Mobile Clinical Decision Support for the Quality Improvement of Maternal, Neonatal and Child Health Services Delivered by Community Health Workers in Kosirai, Kenya

This dissertation presents qualitative and quantitative results from the implementation of a mobile clinical decision support (CDS) system designed to improve the quality of maternal, neonatal and child health (MNCH) services delivered by community health volunteers (CHVs) in a collaboration between the Academic Model for Providing Access to Healthcare (AMPATH) and the Kenya Ministry of Health in Kosirai Division, Nandi North District, Kenya. An initial analysis of the social and technical networks in Kosirai provided a context of the implementation and an evidence basis for the development of a new mobile technology to improve health outcomes. Software was created using an agile development framework over the course of a one year period working directly with CHVs, and was informed through direct observation, regular and informal feedback, in-depth interviews, and focus group discussions with both CHVs and the women they served. The final mobile health tool included an integrated suite of 5 Android applications, 9 electronic encounter forms with embedded decision logic, SQL algorithms on an OpenMRS server, and a customized branch of the CyanogenMod firmware distribution running on Huawei IDEOS u8150 hand sets. We further developed an offline, electronic reference library of information, education, and communication (IEC) materials that included educational websites, a medical dictionary, area maps with overlays of facilities and available clinical services, and more than 2 hours of educational videos covering 50 MNCH topics with content developed by clinicians working in local health centers and hospitals. The effectiveness of the mobile CDS services in facilitating longitudinal, community-based follow-up of pregnant women was evaluated over a 7 month period in a cluster-randomized, controlled trial of 78 CHVs between June 2013 and January 2014. The intervention group (n=39) received `Smart Forms', which integrated diverse sources of information within the workflow of a community-based clinical encounter, including data collected on mobile phones as well as longitudinal health records from the OpenMRS server, and provided CHVs with reminders to visit at-risk clients, complete specific forms, and deliver targeted IEC materials to individual clients. CHVs in the non-intervention group (n=39) were provided phones that offered identical electronic resources, but lacked the CDS features that integrated reminders and decision logic. The CDS intervention did not demonstrate a significant effect on the primary outcome of the proportion of women receiving a postnatal follow-up from CHVs within 48 hours of delivery (n=259) in either unadjusted or adjusted mixed effects logistic regression models OR 0.91 (95% CI 0.11-3.71). Qualitative research on the delivery of community-based MNCH services in Kosirai included 14 focus group discussions held with CHVs (n=24) and women (n=44) who received CHV services in the Kosirai region across three different time points over a two year period. Six primary themes regarding the delivery of MNCH services among CHVs in Kosirai emerged from open coding methods and demonstrated how the mobile CDS system enhanced the quality and accessibility of the primary care and preventive health services that CHVs were able to provide to their communities. Both community volunteers and residents reported that the mobile health intervention enabled CHVs to remain flexible in their workflows, independent of their supervisors, responsive to resident needs, and accessible in their communities. The mobile health platform was also recognized for broadly integrating community and facility data sources and promoting a model of teamwork and coordinated care by relaying information from clinicians to CHVs in the field and returning home-based follow-up data to the clinicians at the facility. Despite improvements in technical capacity, the potential effects were not realized by all CHVs and both qualitative and quantitative results demonstrated that the overall effect of the CDS system was limited by the significant variability in the services delivered by individual CHVs. Based on the results of both the quantitative and qualitative analyses, an ex post theory of how community health volunteers function within Watts-Strogatz small world social networks of their surrounding communities is proposed, and the potential effectiveness of a phone in their delivery of maternal and child health is discussed