 Hi, I'm Rick Nitzel, an associate professor at the University of Michigan School of Public Health. And it's my great pleasure today to talk to you about big data and our Apple hearing study. Before I do so, though, I want to acknowledge my study team at the University of Michigan, which includes Lauren Smith and Linyan Wang, as well as my colleagues at Apple. Let me begin with a quick review of what we know. According to the World Health Organization, at least 430 million people worldwide have moderate or worse hearing loss. Now we know that chronic workplace noise exposure can lead to noise-induced hearing loss. And in fact, our best estimates are that somewhere around 20% of hearing loss globally is caused by workplace noise. However, chronic environmental noise or outside of the workplace noise may also be associated with noise-induced hearing loss. Unfortunately, the data available to date are sparse and inconclusive as to whether non-occupational or environmental noise is associated with noise-induced hearing loss. So what we need to answer that question are really high resolution and long-term data on, number one, personal exposures to environmental noise as well as to headphone sound, which is a potentially important source of noise exposure that hasn't really been well-addressed in previous research. And number two, we need data on hearing impacts, including both hearing loss and tinnitus, again, from non-workplace or non-occupational noise, including environmental sound, as well as headphone sound. So it's my pleasure then to introduce the Apple Hearing Study, which was launched in November 2019 in part to address these very questions. Our study is registered at clinicaltrials.gov, and you can see the website there if you'd like to visit and learn more about the study. Now, the Apple Hearing, excuse me, now the Apple Hearing Study has a couple of different adjectives. We wanna understand typical headphones sound exposures as well as the relationship of those exposures to hearing health. Also, among participants who are Apple Watch users, we also wanna understand environmental sound exposures and the relationship of those exposures to hearing health. And then another objective is to understand how participants actually interact with their own exposure data when it's shared with them. I'll give you a quick overview of the study here. Participants who are all volunteers, this is an opt-in study, choose to download the Apple Research app, and that's actually the only way in which participants interface with the study. Now, here on the U of M side, we have access to all of the data that are collected as part of this study. Apple has built in multiple firewalls and actually can't access personally identifiable information that is shared by the participants. So that was a great concern at the start of the study, but we've designed a system that absolutely guarantees that Apple can't see anything other than de-identified data, but we as the research team and the principal investigator do have access to that information. Now, when participants enroll in the Apple Hearing Study, they're assigned randomly to one of two study arms. The first arm is the basic arm, and this is an arm in which participants can review their headphone and environmental exposures in the health app on their iPhone. The second arm, the advanced arm, participants there actually receive notifications to review their data in the health app rather than just passively or perhaps infrequently remembering to visit that data. Also, participants in this advanced arm are asked to take a survey and complete a brief tone audiometry test after they experience loud exposures as a way to assess potential temporary threshold shifts. What do our participants do? Well, there's a number of different study activities. They complete scheduled surveys at baseline in about every three months afterwards, and this provides us with information on demographics, behaviors, self-reported hearing, as well as perceptions of sound. They complete scheduled hearing tests at baseline and again about every three months afterwards. That includes pure tone audiometry delivered via the research app between 250 and 8,000 Hertz as well as a speech and noise test. They also can choose to share their headphone audio exposures and they do so continuously. This means that their iPhone will share with us information not about what they're listening to but about the durations and levels at which they're listening through their headphones. Finally, they can share environmental noise and heart rate and exercise information continuously if they have an Apple Watch. Again, we get information about the duration as well as the levels of exposure. We're faced with some very complex analytical challenges. We've got multiple, and I know this is not a term of art, but I'll call it massive data streams coming in at different temporal scales. So on the left side here, at the monthly or even yearly level, we're getting data from surveys as well as from hearing tests. On the right hand side, we have very high resolution minute or even sub minute level data including environmental sounds, headphone audio and exercise information. And we're trying to of course merge or integrate all of those things together to better understand the risk of hearing loss and tinnitus. The study also poses some opportunities but also some challenges in terms of assessing potential latency between when an exposure occurs and when the actual health impact results from that. I'll share with you now some very, very preliminary results. We have collected more than 150 million hours of environmental noise and more than 15 million hours of headphone audio exposure from our participants. We've delivered more than 50,000 pure tone audiometry tests and more than 55,000 speech and noise tests and more than 150,000 surveys. And in starting to integrate and combine these data streams, we're doing things like creating the maps that you see on the right here. These can be found on our study website. The map at the top shows environmental sound levels on average by state across the US. This is for the week of January 1st, 2020. The bottom map shows headphone listening levels by state for that same period. And what we've been able to do is actually look at changes in noise exposure, starting at that time, going forward through the pandemic lockdowns to see, for example, what were the impacts of government pandemic-related lockdowns on people's environmental and headphone audio exposures. We've also found that about 25% of our participants are exceeding the environmental exposure limit recommended by the World Health Organization. That's a 24-hour equivalent continuous average of 70 DBA. And just about the same fraction are exceeding their headphone exposure according to the World Health Organization, which has set a limit of a rolling 40-hour average of 80 DBA. About 10% of our participants have reported having a diagnosis of hearing loss from a medical provider. And about 22% of our participants are showing hearing loss via their Purton Audimetric Test delivered by our app. And that's hearing loss defined by the World Health Organization guidelines of greater than 20 decibel hearing threshold level averaged across 0.5, 1, 2, and 4 kilohertz. Again, this is nothing that we set out to do, but this figure here shows you for four different states, California, Florida, New York, and Texas, the impact of the COVID-19 pandemic lockdowns on environmental sound exposure. So in black here, you can see sort of the baseline period from January to February 2020. We have kind of a transition period in gray here where lockdowns were beginning, and then in blue, we have the post-lockdown noise exposure. So we were able to, though we didn't set out to be able to do so, able to identify that there was a three decibel drop on average across all three of these states as a result of pandemic-related behavior changes. So the long-term goals of the study, we wanna characterize environmental noise as well as headphone audio exposures. And we're working right now to identify some common exposure patterns using both supervised and unsupervised machine learning techniques. We're also, of course, looking to evaluate the association between sound exposures and hearing impacts at different temporal scales, and of course, controlling for other important risk factors like age, for example, and evaluating associations between objective and subjective measures. So one way for us to evaluate the validity of our data is to see, well, how to self-reported perceptions of hearing correlate to our measured levels, our pure tone audiometry levels and our speech and noise levels. We're also conducting validation studies, again, to verify the performance of the sensors that are being used. And of course, ultimately, we're hoping to advance personal, in other words, non-research or non-clinical use of these research tools, which may benefit individual patients who are experiencing high noise or hearing loss. So thank you so much for your attention. I do want to acknowledge the study was sponsored and funded by Apple Inc via a grant to the University of Michigan. If you'd like to learn more about the study, you can see a link to our study website in the lower right of this slide. If you have questions, of course, feel free to reach out to me. And if you're interested in learning more about the Apple Research app or even potentially participating in the study, you can see a link to that app at the bottom of the slide. Thank you so much.