This was a prospective, observational, multi-site study conducted April-July 2015 at two tertiary referral hospitals in southern Ghana.(12) Globally accepted STROBE guidelines were used in designing and reporting findings.(13)
Setting: Study sites included the maternal and neonatal units of Komfo Anokye Teaching hospital (KATH) in Kumasi and Cape Coast Teaching Hospital (CCTH) in Cape Coast. KATH is the teaching hospital affiliated with the Kwame Nkrumah University of Science and Technology School of Medical Sciences (KNUST-SMS) and serves as the referral center for most of central Ghana. Each year approximately 11,000 women deliver at KATH. CCTH is the teaching hospital of the University of Cape Coast, School of Medical Sciences (UCC-SMS) and serves as the main referral hospital for much of the rural central and parts of western regions of Ghana. The hospital conducts about 2,800 deliveries per year. At the time of this research, each hospital provided specialty care for sick newborns, with capacity to provide bag and mask ventilation, oxygen and incubator care with radiant warmers (although often shared), and phototherapy. Routine laboratory tests, such as complete blood counts (CBC), serum bilirubin and electrolytes, were available, but delays in processing sometimes limited the utility and utilization of such tests.
Instruments Used: Given the lack of a universally accepted neonatal near-miss screening tool,(3,6,7) we reviewed the existing literature for research assessing neonatal near-misses(4-8, 14, 15) and examined the framework of the WHO Maternal Near-Miss Screening Tool.(16) We also reviewed modifications of the WHO Maternal Near-Miss Screening Tool for use in Ghana.(17) Neonatal tools to date have focused on “pragmatic markers of severity” (e.g. Apgar scores <7 at 5th minute, birthweight <1750g, gestational age <33 weeks) and “management markers of severity” (e.g. use of IV antibiotics, nasal CPAP, any intubation in the first 7 days, use of phototherapy, cardiopulmonary resuscitation, etc.),(3, 4, 15) while the WHO Maternal Near-Miss Tool focuses on symptoms, interventions, and organ system dysfunction to categorize mothers as having experienced a maternal near-miss. We consolidated all criteria found in the literature across the neonatal tools, and then used an interactive heuristic approach consisting of structured interviews with the heads of the departments of pediatrics and the supervisors of the NICUs at KATH, CCTH, and Korle Bu Teaching Hospital in Accra (affliated with the University of Ghana) and an international neonatalogist to determine the appropriateness of the included criteria for the Ghanaian setting. Each criterion was discussed at length, and cut points were agreed upon as most appropriate in this setting. The final tool, termed the NNMAT (Neonatal Near Miss Assessment Tool), was iteratively reviewed and revised, resulting in a tool with four categories: evidence of severe complications (akin to the “pragmatic category” seen in the literature), interventions conducted (akin to “management markers of severity”), organ-based dysfunction, and investigations conducted. Definitions conformed to the International Statistical Classification of Diseases and Health Related Problems (ICD-10). (See Table 1.)
Study Participants: All babies delivered at KATH and CCTH with birth complications indicated in the delivery ledger during the study period (e.g. pre-term delivery, low birthweight (<2500g), Apgar scores <7 at 1 minute, birth asphyxia, congenital malformations, or indications that they were being transferred to the NICU) were identified and screened with the Neonatal Near Miss Assessment Tool (NNMAT) to identify potential near-miss cases. In addition, all babies admitted to the neonatal units at KATH and CCTH – regardless of where they were born – were recruited for participation as soon as possible following admission to the neonatal units. Babies less than 28 weeks gestation (the official age of viability in Ghana, estimated via date of last menstrual period or ultrasound scan if available) or less than 500g birthweight were excluded, as were those who were stillborn.
Data Collection: Trained research assistants reviewed the admission ledgers at each site to identify newborns with complications. Parents of newborns with complications who had been admitted were approached, asked about their willingness to participate in a study and were taken through a written consent process if they agreed. The admission ledger was used to complete the NNMAT, supplementing information from the medical record with information from the physician or nurse on duty as necessary. NNMAT forms were completed using Qualtrics data collection software (Provo, Utah) on a hand-held tablet as soon as possible after admission, typically within 24 hours. Babies were followed at 7, 14, and 28 days to determine survival. For those discharged home, families were contacted via telephone and asked to provide an update on the health of the baby.
Key Variables: In addition to generating the NNMAT, the two primary outcome variables of interest in this study were near-miss categorization and death. Table 1 illustrates the components of the NNMAT. Each item was recorded as yes, no, or don’t know. A positive response to an individual item yielded a positive classification within its overarching category, and any positive classification to any category was considered to be a neonatal near-miss. Thus newborns who had positive answers to any of the screening criteria were considered to be a suspected neonatal near-miss at enrollment, and a confirmed near miss if they survived to 28 days in keeping with the definition used by Pillegi et al.(6) and Say(7). Death was assessed in two ways. First, aggregate, hospital-level statistics regarding institutional neonatal mortality were retrieved for the months of the study to be able to compare rates of near-misses with rates of reported institutional neonatal mortality. In addition, research assistants at each site continued to track every newborn recruited into the study for 7, 14, and 28 days after birth to determine survival. Those who died prior to 28 days were considered neonatal deaths for the purposes of identifying predictors of death vs. survival among those categorized as near-misses.
Data Analysis: Data were exported from Qualtrics into Stata version 13.1 (College Station, Texas) for cleaning and analysis. Frequencies were generated across all indicators assessed. Summary statistics reflecting total number of deliveries, live births, and neonatal deaths for the study period were collected from each hospitals’ records, allowing for the calculation of neonatal near-miss incidence. Neonatal near-miss incidence was calculated by dividing the number of neonatal near-misses recorded by the number of live births recorded, in keeping with the predominant method of the existing literature.(18) Confidence intervals were calculated.
Pearson’s Chi Square analysis was conducted to compare those who died with those who survived the first 28 days after delivery. Multivariate logistic regression was conducted (with death before 28 days as the outcome variable) to determine which categories of the neonatal near-miss tool were most predictive and then which elements within each category were most predictive. Variables significant at p<.05 in these preliminary models were carried forward into a final multivariate model to illustrate the strongest predictors of mortality. For the final model, a conservative p value of 0.01 was taken as statistically significant to account for multiple comparisons. In addition, positive and negative predictive values of the NNMAT were calculated to determine how well the NNMAT classification predicted death.
Ethical Review: Ethical approval was obtained from the Institutional Review Boards of Kwame Nkrumah University of Science and Technology, University of Cape Coast, University of Ghana, and the University of Michigan.