Deep learning reveals what facial expressions mean to people in different cultures.

Cross-cultural studies of the meaning of facial expressions have largely focused on judgments of small sets of stereotypical images by small numbers of people. Here, we used large-scale data collection and machine learning to map what facial expressions convey in six countries. Using a mimicry paradigm, 5,833 participants formed facial expressions found in 4,659 naturalistic images, resulting in 423,193 participant-generated facial expressions. In their own language, participants also rated each expression in terms of 48 emotions and mental states. A deep neural network tasked with predicting the culture-specific meanings people attributed to facial movements while ignoring physical appearance and context discovered 28 distinct dimensions of facial expression, with 21 dimensions showing strong evidence of universality and the remainder showing varying degrees of cultural specificity. These results capture the underlying dimensions of the meanings of facial expressions within and across cultures in unprecedented detail.

Coverage of antenatal, intrapartum, and newborn care in 104 districts of Ethiopia: A before and after study four years after the launch of the national Community-Based Newborn Care programme.

BACKGROUND: Access to health services across the continuum of care improves maternal and newborn health outcomes. Ethiopia launched the Community-Based Newborn Care programme in 2013 to increase the coverage of antenatal care, institutional delivery, postnatal care and newborn care. The programme also introduced gentamicin and amoxicillin treatment by health extension workers for young infants with possible serious bacterial infection when referral was not possible. This study aimed to assess the extent to which the coverage of health services for mothers and their young infants increased after the initiation of the programme. METHODS: A baseline survey was conducted in October-December 2013 and a follow-up survey four years later in November-December 2017. At baseline, 10,224 households and 1,016 women who had a live birth in the 3-15 months prior to the survey were included. In the follow-up survey, 10,270 households and 1,057 women with a recent live birth were included. Women were asked about their experience of care during pregnancy, delivery and postpartum periods, as well as the treatment provided for their child's illness in the first 59 days of life. RESULTS: Between baseline and follow-up surveys the proportion of women reporting at least one antenatal care visit increased by 15 percentage points (95% CI: 10,19), four or more antenatal care visits increased by 17 percentage points (95%CI: 13,22), and institutional delivery increased by 40 percentage points (95% CI: 35,44). In contrast, the proportion of newborns with a postnatal care visit within 48 hours of birth decreased by 6 percentage points (95% CI: -10, -3) for home deliveries and by 14 percentage points (95% CI: -21, -7) for facility deliveries. The proportion of mothers reporting that their young infant with possible serious bacterial infection received amoxicillin for seven days increased by 50 percentage points (95% CI: 37,62) and gentamicin for seven days increased by 15 percentage points (95% CI: 5,25). Concurrent use of both antibiotics increased by 12 percentage points (95% CI: 4,19). CONCLUSION: The Community-Based Newborn Care programme was an ambitious initiative to enhance the access to services for pregnant women and newborns. Major improvements were seen for the number of antenatal care visits and institutional delivery, while postnatal care remained alarmingly low. Antibiotic treatment for young infants with possible serious bacterial infection increased, although most treatment did not follow national guidelines. Improving postnatal care coverage and using a simplified antibiotic regimen following recent World Health Organization guidelines could address gaps in the care provided for sick young infants.

District decision-making for health in low-income settings: a case study of the potential of public and private sector data in India and Ethiopia.

Many low- and middle-income countries have pluralistic health systems where private for-profit and not-for-profit sectors complement the public sector: data shared across sectors can provide information for local decision-making. The third article in a series of four on district decision-making for health in low-income settings, this study shows the untapped potential of existing data through documenting the nature and type of data collected by the public and private health systems, data flow and sharing, use and inter-sectoral linkages in India and Ethiopia. In two districts in each country, semi-structured interviews were conducted with administrators and data managers to understand the type of data maintained and linkages with other sectors in terms of data sharing, flow and use. We created a database of all data elements maintained at district level, categorized by form and according to the six World Health Organization health system blocks. We used content analysis to capture the type of data available for different health system levels. Data flow in the public health sectors of both counties is sequential, formal and systematic. Although multiple sources of data exist outside the public health system, there is little formal sharing of data between sectors. Though not fully operational, Ethiopia has better developed formal structures for data sharing than India. In the private and public sectors, health data in both countries are collected in all six health system categories, with greatest focus on service delivery data and limited focus on supplies, health workforce, governance and contextual information. In the Indian private sector, there is a better balance than in the public sector of data across the six categories. In both India and Ethiopia the majority of data collected relate to maternal and child health. Both countries have huge potential for increased use of health data to guide district decision-making.

Using Health Extension Workers for Monitoring Child Mortality in Real-Time: Validation against Household Survey Data in Rural Ethiopia.

BACKGROUND: Ethiopia has scaled up its community-based programs over the past decade by training and deploying health extension workers (HEWs) in rural communities throughout the country. Consequently, child mortality has declined substantially, placing Ethiopia among the few countries that have achieved the United Nations' fourth Millennium Development Goal. As Ethiopia continues its efforts, results must be assessed regularly to provide timely feedback for improvement and to generate further support for programs. More specifically the expansion of HEWs at the community level provides a unique opportunity to build a system for real-time monitoring of births and deaths, linked to a civil registration and vital statistics system that Ethiopia is also developing. We tested the accuracy and completeness of births and deaths reported by trained HEWs for monitoring child mortality over 15 -month periods. METHODS AND FINDINGS: HEWs were trained in 93 randomly selected rural kebeles in Jimma and West Hararghe zones of the Oromia region to report births and deaths over a 15-month period from January, 2012 to March, 2013. Completeness of number of births and deaths, age distribution of deaths, and accuracy of resulting under-five, infant, and neonatal mortality rates were assessed against data from a large household survey with full birth history from women aged 15-49. Although, in general HEWs, were able to accurately report events that they identified, the completeness of number of births and deaths reported over twelve-month periods was very low and variable across the two zones. Compared to household survey estimates, HEWs reported only about 30% of births and 21% of under-five deaths occurring in their communities over a twelve-month period. The under-five mortality rate was under-estimated by around 30%, infant mortality rate by 23% and neonatal mortality by 17%. HEWs reported disproportionately higher number of deaths among the very young infants than among the older children. CONCLUSION: Birth and death data reported by HEWs are not complete enough to support the monitoring of changes in childhood mortality. HEWs can significantly contribute to the success of a CRVS in Ethiopia, but cannot be relied upon as the sole source for identification of vital events. Further studies are needed to understand how to increase the level of completeness.