The Solomon Islands is an archipelago of approximately 1000 islands. With an estimated population of only 700,000, much of the country is remote and sparsely populated, with travel occurring largely by boat and air. Out of an estimated 3500-4000 annual deaths, nearly 75% occur in the community and therefore do not receive a medical certification of cause of death (MCCOD). In order to explore the causes of these deaths, automated verbal autopsies (VAs) were introduced to the Solomon Islands through the University of Melbourne Data for Health (UoM D4H) initiative in 2016. VA was initially implemented in Guadalcanal and Western provinces and was scaled up to national level by 2018.
Data for Health (D4H) has been working with Solomon Islands Ministry of Health and Medical Services (MHMS) to implement an automated SmartVA system. This system involves the use of tablet-based digital forms for data collection at hospitals, for deaths on arrival, and Area Health Centres (AHCs) for community deaths. Additionally, SmartVA was implemented at some Rural Health Clinics (RHCs) in the Malaita and Choiseul provinces. This system involves the use of tablet-based digital forms for data collection at hospitals, for deaths on arrival, and Area Health Centres (AHCs) for community deaths.
This system involves the use of tablet-based digital forms for data collection at hospitals, for deaths on arrival, and Area Health Centres (AHCs) for community deaths. Additionally, SmartVA was implemented at some Rural Health Clinics (RHCs) in the Malaita and Choiseul provinces. This approach is generally considered best practice as it minimises errors in data and the “double handling” associated with paper-based data collection. However, in the Solomon Islands, country staff and health workers rolling out automated tablet-based Smart VA have been met with significant geographical, technological and human resource challenges.
Due to difficulties associated with transporting tablets for repairs when damaged, there have been frequent delays resulting in a missing of deaths which were otherwise eligible to undergo VA interviews. It is also difficult to support the use of tablets in health facilities below the AHC level due to inconsistencies in power supply, mobile networks and internet coverage. These smaller facilities also do not have many deaths occurring annually in the catchment area, making use of tablets for verbal autopsy uneconomical. Instead, AHC nurses are trained to conduct VA interviews on supervisory visits to rural health centres and Nurse Aid Posts (NAPs). Due to frequent difficulties in contacting next-of-kin on these visits, some AHC nurses use local nurses as secondary informants for SmartVA interviews. Due to long recall times and respondent nurses’ lack of complete information on circumstances surrounding community deaths, VAs submitted from these sources often contain errors or missing information, which increases errors and uncertainty in cause of death estimates.
UOM D4H staff worked with MHMS to investigate these issues and identify potential solutions. They observed that the nurses at rural health facilities often kept their own notes on deceased community members to aid their recall when acting as informants during AHC nurses’ supervisory visits. While note-taking appeared to be common, it was unclear whether this practice aided the provision of more accurate and complete data. MHMS and the UOM D4H team decided to formalise the practice through the trialling of a complementary paper-based VA system, with the primary objective of improving the quality of data collected and the secondary objective of increasing coverage. Nurses at lower-level health facilities are typically embedded within their communities, including attendance at functions such as funerals. They are also well placed to identify a “best respondent”, the person best placed to provide accurate information on the deceased and the circumstances surrounding their death. With this in mind, UOM D4H staff developed an easy-to-carry paper-based template to act as a helpful prompt for community nurses to collect VAs close to the time of death.
The chosen sites for piloting a complementary paper-based verbal autopsy data collection method.
Local team members selected the sites for the paper-based VA pilot, considering factors including feasibility of supervision and relative ease of access. In total, nine clinics within Gaudalcanal province were chosen as pilot sites: Marara (AHC), Tamboko (NAP), Kohimarara (NAP), Visale (RHC), Selwyn College (NAP), Lambi (RHC), Lunga (NAP), Numbu (NAP) and New Tenabuti (RHC).
In designing the paper VA form, UOM D4H staff considered how it could be as user-friendly as possible. “We recognised that what we were designing was essentially a quality improvement process for something already happening,” said Buddhika Mahesh, UOM D4H Technical Advisor. “We wanted to ensure that we didn’t create a complicated or burdensome process for those collecting the data.” The UOM D4H team worked with local health officials to develop and trial a prototype form at two local health facilities. Nursing staff were asked for feedback on the instructions provided, the wording of questions and the layout of the form. This information was fed into the final design. Thought was also given to how some of the functionality of the tablet-based form could be incorporated into the paper version. With the tablet version’s “skip patterns” in mind – automatic skipping of irrelevant questions – the team developed three separate paper forms for adults, children and neonates, to reduce the likelihood of forms being filled out incorrectly or left incomplete.
We recognised that what we were designing was essentially a quality improvement process for something already happening. We wanted to ensure that we didn’t create a complicated or burdensome process for those collecting the data.
Buddhika Mahesh, UoM D4H Technical Advisor
As a solution to the difficulty of regularly transporting paper-based forms back to a central location, the team arranged for the forms to be included in an existing “monthly return” of official documents to Honiara. Supervisors were assigned to enter the data of the completed paper-based VA questionnaires into the tablet devices and to upload into an electronic database. Training was provided around rules developed to resolve any data inconsistencies when entering data form the forms.
Distribution of paper-based VA forms has now been completed in five of the nine pilot sites, with the remaining four on hold due to COVID-19.
As of early November, 19 paper-based VAs have been completed and analysed, with 18 having been assigned a specific cause of death.Once a larger number of paper-based forms have been completed, VAs will be added to the electronic database and analysed using SmartVA-Analyse software. The team will compare this data with data from the fully automated process to assess whether the paper-based approach is as successful at assigning a cause of death. Paper VA data and the digitised VAs derived from them will both be reviewed for data entry errors to establish whether the hybrid system is able to maintain good data quality.
It is hoped that this system where data are first recorded on paper and then entered into a digital questionnaire, with its associated error checks, constraints and automated skips will mitigate the disadvantages in data quality usually associated with paper-based VAs. “It’s early days, but these initial results are promising,” said Hafiz Chowdhury, UOM D4H Technical Advisor. “We are already starting to see VAs from areas where we previously lacked data. If we become confident that this process produces good quality VA data, we will then look at scaling up the initiative to whole provinces.”
The UOM D4H team meets fortnightly with the ministry’s Chief Medical Statistician and CRVS Coordinator to monitor the progress of the trial and discuss any potential issues. So far, the feedback is positive, with local nursing and data entry staff reporting that the paper-based forms are easy to use. UOM D4H staff believe this speaks to the value of adapting approaches to the local context. “While digital VA is still best practice for VA data collection wherever practicable, I think our approach shows the importance of flexibility and responsiveness,” said Matthew Reeve, UOM D4H Senior Technical Advisor. “We hope that, until the country is ready to scale up to a fully digital process, this complementary paper-based system will be able to support accurate data collection that can inform effective health planning.”
I think our approach shows the importance of flexibility and responsiveness. We hope that, until the country is ready to scale up to a fully digital process, this complementary paper-based system will be able to support accurate data collection that can inform effective health planning.
Matthew Reeve, UoM D4H Senior Technical Advisor