Elsevier

The Lancet

Volume 362, Issue 9379, 19 July 2003, Pages 192-197
The Lancet

Articles
Validation of the paediatric logistic organ dysfunction (PELOD) score: prospective, observational, multicentre study

https://doi.org/10.1016/S0140-6736(03)13908-6Get rights and content

Summary

Background

Multiple organ dysfunction syndrome is more frequent than death in paediatric intensive care units. Estimation of the severity of this syndrome could be a useful additional outcome measure in clinical trials in such units. We aimed to validate the paediatric logistic organ dysfunction (PELOD) score and estimate its validity when recorded daily (dPELOD).

Methods

We did a prospective, observational, multicentre cohort study in seven multidisciplinary, tertiary-care paediatric intensive care units of university-affiliated hospitals (two French, three Canadian, and two Swiss). We included 1806 consecutive patients (median age 24 months; IQR 5–90). PELOD score includes six organ dysfunctions and 12 variables and was recorded daily. For each variable, the most abnormal value each day and during the whole stay were used in calculating the dPELOD and PELOD scores, respectively. Outcome was vital status at discharge. We used Hosmer-Lemeshow goodness-of-fit tests to evaluate calibration and areas under receiver operating characteristic curve (AUC) to estimate discrimination.

Findings

370 (21%) patients had no organ dysfunction, 471 (26%) had one, 457 (25%) had two, and 508 (28%) had three or more. Case fatality rate was 6·4% (115 deaths). PELOD score was significantly higher in non-survivors (mean 31·0[SE 1·2]) than survivors (9·4[0·2]; p<0·0001). Calibration (p=0·54) and discrimination (AUC=0·91, SE=0·01) of PELOD and dPELOD (p≥0·39; AUC≥0·79) scores were good.

Interpretation

PELOD and dPELOD scores are valid outcome measures of the severity of multiple organ dysfunction syndrome in paediatric intensive care units; their use should significantly reduce the sample size required to complete clinical trials in critically ill children.

Introduction

Death is thought by many intensive care consultants to be the most reliable endpoint for clinical trials in intensive care units. However, with death as the primary outcome measure, the low death rate in paediatric intensive care units (about 6%1, 2, 3 vs >20% in adult intensive care units4, 5) increases the sample size required to complete a clinical trial, which reduces the feasibility of doing clinical trials in such units. A surrogate outcome can be substituted (or at least added) to a gold standard such as death rate, if its relation to the gold standard is very good, and if the prevalence of what it measures is substantially greater than that of the gold standard. Multiple organ dysfunction syndrome is more frequent than death in paediatric intensive care units—the rate ranges from 11% to 27%6, 7, 8, 9 Thus, a score that could be used to estimate the severity of multiple organ dysfunction syndrome could be an additional outcome measure to death in critically ill patients.10 The difference in rate should significantly reduce the sample size required to complete a clinical trial if multiple organ dysfunction syndrome is chosen as the primary outcome measure rather than death. Three multiple organ dysfunction syndrome scores have been validated in adults for this purpose,4, 5, 11 but none in children. In a prospective, multicentre study, we developed a paediatric multiple-organ dysfunction syndrome score, the paediatric logistic organ dysfunction (PELOD) score.12 This score was developed against mortality as in all such studies in adults. In this study, we aimed to validate the PELOD score, which is based on the most abnormal values of variables measured during the entire paediatric intensive care unit stay; and to estimate the validity of the PELOD score when recorded daily (dPELOD).

Section snippets

Participants

We prospectively included all consecutive patients admitted to seven multidisciplinary, tertiary-care paediatric intensive care units of university-affiliated hospitals (two French, three Canadian, and two Swiss). Exclusion criteria were age 18 years or older, premature, pregnant, length of stay in unit less than 4 h, admission in a state of continuous cardiopulmonary resuscitation without achieving stable vital signs for at least 2 h, transfer to another paediatric intensive care unit, and

Statistical analyses

An estimated 1422 patients (203 patients or 14 deaths by centre) were needed for 80% power with significance at 5% for an estimated mortality rate of 7% and relative risk worth detecting at 1·35 (PELOD odds ratio=0·30). We assessed interobserver reliability of data collection by checking a random selection of 3% of included patients and using the κ statistic for each organ dysfunction and the intraclass correlation coefficient for PELOD and dPELOD scores to estimate agreement.

The dependent

Results

Between September, 1998 and February, 2000, there were 2021 consecutive admissions. 215 patients were excluded for the following reasons: incomplete records (13); still cared for in paediatric intensive care units (two); palliative care (two); and exclusion criteria (in order listed in methods) 12, 55, 0, 13, 0, one, and 117, respectively. Thus, we included 1806 patients. Median length of study period in the units was 5 (IQR 4–6) months. The case-fatality rate was 6·4% (115 deaths). Table 2

Discussion

Our results validate the PELOD score. We have also validated the dPELOD score for the first 5 days of stay in paediatric intensive care units, which should enable increased sensitivity in assessment of changes in a patient's condition.

A scoring system should have construct and content validity. It must be able to be reproduced over time and across geographic and methodological boundaries and be accurate (calibration and discrimination) and clinically meaningful.14 The construct validity of the

References (31)

  • JD Wilkinson et al.

    Outcome of pediatric patients with multiple organ system failure.

    Crit Care Med

    (1986)
  • JL Vincent et al.

    The SOFA (sepsis-related organ failure assessment) score to describe organ dysfunction/failure.

    Intensive Care Med

    (1996)
  • S Leteurtre et al.

    Development of a pediatric multiple organ dysfunction score: use of two strategies.

    Med Decis Making

    (1999)
  • MM Pollack et al.

    Pediatric risk of mortality (PRISM) score.

    Crit Care Med

    (1988)
  • DG Altman et al.

    What do we mean by validating a prognostic model?

    Stat Med

    (2000)
  • Cited by (553)

    View all citing articles on Scopus
    View full text