+Model

ARTICLE IN PRESS
An Pediatr (Barc). 2019;xxx(xx):xxx---xxx

www.analesdepediatria.org

ORIGINAL ARTICLE

Paediatric sequential organ failure assessment (pSOFA)

score: a new mortality prediction score in the
paediatric intensive care unit夽
Ghada Mohamed El-Mashad a , Muhammad Said El-Mekkawy a,∗ , Mohamed Helmy Zayan b

a
Departamento de Pediatría, Facultad de Medicina, Universidad de Menufia, Menufia, Egypt
b
Hospital Atfal Misr, El Cairo, Egypt

Received 18 January 2019; accepted 11 May 2019

KEYWORDS Abstract
Sequential organ Objectives: To assess performance of the age-adapted SOFA score in children admitted into
failure assessment; Paediatric Intensive Care Units (PICUs) and whether the SOFA score can compete with the
Systemic systemic inflammatory response syndrome (SIRS) in diagnosing sepsis, as recommended in the
inflammatory Sepsis-3 consensus definitions.
response syndrome; Methods: Two-centre prospective observational study in 281 children admitted to the PICU.
Prognosis; We calculated the SOFA, Pediatric Risk of Mortality (PRISM), and Pediatric Index of Mortality-2
Paediatric; (PIM2) scores and assessed for the presence of SIRS at admission. The primary outcome was
Sepsis 30-day mortality.
Results: The SOFA score was higher in nonsurvivors (P < .001) and mortality increased pro-
gressively across patient subgroups from lower to higher SOFA scores. The receiver operating
characteristic (ROC) curve analysis revealed that the area under the curve (AUC) of the SOFA
score for predicting 30-day mortality was 0.89, compared to AUCs of 0.84 and 0.79 for the
PRISM and PIM2 scores, respectively. The AUC of the SOFA score for predicting a prolonged stay
in the PICU was 0.67. The SOFA score was correlated to the PRISM score (rs = 0.59) and the
PIM2 score (rs = 0.51). In children with infection, the AUC of the SOFA score for predicting
mortality was 0.87 compared to an AUC of 0.60 using SIRS. The diagnosis of sepsis applying a
SOFA cutoff of 3 points predicted mortality better than both the SIRS and the SOFA cutoff of 2
points recommended by the Sepsis-3 consensus.

夽 Please cite this article as: Mohamed El-Mashad G, Said El-Mekkawy M, and Helmy Zayan M. La escala pediátrica de evaluación del fallo

multiorgánico secuencial (pSOFA): una nueva escala de predicción de la mortalidad en la unidad de cuidados intensivos pediátricos. An
Pediatr (Barc). 2019. https://doi.org/10.1016/j.anpedi.2019.05.018
∗ Corresponding author.

E-mail address: [email protected] (M.S. El-Mekkawy).

2341-2879/© 2019 Published by Elsevier España, S.L.U. on behalf of Asociación Española de Pediatrı́a. This is an open access article under
the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

ANPEDE-2732; No. of Pages 9

+Model
ARTICLE IN PRESS
2 G.M. El-Mashad et al.

Conclusions: The SOFA score at admission is useful for predicting outcomes in the general PICU
population and is more accurate than SIRS for definition of paediatric sepsis.
© 2019 Published by Elsevier España, S.L.U. on behalf of Asociación Española de Pediatrı́a.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

PALABRAS CLAVE La escala pediátrica de evaluación del fallo multiorgánico secuencial (pSOFA): una
Escala SOFA; nueva escala de predicción de la mortalidad en la unidad de cuidados intensivos
Síndrome de pediátricos
respuesta
Resumen
inflamatoria
Objetivos: Analizar el rendimiento de la escala SOFA adaptada por edad en niños ingresados
sistémica, prognosis,
en la unidad de cuidados intensivos pediátricos (UCIP) y establecer si la escala SOFA puede
sepsis pediátrica.
competir con el síndrome de respuesta inflamatoria sistémica (SRIS) para el diagnóstico de
sepsis, de acuerdo con las recomendaciones del consenso Sepsis-3.
Métodos: Estudio prospectivo observacional en 2 centros en 281 niños ingresados en la UCIP. Se
calcularon las puntuaciones de las escalas SOFA, Pediatric Risk of Mortality (PRISM) y Pediatric
Index of Mortality-2 (PIM2) y se evaluó la presencia de SRIS al ingreso. La variable primaria fue
la mortalidad a los 30 días.
Resultados: la puntuación SOFA fue más alta en los no supervivientes (p < 0,001) y la mortalidad
incrementó progresivamente de los subgrupos con las puntuaciones SOFA más bajas a aquellos
con las puntuaciones más altas. El análisis de las curvas de las características operativas del
receptor (ROC) mostró que el área bajo la curva (AUC) para la predicción de la mortalidad a
30 días con la puntuación SOFA fue de 0,89, comparado con 0,84 y 0,79 en las escalas PRISM y
PIM2, respectivamente. La AUC de la puntuación SOFA para la predicción de estancia prolongada
en la UCIP fue de 0,67. La escala SOFA se correlacionó con las escalas PRISM (coeficiente de
correlación de Spearman rs = 0,59) y PIM2 (rs = 0,51). En niños con infección, la AUC de la escala
SOFA para la predicción de la mortalidad fue de 0,87, mientras que la AUC del SRIS fue de 0,60.
El diagnóstico de sepsis definido como una puntuación SOFA de 3 o más predijo la mortalidad
mejor que el SRIS y que la escala SOFA con el punto de corte de 2 puntos recomendada en el
consenso Sepsis-3.
Conclusiones: La puntuación SOFA al ingreso es útil como predictor de la evolución en la
población general de la UCIP y es más apropiada que el SRIS para definir la sepsis pediátrica.
© 2019 Publicado por Elsevier España, S.L.U. en nombre de Asociación Española de Pediatrı́a.
Este es un artı́culo Open Access bajo la licencia CC BY-NC-ND (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

Introduction The Pediatric Risk of Mortality (PRISM) and the Pediatric

Index of Mortality (PIM), as well as their updated versions,
In critically ill children admitted to the paediatric inten- are used widely to predict mortality.6,7 Other scores use
sive care unit (PICU), the risk of morbidity and mortality is organ dysfunction as a surrogate for mortality. Multiple
substantial in both developed and developing countries.1,2 organ dysfunction syndrome (MODS) is associated with a sig-
Several prognostic scores have been developed with the nificantly higher risk of mortality,8 making it an excellent
aim of objectively quantifying the severity of disease and candidate marker of disease severity, with one study report-
predicting the risk of death at the time of PICU admission, ing a mortality rate of 1 % with dysfunction of 1 organ system
which can be very useful in treatment planning. Further- with a progressive increase to up to 75 % in the presence of
more, such scores are essential to assess the quality of dysfunction of 4 organ systems.9
medical care, as much of the variation in mortality rates The pediatric organ dysfunction scores include the
between PICUs is influenced by factors other than medical Pediatric logistic organ dysfunction score (PELOD)10 ,
management, including the primary diagnosis and disease PELOD-2, and Pediatric Multiple Organ Dysfunction Score (P-
severity at admission.3,4 MODS).11,12 The Sequential Organ Failure Assessment (SOFA)
The ideal prognostic score should be accurate, simple, score (originally known as Sepsis-Related Organ Failure
easy to use, minimally invasive and inexpensive.5 However, Assessment) is the score used most widely to quantify organ
no paediatric mortality prediction score is completely sat- dysfunction in critically ill adults13 and has been recently
isfactory at present, and therefore researchers continue to considered to be intimately associated with the diagnosis of
devote significant effort to improving the accuracy of cur- sepsis in adults (Third International Consensus Definitions for
rently available scores and developing new ones. Sepsis and Septic Shock [Sepsis-3]).14 This has triggered an
+Model
ARTICLE IN PRESS
Paediatric SOFA score 3

eager interest in researchers to adapt the SOFA to the paedi- ordered. In the absence of documentation of previous organ
atric population. Recently, a paediatric version of the SOFA dysfunction, we assigned a baseline SOFA score of zero.
score (pSOFA) was developed and validated retrospectively We followed up patients until they were discharged from
in critically ill children.15 However, there is limited data on the PICU. The primary outcome was 30-day mortality. One
the paediatric population and the pSOFA score has not yet secondary outcome was the composite variable 30-day mor-
been validated by means of prospective cohort studies or in tality or PICU length of stay of 3 or more days. Another
low-income countries. The aim of our study was to assess secondary outcome was the length of PICU stay in survivors.
the prognostic utility of the SOFA score in a heterogeneous
cohort of critically ill children in a resource-limited country,
and to determine whether the SOFA score can replace the Statistical methods
long-held concept of Systemic Inflammatory Response Syn-
drome (SIRS) as the cornerstone of sepsis diagnosis in the We have expressed categorical data as absolute frequen-
paediatric age group. cies and percentages. We summarized continuous data as
mean ± standard deviation (SD) and compared them by
means of the t test if they were normally distributed, and
otherwise summarized them as median and range and com-
Patients and methods pared them with the Mann-Whitney U test We used the chi
square test to assess the association between categorical
We conducted a prospective observational study between variables, and the Spearman correlation coefficient to assess
March and November 2018 in the PICUs of 2 tertiary care hos- the association between the SOFA score and different varia-
pitals in Egypt (Menoufia University Hospital and Atfal Misr bles. We used univariate binary logistic regression was used
Hospital). The study protocol was approved by the Commit- to analyse the association of the SOFA score and other varia-
tee for Medical Research Ethics of the Menoufia University bles with outcomes, and multivariate logistic regression was
School of Medicine, and we obtained informed consent from used to adjust for confounding variables.
the parents. We used receiver operating characteristic (ROC) curve
We enrolled critically ill patients aged 1 month to 18 analysis to assess the power of the SOFA score to discrim-
years by consecutive sampling. We excluded patients aged inate between survivors and non-survivors, calculating the
less than 1 month or more than 18 years. At the time of Youden index to determine the optimal cut-off points for
admission, the patient underwent a physical examination discrimination. The ‘‘area under the curve’’ (AUC) ranges
and a thorough history was taken. The laboratory tests per- between 0 and 1, and the closer the AUC of a variable is to
formed included a complete blood count (CBC), blood gas 1, the better the performance of that variable. We assessed
analysis, measurement of levels of C-reactive protein (CRP), the calibration of SOFA score with the Hosmer-Lemeshow
blood glucose and serum electrolytes, a coagulation profile goodness-of-fit test to determine whether the predicted
and liver and kidney function tests. Cultures of body fluids, frequencies deviated from the observed frequencies of sur-
including blood, urine, cerebrospinal fluid (CSF) or pleural vivors and non-survivors. A p-value of 0.05 or greater in the
fluid, were performed based on the judgment of the clini- goodness-of-fit test indicates that the model is well cali-
cian. Other laboratory and radiological investigations were brated.
performed where indicated based on the clinical condition All statistical tests were two-tailed and we considered
of the patient. p-values of less than 0.05 statistically significant. In this doc-
In all patients, the paediatric SOFA score15 was calculated ument, we give the exact p-value unless it was extremely
within 24 h of admission. This score assesses 6 organ systems: low, in which case we have expressed it as P < .0001. We
respiratory, hematological, hepatic, cardiovascular, neuro- performed the statistical analyses with the IBM Statisti-
logical and renal. A subscore of 0---4 points is calculated for cal Package for the Social Sciences, version 23 (SPSS Inc;
each component, (Supplemental online content 1). We used Chicago, IL, USA).
the lowest value of each subscore in the first 24 h to calcu-
late total score. We gave a subscore a value of 0 if any of the
data were missing. In addition to the SOFA, we calculated Results
the PRISM6 and PIM-216 scores.
The number of SIRS criteria met by each patient was Patient characteristics
determined on admission, and we considered SIRS was
present if a patient met 2 of the 4 criteria for SIRS, with We enrolled 281 critically ill children in the study. Table 1
1 being either the change in the white blood cell count or in shows their basic characteristics. The 30-day mortality rate
body temperature. The classic diagnosis of sepsis required was 28.1 %. The standardized mortality ratio (SMR) was
the presence of SIRS in addition to suspected or proven 3.09 (SMR = observed/expected mortality; expected mortal-
infection. Severe sepsis was defined as sepsis with cardio- ity = median PRISM).
vascular dysfunction, acute respiratory distress syndrome, In the sample under study, 150 patients had suspected
or dysfunction in 2 other organ systems.17 We also applied infection, which was confirmed by culture in only 34 %. Of
the Sepsis-3 recommendations,18 according to which a diag- the patients with infection, 96 had SIRS and were classified
nosis of sepsis requires an acute increase in the SOFA score as cases of sepsis. Of the patients with sepsis, 40 were clas-
of 2 or more points in association with proven or suspected sified as having severe sepsis. Ninety-four patients had SIRS
infection. A patient was considered to have a suspected in the absence of infection (noninfectious SIRS). Ninety-one
infection if antibiotherapy had been prescribed and cultures patients did not have SIRS (non-SIRS).
+Model
ARTICLE IN PRESS
4 G.M. El-Mashad et al.

Table 1 Demographic, clinical and laboratory data of the patients under study.
Survivors(n = 202) Nonsurvivors (n = 79) Total sample(n = 281) P
Age, months 27.5 (1.3---216) 11 (1.5---180) 11 (1---216) <.0001a
Male sex 100 (49.5 %) 48 (60.8 %) 148 (52.7 %) .089
Weight, kg 11 (2---83) 6.5 (2.5---57) 9.5 (2---83) <.0001a
Primary reason for PICU admission
Primary reason for PICU
admission
Gastrointestinal 22 (10.9 %) 3 (3.8 %) 25 (8.9 %)
Respiratory 40 (19.8 %) 24 (30.4 %) 64 (22.8 %)
Cardiac 16 (7.9 %) 15 (19 %) 31 (11 %)
Neurologic 24 (11.9 %) 9 (11.4 %) 33 (11.7 %)
Haematological/Oncological 15 (7.4 %) 4 (5.1 %) 19 (6.8 %)
Infectionb 9 (4.5 %) 10 (12.7 %) 19 (6.8 %)
Metabolic (DKA) 36 (17.8 %) 0 (0 %) 36 (12.8 %) <.0001a
Postsurgical recovery 20 (9.9 %) 4 (5.1 %) 24 (8.5 %)
Trauma 7 (3.5 %) 1 (1.3 %) 8 (2.8 %)
Otherc 13 (6.4 %) 9 (11.4 %) 22 (7.8 %)
Suspected infections 93 (46 %) 57 (72.2 %) 150 (53.4 %)
Positive culture 27 (29 %) 22 (38.6 %) 49 (32.7 %) .48
Negative culture 66 (71 %) 35 (61.4 %) 101 (67.3 %)
Category according to SIRS
Sepsis 46 (22.8 %) 50 (63.3 %) 96 (34.2 %)
Non-infectious SIRS 77 (38.1 %) 17 (21.5 %) 94 (33.5 %) <.0001a
No SIRS 79 (39.1 %) 12 (15.2 %) 91 (32.4 %)
MODS 44 (21.8 %) 64 (81 %) 108 (38.4 %) <.0001a
MV use 47 (23.3 %) 70 (88.6 %) 117 (41.6 %) <.0001a
SOFA score 4 (0---12) 10 (0---21) 5 (0---21) <.0001a
PRISM mortality risk% 6.2 (0.6---98.9) 44.1 (1.4---98.9) 9.1 (0.6---98.9) <.0001a
PIM2 mortality risk% 1.6 (0.1---99) 25 (0.1---99.5) 2.3 (0.1---99.5 %) <.0001a
WBC count, 1000/␮L 11.85 (1.9---31) 13 (0.6---115) 12.7 (0.6---115) .93
Haemoglobin, g/dL 10.7 ± 2.64 9.7 ± 2.93 10.25 ± 2.84 .13
Platelet count, 1000/␮L 300 (2---1047) 181 (11---598) 274 (2---1047) <.0001a
Data expressed as median (minimum-maximum), mean ± standard deviation, or n (%).
DKA, diabetic ketoacidosis; MODS, multiorgan dysfunction syndrome (defined as simultaneous dysfunction of ≥ 2 organ systems); MV,
invasive mechanical ventilation; PICU, paediatric intensive care unit; PIM2, Pediatric Index of Mortality score 2; PRISM, Pediatric Risk of
Mortality score; SIRS, systemic inflammatory response syndrome; SOFA, Sequential Organ Failure Assessment Score; WBC, white blood
cell; PICU, Pediatric Intensive Care Unit.
a Statistically significant.
b These include blood stream infection without a source, septic arthritis, orbital cellulitis and severe skin infections, and exclude

respiratory, gastrointestinal, or nervous system infections.

c These include poisoning, autoimmune diseases, renal diseases, drowning, corrosive ingestion, and anaphylaxis.

Admission to the PICU was most frequently due to res- only variable that remained independently associated with
piratory disorders, especially pneumonia. Most patients mortality.
admitted for postoperative recovery had undergone surgery
for correction of congenital heart defects.

SOFA score and risk of 30-day mortality

Discriminative power of the SOFA score
The SOFA score was significantly higher in non-survivors com-
pared with survivors (Table 1). When we grouped patients The ROC curve analysis (Table 3), showed that the AUC of
based on the 25th , 50th and 75th percentiles of the SOFA the SOFA score for discrimination between survivors and non-
score distribution (Fig. 1), we found that mortality increased survivors was larger compared to the AUC of the PRISM and
progressively with the SOFA score (P < .001). PIM2 and other laboratory markers.
The univariate logistic regression analysis (Table 2) In the subset of patients with suspected infection, the
revealed that the SOFA score was positively associated with SOFA score discriminated between non-survivors and sur-
mortality, while weight and age exhibited a negative associ- vivors better than the PRISM and PIM2 scores and the
ations. In the multivariate analysis, the SOFA score was the presence of SIRS (Fig. 2).
+Model
ARTICLE IN PRESS
Paediatric SOFA score 5

Table 2 Logistic regression analysis for prediction of 30-day mortality by SOFA score and other variables among the general
PICU population.
Univariate analysis Multivariate analysis

OR 95 % CI P Adjusted OR 95 % CI P
a
Age 0.990 0.985---0.995 < .0001 1.005 0.992---1.019 .43
Weight 0.957 0.933---0.981 < .0001a 0.957 0.902---1.015 .14
SOFA 1.61 1.43---1.80 < .0001a 1.59 1.42---1.79 <.0001a
CI, confidence interval; OR, odds ratio; SOFA, Sequential Organ Failure Assessment score.
a Statistically significant.

Table 3 Receiver operating characteristic curve analysis for prediction of 30-day mortality with the SOFA score and other
predictors in the general PICU population.
AUC 95 % CI P value Cutoff level Sensitivity Specificity
SOFA 0.886 0.840---0.931 <.0001a >6.5 points 80.9 % 81.8 %
PRISM 0.840 0.780---0.900 <.0001a >17.2 % 70.6 %% 82.3 %%
PIM2 0.792 0.717---0.867 <.0001a >10.55 % 60.3 % 93.4 %
Platelet count 0.646 0.568---.725 .0004a <266 500/␮L 64.7 % 57.5 %
WBC count 0.504 0.419---0.588 .93 >21 900 /␮L 19.1 %% 89 %
AUC, area under the receiver operating characteristic curve; CI, confidence interval; PIM2, Pediatric Index of Mortality score 2; PRISM,
Pediatric Risk of Mortality score; SIRS, systemic inflammatory response syndrome; SOFA, Sequential Organ Failure Assessment Score;
WBC, white blood cell.
a Statistically significant.

1.0
60 Source of the
curve
SOFA
PIM2
Percent of dead patients

50 PRISM
0.8
SIRS
Reference line
40
0.6
Sensitivity

30

20 0.4

10
0.2

0
SOFA < 3 5 > SOFA > 3 8 > SOFA > 5 SOFA > 8
(n=83) (n=74) (n=65) (n=59) 0.0
0.0 0.2 0.4 0.6 0.8 1.0
1 - Specificity
Fig. 1 Distribution of 30-day mortality among the various
SOFA subgroups: The x-axis shows the different patient sub- Fig. 2 Analysis of ROC curves for prediction of 30-day mor-
groups based on the SOFA score. Patients were divided based tality using the SOFA score and other variables in patients with
on cut-off values corresponding to the 25th, 50th and 75th per- infection: In patients with infection, the AUC and 95 % CI for
centiles in the distribution of SOFA scores. The y-axis shows the predicting mortality using the SOFA score were 0.871 (0.808 ---
percentage of patients that died in each subgroup relative to 0.933), with P < .001, compared to an AUC of 0.602 using the
the total number of deaths. SIRS (95 % CI, 0.502---0.702; P = .056), an AUC of 0.859 using the
First subgroup (SOFA ≤ 3): 4 (4.8 %) of 83 patients died (5.1 % of PRISM score (95 % CI, 0.792---0.925; P <.001) and an AUC of 0.738
the total number of deaths). using the PIM2 score (95 % CI, 0.637---0.839; P < .001). Applying
Second subgroup (5 ≥ SOFA > 3): 10 (13.5 %) of 74 patients died a cutoff of 6.5 points, the SOFA score had a sensitivity of 81.3 %
(12.6 % of the total number of deaths). and a specificity of 77.9 % for discriminating nonsurvivors from
Third subgroup (8 ≥ SOFA > 5): 20 (30.8 %) of 65 patients died survivors. Applying a cutoff of 2.5 met criteria, the SIRS has a
(25.3 % of the total number of deaths). sensitivity of 45.8 % and a specificity of 66.2 %.
Fourth subgroup (SOFA > 8): 45 (76.3 %) of 59 patients died (57
% of the total number of deaths).
Calibration of the SOFA score
The differences between the 4 subgroups were statistically sig-
nificant (P <.001)
The Hosmer-Lemeshow goodness-of-fit test yielded a p-
value of 0.65, which indicates that the SOFA score is well
+Model
ARTICLE IN PRESS
6 G.M. El-Mashad et al.

Table 4 Spearman correlation between the SOFA score and

SOFA, SIRS, and the definition of sepsis
other variables in the patients under study.
In the patients with proven or suspected infection, the AUC
SOFA of the SOFA score for predicting mortality was higher com-
pared to the AUC of SIRS (Fig. 2).
Spearman correlation P
Defining sepsis (applying the Sepsis-3 criteria) as the pres-
coefficient (rs )
ence of infection plus a SOFA score of 2 points or greater,
Weight ---0.318 <.0001a 93.3 % of patients with infection had sepsis, but the associ-
Age ---0.332 <.0001a ation of sepsis with mortality was not significant (Table 5).
PRISM 0.59 <.0001a Applying this cutoff for the SOFA score, the mortality rate
PIM2 0.51 <.0001a was 40 %, the sensitivity 100 % and the specificity 5.2 %.
PICU stay 0.214 .002a Defining sepsis as a SOFA score cutoff of 3 points com-
MV duration ---0.09 .55 bined with infection, 84.7 % of patients with infection had
WBC count 0.16 .014a sepsis, the sensitivity was 100 % and the specificity 18.2 %,
Platelet count ---0.31 <.0001a and the association between sepsis and mortality became
Haemoglobin ---0.24 .0001a statistically significant (OR = 8.02; P = .006).
MV, mechanical ventilation; PICU, Pediatric Intensive Care Unit; Defining sepsis as a SOFA score of at least 7 points (the
PIM2, Pediatric Index of Mortality 2; PRISM, Pediatric Risk of Mor- optimal cutoff for predicting mortality based on the ROC
tality; SOFA, Sequential Organ Failure Assessment Score; WBC, curve), 40 % of patients with infections had sepsis, and they
white blood cell. were at higher risk of dying (OR = 13.73; P <.001).
a Statistically significant.
Defining sepsis based on meeting the criteria for SIRS (at
least 2 criteria, with 1 being an abnormal temperature or
WBC count) in the context of infection, the odds ratio for
mortality was 4.46 (P <.001).
calibrated. The p-values for the PRISM and PIM2 scores were
0.25 and 0.47, respectively. Discussion

The development of an ideal paediatric prognostic score

remains a challenging objective. In our study, we found
Correlation between SOFA score and other that the SOFA score at admission was a good predictor of
variables mortality in the overall PICU population. The median SOFA
score was significantly higher in nonsurvivors compared to
We found that the SOFA score was positively correlated to survivors and the mortality increased in a stepwise manner
the PRISM and PIM2 scores and the white blood cell count across patient subgroups from lower to higher SOFA scores.
(WBC), and negatively correlated to age, weight, platelet In addition, the association of the SOFA score with mortal-
count and hemoglobin concentration (Table 4). ity continued to be significant in the multivariate analysis,
while the association of age and weight with mortality did
not. Moreover, the SOFA score had a good AUC of 0.886 for
prediction of 30-day mortality, a finding consistent with a
SOFA score and length of PICU stay retrospective study in 8711 critically ill children in which
the AUC of the SOFA score for predicting 28-day mortality
The SOFA score was positively correlated to the length of was nearly identical at 0.88.15
stay in the PICU stay (Table 4). When we divided survivors Based on our data, the optimal SOFA cutoff for discrim-
into subgroups according to the median length of stay in the inating nonsurvivors from survivors was 7 points, compared
PICU into those with prolonged stays (>5 days) and those to the cutoff of 8 points reported in a different paediatric
with short stays (≤ 5 days), we found that the SOFA score study15 and the cutoff of more than 8 points described in a
was significantly higher in the prolonged stay group (median, study in adults.19
5; range, 0---12) compared to the short stay group (median, 4; Furthermore, we found that the SOFA score performed
range, 0---10) (P < .001). The AUC for the SOFA score for pre- better compared to the PRISM and PIM2 scores in predicting
dicting a PICU stay in survivors of more than 5 days was 0.67 mortality in the general PICU population, whereas another
(95 % CI = 0.59---0.76; P <.001) compared to AUCs of only 0.48, paediatric study reported an equal AUC for the SOFA score
0.52, and 0.52 for the PRISM, PIM2 and SIRS, respectively. and a more recent version of the PRISM score (PRISM III).15
We also divided the whole cohort according to the compo- While the SOFA score performed well in predicting mor-
site secondary outcome of death within 30 days or a length tality, it was less successful in predicting the length of stay
of PICU stay of 3 or more days. The SOFA score was signifi- in the PICU in survivors and the combined outcome of mor-
cantly higher in the subgroup with the composite secondary tality and length of stay in the PICU of 3 or more days. These
outcome compared with the other subgroup (median and findings are consistent with those of previous studies.20,21
range, 5 [0---21] vs 3 [1---7]; P < .001). The AUC of the SOFA for Besides the superior performance observed our study,
prediction of the composite secondary outcome was 0.77 (95 the SOFA score seems to offer additional advantages over
% CI, 0.70 --- 0.84; P < .001), compared to AUCs of only 0.45, the PRISM score. The SOFA score is available for free and
0.59, and 0.66 for the PRISM, PIM2, and SIRS, respectively. does not require special software for its calculation. Fur-
+Model
ARTICLE IN PRESS
Paediatric SOFA score 7

Table 5 Association of various sepsis definitions with 30-day mortality.

OR (95 % CI) P
Sepsis defined as SOFA ≥ 2 points 6.00 (0.74---48.67) .093
Sepsis defined as SOFA ≥ 3 points 8.02 (1.80---35.67) .006a
Sepsis defined as SOFA ≥ 7 points 13.73 (6.17---30.56) <.0001a
Sepsis defined as SIRS 4.46 (2.06---9.67) .0002a
CI, confidence interval; NA, not applicable; OR, odds ratio; SOFA, Sequential Organ Failure Assessment Score; SIRS, systemic inflammatory
response syndrome.
a Statistically significant.

thermore, the SOFA score requires measurement of fewer whereas the classic SIRS-based sepsis diagnosis predicted a
parameters than the PRISM (6 vs 14). Also, the SOFA may be lower risk of mortality (OR, 4.46).
calculated daily, offering a dynamic assessment of disease Our findings, therefore, suggest that SOFA is superior
progression, while the PRISM is only calculated at admission. to SIRS in defining paediatric sepsis. In fact, the concept
The SOFA also does not require arterial blood gas measure- of SIRS has been recently challenged, as it was found to
ments, which are difficult to obtain in children, as it was suffer from excessive nonspecificity. In one study, nearly
adapted to use SPO2 instead of PaO2 values. half of adult patients admitted to the ward developed
On the other hand, one disadvantage of SOFA is that it SIRS at least once, making it impractical to use the SIRS
is not calculated immediately on PICU admission and there criteria for early identification of sepsis.23 Another study
is a turnaround time of 24 h to get the worst value for each described that 92 % of children visiting the emergency
subscore. In this regard, the PIM2 score offers an advan- department with a fever of more than 38.5 ◦ C met the
tage over the SOFA score, as the former is calculated within SIRS criteria, but most of them were discharged with-
one hour of PICU admission, allowing earlier identification of out requiring intravenous therapy or readmission.24 On
high-risk patients that need more aggressive management. the other hand, researchers realized that SIRS should not
In addition, the SOFA score is relatively complex, which has be the sole parameter for sepsis, as the pathophysiology
led researchers to develop a quick version (qSOFA), but the of sepsis involves aspects other than excessive inflamma-
latter seems to be less accurate.20,22 Unfortunately, it seems tion, such as activation of the anti-inflammatory response
that improving one aspect of a score often comes at the and changes in non-immunological pathways, including hor-
expense of other aspects. There is hope that the discov- monal, metabolic, autonomic, and coagulation pathways, all
ery of more accurate prognostic biomarkers may improve or of which have prognostic significance.14
replace the currently available scores. Even with such devel- In any case, our study found that a SOFA cutoff of 2 points,
opments, it will remain difficult to perfectly predict the proposed by the Sepsis-3 group for diagnosis of sepsis, was
disease course of each patient owing to the extreme com- not significantly associated with mortality and also had a
plexity of the underlying immune, metabolic, and endocrine very low specificity, which suggests that it is not possible to
mechanisms, in addition to the effect of the genetic makeup directly extrapolate the Sepsis-3 recommendations to the
on the individual’s response to treatment. Additionally, paediatric population, although other paediatric studies15,20
there is no guarantee that mildly ill patients will not dete- have confirmed the usefulness of this SOFA cutoff. Differ-
riorate as a result of healthcare-related infections, adverse ences in sample size, setting or study design may explain
drug reactions or other care-related problems. this discrepancy, but it is important to keep in mind that the
There has been a growing interest in the adaptation of SOFA threshold of 2 points adopted in the Sepsis-3 criteria
the SOFA score to the paediatric age group following the was selected a priori and not derived from ROC curve anal-
recent change in the definition of sepsis in adults (Sepsis-3), ysis. Furthermore, an important aspect has been criticised
by which the long-held construct of SIRS has been aban- about the Sepsis-3 definition, which is that a patient is only
doned in favour of the SOFA score.14 In our study, the considered to have sepsis when infection has reached a high
SOFA score proved to be more accurate in predicting mor- degree of severity associated with organ dysfunction, which
tality in patients with infection than SIRS (AUC, 0.87 vs has a negative impact on the outcome.25 In other words, the
0.60). The SOFA score also proved superior to the PRISM new sepsis definition has improved specificity at the cost of
and PIM2 scores. These findings are consistent with those sensitivity.26 This issue should be taken into consideration in
of a recent multicentric retrospective study where the age- establishing the SOFA cutoff to be used for diagnosis of pae-
adapted SOFA score, calculated on admission, had an AUC diatric sepsis, and suggests that a SOFA cutoff of 7 points
of 0.83 in predicting mortality in children admitted to the should not be used to diagnose paediatric sepsis, even if
PICU with proven or suspected infection, compared to an this was the optimal cutoff for predicting mortality. Instead,
AUC of 0.73 for the SIRS.20 Similarly, another retrospective based on the findings of our study, it seems more prudent to
study reported that the SOFA score was more accurate than diagnose paediatric sepsis applying a SOFA cutoff of 3 points,
SIRS in predicting mortality in children admitted to PICU for which is associated with a greater likelihood of mortality
infection.21 compared to SIRS and at the same time allows physicians
When we compared the SOFA score and SIRS specifically, to diagnose sepsis relatively early, before organ dysfunction
we found that the odds ratio for the diagnosis of sepsis apply- has progressed to higher levels of severity. We do not claim
ing a SOFA cutoff of 7 for predicting mortality was 13.73, that the SOFA score has resolved the controversy surrounding
+Model
ARTICLE IN PRESS
8 G.M. El-Mashad et al.

the definition and diagnosis of paediatric sepsis. The SOFA [4]. Straney L, Clements A, Parslow RC, et al. Paediatric index of
score should be seen as a ‘‘step forward’’ rather than ‘‘a mortality 3: an updated model for predicting mortality in pedi-
final solution’’ to the problem. atric intensivecare. Pediatr Crit Care Med. 2013;14(7):673---81.
Overall, our findings suggest that the SOFA score should [5]. Pollack MM, Patel KM, Ruttimann UE. PRISM III: an update pedi-
replace SIRS for the diagnosis of sepsis, although the applied atric risk of mortality score. Crit Care Med. 1996;24:743---52.
[6]. Pollack MM, Ruttimann UE, Getson PR. Pediatric risk of mortality
cutoff in the SOFA should be slightly higher than the one
(PRISM) score. Crit Care Med. 1988;16:1110---6.
proposed in the Sepsis-3 criteria for adults. [7]. Shann F, Pearson G, Slater A, et al. Paediatric index of mortal-
The limitations of our study include its small sample size, ity (PIM): a mortality prediction model for children in intensive
although this is counterbalanced by the very small p-values care. Intensive Care Med. 1997;23:201---7.
in our main results. We also used older versions of the PRISM [8]. Typpo KV, Petersen NJ, Hallman DM, et al. Day one MODS
and PIM2 scores. Furthermore, we did not calculate the SOFA is associated with poor functional outcome and mortality
score at different time points, when previous studies have in the pediatric intensive care unit. Pediatr Crit Care Med.
reported that serial SOFA scores have predictive value.19 2009;10(5):562---70.
It makes sense for serial SOFA scores to predict mortality [9]. Wilkinson JD, Pollack MM, Ruttimann UE, et al. Outcome of pedi-
atric patients with multiple organ system failure. Crit Care Med.
more accurately, as changes in the score would parallel the
1986;14:271---4.
progressive increase in severity that precedes death. How-
[10].Leteurtre S, Martinot A, Duhamel A, et al. Development of a
ever, an earlier prediction of outcomes before the patient pediatric multiple organ dysfunction score: use of two strate-
deteriorates to an irreversible state seems to be crucial. gies. Med Decis Making. 1999;19:399---410.
Last of all, the mortality in our cohort was high, presum- [11].Leteurtre S, Duhamel A, Salleron J, et al. PELOD-2: an update of
ably due to a lack of resources that frequently hindered the the Pediatric Logistic Organ Dysfunction score. Crit Care Med.
implementation of appropriate diagnostic and therapeutic 2013;41(7):1761---73.
interventions. Therefore, some of the findings of our study [12].Graciano AL, Balko JA, Rahn DS, et al. The pediatric multiple
might to be generalizable to centres with low mortality organ dysfunction score (P-MODS): development and valida-
rates. tion of an objective scale to measure the severity of multiple
organ dysfunction in critically ill children. Crit Care Med.
2005;33:1484---91.
Conclusion [13].Vincent JL, Moreno R, Takala J, et al. The SOFA (Sepsis-related
Organ Failure Assessment) score to describe organ dysfunc-
The paediatric SOFA score was useful for predicting 30-day tion/failure. On behalf of the Working Group on Sepsis-Related
mortality in the general PICU population and performed bet- Problems of the European Society of Intensive Care Medicine.
ter than the PRISM and PIM2 scores in this regard, but its Intensive Care Med. 1996;22(7):707---10.
[14].Singer M, Deutschman CS, Seymour CW, et al. The Third Inter-
performance was only fair in predicting a prolonged PICU
national Consensus Definitions for Sepsis and Septic Shock
stay. Additionally, the SOFA score should replace the criteria
(Sepsis-3). JAMA. 2016;315(8):801---10.
for SIRS for diagnosis of sepsis in the paediatric population, [15].Matics TJ, Sanchez-Pinto LN. Adaptation and validation of a
as it proved to be superior in predicting 30-day mortality in Pediatric Sequential Organ Failure Assessment Score and eval-
patients with infection. However, contrary to what is estab- uation of the Sepsis-3 definitions in critically ill children. JAMA
lished in the Sepsis-3 definition, the optimal SOFA score Pediatr. 2017;171(10):e172352.
cutoff for diagnosis of paediatric was 3 points rather than [16].Slater A, Shann F, Pearson G. PIM2: a revised version
2 points. Prospective multicenter studies are required for a of the Paediatric index of mortality. Intensive Care Med.
better assessment of the paediatric SOFA score. 2003;29(2):278---85.
[17].Goldstein B, Giroir B, Randolph A. International pediatric sepsis
consensus conference: definitions for sepsis and organ dysfunc-
Appendix A. Supplementary data tion in pediatrics. Pediatr Crit Care Med. 2005;6:2---8.
[18].Seymour CW, Liu VX, Iwashyna TJ, et al. Assessment of
Supplementary material related to this article can be found, clinical criteria for sepsis: for the third international consen-
in the online version, at doi:https://doi.org/10.1016/j. sus definitions for sepsis and septic shock (Sepsis-3). JAMA.
anpedi.2019.05.018. 2016;315(8):762---74.
[19].Ferreira FL, Bota DP, Bross A, et al. Serial evaluation of the
SOFA score to predict outcome in critically ill patients. JAMA.
References 2001;286(14):1754---8.
[20].Schlapbach LJ, Straney L, Bellomo R, et al. Prognostic accuracy
[1]. Pollack MM, Holubkov R, Funai T, et al. Simultaneous prediction of age-adapted SOFA, SIRS, PELOD-2, and qSOFA for in-hospital
of new morbidity, mortality, and survival without new morbidity mortalityamong children with suspected infection admitted
from pediatric intensive care: a new paradigm for outcomes to the intensive care unit. Intensive Care Med. 2018;44(2):
assessment. Crit Care Med. 2015;43(8):1699---709. 179---88.
[2]. Bekhit Oel S, Algameel AA, Eldash HH. Application of pedi- [21].Wu Z, Liang Y, Li Z, et al. Accuracy comparison between age-
atric index of mortality version 2: score in pediatric intensive adapted sofa and sirs in predicting in-hospital mortality of
care unit in an African developing country. Pan Afr Med J. infected children at China’s PICU. Shock. 2018. Sep 11. doi:
2014;17:185. 10.1097/SHK.0000000000001261. [Epub ahead of print].
[3]. Pollack MM, Ruttimann UE, Getson PR. Accurate prediction of [22].van Nassau SC, van Beek RH, Driessen GJ, et al. Translating
the outcome of pediatric intensive care. A new quantitative Sepsis-3 Criteria in Children: Prognostic Accuracy of Age-
method. N Engl J Med. 1987;316:134---9. Adjusted Quick SOFAScore in Children Visiting the Emergency
+Model
ARTICLE IN PRESS
Paediatric SOFA score 9

Department With Suspected Bacterial Infection. Front Pediatr. signs in a pediatric emergency department. Acad Emerg Med.
2018;6:266. 2015;22:381---9.
[23].Churpek MM, Zadravecz FJ, Winslow C, et al. Incidence and [25].Sprung CL, Schein RMH, Balk RA. The new sepsis consensus def-
prognostic value of the systemic inflammatory response syn- initions: the good, the bad and the ugly. Intensive Care Med.
drome and organ dysfunctions in ward patients. Am J Respir 2016;42:2024---6.
Crit Care Med. 2015;192(8):958---64. [26].A Cabrita J, Pinheiro I, Menezes Falcão L. Rethinking the
[24].Scott HF, Deakyne SJ, Woods JM, et al. The prevalence and diag- concept of sepsis and septic shock. Eur J Intern Med.
nostic utility of systemic inflammatory response syndrome vital 2018;54:1---5.