I'm a physician specializing in infectious diseases and medical microbiology. I split my research between cough analysis and using DNA sequencing to study tuberculosis transmission. My interest in cough measurement began during my postdoc work in Madagascar, focusing on tuberculosis transmission. We saw widespread coughing in remote communities with limited access to care and wondered if technology could identify households at higher risk for active TB. Initially, the idea was to use syndromic surveillance of cough sounds to target high-risk households. Privacy and ethical concerns prevented this project, but it evolved into various collaborations with Hyfe.
Watch the full interview here, or scroll down to continue reading.
The bigger trial is called "Making Cough Count for Tuberculosis." It has two main components:
The goal is to develop cough as a biomarker that can be used not only for diagnosing diseases but also for supporting clinical management throughout the course of the disease.
Primarily scientists and clinicians working in high TB-burden areas with limited healthcare resources. The cough data can help in triaging patients and improving clinical management by identifying those who need more urgent care. The real value is that it’s highly accessible and virtually cost-neutral because once the model is developed it can be made freely accessible.
We use several vital signs and different markers on a longitudinal basis to care for many diseases. In the end, cough could become one of these metrics—something clinicians use to ensure their patient is fine, the treatment is working, and there is no exacerbation. The challenge in transitioning to clinical use is to increase our understanding of what meaningful clinical signals are in cough trends, because it’s a lot of data. We have good tools, we can generate reliable data, and we see value in long-term clinical management, but we need to find the meaningful and usable signals in that dataset so the clinical team knows what to do with the information.
Wearables could be a better option in some situations. They might be more comfortable to wear continuously compared to phones, which some users might be hesitant to keep with them at all times due to security reasons. Wearables offer practical advantages and can fill gaps in data collection, making monitoring more consistent. However, it's about having various tools for different contexts rather than one superior device.
There are two components here: improving the quality of data and enhancing the prediction potential of the model. To improve data quality, we could use an accelerometer to ensure the patient is wearing the device. This would help us determine if the absence of data is due to the absence of cough or the device not being used.
Additionally, what other biomarkers could we include in the models? The field of remote patient monitoring is rapidly evolving, with metrics like temperature, breathing rate, and step count being used. We haven't made significant progress in combining these with cough data. This is an area we’d like to explore, but I don’t have data to confirm that cough should be combined with specific metrics for particular clinical applications. There’s certainly room to grow in this area.
The true answer is, I don’t know. As clinicians, we ask patients questions like, “Is your cough productive? Is it increasing or decreasing?” but we don’t have a differential way of analyzing these answers in clinical practice. When we use these data points as patient-reported, they don’t lead to a stratified approach to managing patients. Even if we could get this information more reliably, I’m not sure a clinician would make a different call based on it.
For longitudinal monitoring, such as in chronic conditions like COPD, having data on whether a patient’s cough is becoming more productive could indicate an exacerbation requiring early intervention. In these cases, measuring these additional data points objectively and longitudinally might help us act earlier and achieve better patient outcomes.
I see cough monitoring reaching a crossroad. Are we going to keep pushing and improving the technology to get more reliable data without addressing the other component: what we do with the data? If we provide healthcare workers with comprehensive, 24/7 cough time series for a year, but it leads to no action, it’s not changing patient management. I'd like to see more evidence of actionable information from the cough signal that can have a clinical impact. We need to start measuring impact in terms of how it benefits the patient, rather than just improving the quality of the data we generate.
- Q&A with Alex Zimmer, McGill University
- Q&A with Dr. Jane Reynolds, University of Montana
- Q&A with Dr. Marc Judson
- Q&A with Dr Dominic Sykes- Q&A with Dr. Emily Henkle
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Non eget pharetra nibh mi, neque, purus.