Improving neonatal health in low resource settings using mobile health technology
To attain the Sustainable Development Goal of ending preventable new-born deaths and reducing neonatal deaths to at least 12 per 1000 births, urgent attention is required in low- and middle-income countries which contribute the most to the global burden of neonatal deaths. Mobile health (mHealth) can potentially improve neonatal health outcomes. We designed and implemented an mHealth intervention that provided easy access to neonatal health guidelines for clinical decision-making for health workers in the Eastern Region of Ghana. We evaluated the impact of the intervention on institutional neonatal mortality and investigated the ‘how and why’ of the observed intervention effect. Sixteen districts in the Eastern Region were randomized into a 2-arm cluster-randomized controlled trial (8 intervention and 8 control clusters per arm) that assessed the impact of the intervention on neonatal mortality over an 18-month period. To understand the possible explanatory mechanism for the observed intervention effect, three sub-studies were undertaken. Firstly, we assessed utilization of the intervention and measured the correlation between the requests made and the incidence of deliveries and neonatal morbidities in the intervention clusters. Secondly, we assessed health worker adherence to neonatal protocols before and during the trial period using in-patient clinical records. Thirdly, to understand how and why the intervention was utilized as observed, we performed a single case study with each cluster as an embedded sub-unit of analysis using key informant interviews and focus group discussions with the intervention users, and manually analysed the data for themes. This thesis showed that health workers readily used the intervention to access neonatal health guidelines. The use of the intervention however, declined over time due to individual health worker, organizational, and technological factors as well as client perception of health worker intervention usage. During the trial period, there was a raise in neonatal deaths in both arms study arms. The odds of neonatal death was 2.09 (95% CI (1∙00;4∙38); p=0∙051) times higher in the intervention arm compared to the control arm (adjusted odds ratio). The correlation between the number of protocol requests and the number of deliveries per intervention cluster was 0∙71 (p=0∙05). The higher odds of neonatal death in the intervention arm is possibly due to unmeasured and unadjusted confounding due to limitations in the data structure of the national health database of Ghana, unintended use of the intervention and problems with births and deaths registration at the study sites. Many other neonatal health improvement programmes (unrelated to our mHealth intervention) were observed in the control arm compared to the intervention arm during the trial period. Adherence to neonatal health protocols improved in both study arms and this may be related to these other neonatal improvement programmes that took place particularly in the control arm clusters. Technological factors alone are unlikely to influence outcomes. This thesis highlights the importance of validating successful programmes and interventions in settings where they are to be implemented. Harmonized rather than fragmented efforts are needed to scale up mHealth interventions that have been proven to be effective in a given context.