Assessing the Effect of the Facilitated Implementation of a Co-designed Newborn Clinical Audit Tool on Overcoming the Modifiable Factors in the Feeding Practices of Low Birth Weight Newborns: a Mixed Methods Study

Abstract

Background Prematurity and its complications account for the majority of the deaths during the newborn period in sub-Saharan Africa. The majority of these deaths are from preventable causes, indicating poor quality of in-hospital care. Clinical audits are proven to be an important quality improvement intervention that enable health workers to reflect on their practice and identify and act on modifiable gaps in the care provided to the small and sick newborns. The challenge however lies in the implementation strategy used that can be flexible enough to identify the requirements of an individual setting and adapt to them. Facilitation utilises such an approach and has proved to be effective for implementing complex interventions. Study objective: To design and introduce the use of a newborn audit tool and its implementation guide to reduce the time to regain birth weight of LBW newborns through improved feeding practices in County Hospitals in Kenya. Study methods Study design: A quasi-experimental study design of the implementation science format that uses a mixed methods approach. Study methods: The study was conducted in six hospitals that are part of a Clinical Information Network. These included Pumwani Maternity Referral Hospital (PMRH), Kenyatta National Hospital (KNH), Mama Lucy Kibaki Hospital, Kiambu County Referral Hospital, Thika Level 5 Hospital and Machakos Level 5 Hospital. The study population included two arms; the health workers who were involved in the audit meetings and the low birth weight and very low birth weight newborns who were used to evaluate the outcomes. The co-design of a newborn clinical audit tool and audit implementation guide using a threestep Human Centred Design (HCD) approach. The three steps included; (1) understanding the context,

the users and the available audit tools, (2) the cognitive walkthrough where the prototype audit tool was taken through several cycles of reviewing with users on real cases in KNH and PMRH newborn units and refining it based on their feedback, and (3) usability testing where the final prototype tool and the implementation guide were tested in two high volume newborn units to determine their usability. The integrated Promoting Action on Research Implementation in Health Sciences (iPARIHS) framework was used as the conceptual framework to guide the implementation of the final audit tool and implementation guide in four hospitals using facilitation as a strategy with the intended recipients in their contextual setting. The four hospitals were randomly assigned to experimental and control arms and facilitation was used as an implementation strategy in the experiment arm. The secondary outcome of interest was the effect of facilitation on mortality of the LBW newborn. Evaluation of the effect of facilitation was done using a mixed methods approach. A controlled before and after (CBA) study was used to evaluate the effect of facilitation on improving newborn feeding practices and subsequently, reducing their time to regain birth weight. Participant observation was used to understand why and how facilitation worked or did not work. Data analysis: Clinical characteristics were summarised descriptively. The inferential statistics used a competing risk survival analysis as the univariate analysis to estimate the probability of regaining birth weight with death as the competing risk. A Cox proportional cause-specific hazard regression analysis was used for the multivariable analysis. Qualitative data were managed on NVivo 12 software and thematic analysis was used for the qualitative analysis. Results Quantitative results: I included a total of 2956 low birth weight (LBW) (1500 – 2499g) and very low birth weight (VLBW) (1000 – 1499g) newborns from both study arms throughout the study period. The cumulative incidence function (CIF) curves showed that the probability of regaining birth weight compared to death was comparable between experimental (facilitation) and control study arms (non facilitated sites) and across the study periods. There was, however, a significant difference in overall mortality between the experiment and control hospitals in the post-intervention period with the probability of death peaking at 13% (95% CI 0.1 – 0.16) in the control arm compared to 6.6% (95% CI 0.04 – 0.09) in the experiment arm. The cause-specific hazard regression demonstrated no difference in the hazard of regaining birth weight in the experiment arm compared to the control arm (HR 0.95, p = 0.75) after adjusting for all the covariates. There was, however, a significant decrease in the hazard of death among the newborns in the experiment arm compared to the control arm (HR 0.64, p = 0.019). Qualitative results: The factors supporting the implementation of the newborn clinical audit tool using facilitation included: Leadership with a dynamic mindset, ownership of the clinical audit, availability of infrastructure and the interrelatedness of the departments. The barriers included: Limited leadership support, resistance to change in how clinical audits were conducted, lack of a shared vision, power dynamics revealing team-based hierarchies, infrastructural constraints, hindrances to effective feed and fluid management and slow organizational adoption of digital technology. Conclusion Facilitation as an implementation strategy recognizes the complex, unpredictable and non-linear relationship between an innovation, the recipients and context making it an effect implementation strategy for the clinical audit. Facilitation was effective in enabling more use of audit processes. However, I observed no difference in the primary outcome but did observe a difference in mortality that likely requires further investigation.