16 Workgroup. Nat Rev Nephrol 13:241–257. 10.1038/nrneph.2017.2 [DOI] [PubMed] [Google Scholar]
- 32.Akcan-Arikan A, Zappitelli M, Loftis LL, et al. (2007) Modified RIFLE criteria in critically ill children with acute kidney injury. Kidney Int 71:1028–1035. 10.1038/sj.ki.5002231 [DOI] [PubMed] [Google Scholar]
- 33.Sutherland SM, Ji J, Sheikhi FH, et al. (2013) AKI in Hospitalized Children: Epidemiology and Clinical Associations in a National Cohort. Clin J Am Soc Nephro 8:1661–1669. 10.2215/cjn.00270113 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Sutherland SM, Byrnes JJ, Kothari M, et al. (2015) AKI in Hospitalized Children: Comparing the pRIFLE, AKIN, and KDIGO Definitions. Clin J Am Soc Nephro 10:554–561. 10.2215/cjn.01900214 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Fleming GM, Sahay R, Zappitelli M, et al. (2016) The Incidence of Acute Kidney Injury and Its Effect on Neonatal and Pediatric Extracorporeal Membrane Oxygenation Outcomes. Pediatr Crit Care Med 17:1157–1169. 10.1097/pcc.0000000000000970 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Jetton JG, Boohaker LJ, Sethi SK, et al. (2017) Incidence and outcomes of neonatal acute kidney injury (AWAKEN): a multicentre, multinational, observational cohort study. Lancet Child Adolesc Heal 1:184–194. 10.1016/s2352-4642(17)30069-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Fitzgerald JC, Basu RK, Akcan-Arikan A, et al. (2016) Acute Kidney Injury in Pediatric Severe Sepsis. Crit Care Med 44:2241–2250. 10.1097/ccm.0000000000002007 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Fitzgerald JC, Ross ME, Thomas NJ, et al. (2018) Risk factors and inpatient outcomes associated with acute kidney injury at pediatric severe sepsis presentation. Pediatr Nephrol 33:1781–1790. 10.1007/s00467-018-3981-8 [DOI] [PubMed] [Google Scholar]
- 39.Hessey E, Morissette G, Lacroix J, et al. (2018) Long-term Mortality After Acute Kidney Injury in the Pediatric ICU. Hosp Pediatrics 8:260–268. 10.1542/hpeds.2017-0215 [DOI] [PubMed] [Google Scholar]
- 40.Weiss SL, Asaro LA, Flori HR, et al. (2017) Multiple Organ Dysfunction in Children Mechanically Ventilated for Acute Respiratory Failure. Pediatr Crit Care Med 18:319–329. 10.1097/pcc.0000000000001091 [DOI] [PMC free article] [PubMed] [Google Scholar]