Among 17,530 eligible livebirths, 17,198 (98.1 percent) were enrolled in the study, received one or more home visits, and are included in the analysis. There were 157,834 home visits to assess infants for signs of umbilical cord infection, and the mean number of visits per child was 9.2 (standard deviation: 2.1). Visits were missed during periods of uncertain security, military curfews, nationwide strikes, and religious holidays or in the case of late arrival at the home after the birth or death of the infant. In the study population, 91.9 percent of infants were born in the home, there were slightly more males (51.1 percent) than females, 29.5 percent of the sample were low birth weight, and more than 75 percent of newborns' mothers had received no formal education. Additional details on the characteristics of the study population were reported previously (4
Among all umbilical cord assessments, there were 1,290 (1.0 percent) assessments that met the cord infection definition among 954 individual infants, for a total incidence of 5.5 per 100 neonates. The median time from birth to infection was 2.5 days. More than 75 percent of infections occurred within the first 5 days, and 90 percent occurred in the first week of life ().
Distribution of age at onset of cord infection, Sarlahi, Nepal, 2002−2005. The total number of cases for each day is shown at the top of each bar.
At the time of birth, one or more substances were applied to the cords of 3,236 of 15,755 (20.1 percent) newborns. Throughout the newborn period, 13,827 (80.4 percent) infants received at least one application of mustard oil. Ash (7.1 percent), mud (6.8 percent), and other substances (5.5 percent; breast milk, saliva, water, other oils, herbs, spices, curry) were almost equally applied, while 2,653 (15.4 percent) infants received one or more home-delivered non-study “antiseptics” (). Data on the specific type of reported antiseptic were not collected. Comparisons between cord infection and these and other potential risk factors, adjusted only for treatment group in the randomized trial, are shown in .
Tabulation of topical applications to the umbilical cord during the neonatal period, by time of application (more than one substance per infant possible), Sarlahi, Nepal, 2002−2005
Analysis of potential risk factors for cord infection, adjusted for treatment group, Sarlahi, Nepal, 2002−2005
Topical applications to the cord may have increased as a result of infection; the use of ash, mud, and home-delivered antiseptics peaked after the appearance of local signs of infection and can lead to spurious associations. For example, infection risk was slightly higher among infants receiving applications of non-study antiseptics but was potentially protective when delivered within the first 48 hours (relative risk (RR) • 0.74, 95 percent confidence interval (CI): 0.55, 0.98). Thus, in , exposure to topical applications was restricted to those occurring within the first 48 hours after birth, when the cord is still attached in most infants. Although neither ash nor mud was associated with cord infection, mustard oil and other substances (breast milk, saliva, water, other oils, herbs, spices, curry) increased the risk of infection. There was an association between birth weight and risk of infection, with more infections in smaller babies (p • 0.03).
A number of intermediate determinants were associated with umbilical cord infection. These included a protective benefit of hand washing, by both the birth assistant (with soap) before delivery (RR • 0.69, 95 percent CI: 0.61, 0.79) and the mother during the first 14 days of life (RR • 0.71, 95 percent CI: 0.56, 0.91), and the reported correct !use of the soap in the clean delivery kit (RR • 0.51, 95 percent CI: 0.45, 0.58). Other items in the kit, such as the new blade and clean string, were not associated with decreased infection, but the almost universal use of these items led to low statistical power to detect any true differences. Thermal care practices such as skin-to-skin contact between mother and newborn (RR • 0.67, 95 percent CI: 0.46, 0.97) and the regular use of a hat (RR • 0.82, 95 percent CI: 0.71, 0.96) were also associated with decreased risk of omphalitis, while infants whose mothers reported regularly warming the room in which the infant was cared for were at slightly elevated risk. There was no evidence that breastfeeding (either feeding colostrum or initiation on the first day of life) or bathing practices protected the newborn from infection. Among the distal determinants assessed, more infections occurred among infants born in the hot season, and newborns born in lower castes (non-Brahmin/Chhetri) were at higher risk of infection. Infants born to mothers in the highest educational category were at slightly less risk, while paternal education and overall ethnic group (Pahadi (originating from the hills of Nepal) vs. Madeshi (originating from the plains)) were not associated with infection risk.
Multivariable models () were constructed on the basis of these results. Variables showing at least modest statistical evidence for an association (p < 0.10) were included. In the first model, only home-delivered applications to the cord were included. After adjustment, there was little change in the estimates of association for mustard oil, non-study antiseptics, and “other” substances delivered in the first 2 days. Hand-washing and thermal-care variables were added to the second model; the adjusted risk of infection was 27 percent (95 percent CI: 17, 36) lower among infants where the birth assistant washed her hands with soap before delivery. The adjusted risk of infection was also significantly lower when mothers reported “always” washing their hands with soap before handling the newborn (RR • 0.75, 95 percent CI: 0.59, 0.96) or when skin-to-skin contact was practiced (RR • 0.65, 95 percent CI: 0.44, 0.96). After the addition of maternal education, birth weight category (<2,500 g vs. • 2,500 g), season of birth, and caste to the third model, evidence for a protective benefit of non-study antiseptics, inconsistent hand washing by mothers, and the use of a hat to keep the infant warm was low; the latter two variables were removed from the final model. Being born during the hot season, low birth weight, lower caste, and harmful applications to the cord (mustard oil and other substances) were significant risk factors for infection, while hand washing and skin-to skin contact provided protection. Regular room warming remained associated with a higher risk of infection, even after adjustment.
Multivariate analysis of potential risk factors for cord infection, Sarlahi, Nepal, 2002−2005
The use of the soap in the clean delivery kit was not included in the multivariable models, because of colinearity with the birth assistant hand-washing variable. When that variable, however, was replaced with the clean delivery kit soap variable, the estimated adjusted risk ratio was 0.49 (95 percent CI: 0.43, 056), further confirming the importance of this hygienic care practice.
Given the potential of the treatment protocol provided in the overall trial to modify the associations between these covariates and the infection outcome, we constructed the final multivariate model separately for each of the treatment methods (chlorhexidine, soap/water, dry cord care). Apart from the association between hand washing and infection, there were no other statistically significant differences in the risk ratio estimates across treatment groups. For hand washing by the birth assistant, the adjusted risk ratio for infection was 0.60 (95 percent CI: 0.49, 0.75) in the dry cord care group compared with 0.87 (95 percent CI: 0.70, 1.08) and 0.79 (95 percent CI: 0.60, 1.03) in the soap/water- and chlorhexidine-cleansing groups, respectively (tests for interaction: p • 0.02, p • 0.10, respectively).
To further examine the potential for reverse-causation bias in the analyses of cord infection and home-delivered applications to the cord, we conducted a nested case-cohort analysis adjusting for timing of exposure, with results shown in . There were no important differences in the estimates of association of the conclusions drawn; potential unclean substances were associated with case status, while there was moderate evidence that non-study antiseptics decreased risk.
Nested case-cohort analysis of home-delivered applications to the cord and cord infection status, Sarlahi, Nepal, 2002−2005