Reliability of visual assessment of neonatal jaundice among neonates of black descent: a cross-sectional study from Tanzania

A hospital-based cross-sectional study aimed at assessing the reliability of the Kramer’s method in detecting NJ was conducted between June and July 2020 at Muhimbili National Hospital (MNH), Tanzania. MNH is a national hospital located in Dar es Salaam center, receiving patients from different parts of the city and regions of Tanzania. The neonatal unit at MNH is one the biggest in the country, serving up to 500 neonates per month. The unit has adequate numbers of skilled pediatricians, medical doctors and nurses. In addition, the MNH has a well-equipped central laboratory with capacity of performing numerous tests, including total serum bilirubin.

Black descent neonates aged less than 28 days who were admitted at MNH and whose mothers/guardians provided written consent to participate in the study were enrolled. Neonates with rashes and those who were severely sick (in life support, in oxygen masks) were excluded because assessment of NJ using Kramer’s method cause physical discomfort to the babies.

A total of 315 neonates were recruited into this study. The formula for ‘diagnostic sample size calculation’ was used to calculate the sample size [10] (i.e. n = Z² x Specificity x (1-specificity) / e² x (1-prevalence)). Specificity of Kramer’s method of 89% from a study conducted in Indonesia [8], prevalence of NJ of 50% in Tanzania (unpublished article), precision of 5 and 95% (z = 1.96) confidence interval were used to obtain 283.5. Considering 10% non-respondent rate, a total of 315 neonates were obtained. Participants were recruited consecutively.

Data were collected using case report form (CRF) comprising of 3 sections; demographic information (age, gender, birthweight, gestational age, and delivery mode), Kramer scores and TSB results. The CRF was developed following literature review and experts’ consultation. Each neonate was subjected to visual assessment using Kramer’s method and TSB test. Neonatal jaundice was defined by yellowish discoloration of the skin by visual assessment or TSB level of ≥ 85 μmol/l by TSB [11].

Two medical doctors and 1 pediatrician with experience in neonatal assessment performed the visual assessment on routine ward round. For each neonate, visual assessment using Kramer’s method was performed by the 2 doctors separately and discrepancies were resolved by the pediatrician. Visual assessment of the 5 dermal zones were scored as no jaundice =0, jaundiced at level of head and neck = 1, jaundiced at upper trunk (above umbilicus) =2, jaundiced at lower trunk and thighs (below umbilicus) =3, jaundiced at arms and lower legs =4, jaundiced palms and soles = 5 [12]. The discrepancy of one value in Kramer score was acceptable. Discrepancies of more than one value in Kramer score were resolved by consulting a pediatrician. Kramer score recorded for each neonate was the average score of the two examiners. During assessment, the neonate was fully undressed, and assessment was conducted under blue fluorescent light. Besides, by using the thumb, the skin of neonate was blanched to observe the underlying skin color from the head to toes following standard procedure described by Devi et al. [6].

About 2–2.5 mls of blood sample was withdrawn from each neonate’s femoral vein using 2.5 cc syringe into a green topped heparinized vacutainer tube. The collected sample was transported to the laboratory for processing within 2 h using a cooler box containing ice packs. In the laboratory, the sample was centrifuged using a centrifuging machine to obtain serum. The obtained serum was analyzed for bilirubin concentration using the spectrophotometric chemistry analyzer (Architect chemistry analyzer, USA). Bilirubin determination was based on the run of bilirubin with diazotized sulfanilic acid. Azobilirubin concentration was measured at an absorbance of 500-600 nm which is proportionate to bilirubin concentration.

Data was entered and analyzed by SPSS version 23 software. Frequencies and percentages were used to summarize categorical variables and continuous variables were summarized by using median (interquartile ranges (IQR)). For Kramer’s method, those who scored 0 were grouped as no jaundice while 1–5 were considered to have jaundice. TSB test of ≥85 μmol/l were considered to have jaundice and below that as no jaundice. Sensitivity, specificity, predictive values and diagnostic effectiveness (accuracy) were obtained by contingency Tables (2 × 2 table for diagnostic test) [13]. Diagnostic accuracy results were defined as very weak (> 50–60%), weak (> 60–70%), moderate (> 70–80%), good (> 80–90%), or very good (> 90–100%). Positive likelihood ratio (+LR) >10 and negative likelihood ratio (−LR) < 0.1 were considered to be a good test [14]. Cohen kappa statistic (κ) was used to determine agreement of Kramer’s method and TSB results, with (κ) of 0.21–0.40, 0.41–0.60, 0.61–0.80 and 0.81–1.00 being fair, moderate, good and very good agreement, respectively [13]. Association of independent variables and NJ was done using the chi-square test (X² test) and p < 0.05 was considered to be statistically significant.

Of 315 respondents with complete data, 182 (57.8%) were males with the median (IQR) age of 4 [2‐8] days. More than half (58.4%) of the neonates were delivered by spontaneous vaginal delivery and half of the neonates 158 (50.2%) were born at preterm with median (IQR) birth weight of 2500 (1700–3000) grams Table 1.

The prevalence of NJ was 63.5% by TSB and 49.8% by Kramer’s method. The median (IQR) TSB for neonates who were clinically not jaundiced was 65(40–106) μmol/l and 178 (121–238) μmol/l for neonates who were clinically jaundiced. Following chi-square test, all variables including age (p = 0.46), gender (p = 0.77), delivery mode (p = 0.13), birthweight (p = 0.19) and gestational age (0.78) were not significantly associated with the occurrence of NJ.

Kramer’s method had a Sensitivity-70.5%, Specificity-86.1%, PPV-89.8%, and NPV-62.6% in detecting NJ. It had 5.07 LR of detecting jaundice and 0.34 LR of not detecting jaundice among neonates with a diagnostic accuracy of 0.761. Kramer’s method had moderate agreement with TSB test (κ = 52.4, P < 0.01) Table 2.

For neonates born at term, the Kramer test had 75.3% sensitivity, 82.8% specificity, 86.4% PPV, 69.7% NPV and 56.4% Cohen kappa in comparison to 66.4% sensitivity, 82.8% specificity, 93.0% PPV, 56.0NPV and 48.8% Cohen kappa for preterm neonates, P < 0.001 Table 3.