Overview

The MHGH Lab has a demonstrated track record of success operating at the unique interface of mobile and global health software development. We have ongoing collaborations outside of the Chemistry Department at Vanderbilt (Biomedical Engineering, Computer Science, Data Science Institute, Medical Center, Institute for Global Health), as well as beyond campus (Columbia University, Lurie’s Children’s Hospital in Chicago, Oxford University, Leiden University Medical Center, the Kenya Medical Research Institute, the Macha Research Trust). Our full-stack mHealth applications have solved problems in infectious disease surveillance (e.g., malaria, HIV, COVID-19, schistosomiasis), global health logistics (e.g., personnel, resource, scheduling management), data science (e.g., creation of “data-rich” environments with data capture apps, large-scale structured and unstructured data analysis, machine learning and model development), and mobile health infrastructure. We have received research support from the National Science Foundation, the National Institutes of Health, the Tennessee Center for AIDS Research, the Burroughs Wellcome Fund, Amazon Web Services, and Google Cloud. Software from our lab has been deployed domestically in the US (Nashville, Baltimore, New York, Chicago) and globally (Kenya, Zambia, South Africa, Côte d’Ivoire, Brazil).

Connected Diagnostics

Perhaps our most advanced research initiative is centered on the concept of “connected” diagnostics. One significant issue in the field of diagnostics is that point-of-care tests are made to be inexpensive and portable, but this comes with drawbacks that can be readily addressed by mobile devices. First, visual inspection, the most common approach for determining the result of a point-of-care diagnostic test, is subjective and ambiguous. This is particularly true of tests that are designed to be semi-quantitative. Lastly, the results from these tests have no integration into electronic health records or epidemiology surveillance platforms — they rely on manual reporting of the results, which is slow and error prone.

Through a combination of novel digital image processing algorithms, app development, and cloud software design, we have addressed each of these challenges. The web app version of our software, mLab, can be accessed on any device with a web browser, and can interpret point-of-care test results for a variety of infectious diseases (e.g., malaria, COVID-19, schistosomiasis, and HIV). In addition, we have developed native mobile applications (i.e., iOS and Android) for regions where network connectivity is intermittent or non-existent.

Finally, we have begun to integrate features that more closely couple mobile devices into the diagnostic tests. Primarily, this has been through patterning assay bioreagents in the shape of barcodes directly onto the flow membrane. This allows mobile devices to natively recognize the test zone of the diagnostic test, and incorporates both standardization and security, improves quantification, and enables higher-degrees of multiplexing than what is currently attainable on available diagnostic tests.

Epidemiology Surveillance Infrastructure

Infectious disease surveillance infrastructure is often composed of fragmented systems with poor interoperability. There are software platforms for individual end-users, but these do not integrate with higher-level, national-scale surveillance programs. Meanwhile, national-scale surveillance software often fails to provide meaningful insight and can cause considerable lag between data stream updates, thus delaying critical action in the face of disease outbreaks. In low- and middle-income countries (LMICs), many of these platforms are legacy systems that were selected by governments and NGOs long ago because they were the best option available at the time, and their continued usage is hindered by their limited feature set and poor usability. In the US, the competition between systems is strong which results in improved but proprietary features that usually don’t communicate with each other. Our lab is working on improved software platforms that provide the right information to the right people at the right time. Our software in this space takes a holistic approach, integrating components for patients/end-users, field healthcare teams, infectious disease program managers, funding organizations, and government officials. This work started with the development and deployment of COVID-19 symptom checkers, which were integrated with diagnostic testing and healthcare resource management, and eventually added in a full suite of contact tracing tools. We continue to pursue opportunities in this space, agnostic of disease and geography.

mHealth Usability

With the surge in popularity in mobile applications for healthcare, there has been a growing effort to understand why certain applications see successful user adoption and others do not. We work closely with colleagues that perform focus groups to understand motivators and deterrents for using features of, or entire, mobile health applications. While we occasionally take on that role ourselves, our primary contribution to this research is in building ways to observe and analyze user behavior from within mHealth apps. Primarily to date, this has stemmed from a behind-the-scenes, automatic paradata collection library. This library allows for the analysis of user workflows within an app, grouping together users for analysis based on actions taken or outcomes within an app, and correlation between quantitative and semi-quantitative feedback.

Program Management Software

Managing a research field trial goes beyond just the science. Principal Investigators and field trial managers have to become experts in logistics planning – monitoring stocks and locations of critical inventory, understanding personnel workloads and performance, and data collection/aggregation/analysis — all while staying under a strict operating budget with a hard ceiling. This logistics management is even more challenging in LMICs, where plans must be fluid due to unexpected events and unplanned disruptions. Our group is building software that manages and sometimes automates field trial management.

Data Curation and Analysis

In LMICs, decisions are often made based on anecdotal evidence. One initiative that typically makes its way into all of our projects is the acquisition of data to drive informed decision-making. We have built custom applications for the sole purpose of collecting data, turning what was once a void into a data-rich environment. Often, these datasets are the first of their kind – naturally, we develop thorough analysis algorithms that set the standard for how data of this type can be analyzed. In certain instances, and after careful data cleaning, we openly share the data repository for various analyses by other researchers.

Standardization

Many new research fields see an initial rapid expansion where technology outpaces the ability to understand its implications. Mobile health has been no different. But as we near the end of the dawn of mHealth, two things are clear: 1) it is expanding, and certainly not a fad that will disappear, 2) it has become a bit of the “Wild West”. The US Food & Drug Administration regulates diagnostics and therapeutics in the United States, and is only now starting to regulate software applications in healthcare. Similarly on the global stage, the World Health Organization has just recently released guidelines for digital health applications. Still, the current dialogue is nascent and opportunities for standardization are ripe. To this end, we have served as contributors to multiple Target Product Profiles and protocol standardizations related to diagnostics and mobile health, including ones hosted by the World Health Organization and FIND (the global alliance for diagnostics).

Anti-tampering and Authenticity Verification

Through our many collaborations, we have learned of challenges in point-of-care diagnostics related to security, anti-tampering, and counterfeit tests. The combination of our experience in bioreagent patterning and assay development, our expertise in computer vision, and advances in secure ledgers (i.e., blockchain), allows us to build anti-tampering and anti-counterfeit features into the diagnostic assays themselves that allow users and surveillance agencies to verify their authenticity and uncompromised status. This has potential applications in regulated industries where testing is required (i.e., transportation workers), situations where there is a social or economic incentive for modifying the outcome of a diagnostic test (i.e., COVID-related access restrictions), as well as defense applications (i.e., verification of resources or personnel).

Machine learning for predictive modeling

One thing that we’ve learned is that infectious disease trajectory and outcomes can be predicted. If you have enough data describing what is currently happening in a certain place, machine learning models can identify other locations where a similar outcome will occur. Given our ongoing efforts in high-resolution infectious disease surveillance data collection, it makes sense for us to also leverage that data for predictive modeling.

Democratization of software for global health

It is widely agreed that “resources” in LMICs are harder to come by. But what exactly constitutes a resource is less clear — money, stable electricity, continuous and widespread internet connectivity, expensive lab equipment, and expert-level staff are less readily available in low-resource settings. That said, there is not a shortage of intelligent, creative, and highly-motivated people living in these locations that are passionate about improving the health of those around them. These local innovators may have an idea about how computer vision, machine learning, or mHealth apps could be used — they just might lack the time or resources to learn these skills. To this end, our lab is focusing on ways to democratize access to these tools through open-source and low-code/no-code solutions. These tools will empower local teams on the ground with customizable software, from plug-and-play modules to fully-featured applications to web-hosted application programming interfaces, to build the solutions to problems that they face.