WEBVTT - The Covid Symptoms Hidden in Our Voices 0:00:01.320 --> 0:00:07.240 Welcome to Prognosis. I'm Laura Carlson. It's day three since 0:00:07.280 --> 0:00:13.319 coronavirus was declared a global pandemic. Today's main story. New 0:00:13.360 --> 0:00:18.919 technology promises it can spot early COVID nineteen symptoms simply 0:00:18.960 --> 0:00:23.760 by listening to the sound of someone's voice. But first, 0:00:24.640 --> 0:00:36.519 here's what happened in virus moons today. President Joe Biden's 0:00:36.560 --> 0:00:40.440 pushed for another large COVID nineteen relief package got some 0:00:40.520 --> 0:00:45.800 help today. A surprisingly weak January employment report helped make 0:00:45.840 --> 0:00:49.440 the case for substantial aid, and the Senate took action 0:00:49.880 --> 0:00:54.240 to smooth the legislative path for his proposal. Private sector 0:00:54.280 --> 0:00:58.840 payrolls barely grew in January, while the restaurant and lodging 0:00:58.880 --> 0:01:03.640 sector log almost six hundred thousand in job losses over 0:01:03.680 --> 0:01:07.319 the past two months. That's according to the monthly jobs 0:01:07.319 --> 0:01:11.720 report from the Labor Department out today. While the jobless 0:01:11.800 --> 0:01:15.120 rate dipped to six point three percent, that was partly 0:01:15.280 --> 0:01:20.520 because some Americans gave up looking for work. The United 0:01:20.640 --> 0:01:24.120 Kingdom said it plans to offer everyone over the age 0:01:24.120 --> 0:01:28.480 of fifty a first dose of coronavirus vaccine by May. 0:01:28.920 --> 0:01:32.479 It's the first time that Boris Johnson's government has set 0:01:32.480 --> 0:01:37.279 a target for the next stage of its vaccination program. 0:01:37.280 --> 0:01:40.440 The plan was included in an announcement that also said 0:01:40.720 --> 0:01:44.479 local and mayoral elections will go ahead on May six, 0:01:45.200 --> 0:01:49.960 after they were delayed by a year due to the pandemic. Finally, 0:01:50.600 --> 0:01:54.320 a study has found a surprisingly strong indicator of whether 0:01:54.440 --> 0:01:58.080 someone infected with COVID nineteen has a higher risk of 0:01:58.160 --> 0:02:02.880 severe disease and death. That indicator is whether they've had 0:02:03.080 --> 0:02:08.760 pneumonia in the past. According to researchers at Harvard University, 0:02:09.240 --> 0:02:12.720 a prior episode of pneumonia was the second greatest overall 0:02:12.840 --> 0:02:16.720 risk factor for death from COVID nineteen. The study showed. 0:02:17.120 --> 0:02:21.360 The researchers pointed out that by itself, a single pneumonia 0:02:21.440 --> 0:02:25.240 case probably doesn't put someone at high risk. It's more 0:02:25.360 --> 0:02:29.840 likely to be an indicator of underlying chronic disease that's 0:02:29.880 --> 0:02:39.120 gone undiagnosed. And now for today's main story. When we're 0:02:39.160 --> 0:02:41.960 coming down with a cold, or are feeling a bit stressed, 0:02:42.360 --> 0:02:45.880 or perhaps even exhibiting the first symptoms of COVID nineteen, 0:02:46.760 --> 0:02:50.079 minute changes to our voice are often one of the 0:02:50.120 --> 0:02:55.240 first indicators that something is wrong. These vocal bio markers 0:02:55.320 --> 0:02:58.600 are often beyond what a human can detect, But what 0:02:58.680 --> 0:03:03.760 if an app on your phone could. Health reporter Michelle 0:03:03.760 --> 0:03:08.440 fack Cortez recently spoke to David Luke, CEO of sound Health, 0:03:08.880 --> 0:03:11.640 which is released an app that uses a person's voice 0:03:11.840 --> 0:03:17.200 to detect early symptoms of respiratory illnesses, including COVID nineteen. 0:03:18.120 --> 0:03:21.680 I asked Michelle what vocal bio markers can tell us. 0:03:22.880 --> 0:03:26.040 You recently had a chance to speak with the CEO 0:03:26.280 --> 0:03:30.400 of sound Health, which is released an app called sold one. 0:03:30.520 --> 0:03:32.560 I was wondering if you could tell us a bit 0:03:32.600 --> 0:03:36.600 about this app and how it relates to COVID nineteen. 0:03:38.080 --> 0:03:41.520 This app is looking for vocal bio markers indications that 0:03:41.560 --> 0:03:44.960 you might be actively infected with coronavirus and that the 0:03:45.040 --> 0:03:49.240 illness could be having an impact on your voice. It 0:03:49.440 --> 0:03:52.920 is something that is relatively easy to use. It's something 0:03:52.960 --> 0:03:57.920 that's being adopted by companies and schools and other outlets 0:03:57.960 --> 0:04:00.320 as a way to get an early heads up on 0:04:00.440 --> 0:04:03.880 potential risk. David Leeu is the CEO of sound Health, 0:04:03.960 --> 0:04:06.240 and he laid out how it works for us. Song 0:04:06.360 --> 0:04:10.320 One is a app based product that can be downloaded 0:04:10.360 --> 0:04:13.800 from any smartphone with that that app, you're able to 0:04:13.880 --> 0:04:16.400 record six seconds of voice and we can give you 0:04:16.560 --> 0:04:21.359 then a reading on your risk for having symptoms of 0:04:21.400 --> 0:04:25.160 respiratory disease like COVID nineteen. In terms of how this 0:04:25.279 --> 0:04:29.680 app can actually detect whether or not you have COVID 0:04:29.760 --> 0:04:34.000 nineteen or another respiratory illness, I mean, how exactly is 0:04:34.000 --> 0:04:37.240 it doing that through your voice? Your voice is an 0:04:37.320 --> 0:04:40.800 integral part of you, and you know from your own 0:04:40.800 --> 0:04:43.440 personal experience listening to your loved ones and people that 0:04:43.480 --> 0:04:47.039 you know well, sometimes you can tell if somebody is 0:04:47.120 --> 0:04:49.960 not feeling great, if there might be something going on 0:04:50.080 --> 0:04:54.599 with them in terms of illness or perhaps anxiety or stress. 0:04:55.200 --> 0:04:59.120 We're talking about things like the patterns in the rhythms, 0:04:59.480 --> 0:05:04.320 the will intonation, any kind of variation in that process 0:05:04.440 --> 0:05:07.280 is what they're looking for. And as you do it 0:05:07.720 --> 0:05:10.760 over and over again, the app can get better at 0:05:10.839 --> 0:05:15.240 knowing your own individual patterns comparing that to yourself and 0:05:15.400 --> 0:05:18.240 to other people. David laid out for us the way 0:05:18.240 --> 0:05:23.080 that it works. Bio Markers are any any information that 0:05:23.200 --> 0:05:27.000 is given off by the body. Could be your come 0:05:27.080 --> 0:05:31.120 from blood, could be come from saliva, your your heart rate. Um. 0:05:31.240 --> 0:05:33.240 These are signals that are coming from your body that 0:05:33.279 --> 0:05:36.440 can be measured right and can be recorded, and so 0:05:36.800 --> 0:05:41.719 vocal bio markers are field that has really begun to 0:05:41.880 --> 0:05:44.320 grow and explode in the last five years, I would say, 0:05:44.360 --> 0:05:47.760 so it's fairly new when used in the health context. 0:05:48.240 --> 0:05:52.279 Our company son has has already launched vocal bio markers 0:05:52.320 --> 0:05:56.360 for detection of mental health conditions such as depression and 0:05:56.440 --> 0:05:59.960 now respiratory illness. There are other companies in the space 0:06:00.040 --> 0:06:03.320 stead of also UH used vocal bio markers for other 0:06:03.360 --> 0:06:08.240 health conditions. So what is the science or you know, 0:06:08.360 --> 0:06:13.360 the research that underpins this app How have they constructed 0:06:13.400 --> 0:06:18.480 away that something on your phone can detect symptoms of 0:06:18.520 --> 0:06:22.360 an illness or even COVID nineteen. There has been some 0:06:22.680 --> 0:06:26.720 research done on this in order to determine how effective 0:06:26.760 --> 0:06:29.800 the approaches. But of course it's important to realize this 0:06:29.920 --> 0:06:32.800 isn't something that's approved by the Food and Drug Administration, 0:06:33.200 --> 0:06:35.880 and you're ensurre isn't going to pay for it, but 0:06:36.000 --> 0:06:39.320 it is an early indicator of potential danger. You can 0:06:39.320 --> 0:06:43.279 think of it kind of like a thermometer or your 0:06:43.360 --> 0:06:46.240 mom putting her hand on your forehead to see if 0:06:46.279 --> 0:06:49.279 there's something going on with a layer of science and 0:06:49.320 --> 0:06:53.760 technology on top of that. Really drilling down into the details. 0:06:54.920 --> 0:06:56.520 You need to have a lot of experience and a 0:06:56.520 --> 0:06:59.280 lot of data to have been and to have studied 0:06:59.320 --> 0:07:03.599 this data. Voice data that is health condition labeled, meaning 0:07:03.760 --> 0:07:06.520 we have now over a million voice samples that we 0:07:06.560 --> 0:07:11.360 have collected and studied must have confirmed labels stating that 0:07:11.440 --> 0:07:15.320 this person had the disease or the the health condition, 0:07:15.800 --> 0:07:18.040 and so they've been diagnosed with that condition, and then 0:07:18.080 --> 0:07:20.840 we study their voice from there. We also study people's 0:07:20.880 --> 0:07:24.840 voices who are completely healthy and we compare those those voices. 0:07:25.360 --> 0:07:29.240 When you have that larger volume of data, you're then 0:07:29.320 --> 0:07:33.800 able to look at different groups of people men, women, uh, 0:07:33.840 --> 0:07:37.400 You're able to look at different ages of people, different devices. 0:07:37.480 --> 0:07:41.200 So many things come into consideration when you're analyzing voice. 0:07:41.640 --> 0:07:44.000 When we're looking at different health conditions, then we can 0:07:44.080 --> 0:07:47.560 zero in, for example, in respiratory disease or even depression, 0:07:47.880 --> 0:07:50.960 we can zero in on the vocal features that are 0:07:51.000 --> 0:07:55.960 most sensitive to change when somebody displays a symptom of 0:07:56.000 --> 0:07:59.080 that disease. And again we are not diagnosing for disease. 0:07:59.360 --> 0:08:04.000 We are comply monitoring and detecting the change in your voice. 0:08:04.240 --> 0:08:06.800 Vocal features that are not able to be picked up 0:08:06.840 --> 0:08:11.200 by the human ear, such as prosody, such as tonality, 0:08:11.920 --> 0:08:14.960 even breaks and speech um. There are so many of 0:08:15.000 --> 0:08:17.559 these vocal features that we do examine and put into 0:08:17.600 --> 0:08:21.840 our machine learning models that help us then predict based 0:08:21.920 --> 0:08:26.000 upon the data, who might have who has a higher 0:08:26.120 --> 0:08:30.080 probability of having these symptoms of disease versus people who 0:08:30.120 --> 0:08:33.839 are completely fine. Just to be clear, this is something 0:08:33.880 --> 0:08:38.640 that can only pick up symptoms when you're already manifesting 0:08:38.679 --> 0:08:42.480 symptoms of COVID nineteen or another respiratory illness. This isn't 0:08:42.520 --> 0:08:45.880 something that could pick up whether or not you have 0:08:45.960 --> 0:08:49.880 COVID nineteen if you happen to be asymptomatic. It's not 0:08:50.120 --> 0:08:53.680 a diagnostic. It doesn't tell you that you are definitively 0:08:53.800 --> 0:08:58.880 infected with coronavirus. But it is another layer, another indicator 0:08:59.160 --> 0:09:03.160 that could help well determine if they are at risk. 0:09:03.520 --> 0:09:07.560 The company is not claiming that it is something that 0:09:07.600 --> 0:09:11.960 should be used in order to determine who is safe 0:09:11.960 --> 0:09:15.360 for high risk situations in terms of you know, visiting 0:09:15.440 --> 0:09:18.079 friends and family, are going to a nursing home, flying, 0:09:18.160 --> 0:09:21.640 that sort of a thing. But it is another piece 0:09:21.760 --> 0:09:26.160 of the puzzle. You know, we've been focusing on COVID nineteen, 0:09:26.320 --> 0:09:30.240 but what else can these types of apps like so 0:09:30.320 --> 0:09:35.560 on one detect using these vocal bio markers. Previous research 0:09:35.600 --> 0:09:39.280 has looked at vocal bio markers for indications of other 0:09:39.960 --> 0:09:46.040 potential health conditions, most specifically mental health issues like anxiety 0:09:46.160 --> 0:09:49.600 and depression. Those are areas where we do get a 0:09:49.640 --> 0:09:54.720 lot of change that happens, that comes through your voice. 0:09:55.440 --> 0:09:59.800 Sometimes looking intently at that particular piece can give you 0:09:59.840 --> 0:10:02.520 in it into how you're feeling and where your health 0:10:02.600 --> 0:10:06.880 is going. I'm wondering about some of the potential risks 0:10:07.040 --> 0:10:10.000 or problems. I mean, being able to speak into your 0:10:10.040 --> 0:10:12.480 phone and it tell you whether or not you have 0:10:12.720 --> 0:10:15.960 symptoms of either COVID nineteen or just a respiratory illness 0:10:16.000 --> 0:10:21.199 could be very useful, but what are the risks, say 0:10:21.240 --> 0:10:24.560 I'm thinking of um in terms of either patient privacy 0:10:24.720 --> 0:10:31.320 or just misdiagnosis. Privacy and misdiagnosis are actually critical issues, 0:10:31.520 --> 0:10:34.280 especially in this period of time that we're in right now. 0:10:34.760 --> 0:10:40.240 Everyone is talking about privacy across every platform as we're 0:10:40.280 --> 0:10:44.640 all interacting with each other online, but there are absolutely 0:10:44.840 --> 0:10:49.440 issues about where this information is going, especially because in 0:10:49.480 --> 0:10:54.120 many cases it's employers or other organizations that are gathering 0:10:54.120 --> 0:10:59.600 this information. There is room for things to go sideways. Now, Michelle, 0:10:59.679 --> 0:11:03.080 you've actually have the opportunity to try out sound one 0:11:03.120 --> 0:11:07.440 for yourself. Walk us through what the process is of 0:11:07.440 --> 0:11:11.600 of using the app. So I downloaded the app this morning, 0:11:12.400 --> 0:11:15.920 and I used a key that the company gave me 0:11:16.160 --> 0:11:20.200 full disclosure, and I answered a couple of questions including 0:11:20.200 --> 0:11:24.640 my email address, my gender, my birthdate, and now I'm 0:11:24.679 --> 0:11:27.120 ready to give it a shot. All right, walk us through, 0:11:27.640 --> 0:11:30.240 you know, submitting a vocal sample. Here we go. So 0:11:30.320 --> 0:11:33.600 I'm looking at a screen. It says sound one tap 0:11:33.679 --> 0:11:39.400 here to get started inside concentric white circles. It's loading. 0:11:39.880 --> 0:11:42.959 You have a questionnaire and a voice activity to complete. 0:11:43.400 --> 0:11:45.360 My voice does feel a little scratchy today, so we'll 0:11:45.360 --> 0:11:48.880 see how that goes. I definitely feel healthy, though, and 0:11:48.920 --> 0:11:51.000 I have been pretty much in my basement for the 0:11:51.080 --> 0:11:53.200 last month, so I don't think I'm high risk. Here 0:11:54.080 --> 0:11:55.960 in this activity, I want you to take a breath 0:11:56.000 --> 0:11:59.079 and hold the vowel sound ah like in the word father. 0:12:00.040 --> 0:12:02.360 Please hold that sound for six seconds or until you 0:12:02.440 --> 0:12:10.760 run out of breath, let's begin. Uh wof, I got 0:12:10.800 --> 0:12:14.520 it done. That's a little bit harder than I thought. Okay, 0:12:14.559 --> 0:12:18.240 my health score, I'm a low risk, I'm a fifty 0:12:19.000 --> 0:12:21.960 means that today's score is in your normal range. And 0:12:22.000 --> 0:12:25.960 now I'm getting my screening results. Sounds like you're feeling okay. 0:12:26.080 --> 0:12:28.680 Watch for COVID nineteen symptoms and call your medical provider 0:12:28.720 --> 0:12:31.880 if any symptoms develop. Follow your local and state public 0:12:31.880 --> 0:12:35.680 health guidelines to keep yourself and others safe. So it 0:12:35.720 --> 0:12:38.200 occurs to me, Laura, that part of the benefit of 0:12:38.200 --> 0:12:40.840 a screening like this is that you just literally have 0:12:41.000 --> 0:12:44.679 somebody every day walking through the process of thinking about 0:12:44.840 --> 0:12:49.280 how they're feeling. Do I feel hot? Did I sleep poorly? 0:12:49.760 --> 0:12:53.360 Am I losing my sense of taste or smell? Those 0:12:53.360 --> 0:12:55.520 are things we should all be asking ourselves, but I 0:12:55.520 --> 0:12:58.720 don't know that we do it intentionally every day. So 0:12:58.760 --> 0:13:00.959 if you do have an app that's walking you through this, 0:13:01.559 --> 0:13:05.000 that could be a big part of the benefit right there, 0:13:05.040 --> 0:13:08.280 regardless of what you learn from your vocal bio markers. 0:13:11.640 --> 0:13:14.320 That was Michelle fa Cortez and that's it for our 0:13:14.320 --> 0:13:17.520 show today. For coverage of the outbreak from one twenty 0:13:17.559 --> 0:13:21.920 bureaus around the world. Visit bloomberg dot com, slash Coronavirus, 0:13:22.840 --> 0:13:25.040 and if you like the show, please leave us a 0:13:25.040 --> 0:13:28.960 review and a rating on Apple Podcasts or Spotify. It's 0:13:29.000 --> 0:13:32.160 the best way to help more listeners find our global reporting. 0:13:33.360 --> 0:13:37.360 The Prognosis Daily edition is produced by tophor Forheas, Magnus 0:13:37.360 --> 0:13:41.960 Hendrickson and me Laura Carlson. Today's main story was reported 0:13:41.960 --> 0:13:46.600 by Michelle fay Cortez. Original music by Leo Sidran. Our 0:13:46.760 --> 0:13:50.800 editors are Rick Shine and Francesco Leady. Francesco Levi is 0:13:50.840 --> 0:14:02.320 Bloomberg's head of podcasts. Thanks for listening well.