Early application of generic mortality risk scores in presumed meningococcal disease*

Abstract

Objective: Mortality from meningococcal disease typically occurs within 24 hrs of intensive care unit (ICU) admission. An early, accurate mortality-risk tool may aid in trial design for novel therapies. We assessed the performance of two generic scores that assign mortality risk within 1 hr of ICU admission: the Preintensive Care Pediatric Risk of Mortality (Pre-ICU PRISM) and Pediatric Index of Mortality (PIM). Design: Prospective, observational study over 21 months. Setting: Two tertiary pediatric ICUs accepting referrals from southeast England. Patients: Patients were 165 consecutive children with meningococcal disease. Ages ranged from 0.1 to 17 yrs (median 2.3 yrs). Interventions: None. Measurements and Main Results: PIM demonstrated greater sensibility, with complete data collected in 93% of cases, compared with 35% for the pre-ICU PRISM. Both scores discriminated well. The area under the receiver operating characteristic curve was 0.90 (95% confidence interval, 0.81–1.00) for PIM and 0.94 (95% confidence interval, 0.88–0.98) for Pre-ICU PRISM; this did not change when applied to the subgroup of patients with complete data. Both scores calibrated poorly, overestimating mortality in the medium-risk strata (and also in the high-risk stratum in the case of Pre-ICU PRISM). When used as a stratification tool for a hypothetical trial (60% reduction in mortality, 80% power), the scores allowed for a reduction in study size by 50% (PIM) and 43% (pre-ICU PRISM). Conclusions: Pre-ICU PRISM and PIM both discriminate well but calibrate poorly when applied to a cohort of children with meningococcal sepsis. Both scores provide an effective means of stratification for clinical trial purposes. The main advantage for PIM appears to be ease of data collection.