In low- and middle-income countries, community health workers (CHWs) have been mobilized to bridge the gap in delivering healthcare to marginalized populations. However, with the rising burden of noncommunicable diseases, these workers need to rapidly expand their skills and access effective diagnostic tools to successfully avert preventable morbidity and mortality.
My research in this area has explored:
how to leverage smartphone hardware to detect blood clotting. I collaborated to create the first contactless system that can determine fluid properties using the LiDAR sensors on a modern smartphone. By targetting a coherent laser beam at a drop of blood, and characterizing the Brownian motion of the resulting speckle pattern, we can distinguish between uncoagulated and coagulated samples. This sort of mHealth tool could enable CHWs to detect blood clotting status with no attachments on their extant smartphone hardware. More details on this work can be found in our publication in UbiComp ‘22.
how to use AI models to detect cardiovascular disease before symptoms present. Typically, diagnosing Atrial Fibrillation (AF) requires having a trained cardiologist use a 12-lead ECG machine to capture and interpret an ECG during an AF episode. These experts and tools are absent in places like Nepal, that have a high cardiovascular disease burden. Using mobile, 6 lead ECGs from low cost, portable ECG devices, I built an AI system to predict whether AF has happened or will happen to a patient in a given 60-day window. This paradigm of diagnosis surpasses the current capacity of cardiologists and fits perfectly into the workflow of CHWs that may visit patients once in a two-month period. More details on this work can be found in our publication in Cardiovascular Digital Health ‘23.
what it takes to integrate mHealth and AI tools into CHWs’ workflows within country health systems. Although several mHealth and AI tools are being designed to fill diagnostic gaps, little is understood a priori about LMICs governments’ vision for such tools, and whether this vision can become a reality in the context of existing gaps in the health system. By collecting the “imaginaries” of diagnostic AI tools and juxtaposing them with the realities of creating and maintaining the tools within the extant health system, I develop a framework to assess a priori whether a conceived AI has the capacity to impact human outcomes in the desired manner. More details about this work can be found in a forthcoming publication.
Overall, I have built expertise designing and developing AI-enabled tools that amplify the critical care work that CHWs perform, with careful importance placed on how such tools could be systematically integrated into country health systems.
