Risk factors and practices contributing to newborn sepsis in a rural district of Eastern Uganda, August 2013: a cross sectional study

Study design

This was a cross-sectional study conducted at Kidera Health centre between January
and August 2013. Kidera Health Centre is a level four facility located in Kidera County,
Buyende District Eastern Uganda. It is the main referral unit for Buyende District.
The Health Centre serves the five counties in the district with an estimated population
of 248,000 people. The study population was mother and sick newborn pairs admitted
at the health facility during the study period. The sick newborns were those admitted
with signs and symptoms of sepsis. The definition of neonatal sepsis was adopted from
the International Paediatric Sepsis Consensus criteria (PSC) and the Intensive care
chapter of Indian Academy of Paediatrics (IAP) 23], 24]. We excluded cases where mothers or newborns were too ill to participate because
they had to be referred to Kamuli district hospital for emergency medical care.

Sample size

The estimated sample size using the formula by Kish Leslie (1965), assuming a prevalence
of sick newborns with sepsis to be 37 % 25] and a maximum error of 5 % within a 95 % confidence interval was 183 mother and sick
newborn pairs. The level of significance was set at p  0.05.

Sampling procedure

All sick newborns who presented to the health facility were screened for signs and
symptoms of sepsis by the health workers. The newborns that fitted the selection criteria
for having signs or symptoms of sepsis were moved to the admission ward where the
health workers first provided routine investigations and provided treatment according
to the national guidelines. After the health workers had provided the treatment, the
study team sought informed written consent from the mothers to take part in the study.
One of the routine investigations was blood culture to confirm sepsis and determine
the aetiological agent.

Qualitative data were gathered using a key informant interview question guide for
the eight health service providers. They were purposefully selected from the health
facility department providing newborn care services and district health team. These
included the midwives, nurses and clinical officers at the health facility plus the
district focal person for maternal and child health activities.

Collection and processing of samples procedure

The standard diagnostic test for newborn sepsis in this study was a positive blood
culture according to World Health Organization standard guidelines 26]. Blood cultures have a sensitivity that ranges between 80 and 90 % 27], 28]. However in Uganda, most diagnoses for newborn sepsis at lower level facilities (Health
Centres) depend on clinical signs and symptoms alone, which results in misclassification
of cases 17]. This study used clinical examination to identify the cases and blood culture investigations
to confirm the sepsis cases at a Health Centre. The blood cultures identified the
aetiological agents while the drug sensitivity tests were done to help determine the
appropriate antibiotic treatment regimens to use. The blood samples were collected
and processed by a trained district laboratory technician at the health facility.
The study provided commercially prepared blood cultures, drug sensitivity discs and
an incubator so that the investigations could be done at the health facility. The
study used a BBL Septi-Chek manual blood culture system from Dicfo Laboratories.

Approximately 2 ml venous blood was obtained after thorough cleansing of the patients’
skin for 2 min with povidine iodine and allowing the skin to dry before taking blood.
One millilitre of blood was collected in each of two bottles containing brain heart
infusion (BHI) in a ratio of blood: BHI of 1:10 and then taken to the laboratory within
30 min. Each bottle was incubated at 35 °C for 24 h following which it was examined
for visible growth and Gram stain was done. Subcultures were done on blood, Chocolate
and MacConkey agar from those blood culture bottles which showed presence of bacteria
on physical examination and Gram stain. The agar plates were incubated under aerobic
conditions 29]. The identification of the microbial isolates was by a combination of physical and
biochemical methods that included gram stain, catalase reaction, coagulase reaction,
colony morphology and haemolytic activity on blood agar 30]. Oxidase and citrate tests were done for identified gram negative isolates 29].

For a blood cultures which showed no visible growth and was negative on Gram stain,
three subcultures were done on blood, Chocolate and MacConkey agar and observed for
a maximum of 7 days before being discarded as negative if there was no growth. All
culture bottles with mixed growths (defined as more than 2 types of bacteria) were
discarded.

The disk diffusion method adopted from the Clinical laboratory Institute was used
to assess the antimicrobial susceptibility of all the isolates 31]. We used the commonly used antimicrobial agents in the Uganda National treatment
guidelines for newborn sepsis to assess the susceptibility of the isolates.

Study tools

Pre-coded, pre-tested, semi-structured questionnaires were used to collect information
from the mothers of newborns. The key informant interview (KI) guide was used to interview
health providers. The questionnaires and interview guide were pre-tested by the Principal
Investigator at Kamuli district hospital.

Data collection

The semi-structured questionnaires were administered by four trained research assistants
who were supervised by the Principal Investigator. The questionnaires used to interview
mothers of newborn were translated into the local language (Lusoga). The interviews
were conducted in the local language and immediately transcribed to English. The Principal
investigator and research assistants edited the data collected to ensure completeness
and consistency. The data were cleaned and stored by the Principal investigator. The
Principal Investigator conducted face to face interviews of health providers using
the Key Informant interview guide.

The health workers were interviewed to obtain information about challenges in prevention
and management of newborn sepsis practices within the facility. The Principal Investigator
observed for the availability of equipment and drugs in the facility used in providing
appropriate care of newborns.

Data analysis

The quantitative data collected using the questionnaire were coded, entered, cleaned
using Epi data and exported to SPSS version 16 software where additional variables
were created for analysis 32]. The primary outcome was a newborn with laboratory confirmed sepsis. The univariate
analysis of demographic factors was carried out using frequencies and means to describe
the study participants. Bivariate analysis was done to assess the associations between
the primary outcome and the independent exposure variables. The independent variables
included;

Social demographics such as Age, Education level and Source of income of mother.

Antenatal care (ANC) history of mother such as number of ANC visits attended in last pregnancy, history of being screened
for infections, history of receiving health education, Tetanus Toxoid immunisation
and history of bacterial infection during pregnancy as well as, treatment received.

History of birth circumstances including place of delivery, type of delivery and type of health provider who assisted
in delivery.

History of birth outcomes such as Birth weight and gestational age.

Postnatal care (PNC) history such as history of illness or complications related
to delivery and treatment.

Newborn care practices such as Cord care, feeding practice, and cleaning practice.

The continuous independent variables were categorized and their association with the
primary outcome was established using Chi square tests or Fisher’s exact test (for
tables where the expected cell values were less than 5). The measure of association
for categorical variables was Odds Ratios. The P values as well as 95 % confidence
intervals for the Odds Ratios were determined.

Variables found to be significantly associated with laboratory confirmed newborn sepsis
at 95 % level in bivariate analysis were included in the multivariate logistic regression
analysis using 0.05 for entry and 0.1 for exit P values. Confounding and interaction
between the various independent variables was assessed using logistic regression.
A stepwise logistic regression model was constructed using a backward elimination
approach. Multivariate logistic regression was applied to produce odds ratios of associations
between the independent variables and primary outcome of laboratory confirmed sepsis,
and adjusted to address the influence of other significant variables. The associations
in the multivariate logistic regression analysis that were statistically significant
(p  0.05) were included in the final results.

Qualitative data from key informant interviews were transcribed, coded, analysed and
separated into themes. It was triangulated with the findings from the questionnaires
to gain a deeper understanding of the information observed.

Ethical considerations

The ethical approval was obtained from Makerere University School of Public Health
Institutional Review Board, the Higher Degrees and Ethics Committee and the National
Council of Science and Technology. Buyende District health authorities and in-charge
of Kidera health centre were asked for permission to use their facilities. Informed
written consent was obtained from all the participants after explaining the risks
and benefits of the study before they were interviewed. We used anonymous identifiers
on the questionnaires to ensure privacy of the participants. All the members of the
study team complied with good clinical practices.