Community health workers trained to conduct verbal autopsies provide better mortality measures than existing surveillance: Results from a cross-sectional study in rural western Uganda

Background: In much of sub-Saharan Africa, health facilities serve as the primary source of routine vital statistics. These passive surveillance systems, however, are plagued by infrequent and unreliable reporting and do not capture events that occur outside of the formal health sector. Verbal autopsies (VA) have been utilized to estimate the burden and causes of mortality where civil registration and vital statistics systems are weak, but VAs have not been widely employed in national surveillance systems. In response, we trained lay community health workers (CHW) in a rural sub-county of western Uganda to conduct VA interviews in order to assess the feasibility of leveraging CHW to measure the burden of disease in resource limited settings. Methods and findings: Trained CHWs conducted a cross-sectional survey of the 36 villages comprising the Bugoye sub-county to identify all deaths occurring in the prior year. The sub county has an estimated population of 50,249, approximately one-quarter of whom are children under 5 years of age (25.3%). When an eligible death was reported, CHWs administered a WHO 2014 VA questionnaire, the results of which were analyzed using the InterVA-4 tool. To compare the findings of the CHW survey to existing surveillance systems, study staff reviewed inpatient registers from neighboring referral health facilities in an attempt to match recorded deaths to those identified by the survey. Overall, CHWs conducted high quality VA interviews on direct observation, identifying 230 deaths that occurred within the sub-county, including 77 (33.5%) among children under five years of age. More than half of the deaths (123 of 230, 53.5%) were reported to have occurred outside a health facility and thus would not be captured by passive surveillance. More than two-thirds (73 of 107, 68.2%) of facility deaths took place in one of three nearby hospitals, yet only 35 (47.9%) were identified on our review of inpatient registers. Consistent with previous VA studies, the leading causes of death among children under five years of age were malaria (19.5%), prematurity (19.5%), and neonatal pneumonia (15.6%). while among adults, HIV/AIDS-related deaths illness (13.6%), pulmonary tuberculosis (11.4%) and malaria (8.6%) were the leading causes of death. No child deaths identified from inpatient registers listed HIV/AIDS as a cause of death despite 8 deaths (10.4%) attributed to HIV/AIDS as determined by VA. Conclusions: Lay CHWs are able to conduct high quality VA interviews to capture critical information that can be analyzed using standard methodologies to provide a more complete estimate of the burden and causes of mortality. Similar approaches can be scaled to improve the measurement of vital statistics in order to facilitate appropriate public health interventions in rural areas of sub-Saharan Africa. The study was conducted in Bugoye sub-county in the Kasese District of western Uganda (0o 18’ North, 30o 5’ East). The sub-county is comprised of five parishes and 36 villages spanning an area of 55 km2 with an estimated population of 50,249, approximately one-quarter of whom are children under 5 years of age (25.3%) [23]. There are nearly 7,000 households in the sub-county, with an average size of 7.3 persons per household. The terrain in the sub-county is rugged and highly varied. Village elevations range from 1,100m along the river basins to upwards of 1,800m near the western border with the Rwenzori National Park. The area is predominantly rural and most residents work as subsistence farmers. Road access in many villages is limited, especially during the rainy seasons. The main means of transport is via motorcycle taxi, although some villages can only be reached on foot. Although no sub-county specific data is available, it is assumed that like much of Uganda, malaria, along with respiratory illnesses and diarrheal diseases, accounts for the majority of deaths in children under five years of age. Local adult mortality statistics, including those describing maternal mortality, are also limited to generalizations drawn from national surveys. There are seven public health facilities distributed throughout Bugoye sub-county. These include five Level II Health Centers staffed by nurses and midwives, and two Level III Health Centers, which are led by a clinical officer, and include inpatient and labor and delivery wards, in addition to limited laboratory services. The nearest referral hospital is Kilembe Mines Hospital, which is a private not-for-profit facility located approximately 25 km from the sub-county. Public transport to Kilembe Mines Hospital typically takes 45 minutes at costs that can represent the weekly income for a typical subsistence farmer. This was a population-based, cross-sectional study of Bugoye sub-county to assess the feasibility of leveraging CHW to measure the burden of disease in resource-limited settings using a validated WHO 2014 verbal autopsy tool. Trained VHTs surveyed all households in their respective areas of responsibility to identify those households where a death occurred in the one-year period between January 1, 2016 and December 31, 2016. Verbal autopsy results were analyzed with the InterVA model to estimate the cause of death [24, 25]. Results of the survey were compared to routinely collected records from the Kasese District Health Office and clinical records from the nearest referral centers. One VHT member with at least one year of secondary education, and competent in both reading and writing, was identified by the VHT Coordinator at BHCIII from each of the 36 villages of Bugoye sub county. The thirty-six VHTs were trained for one week at Bugoye Health Center III (BHC) in the conduct of verbal autopsy interviews, conducted in February 2018 and led by the first author. During the training, Trainers reviewed all elements of the questionnaire, as well as the flow, meaning, and purpose of each question with the participants The VHTs also learned about the cardinal signs and symptoms, Description of the different diseases in the local language, community entry skills among others, the Techniques utilized in the training sessions included lectures, group discussions, practical exercises, and role play. The trainers conducted quality assurance exercises including direct observation of VA interviews and inspection of completed questionnaires to confirm adherence to study protocols and ensure the completeness and accuracy of the data. Approximately 20% of the VHT interviews were directly observed by the trainers. In rural settings of Uganda, the village council chairman (LC) and respective VHT are generally invited to attend every burial ceremony in their village. Using this information, the VHTs in collaboration with the LCs, identified all households where a death occurred in the preceding twelve months. A next of kin or caretaker who was present with the deceased was identified and an appointment made for the interview. If a household had multiple deaths, data was collected on each individual death in that household using a separate survey. For the households which had reported a death and an appointment had been made, a trained VHT of that respective village conducted a verbal autopsy interview using a WHO 2014 VA questionnaire with the primary caregiver. WHO 2014 VA questionnaires [24] were used to collect data on the identified deaths. The questionnaire captures information relevant to assessment of causes of death and the context including demographic information of the deceased, medical history, general signs, symptoms, risk factors, service utilization and reported cause of death. This tool also includes an open narrative text field. In an attempt to capture information on deaths that occurred in referral health facilities outside of the sub-county, we also reviewed the records of three health facilities where the majority of the facility deaths were reported to have occurred. These included Kilembe Mines Hospital, Kagando Hospital, and St. Paul’s Level IV Health Center. Using basic demographic data (i.e. age, sex, and village of residence) we attempted to identify any deaths from Bugoye sub-county and link them to those identified in the VHT survey. Data from VA interviews was assessed for quality and extracted from the questionnaire into an EpiData database (EpiData Association, Odense, Denmark). The results were analyzed using the InterVA4 (www.interva.net) to estimate the cause of death from verbal autopsy [25]. The InterVA-4 computer program was chosen because it presents a relatively simple, fast, cheap, very accessible, do not require training of physicians to review and ideal for determining cause of death on a large scale. Cause. Specific mortality fractions were determined as the proportion of all deaths that were attributable to any specific cause of death. We stratified mortality analysis by age. All statistical analysis was carried out using Stata 14 (College Station, TX). Ethical approval of the study was provided by the institutional review committee of the Mbarara University of Science and Technology and the Uganda National Council for Science and Technology. A Meeting was held with the DHO, sub county leaders and LC1 chairmen and verbal ascent was obtained on behalf of the community after the researcher gave them a presentation about the study, In the villages the VHTs consented the participants by briefing them on the objectives of the study and procedure, benefits for participating, what happens in case one does not participate, contacts in case of a problem arising from the study among others. After the brief, participants who verbally consented were requested to sign or put a thumb print on the consent form. All the participants consented.