External validation of the clinical score system for early detection of late–onset neonatal sepsis

Highlights

1. Introduction

Neonatal sepsis is the main cause of newborn death worldwide and one of the main factors contributing to the high neonatal mortality rate.¹ The clinical signs and symptoms of neonatal sepsis are non-specific, and the differential diagnosis is vast.² Many diagnostic biomarkers have been studied, but none are of sufficient quality to be applied in daily clinical practice.³ A clinical scoring system approach makes it possible for clinicians, especially in resource-limited settings, to make diagnoses other than microbiological culture.⁴ Husada D. et al. developed a very practical late-onset neonatal sepsis score system using six parameters that was published in 2020, based on the data from Queen Sirikit National Institute of Child Health in Bangkok. This scoring system has never been validated externally.

The number of neonatal sepsis cases worldwide is estimated to be around three million per year.⁵ According to the World Health Organization (WHO), neonatal sepsis caused approximately 430,000 neonatal deaths in 2013, accounting for an estimated 15% of all neonatal deaths worldwide.⁶ Meanwhile, Nyma Z. et al. (2020) reported that neonatal sepsis is currently responsible for approximately 1.6 million neonatal deaths worldwide, with developing countries accounting for 99% of cases (an incidence rate of 38 per 1000 live births).⁷ Blood culture has a 60% false-negative rate; hence, it cannot be utilized as a sole diagnostic tool. In contrast, Wirschafter DD. et al. (2011) discovered that the ratio of antibiotics given to positive cultures was 14:1, leading to an overuse of antibiotics.⁸

There are two types of neonatal sepsis based on the age of presentation: early-onset and late-onset neonatal sepsis.² The age limit to differentiate between early-onset and late-onset neonatal sepsis ranges from 3 to 7 days.^2,9–14 Some literature uses 7 days as the limit.^2,14,15 Early-onset neonatal sepsis occurs before 7 days of age, while LONS occurs after 7 days of age.² LONS is typically caused by organisms inferred after delivery and is categorized as a nosocomial community-acquired infection.^{2,11,13,15,16} LONS often presents with more subtle symptoms, such as feeding intolerance, vomiting, diarrhoea, abdominal distention, ABD (apnea, bradycardia, desaturation) spells, and hypoglycaemia. Focal infections such as omphalitis, meningitis, or osteomyelitis may precede or accompany LONS.² The most common pathogens causing LONS are Pseudomonas, Serratia, Escherichia coli, and Staphylococcus aureus.^2,13,15

Neonates are very susceptible to infection due to their immature immune system, and they become even more vulnerable at a smaller gestational age due to a lack of maternal antibody transfer in the last trimester.¹⁷ The diagnosis of neonatal sepsis is difficult because the clinical presentations are non-specific. In addition, specific laboratory tests for sepsis markers are still not available in all health facilities. The clinical score system is a method used to assess the likelihood of neonatal sepsis based on clinical indicators.⁴ The application of a clinical score system based on clinical presentations was shown to reduce excessive laboratory examination and antibiotic administration.¹³ Husada D. et al. (2020) have published a scoring system with six clinical parameters. This clinical score system is simple and practical for a clinician to use in making medical decisions. However, external validation of this scoring system has not been carried out. This prompted the idea for this study, which aimed to validate the score system as a diagnostic tool for LONS.

2. Methods

3. Results

3.1 Subject enrolment

During the research period, 370 neonates were treated at the NICU of Dr. Soetomo Academic General Hospital. There were 31.9% (118/370) neonates suspected of having late-onset neonatal sepsis. Twenty-eight of the 118 neonates suspected of having sepsis were excluded from this study, namely 19 neonates with major congenital malformations and 9 neonates who had surgery prior to the diagnosis of sepsis, leaving 90 neonates who met both the inclusion and exclusion criteria ( Figure 1).

Figure 1. Enrolment of subjects.

3.4 Validity results

The six parameters of the Husada D. et al. (2020) score are used to assess neonates suspected of having sepsis. The majority of the participants in this study, 66.7% (60/90), had a sepsis score <5, while the remaining 33.3% had a score ≥5. Table 5 revealed that neonates with low and medium scores had culture-negative results. Contrary to this, neonates with high and very high scores had culture-positive results. The results showed that this scoring system had a very good sensitivity value of 100.0% (87.6-100%), very good specificity of 96.8% (88.8-99.6%), PPV 93.3% (78.1-98.2%), NPV 100.0%, LR+ 31.0 (7.9-121.1), and LR– 0.0 with accuracy of 97.8% (92.2-99.7%).

Table 5. Distribution of score criteria based on blood culture results.

Score criteria	Frequency (n = 90)	Positive-culture result (n = 28)	Negative-culture result (n = 62)	p value
Low (0-2), n (%)	30 (33.3)	0 (0.0)	30 (48.4)	< 0.001*
Medium (3-4), n (%)	30 (33.3)	0 (0.0)	30 (48.4)
High (5-6), n (%)	9 (10.0)	8 (28.6)	1 (1.6)
Very high (7-14), n (%)	21 (23.4)	20 (71.4)	1 (1.6)

* Statistically significant at p < 0.05.

The ROC curve demonstrates that Husada D. et al.’s (2020) late-onset neonatal sepsis score has a good diagnostic value because the curve is far from the 50% line and approaches 100%. The AUC in this study was 98.8% (95% CI: 96.9-100.0, p-value < 0.001), as shown in Figure 2. The AUC value of 98.8% is statistically classified as very strong.

Figure 2. ROC curve of Husada D. et al.’s sepsis score.

4. Discussion

During the study period, 31.1% (28/90) sepsis neonates were found in the suspected-sepsis population. There are more males than females. According to a study conducted in Lahore by Sheikh AM. et al. (2010), male neonates are more likely to have neonatal sepsis. This could be due to sex-related factors in host vulnerability to infection.¹² Jaya IGA. et al. (2019) stated that the X chromosome has genes that influence the function of the thymus gland and the synthesis of immunoglobulins. Males are more susceptible to infection since they have only one X chromosome as opposed to two in females.²⁰ Xiao T. et al. (2017) revealed that the incidence of sepsis was higher in male neonates, particularly 52.1% neonates (100/192, p-value 0.930).²¹

In this study, the culture-positive group had a lower average birth weight (1827.14 ± 798.63 grams) than the culture-negative group (2043.55 ± 785.98 grams). Verstraete EH. et al. (2015) found that very-low-birth-weight was not a significant predictor (p-value 0.205) despite being related to nosocomial bloodstream infections.²² Belachew A. and Tewabe T. (2020) reported that neonates with a birth weight less than 2.5 kg are 1.42 times more likely to have sepsis than neonates with a birth weight greater than 2.5 kg.²³

Raguindin PFN. et al. conducted a validation study on the Okascharoen score in 2014.^16,24 The Okascharoen score predicts late-onset neonatal sepsis infection based on three clinical parameters (hypotension, abnormal temperature, and poor breathing) and three laboratory parameters (neutrophil bandemia > 1%, thrombocytopenia < 150000, and use of an umbilical catheter).²⁴ The validation study was performed on neonates aged 2 to 90 days in the NICU of Philippine General Hospital. According to the study, 50% (59/119) of the research subjects experienced sepsis. Males predominate in the culture-negative group. The average birth weight in the culture-negative group was lower than in the culture-positive group (1624 ± 942 grams vs. 1905 ± 887 grams). The gestational age in the two groups was relatively close, namely 33.93 ± 3.9 weeks in the culture-negative group and 34.08 ± 3.8 weeks in the culture-positive group.¹⁶ The results of the study reported by Raguindin PFN. et al. (2014) are comparable to those of this study in that LONS was not significantly correlated with gender, gestational age, or birth weight.

This study found 28 positive-culture results, with the majority of late-onset neonatal sepsis pathogens being gram-positive (19/28, 67.9%), and the remainder being gram-negative (32.1%). According to NICHD Neonatal Research Network data, gram-positive bacteria cause around 79% of LONS, while gram-negative bacteria account for 18%.^15,17 Staphylococcus spp. is the most prevalent type of bacteria (67.86%) in the gram-positive group, which is dominated by Staphylococcus epidermidis and Staphylococcus haemolyticus. Staphylococcus epidermidis is the most common pathogen detected in hospital-acquired infections, followed by Staphylococcus aureus, Staphylococcus capitis, Staphylococcus haemolyticus and Staphylococcus homini.²⁵ Coagulase-negative staphylococci are the most abundant gram-positive bacteria (68%).¹⁵ Staphyloccus epidermidis is a common colonizer of human skin and mucous membranes that seldom causes infection in healthy tissue. However, it can attach and reproduce on the plastic surface of medical devices, forming a biofilm that is resistant to antibiotics and the immune system.²⁶

Clinical score system have various performance in different settings and should be validated locally before being used routinely. A clinician must consider the diagnostic accuracy of a clinical score system while utilizing it to avoid improper or unnecessary antibiotic therapy, which may increase the development of antimicrobial resistance, gastrointestinal imbalance, and adverse clinical consequences. Numerous clinical score systems are available as screening tests for hospital-acquired infections in neonates. Lloyd LG. et al. (2022) discovered eleven clinical score systems in a literature review.¹¹

Blood culture, as the imperfect gold standard, acts as the strength as well as the limitation of this study. A negative blood culture cannot rule out the diagnosis of neonatal sepsis if clinical and other examinations are supportive.²⁷ The ideal gold standard always conveys positive results to all subjects with disease and negative results to all subjects without disease.²⁸ Even though blood culture yields positive results in only 41.4% of sepsis cases, it remains the best tool for diagnosing neonatal sepsis as it serves as a strong reliance for the sepsis group.¹⁰ Furthermore, blood cultures were only conducted on one side in this study. The recommendation for blood sampling should be at two different locations. Pathogens are more likely to be detected when two specimens are collected. As central venous catheters are currently often utilized in level III NICUs, it is favourable to obtain blood cultures simultaneously in peripheral and central vascular catheters for more accurate results.⁹

5. Conclusion

Husada D. et al.’s (2020) sepsis score is valid as a predictor of late-onset neonatal sepsis at Dr. Soetomo Academic General Hospital. Husada D. et al.’s sepsis score can be useful in the diagnosis of LONS. Further research on Husada D. et al.’s sepsis score with five parameters (without blood pH) is needed to develop a clinical score system of LONS in limited facilities.

Ethical statement