Design
This present analysis on neonatal asphyxia utilizes a subset of data collected for the evaluation of SMI, a multinational maternal and child health initiative based in low-income sectors of Mesoamerica. Using a pre-post study design, data collection occurred at the baseline (March 1, 2011, to August 31, 2013) and second-phase follow-up (June 1, 2017, to August 30, 2018) after quality-of-care interventions were implemented. Under SMI, health system strengthening was performed in a stepwise process from 2012 to 2018. For asphyxia care, the first phase of SMI focused on availability of inputs, such as staffing skilled providers and having the medications and equipment necessary for neonatal resuscitation. Highlighted in this study, the second phase of SMI sought to increase quality and coverage of services. Interventions focused on training within broader quality improvement strategies.
Qualified personnel were trained in neonatal resuscitation concentrating on the four multidimensional asphyxia care components measured in this analysis, alongside general newborn care such as temperature control and umbilical cord care, within 3 years prior to second-phase follow-up. Quality improvement strategies sought to translate simulation-based training into real-life settings, with monitoring and evaluating using a Plan-Do-Study-Act cycle. Quality improvement steps involved creating flowcharts to map steps needed to provide care and designate personnel roles, standardizing work processes, measuring performance of the four multidimensional components of asphyxia care, and investigating root causes of underperformance. Technical assistance was provided by SMI to hospitals, with coaching and support throughout quality improvement cycles. Neonatal interventions were customized to the needs of each hospital. Health facilities measured the four multidimensional care components in this analysis in order to timely diagnose and provide asphyxia care at bedside. In most settings, interventions also included updating national protocols based on the latest evidence-based asphyxia practices, certification of general physicians and nurses to manage neonatal asphyxia to address shortages in availability of pediatricians and neonatologists, policy to preemptively request personnel to manage asphyxia for high-risk deliveries, and availability of pre-made asphyxia kits in delivery rooms.
Population and sample
Fifty-eight health facilities at baseline and 71 health facilities at second follow-up were visited in Mexico (health jurisdictions of Ocosingo, Palenque, Pichucalco, and San Cristobal De Las Casas); Nicaragua (departments of Bilwi, Jinotega, Las Minas, Matagalpa, and North Atlantic Region); Honduras (departments of Choluteca, Copan, Intibuca, La Paz, Lempira, Octoepeque, and Olancho); and Guatemala (departments of Huehuetenango and San Marcos) participated in this study. Due to the limited number of facilities, all facilities serving the lowest income (one-fifth of the population) per census data that met the EmONC criteria of basic-level facilities (health centers with basic neonatal resuscitation with bag-mask ventilation and routine maternal delivery care) or comprehensive-level facilities (referral hospitals with surgical and transfusion capabilities) were included. Ambulatory-level facilities (without neonatal asphyxia care capability), as well as care started elsewhere (transferred care) were excluded.
In terms of the medical record sample selection, records were first identified by the International Classification of Disease coding (ICD-10) of newborn asphyxia diagnosis and other neonatal conditions at the time of delivery. Due to this collection method, deliveries with healthy outcomes fell under a separate survey module and were not captured for this analysis. Next, records at each facility were sampled at random using a systematic sampling method, with the sampling interval corresponding to the sampling fraction. A random starting point was chosen, marking a two-year timeframe. Sampling quotas varied, impacted by facility and country resources. Sample sizes had enough power to detect differences in evaluation indicators, such as treatment of neonatal complications according to national norms.
Procedure
During data collection rounds, a health facility survey was performed, consisting of an interview questionnaire to assess facility workforce and infrastructure; an observation checklist to document available medical equipment and drug stock; and a medical record review to determine neonatal care coverage. Hired as independent surveyors, local physicians and nurses were instructed on data collection, and subsequently conducted the described three-part survey. The survey software DatStat Illume was used to electronically upload the data. No personal information that would identify the patient was collected via the electronic survey module.
Study variables
Multidimensional care (MDC) for neonatal asphyxia in health facilities, as measured in this analysis (Fig. 1), required four parts: 1) skilled provider present at delivery (physician, or nurse in Guatemala); 2) immediate assessment of the newborn (Apgar scoring at 1 and 5 min); 3) initial stabilization of the infant (drying or stimulation, heat application); 4) basic and advanced resuscitation if Apgar score ≤ 3 at 1 min (positive-pressure ventilation with self-inflating bag, or chest compressions, or endotracheal intubation, oxygen use if < 32 weeks’ gestational age). Additional resuscitation requirements were based on facility type (referral if delivery occurred at basic-level facility, pulse oximeter monitoring if delivery occurred at comprehensive-level facility). Multidimensional care for neonatal asphyxia was structured according to national norms from the ministries of health of each country and WHO international guidelines [1, 2, 11].
Covariates studied in this analysis were timing of data collection (baseline versus second-phase follow-up), country (Mexico as reference country, Nicaragua, Honduras, Guatemala), EmONC facility type (basic-level versus comprehensive-level), and disease severity (low birthweight, prematurity, and sepsis comorbidities versus asphyxia alone). Information on gestational age, maternal complications, and Apgar scores was obtained from the medical records.
Statistical analysis
We conducted descriptive analysis and multivariable logistic regression analysis to examine potential patient-level, facility-level, and country-level factors associated with the primary outcome of multidimensional care coverage for neonatal asphyxia. Covariates included in the regression analysis were timing of data collection, country, EmONC facility type, and disease severity. We explored gestational age as a separate covariate but kept this integrated into the disease severity covariate. We also explored low Apgar score as a covariate but dropped this as it is part of the primary outcome. Regression models were adjusted for clustering of observations at the health facility level. No missing data were noted for the regression models. P values < 0.05 were considered significant. Stata 14.2 (StataCorp LP, College Station, TX, USA) was used for the analysis.
Ethical considerations
This study received institutional review board approval by the University of Washington (exemption as non-human-subject research determination). The project was approved by data collection agencies (El Colegio de la Frontera Sur, Mexico), and by the ministries of health of participant countries (Guatemala, Honduras, Mexico, Nicaragua). Collaboration with the ministry of health, data-collection agencies, and indigenous communities within each country was maintained throughout the study. For the interview questionnaire and observation checklist sections of the survey, each facility health administrator participated in informed consent. The need for informed consent from patients for the study was waived by the University of Washington Institutional Review Board. For medical record extraction, patient information remained anonymized, and data were collected by trained professionals and uploaded electronically with DatStat Illume software.