Quality improvement strategies to improve inpatient management of small and sick newborns across All Babies Count supported hospitals in rural Rwanda

The hospitals for implementation of ABC during scale-up were identified jointly by the Rwandan MOH and PIH/IMB based on the following criteria: high neonatal mortality rate and no partner organizations to improve neonatal care in their areas. The hospitals were Nemba and Ruli in Gakenke District, Kinihira and Rutongo in Rulindo District, Gakoma and Kibilizi in Gisagara District, and Mibilizi in Rusizi District, in the Northern, Southern and Western Provinces of Rwanda respectively (Fig. 1). All hospitals were Rwanda MOH operated facilities and they supervised 69 referring-health centers serving a population of approximately 1,349,280 people. A district-hospital NCU in Rwanda provides phototherapy, oxygen, intravenous fluids, nutrition and breastfeeding, vital signs monitoring, neonatal resuscitation, Kangaroo Mother Care (KMC), and drug administration. The units are led by a nurse in-charge with a supervising medical doctor.

The ABC approach was an 18-month change acceleration process to reduce neonatal mortality through improved quality of care, which is described in depth elsewhere [19]. Before the launch of ABC in each district, a Memorandum of Understanding describing the role of each partner was signed by the MOH, individual participating hospitals, district leadership officials, and PIH/IMB. The ABC intervention package included: (1) clinical training and monthly mentorship, (2) provision of essential equipment and supplies, and (3) a district-wide QI strategy to promote data-use and peer-to-peer learning. A full-time ABC QI Advisor (who was an expert nurse/midwife) was hired as a clinical mentor and QI coach that was responsible for coordinating ABC activities in each participating hospital catchment area. Clinical trainings for neonatal unit staff at the hospital included an on-site five-day training on the national neonatal protocol that was conducted at the very beginning of ABC implementation and a two-day on-site training on Managing Infants with Feeding Difficulties [20] in the NCU to improve breastfeeding that was conducted around the end of the first year of implementation. Trainings included a combination of didactic lectures and hands-on mentored practice during the training using mannequins and direct practice with patients. At the hospital level, trainings targeted nurses, midwives, doctors, and anesthetists who had regular interaction with newborns primarily in the NCU, maternity, and emergency units. The QI Advisor then provided regular mentorship, at least one day per month, for the hospital maternity and NCU as well as all health centers in the hospital catchment area. The package of essential equipment provided was based on a baseline assessment of gaps in standards for newborn care equipment and supplies. Each hospital NCU received equipment based on their needs, most often containing: radiant warmers, Continuous Positive Airway Pressure (CPAP) machines, vital sign monitors, neonatal resuscitation equipment, and electric syringe pumps. Quarterly Learning Collaborative Sessions (LCS) were organized to accelerate uptake of QI through promotion of peer-to-peer learning and data use. The first LCS was organized in the second quarter of ABC intervention, following the initial clinical trainings and mentorship visits. Through the quarterly LCS and monthly QI coaching visits from the QI advisors, facility QI teams learned QI methods to identify gaps in care, experimentation of QI change ideas to improve processes and systems, PDSA cycles, and QI measurement. The LCS were attended by district and health facility leadership, interdisciplinary facility QI teams comprised of nurses/midwives, doctors, supervisors, and monitoring and evaluation staff, and PIH/IMB. In between LCS, facility QI teams were expected to meet monthly. The full model of ABC implementation is described in Fig. 2.

The ABC project instituted multiple strategies to ensure intervention fidelity. ABC QI Advisors submitted a mentor daily log to keep track of mentorship and QI coaching activities. On a monthly basis, the ABC QI Advisor and hospital monitoring and evaluation team made a joint plan of activities which was submitted to the project managers and the hospital leadership. Additionally, there were quarterly check-in meetings with ABC QI Advisors, quarterly reports of ongoing ABC activities to all stakeholders (hospital leadership, district leadership, and the MOH), regular field visits to the ABC sites by a project manager, and quarterly data validation to ensure the quality of data used for evaluation.

The aim of ABC project was to decrease neonatal mortality by 30 % within 18 months of the intervention. This was a pre-post quasi-experimental study conducted in seven rural hospitals that received ABC interventions between 2017 and 2019. Roll-out of ABC occurred in two phases, with Phase I (Gakenke and Rulindo Districts) in June-July 2017 and Phase II (Gisagara and Rusizi Districts) in October-November 2017.

The evaluation of the study included all seven NCUs in the seven public hospitals that received ABC intervention between July 2017 and September 2019. Data were collected through facility audits, mentorship reports, provider surveys, and the national Health Management Information System (HMIS).

Facility audits were conducted by ABC QI Advisors at each hospital to evaluate the number of NCUs staffs and baseline status of essential neonatal care medical equipment. The facility audits were reported in Android tablets using KoboCollect software and conducted prior to the implementation of ABC activities and repeated at the completion of ABC.

The ABC QI Advisor reported (using the KoboCollect software on the Android tablet) all the clinical mentorship and QI coaching visits provided to hospital NCU throughout the course of ABC implementation. These were reported as the average number of mentorship and QI coaching visits reported by ABC QI Advisors to NCUs per month.

We adapted a self-report survey of QI skills and confidence from a Zimbabwe learning collaborative assessment tool [21]. The self-report survey was administered to NCU health care providers at the start of the first and fourth LCS. Participants were asked to rate their own knowledge of different tools used for QI project design and planning (process mapping, root cause analysis, aim setting, prioritization, using PDSA cycles, change ideas development and testing, conducting QI meetings and using QI statistics). Participants rated their knowledge on a Likert scale from 0 (I do not know what this tool is) to 4 (I can teach others this method). At the fourth LCS, we adapted a tool to harvest successful and implemented QI change ideas [17].

The number, reasons, and outcomes of NCU admissions were obtained from HMIS. The data from HMIS were validated through register review by well-trained data officers from PIH/IMB in collaboration with MOH facility teams. HMIS data for Phase I sites were collected from: April – June 2017 (baseline) and April – June 2019 (endline). HMIS data for Phase II sites were collected from: July – September 2017 (baseline) and July – September 2019 (endline).

We analyzed (1) facility characteristics of NCUs before and after ABC intervention, (2) clinical mentorship and QI coaching in NCUs throughout ABC intervention, (3) QI projects and change ideas tested during ABC intervention, (4) QI skills and confidence among NCU providers, and (5) neonatal unit admissions, morbidity, and mortality. For each hospital, we reported the average number of clinical mentorship and QI coaching visits conducted by the ABC mentor in the NCU per month during the 18-month intervention.

QI project change ideas were coded by the study team into thirteen common themes for areas of improvements and we reported the number and percentages of NCUs that implemented each change idea. In addition, we summarized individual scores obtained by NCU providers on their knowledge and confidence in each QI skill. Finally, we compared the proportion of deaths among all neonatal admissions in all seven NCUs from baseline to endline. We used median and interquartile ranges (IQR) to report continuous variables and frequencies, and percentages for categorical variables. Changes from baseline to endline were assessed using Wilcoxon signed-rank test and Chi-squared tests for continuous and categorical variables, respectively.

The study was approved by the Rwanda National Ethics Committee and the Rwandan Ministry of Health provided approval for access to HMIS data. For individual level data collection during LCS surveys, participants provided written informed consent.