Digital Tools for Malaria Management

VectorCam - a handheld tool for rapid morphological identification of mosquito species for community-based malaria vector surveillance: A Summative Usability Assessment

JMIR Human Factors, 2024 

S. Dasari et.al

VectorCam, a low-cost, artificial intelligence (AI)-powered ‘virtual’ entomologist can facilitate community-based vector surveillance. With a two-step image recognition process, the tool uses deep Convolutional Neural Networks (CNNs) to identify species known to the system as well as novel specimens with high identification accuracy. This review describes the development of the tool and its algorithm.

“VectorCam serves as a promising technology that can transcend barriers in traditional vector surveillance through task-shifting malaria prevention efforts in rural Africa and beyond.”

Key Takeaways:

• This preprint describes a study that aimed to assess the effectiveness, efficiency, and user satisfaction with VectorCam by observing and surveying 20 Village Health Team workers demonstrating the tool in rural Uganda.
• A heuristic analysis and cognitive walk-through helped identify critical sub-tasks. These tasks, such as loading the mosquitos on the tray, placing the smartphone in the box, entering information on the VectorCam app, imaging the mosquito, and subsequently off-loading mosquitoes, were used to assess effectiveness and efficiency.
• Efficiency was assessed by how long it took the “Imager” and “Loader” to do their tasks, with each mosquito being imaged in an average of 56.1 seconds and being loaded onto the tray in an average of 55.7 seconds.
• Similarly, investigators assessed effectiveness by calculating the error rate for each critical task. “Imagers” had an error rate of 16.85% and “Loaders” had an error rate of 11.2%.
• The system usability score (SUS) was 70.25 on average (a score >68 meant positive usability), with several participants agreeing they needed “to learn a lot of things” before using the system independently. 

Comment from the Center for Global Digital Health Innovation

AI-based, smartphone-friendly tools such as VectorCam demonstrate the promise of low-cost and effective solutions for vector identification and subsequent control, especially in low-resource settings. Relying on trained entomologists takes significant time, hampering timely interventions in high-vector density areas. Following tool enhancements, VectorCam is undergoing a Randomized Controlled Trial in multiple districts in Uganda and Mozambique, to assess the effectiveness of community-based vector surveillance, by leveraging community health workers(CHWs)  for collection and identification of local mosquito species density and diversity. 

Advances and challenges in automated malaria diagnosis using digital microscopy imaging with artificial intelligence tools: A review

Frontiers in Microbiology, 2022

C.R. Maturana et. al

In this study, the authors discuss the potential of digital tools for malaria diagnosis including computer vision, traditional imaging methods, and deep learning methods. They discuss the pros and cons of various existing malarial parasite detection and novel techniques such as microscope automation and convolutional neural networks (CNN) based image detection, which could make malaria detection faster, especially in low-resource settings.

“A non-precise diagnosis or treatment due to the low availability of resources is a serious issue in endemic areas. Consequently, the implementation of new and affordable diagnostic imaging techniques could help solve this problem.”

Key Takeaways:

• Smartphone integration with AI, such as deep learning detection-based models, could aid in identifying red blood cells infected with malarial parasites, allowing the gold-standard level of malaria detection (microscopic visualization) more cost-effectively in low-resource settings.
• Convolutional neural networks for parasite detection require large datasets with annotations as input to be trained. The results are validated and then the model is tested with data. 
• As such, there is a need to collect labeled digital images of malarial specimens to create a training database for optimal detection models. In low-income settings, these images can be captured using smartphone cameras as opposed to expensive microscope-integrated cameras.

Comment from the Center for Global Digital Health Innovation

The study highlights several advances in technology to aid in malarial parasite detection, such as smartphone-based image acquisition to microscopy automation and links to AI-based image detection models. Further research could assess their diagnostic performance and effectiveness in resource-poor settings. They must also be adapted to local health system structures and policy and regulatory requirements, undergo operational research, and complete quality assurance.

Reducing Stock-Outs of Life-Saving Malaria Commodities Using Mobile Phone Text-Messaging: SMS for Life Study in Kenya

PLOS ONE, 2013

S. Githinji et. al

This study aimed to evaluate the effectiveness of SMS-based reporting of artemether-lumefantrine (AL) and rapid diagnostic tests (RDT) stock levels on the number of health facility stockouts in 87 public health facilities in Kenya over 26 weeks. 

“The high reporting and low formatting error rates observed in our study are consistent with 95% response rates and 7% error rates reported previously in Tanzania, thus confirming the simplicity of data reporting via text-messaging from the peripheral health facilities.”

Key Takeaways:

• The system was composed of two units: 1) a web-based reporting tool accessible to District Health Management Team managers to identify stock-out alerts, and, 2) an SMS management tool that requested stock-level data through free weekly messages and reminders, with airtime incentives for timely reporting. 
• The average response rate was 97.1% with most facilities (85.6%) responding with their stock information within 24 hours. The district managers accessed the web-based reporting system 1037 times, in 82% of the study weeks, with an average of 8 logins per week.
• The SMS data was validated by health facility visits in the first and last month. They correlated the data with manual stock counts and compared delivery notes of stocks received by the health facilities. Overall, 79.1 % of the stock parameters were reported accurately.
• District managers responded to 44% of the AL stock-out alerts and another 43% were resolved by routine delivery of AL to the health facilities. RDT stock-outs were managed by peripheral redistribution and no resupply was made in 4 out of the 5 study districts.

Comment from the Center for Global Digital Health Innovation

This study demonstrates the success of SMS-based reporting, with high response rates and low error rates in the reported data. Future implementation research assessing SMS-based reporting tools can study reporting uptake without incentives. Providing data in real-time to decision-makers and enabling managers to take corrective action can save lives as long as other supporting factors, such as a consistent supply of resources, are in place. 

Use of a health worker-targeted smartphone app to support quality malaria RDT implementation in Busia County, Kenya: A feasibility and acceptability study
PLOS ONE, 2024
M. Skjefte et. al
The study aimed to identify the feasibility and acceptability of a smartphone malaria rapid diagnostic test (mRDT) reader application through a baseline and endline survey, as well as metadata analysis of app usage by 200 public-sector Community Health Volunteers and 23 private clinic healthcare workers in Busia County, Kenya.  

“With access to timely data, HealthPulse has potential as a malaria surveillance tool, enabling identification of malaria hotspots to further aid resource allocation and decision-making, particularly in settings that have not previously been captured in health management information systems.”

Key Takeaways: 

• RDTs are a cost-effective and reliable tool for malaria case identification, but the results are diminished when the RDT is incorrectly administered or interpreted by healthcare workers.
• This mHealth tool aimed to address these hurdles by guiding health workers on proper RDT use, anti-malarial dispensing support, as well as monitoring drug and test kit stock levels. The end-line survey demonstrated an improvement in the knowledge gap.
• The implementation was feasible as the app is compatible with several phone brands, and RDT kit brands and can work without an internet connection.
• The digital tool was also acceptable to the health workers with 99.5% of CHVs and 100% of the private clinic HCWs saying they found the app useful.

Comment from the Center for Global Digital Health Innovation 
The study findings showcase the potential of mHealth tools to support healthcare worker workflows and aid in malaria case identification. The tool was well accepted by healthcare workers and proved feasible to deploy. Future research could assess its potential as a quality improvement tool by allowing supervision and automated AI-based mRDT interpretation. The data collected could also help identify malaria hotspots for rapid control interventions.

What incentives encourage local communities to collect and upload mosquito sound data by using smartphones? A mixed methods study in Tanzania
Global Health Research and Policy 2023
R. Dam et. al
Mosquito species have unique acoustic signatures generated by the frequency of flight tones. This article describes the findings of a multi-site mixed-methods study conducted in rural Tanzania that aimed to assess what incentives encouraged the adoption of a digital tool to recognize mosquitoes by sound signatures. The HumBug sensor can turn any Android smartphone into a mosquito sensor through an app known as MozzWear. The sensor was trained on acoustic data from several wild and laboratory-cultured mosquitoes to develop Bayesian convolutional neural networks, which aid in identifying acoustic signatures. 

“Our findings demonstrate the study communities’ desire to learn more about the mosquitoes that are transmitting malaria… a very positive reaffirmation of how a ‘bottom-up’ approach could significantly impact malaria transmission.”

Key Takeaways:

• Digital tools such as the HumBug sensor can help overcome several shortcomings of traditional surveillance methods—time-consuming, expensive, subject to collector bias, and risky for human collectors.
• The study aimed to assess what motivated the study participants, (n=37 in each of the four villages), to adopt the surveillance tool, incentivizing them through different strategies in each of the four villages: monetary incentives (airtime scratch cards), SMS reminders, both monetary incentives and SMS reminders and no incentives(comparison group)
• Community participants in an initial qualitative phase said they sought to use the tool to be a citizen scientist and learn about the mosquitoes that were causing harm, and to protect themselves and their family.
• Next, a quantitative assessment found significant differences in the level of adoption of the survey tool, with use significantly less in the intervention groups compared to the control.
• Finally, in a feedback survey, 62.8% of the participants found the device easy to use and 98.7% sought feedback on the number of mosquitoes around their homes through the smartphone.

Comment from the Center for Global Digital Health Innovation

The study findings highlighted the importance of valuing intrinsic motivational factors when developing strategies for the effective uptake of community-based interventions. The qualitative analysis revealed that several factors affected study participant motivations. They said helping communities better understand the vectors that transmit malaria would motivate them to take greater ownership of mosquito control measures. This learning can be applied to improve the uptake of other community-led interventions.
 

Digitally managed larviciding as a cost-effective intervention for urban malaria: operational lessons from a pilot in São Tomé and Príncipe guided by the Zzapp system
Malaria Journal, 2023
A. Vigodny et. al
The study describes the use of the digital tool Zzapp to implement and monitor Larval Source Management (LSM), sharing lessons learned from a pilot on two islands in the Democratic Republic of STP, with inferences comparing findings from intervention and comparison districts. 

“Digitization facilitates all aspects of LSM operations… It eases the work of managers and fieldworkers and presents a clear and reliable picture of operations’ progress, expenditure, and outcomes, which can easily be shared with stakeholders and the community.”

Key Takeaways:

• The Zzapp system consists of two parts: an online dashboard for district managers to assign treatment areas and track progress and a GPS-based mobile app to support field workers.
• The app recorded field workers mapping water bodies in their assigned areas, conducting initial sampling to estimate baseline positivity rates, and subsequently treating water with larvicides biweekly.  
• A total of 12,788 water bodies were treated a total of 128,864 times resulting in a 52.5% reduction in malaria incidence. This is comparable to the effectiveness of long-lasting insecticidal-treated bednets (45%) coupled with Insecticide Residual Spray (18%).
• The LSM operation was also cost-effective, with an operational cost of US$0.68 per person protected (PPP) overall, $0.44 PPP in urban areas, and $1.41 PPP in rural areas. 

Comment from the Center for Global Digital Health Innovation

LSM was once the mainstay of malaria control and could achieve complete elimination of local transmission. It has since been marginalized due to operational difficulties that reduced its effectiveness. However, with rising insecticidal resistance and new digital tools like Zzapp, LSM shows renewed promise as a malaria vector control intervention. The collection of granular, accurate data in real-time facilitates ongoing monitoring and evaluation, allowing field workers to flag and rapidly address concerns. Future studies may aim to overcome limitations described in the article, such as unbalanced comparison and intervention regions. They may further explore how vector control could be further optimized by support from drones or satellite imagery.