Search

Computer-coded verbal autopsy (CCVA) algorithms predict cause of death from high-dimensional family questionnaire data (verbal autopsy) of a deceased individual, which are then aggregated to generate national and regional estimates of cause-specific mortality fractions. These estimates may be inaccurate if CCVA is trained on non-local training data different from the local population of interest. This problem is a special case of transfer learning, i.e., improving classification within a target domain (e.g., a particular population) with the classifier trained in a source-domain. Most transfer learning approaches concern individual-level (e.g., a person's) classification. Social and health scientists such as epidemiologists are often more interested with understanding etiological distributions at the population-level. The sample sizes of their data sets are typically orders of magnitude smaller than those used for common transfer learning applications like image classification, document identification, etc. We present a parsimonious hierarchical Bayesian transfer learning framework to directly estimate population-level class probabilities in a target domain, using any baseline classifier trained on source-domain, and a small labeled target-domain dataset. To address small sample sizes, we introduce a novel shrinkage prior for the transfer error rates guaranteeing that, in absence of any labeled target-domain data or when the baseline classifier is perfectly accurate, our transfer learning agrees with direct aggregation of predictions from the baseline classifier, thereby subsuming the default practice as a special case. We then extend our approach to use an ensemble of baseline classifiers producing an unified estimate. Theoretical and empirical results demonstrate how the ensemble model favors the most accurate baseline classifier. We present data analyses demonstrating the utility of our approach. [ABSTRACT FROM AUTHOR]

Background: In 2010, Ethiopia began scaling up the integrated community case management (iCCM) of childhood illness strategy throughout the country allowing health extension workers (HEWs) to treat children in rural health posts. After 2 years of iCCM scale up, utilization of HEWs remains low. Little is known about factors related to the use of health services in this setting. This research aimed to elicit perceptions and experiences of caregivers to better understand reasons for low utilization of iCCM services.Methods: A rapid ethnographic assessment was conducted in eight rural health post catchment areas in two zones: Jimma and West Hararghe. In total, 16 focus group discussions and 78 in-depth interviews were completed with mothers, fathers, HEWs and community health volunteers.Results: In spite of the HEW being a core component of iCCM, we found that the lack of availability of HEWs at the health post was one of the most common barriers to the utilization of iCCM services mentioned by caregivers. Financial and geographic challenges continue to influence caregiver decisions despite extension of free child health services in communities. Acceptability of HEWs was often low due to a perceived lack of sensitivity of HEWs and concerns about medicines given at the health post. Social networks acted both to facilitate and hinder use of HEWs. Many mothers stated a preference for using the health post, but some were unable to do so due to objections or alternative care-seeking preferences of gatekeepers, often mothers-in-law and husbands.Conclusion: Caregivers in Ethiopia face many challenges in using HEWs at the health post, potentially resulting in low demand for iCCM services. Efforts to minimize barriers to care seeking and to improve demand should be incorporated into the iCCM strategy in order to achieve reductions in child mortality and promote equity in access and child health outcomes. [ABSTRACT FROM AUTHOR]