Skip to main content

Table 4 Discriminatory ability of different scores when applied to data from FEAST and Kilifi

From: Predicting mortality in sick African children: the FEAST Paediatric Emergency Triage (PET) Score

   FEAST (data from the control (no fluid bolus) arm only) (n = 1044) Kilifi High Dependency Ward (n = 1058) Kilifi General Admissions (n = 5107)
Score Variables included Number with score (% died) AUROC (95 % CI) e Hosmer-Lemeshow test (P value) Number with score (% died) AUROC (95 % CI) e Hosmer-Lemeshow test (P value) Number with score (% died) AUROC (95 % CI) e Hosmer-Lemeshow test (P value)
FEAST PET score Axillary temperature, heart rate, capillary refill time, conscious level, deep breathing, respiratory distress, lung crepitations, weak pulse 1024 (7 %) 0.82 (0.77–0.87) 0.56 1053 (9 %) 0.77 (0.72–0.82) 0.30 5098 (2 %) 0.86 (0.82–0.89) 0.50
Bedside PEWS score a Heart rate, capillary refill time, respiratory rate, oxygen saturation, systolic blood pressure 1000 (9 %) 0.64 (0.56–0.71) 0.46 1053 (9 %) 0.69 (0.64–0.75) 0.56 5094 (2 %) 0.74 (0.69–0.79) 0.22
PRISM III score b Heart rate, temperature, conscious level, systolic blood pressure, glucose, potassium, PCO 2 , pH, acidosis, pupillary reflexes 627 (6 %) 0.71 (0.61–0.81) 0.26 1056 (9 %) 0.69 (0.64–0.74) 0.10 5099 (2 %) 0.77 (0.73–0.82) 0.01
AQUAMAT score c (overall) Conscious level, chronic disease, convulsions, BUN, and base excess 648 (5 %) 0.74 (0.65–0.83) 0.84 1011 (9 %) 0.62 (0.56–0.68) 0.79 4964 (2 %) 0.73 (0.68–0.78) 0.04
AQUAMAT score c (malaria positive only) Conscious level, chronic disease, convulsions, BUN, and base excess 360 (3 %) 0.80 (0.68–0.93) 0.65 355 (6 %) 0.54 (0.42–0.66) 0.83 781 (3 %) 0.60 (0.49–0.72) 0.41
PEDIA Immediate death score d Anaemia, jaundice, indrawing, deep breathing, conscious level, convulsions/seizures, temperature 1007 (3 %) 0.75 (0.68–0.83) 0.64 680 (4 %) 0.79 (0.71–0.87) 0.47 3504 (1 %) 0.89 (0.84–0.94) 0.15
PEDIA Early death score d Jaundice, indrawing, conscious level, convulsions/seizures, wasting, kwashiorkor 1003 (4 %) 0.70 (0.63–0.77) 0.02 1016 (9 %) 0.69 (0.63–0.76) 0.76 5071 (2 %) 0.84 (0.78–0.89) 0.08
PEDIA Late death score d History >7 days, conscious level, convulsions/seizures, temperature, wasting, kwashiorkor 959 (1 %) 0.55 (0.40–0.69) 0.35 664 (10 %) 0.66 (0.60–0.73) 0.34 3472 (2 %) 0.72 (0.66–0.77) 0.08
LODS Deep breathing, coma and prostration 1038 0.77 (0.72–0.82) 0.38 1057 (9 %) 0.76 (0.71–0.81) 0.62 5103 0.87 (0.83–0.90) 0.74
  1. a Variables in the score but not measured: receipt of oxygen therapy, respiratory effort in four categories (normal, mild increase, moderate increase, severe increase, any apnoea). Underlined variables were available in the FEAST dataset but not in the Kilifi datasets
  2. b Variables in the score but not measured: pupillary reflexes, pH, total CO2, PCO2, arterial PaO3, creatinine, urea, white blood cells, prothrombin time, and platelets. Underlined variables were available in the FEAST dataset but not in the Kilifi datasets
  3. c Underlined variables were available in the FEAST dataset but not in the Kilifi datasets
  4. d Time of death was not available in the Kilifi data. Immediate deaths were defined as those that occurred on the same day as admission to hospital. The early death score was calculated on mortality by two calendar days but not the same day as admission. Late death defined as strictly greater than 2 days after admission. Immediate deaths were not included in the early death analysis, immediate and early deaths were not included in the late death analysis as in the original publication
  5. e The AUROC value for each score was compared to the FEAST PET score for mortality by 48 h. In the FEAST dataset there was no evidence of a difference between the AUROC for the FEAST PET score versus the AQUAMAT score overall (P = 0.19) and in malaria only (P = 0.65), and the FEAST PET score was significantly better than Bedside PEWS (P < 0.001), PRISM III (P = 0.02), LODS (P = 0.05), and PEDIA for immediate (P = 0.002) and early death (P = 0.04). In the Kilifi validation datasets (high dependency/general) there was no evidence of a difference between the AUROC for the FEAST PET score versus LODS (P = 0.67/0.73) or PEDIA for immediate (P = 0.34/0.82) and early (P = 0.63/0.47) death, and the FEAST PET score was significantly better than Bedside PEWS (P = 0.02/<0.001), PRISM III (P = 0.003/<0.001), and the AQUMAT scores (P <0.001/<0.001)
  6. AQUAMAT, African Quinine Artesunate Malaria Trial; AUROC, Area under the receiver operating curve; BUN, Blood urea nitrogen; FEAST, Fluid Expansion as Supportive Therapy; LODS, Lamberéné Organ Dysfunction Score; PEDIA, Pediatric Early Death Index for Africa; PET, Paediatric Emergency Triage; PEWS, Pediatric Early Warning System; PIM, Paediatric Index of Mortality; PRISM, Pediatric Risk of Mortality