We conducted this study to validate the efficacy of CRIB II in predicting pre-discharge neonatal mortality in preterm neonates needing intensive care.

Gestational age was calculated from the first day of last menstrual period (LMP). In cases where LMP was not known, obstetric ultrasonography was used to assess the gestational age. In cases where both of the above were missing a gestational age assessment was made by using the expanded new Ballard score(4). Birthweight was recorded for each baby as soon as they arrived in the nursery or NICU for admission. This was done using an electronic scale having a sensitivity of 10 grams. Arterial blood gas analysis was done in all preterm babies at admission and then as dictated by the clinical condition of the baby. Temperature was recorded using a digital thermometer. All these parameters along with the sex of the baby were assigned scores according to the CRIB II. The final CRIB II score was obtained by the arithmetic sum of the individual scores assigned. The primary outcome measured was in-hospital mortality. Predicted mortality was compared with observed mortality.

Logistic model was used to analyze the prediction of mortality by the CRIB II score at admission. Discrimination – that is, the ability of the score to correctly predict survival or death – was assessed by calculating receiver operating characteristic curves and their associated area under the curve (AUC). An AUC value of 0.5 indicates no ability to discriminate and larger values indicate increasing ability. A value of 0.8 is considered good(5).

Babies discharged against medical advice (LAMA) were also taken into account. Data were analyzed in three ways (i) cases with known outcomes included in the analysis and excluding those to left against medical advice (LAMA); (ii) a comprehensive analysis of all neonates including those who left (LAMA) and assuming all those who left would have died if they stayed back; and (iii) after including neonates who left and assuming all those who left would have survived if they had stayed back.

Separate ROC curves were generated for all the three scenarios and analyzed. The Hosmer-Lemeshow Chi-square test was performed to look for any statistically significant difference between predicted and observed mortality. STATA 9.1 was used for data analysis.

There were 88 infants admitted to NICU at or below 32 weeks gestational age during the study period. Two babies were excluded, one because of congenital heart disease and the other because of mistaken dates. Thus, 86 neonates (males: 51(59.3%), birthweight: 1228 ± 398 g, gestation: 28.3± 2.4 weeks) were enrolled, of which 17 (19.8%) left hospital against medical advise (LAMA). Among 69 neonates completing the study, 24 (34.8%) died and 45 (65.2%) had a favourable outcome. The mean CRIB II score was 8.29 ± 4.35 (median 8, inter quartile range 5-12).

ROC curve analysis shows the area under curve (AUC) was 0.9032 (SE 0.0345, 95% CI: 0.83553-0.97096) suggesting that mortality prediction was 90% accurate for 69 babies who stayed up to the study end point (Fig.1). When the analysis was done assuming all those who left were survivors up to discharge, the AUC was 0.8703 (SE 0.0394, 95% CI: 0.7931- 0.9474) suggesting mortality prediction was correct in 87%. In the analysis which included LAMA cases as died, the area under ROC curve was 0.8314 (SE 0.043, 95% CI: 0.7468 – 0.9158) suggesting mortality prediction was 83.1% correct.

Fig.1 Mortality prediction on ROC curve for different outcomes.

HL goodness of fit test was applied to test the difference between observed and expected outcome. There was no significant difference between expected and observed outcome (P=0.62).

In our study, the area under ROC curve for mortality prediction by CRIB-II was 0.9 and there was no significant difference between predicted and observed mortality. This is similar to the study by Gagliardi, et al.(6), who showed AUC of 0.907. In our study mortality prediction was better than the development study for CRIB II(3), probable reason for this difference was related to higher mortality (33% vs 9%) and small sample size in our study.

CRIB has previously been evaluated at our center, the area under ROC curve was 0.823(7). The CRIB II has performed better than CRIB in our center. A study by Christoph, et al.(8) showed prediction with CRIB II (AUC of 0.69) was less than CRIB (0.82), birthweight (0.74) and gestational age (0.71). The reason for the low prediction of CRIB II in their study is not clear.

None of the babies in our study received surfactant immediately after birth. The fact that the prediction of survival/mortality was excellent using CRIB II suggests that survival depends primarily on the condition of the baby at birth rather than the intervention used. This validates the primary premise of the workers who have developed this severity of illness score. Although CRIB II score is less affected by perinatal factors(6) and despite good mortality prediction, we need further studies to document the influence of various pre and perinatal factors. A study having controlled for variables like antenatal steroids, maternal illness, multiple pregnancy, APGAR score at birth and use of surfactant is needed.

Contributors: PKR: design, data collection and interpretation, manuscript writing; VS: data analysis and interpretation; NK: concept, design, manuscript drafting and will act as guarantor of study.

Funding: None.

Competing interests: None stated.

1. Ridley SA. Uncertainty and scoring system. Anaesthesia 2002; 57: 761-767.

2. International Neonatal Network. The CRIB (Clinical Risk Index for Babies) Score. A tool for assessing initial neonatal risk and comparing performance of neonatal intensive care units. Lancet 1993; 342: 193-198.

3. Parry G, Tucker J, Tarnow Mordi W. CRIB II: An update of the clinical risk index for babies score. Lancet 2003; 361: 1789-1791.

4. Ballard JL, Khoury JC, Wedig K, Wang L, Eilers-Walsman BL, Lipp R. New Ballard’s score expanded to include extremely premature infants. J Pediatr 1991; 119: 417-423.

5. Hanley JA, Mcneil BJ. The meaning and the use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982; 143: 29-36.

6. Gagliardi L, Cavazza A, Brunelli A, Battagliolo M, Merazzi D, Tandoi F, et al. Assessing mortality risk in very low birth weight infants: a comparison of CRIB, CRIB II, and SNAPPE II. Arch Dis Child Fetal Neonatal Ed 2004; 89: F 419-422.

7. Khanna R, Taneja V, Singh SK, Kumar N, Sreenivas V, Puliyel JM. The clinical risk index of babies (CRIB) score in India. Indian J Pediatr 2002; 69: 657-660.