Professor Kate Steinbeck [00:00:06] Hello everyone and welcome to the WH&Y webinar for September, which is part of what the Wellbeing, Health & Youth NHMRC Centre of Research Excellence does each month to look at new research in the area of adolescent health. I do remind everyone that we have several collaborators across Australia, and I acknowledge all the partners who are part of the CRE. 
 
Professor Kate Steinbeck [00:00:40] I also would like to acknowledge the traditional owners of the country throughout Australia and recognise their continuing connection to land, waters and culture. And we pay our respects to their elders past, present and emerging. 
 
Professor Kate Steinbeck [00:01:03] We also invite everyone to be part of our WH&Y Community of Practice. It's a community research practice, but I think we look at it more as a community of practice where we translate research into real life. And just watch our website because there'll be more information coming up soon. 
 
Professor Kate Steinbeck [00:01:27] And just a few housekeeping rules. While the webinar is going up, your microphone will be muted and video switched off. If you have something to say or if you want to ask the presenter a question, then please use the chat panel which is at the top of your screen and questions will generally be answered at the end of the webinar. And please, of course type any comments that you would like to make. 
 
Professor Kate Steinbeck [00:02:02] So, it is my great pleasure today to introduce Professor Corrine Caillaud, who is a professor of digital health and co-chair of the Nano Sensors and Diagnostic Cluster within the University of Sydney's Nano Health Initiative. Corrine also co-chairs the children and adolescent health and wellbeing in the Pacific Node, for the Charles Perkins Centre. Her team co-designed evidence- based digital health interventions to promote more active lifestyles among young people in Australia and across the Pacific. And it is a delight to have you with us here, Corrine. So over to you. 
 
Professor Corrine Caillaud [00:02:50] Thank you so much, Kate, for this introduction. I have to say, I'm thrilled to have the opportunity to present today as part of the WH&Y Webinar series. Before we get started I just wanted to acknowledge the traditional custodians of Australia and would like to share that. Today I'm sitting and speaking from Gadigal lands and I pay my respects to elders past, present and emerging. 
 
Professor Corrine Caillaud [00:03:20] Today, the presentation is about using wearables, sensors, and I'm going to particularly focus on activity trackers in a way to better understand physical activity in young people, children, and adolescents, and to try to understand how we could leverage digital health technologies and sensors to better understand physical activity patterns. I'm going to talk about physical activity guidelines just to give a little bit of a context to why we are interested in physical activity in the young population, physical activity in adolescence and the World Health Organization ambitions for young people worldwide in terms of physical activity. Then, I'm going to switch to talking about wearables and how they are used to evaluate physical activity in people, methods to identify patterns and changes in response to interventions, and then I will finish by presenting one digital health promotion program that we have designed with colleagues from the University of Sydney and the University of New Caledonia, and how we've been able to take advantage of the wearables to support learning and assess exactly what's happening when such an intervention is developed. And then I will say a few words about what's next for us. 
 
Professor Corrine Caillaud [00:04:56] So what about the World Health Organization Guidelines on Physical Activity for Children? Some of you may already be aware of that, but on these guidelines released in 2020, you can see that the recommendation in terms of physical activity for young people is to perform at least 60 minutes a day on average across the week of moderate to vigorous physical activity (MVPA). I just wanted to focus your attention on the fact that in the guidelines, they're not talking just about physical activity, but pointing to the fact that the quality of this physical activity is at the intensity level of moderate to vigorous. On the top of that, on the right side of the slide, you can see that what is recommended is at least three days a week of vigorous intensity. So really trying to get on the high scale of the aerobic intensity activities and as well to think about strengths-based activities. Based on these guidelines, we can imagine that what's going to be important when we try to assess physical activity and movement in young people is to be able to identify that moderate to vigorous physical activity. I'm going to say a few words about how MVPA is assessed and how wearable sensors can potentially add something to that. 
 
Professor Corrine Caillaud [00:06:27] In most of the past studies in the world, physical activity and MVPA have been assessed using surveys, validated questionnaires, and usually those questionnaires have been developed in high income countries and disseminated worldwide. For example, for us, when we came to do some studies in the Pacific region, we realised that at some point those surveys were not really fit for pupils in those populations. In the recent years, if we go back ten years, researchers started to use more and more wearables, particularly wrist activity trackers, to identify when people were active and when they were engaged in moderate to vigorous physical activity. So why are we focusing on this moderate to vigorous physical activity? Well, this is because the research evidence has shown that when young people were doing those 60 minutes a day on average of moderate to vigorous physical activity, there was an impact on physical fitness. And I really would like to point out to you that physical fitness is an important aspect because it is through good physical fitness, that a lot of cardiometabolic health and mental health and the reduction of adiposity you can see on this slide are driven. Physical activity is important, MVPA is the focus and raising the physical fitness is another very important focus. And the ambition for the World Health Organization globally is to reach by 2030 an increase in physical activity, which they expressed as reducing physical inactivity by 15% in 2030. 
 
Professor Corrine Caillaud [00:08:25] If we look at the status worldwide of physical activity, we can see there's a little bit of work, I guess, to do  continuing promoting and trying to increase physical activity in young people. So on this large study published in The Lancet in 2019, almost 300 school-based surveys were studied from 146 countries, which was 1.6 million students aged between 11 and 17 years. And what the study showed was that 81% worldwide of students were not sufficiently active, which leaves us with roughly 20% of young people worldwide being sufficiently active. And you can notice on the slide on the top for the boys and at the bottom for the girls, and when it's green, it means that the physical activities match. When it goes to yellow and orange, it means there's more students not meeting the physical activity guidelines. 
 
Professor Corrine Caillaud [00:09:41] All right, so now let's get into the wearable technologies and trying to understand what's the landscape and what I'm going to talk about today. On that slide, you can see there's a little bit of a bias. I'm sorry I couldn't find something better because we can see a very fit man who's trying to find a woman, but it's a little bit more complicated. Anyway, we have this picture where I think what is interesting is we have a representation of the different sensors existing now and the type of data they can provide. You can see they can provide several different data from the cardiometabolic side of things, from the oxygenation of the blood, the respiratory system, the muscle system, some kind of indication about stress, physical activity, and sleep. And what I'm going to talk about today, mostly are the data collected via activity trackers, Im going to talk essentially about trackers worn on the wrist. And I'm going to talk about how the data have been used so far and what we've developed in the team. I just wanted to point out that there's still a lot of work that can be done to better understand how sensor fusion could potentially be done in a better way to provide information for the young population in terms of their behaviours. 
 
Professor Corrine Caillaud [00:11:17] We did a little bit of a scoping review to try to understand what type of activity trackers were used across different research, and we particularly focused on that research in which people have tried to develop algorithms to either better understand patterns or to try to classify the activities. Being conscious that there's a wider range of studies that have used activity trackers, but here I'm talking about those studies that have tried to go a little bit further than just collecting how many steps a day were performed or how many minutes of moderate to vigorous activities. But basically, you can see that the type of technology that has been used is across two main categories. One category is called the reserve grade sensors, which means the sensors have been validated against laboratory studies. They provide accurate data and usually they provide quite a lot of data because they can register quite a lot of data points per second. The two main sensors that have been used are the GENEActiv Years ago, those sensors were just actual accelerometers, so providing accelerometer in only one direction, but now we have access to triaxial symmetry, which means that it's a bit more complicated. It provides acceleration in different directions, but it then provides opportunities to do a little bit more with the data. 
 
Professor Corrine Caillaud [00:12:57] In those studies, commercial trackers had been used and we've used some in some our studies. Those ones usually are not validated. I'll have to say there's more and more validation study with the Fitbit, and one characteristic is that it's usually extremely difficult to access the raw data and that is because for commercial reasons the raw data are not that easily disseminated. Again, with the Fitbit, there ar now more possibilities to do that. Some  studies have combined accelerometer and heart rate, and I guess the advantage of that is that it provides information on movement, but the heart rate as well is a very good indicator and potentially more precise indicator of the intensity of the exercise or the movement that's being produced by the person. So it provides more precise information on the intensity. 
 
Professor Corrine Caillaud [00:14:04] The point I wanted to really make is about the how the data we potentially have access to when we work with research grade activity trackers, for example, and the type of data we are using, and basically in lots of studies, the acquisition sampling is quite high, so the sample rate can go between 30hzs  to 100hzs.  It means that in some studies people have 100 data points per second for many days, which is huge. The autonomy varies between one day and one year, so this is something that can be taken into consideration. And here you can see that those studies that are looking at developing some algorithms to understand a little bit better what's happening, how the duration can go from four days to one year. So there's a huge number of data that are collected on that table on the left, you can also see that amongst all of those studies, only a small number were performed in young people and adolescents. 
 
Professor Corrine Caillaud [00:15:21] What I wanted to point to as well, is that despite the fact that there is access to that high level of data, most of the time the data are aggregated and returned in terms of research outcome either over the day, so you can see a number of different studies would report physical activity levels in terms of steps or in terms of time spent on MVPA during the day on average. Some used the month, the week, a few of them report data per minute. But there's very few studies that have really taken all the advantage of the granularity of those data that are collected to, I guess, go beyond just expressing how many minutes are spent at a given intensity. We need to try to understand the physical activity patterns. So how often, for how long and in which pattern young people engage in those activities. 
 
Professor Corrine Caillaud [00:16:24] Before we go a little bit further into analysing the data, I just wanted to describe how we bridged the data coming from the sensors to the World Health Organization guidelines. The data coming out of the sensors are acceleratory data. And the recommendations, for example, if we wanted to have a sense of at which level of intensity the person is performing the activities are given in terms of moderate to physical activity, which is referring most of the time to an intensity related to the aerobic capacity. So how do we link data that are an indication that are provided in different volumes. It goes through several different validation studies that have in the lab asked people to do a number of different types of physical activity while they were recording oxygen uptake. And oxygen uptake is going to give an indication on the metabolic equivalent task (met). So, the energy expenditure of the person is going to be related to the intensity of the exercise. For example, one met is when someone spammed basically this is the resting conditions, 3.5 mls per kilo and per minute of oxygen. And the threshold that has been set up for young people if we want to consider moderate to vigorous physical activity, it will have to sit at a value of 3 to 4 mets (metabolic equivalent of task), but it could come between the different applications. 
 
Professor Corrine Caillaud [00:18:17] In those studies, people were wearing activity trackers while oxygen uptake was measured, which was a way to link the oxygen uptake and the metabolic equivalent and then infer on the exercise intensity. You can see on the graph on the left side that we have the risk accelerometer data and mets on the vertical axis. The blue points are the data collected during lab experiments, and the red points are data collected in free living conditions, and everything that's shaded blue would be moderate to vigorous physical activity. So basically acceleration data are matched to mets through what is called cutoff values. We would say, for example, when the accelerometer is about 200 ngs, then it means that the person is performing an activity that is around 3 mets, so we can classify that activity in moderate activity. This is how the accelerometer data coming out of the sensors are connected to energy expenditure. 
 
Professor Corrine Caillaud [00:19:46] But what I wanted to point out, because a number of studies have published different thresholds in terms of accelerometer corresponding to energy expenditure. And I just wanted to point out, when we do a study, we need to try to understand a little bit of context and have a look at the literature and use the right sensor with threshold that have been developed for that sensor with the same population. And even when we do that, we realise here, for example, that we have different studies. And if we use different cutoff values or thresholds in terms of accelerometer to determine the time spent in moderate to vigorous activity, you can see that depending on the studies and the different cutoff for the same dataset, the time spent, and MVP could be quite different. And the difference  can be quite wide. So, it's very important to really understand the context, really understand the population and how the data from the accelerometer are derived to calculate or assess the time spent in moderate to vigorous activity. 
 
Professor Corrine Caillaud [00:21:10] I’ve just talked about the threshold. But another point that is very important is what we call bouts, which is we have data, we have 100 points or 60 data points per second, then we could average the data per second. And then what do we do with that? Are we trying to pick up every second that is MVP or are we trying to potentially aggregate the data or try to identify bouts of physical activity that we're going to focus on? And what I wanted to show on that slide is that I guess we can make any kind of decision, we just need to understand the consequences of making decisions. So, for example, on that slide, these are data collected in primary school children and we have the hours set between 7am and 9am in the morning. The first two figures at the top show data when only 1 second bouts are used, so each second is targeted with the intensity being light, and this is everything that's in green. Moderate is everything that's in blue, or vigorous. On the left side, we have male participants and on the right side we have female participants. And the two figures at the bottom represent exactly the same participants, except that in that situation, what's been done is that data has been averaged over one minute and then each bout of 60 seconds has been targeted with either light, moderate or vigorous intensity.  
 
Professor Corrine Caillaud [00:23:01] So, what do we see? Well, first we can see that when we only work with 1 second bouts, there's lots of blue, which means lots of 1 seconds that are performed at moderate activity. Maybe part of that is potentially agitation and not physical activity, but we can see there's quite a lot, we can see that potentially the boys in that study were quite active just before entering the class, so in that space between 8.30am and 9am there's quite a lot of activity. We can see it's quite different when we go on the right side and we have to look at the data coming from the female and it looks like there's minimal physical activity, but still we can detect a little bit of physical activity. If we go to the 60 bouts, we can see there's a lot of physical activity that's disappearing compared to the graph at the top. But we still retain a number of physical activity in the boys. But you can see that if we go on the right side and we have a look at the female, most of the physical activity completely disappears. So it's not neutral to decide that we're going to look at the data average of 1 second or 15 seconds or 60 seconds. I think what we can see in that slide is that when people are less active, I would say they are disadvantaged by using longer duration for the bouts. So that would be the case here for the females. We could potentially ask if physical activity assessed by trackers has been consistently underestimated in females? This is a question that I think remains open and we can see that if we go to the vigorous which is in red, again it disappears when we go to the 60 second bouts most of the time. So it's not neutral and we need to find a good way to average the data in young people. And it's particularly true because physical activity patterns in young people are completely different from the physical activity patterns in older people. I think you would never see spontaneously a seven year old child or a 12 year old child just going to jog for an hour. And I think it doesn't exist. What they do is a burst of physical activity, moderate to vigorous, but in a very short period of time, very short bouts. So the risk is that if we average over a longer period of time, we're going to miss a lot of that physical activity. 
 
Professor Corrine Caillaud [00:26:03] What I want show on that slide as well is what we have are different participants, male and female. The dark column shows the minutes spent in MVPA, so we can see that there's quite a lot of variability within boys and girls. We can see there's more physical activity, more time in MVPA in the boys. What the pink bars show is the average percentage of the maximal aerobic capacity. So, I wanted to show that even if we take participant No 3, for example, spending quite a lot of time in the MVPA, we can see that on average the exercise is around 40% to 50% of VO2max. So, this is something we need to explore a little bit further in young people, so when we talk about MVPA, where in that MVPA do they do the physical activity? And in our studies, we realise it's mostly within the moderate range, they spend a very small period of time in the vigorous area. 

Professor Corrine Caillaud [00:27:27] Okay this being said, what I wanted now to show you is what we've done with the data we collected and our approach to try to identify short bouts of moderate to vigorous activity in that population. When I talk about research grade activity trackers they are coming from the GENEActiv and this is a triaxial, so we have data acceleration for the three different directions, and we try to extract from that activity vectors. Now the approach we've taken is to tag each second using existing threshold with either light, moderate or vigorous activity, but not to aggregate the data, but instead to try to identify each bout of 3 seconds, and we've chosen 3 seconds because we did some studies and we demonstrated that shorter than that we capture too much noise, but longer than that we miss out on some of the relevant activity bouts. So, we tried to identify all the bouts targeted with moderate and vigorous, lasting at least 3 seconds. 

Professor Corrine Caillaud [00:28:53] Now, what we've done in our approach is, we asked the question, which is what if instead of just trying to identify moderate to vigorous activity, we also try to identify sedentary time. There are no specific guidelines about sedentary time yet, but we know that the time spent in being sedentary is a time that may need to be reduced and to be a kind of tradeoff with physical activity and MVPA. So, we’ve analysed all the data we had collected on a larger group of students aged 12 years old and we had a look at their daily activities. What was happening during the day here, and we tried to identify those physical activity MVPA times that were lasting at least 3 seconds. But for the sedentary time, we said, well, wait, maybe being sedentary for 3 seconds doesn't really mean anything. So we tagged any bout that was identified as sedentary that was lasting at least 2 minutes. And we would say, well, you know, in that situation then we have people who are actually sitting or not doing physical activity, not moving at all for 2 minutes, and we consider that sedentary time.  
 
Professor Corrine Caillaud [00:30:24] What you can see here, is when we analysed the data, we clusterised the different days of the cohort using both the sedentary time and the time spent in MVPA. And we identified that we could meaningfully analyse the data if we had six clusters, so you can see here the six clusters on that slide, and we were able to describe what was happening in those clusters. For example, if you take the first one, there is quite a lot of sedentary time and not a lot of active time. All the days clustered in that cluster are days with minimal MVPA and quite a lot of sedentary time. When we moved towards the higher end of the spectrum, here we can see that from cluster No 4 the days provide at least the 60 minutes of MVPA, just as is recommended. And then when we go up to 5 and 6, then we have more physical activity, and we have less sedentary time. This is really the cluster in which we would like the kids to move because there is less sedentary time and more MVPA. 
 
Professor Corrine Caillaud [00:31:47]  In the next part of this talk, I wanted to present how we have started to integrate the different algorithms we have used to develop a digital platform, which is a health education platform that is aiming to improve physical activity focusing on moderate to vigorous physical activity and really try to engage young people into doing more physical activity. We have used in the platform commercial trackers that will provide step data on every day of the program. And we use the data from the trackers to support learning activity so the young people understand what it means to actually perform moderate to vigorous physical activity. We've used all the data during the program to help the children to understand what it was to do that activity. We also mapped all the different modules and the content of the learning with the Australian Framework for Physical Literacy, and we have also included in the program behaviour change techniques. 
 
Professor Corrine Caillaud [00:33:20] What I also wanted to present is we've developed this platform i-Engage, which has the ten modules containing health education supported by activity trackers, and we developed the program for two different contexts, the Australian context, but we also developed that platform to be implemented in the Pacific region and in that case that was in New Caledonia, in the small island of Lifou. And I wanted to demonstrate that technology can be used in remote locations providing an amazing opportunity to assess implementation of the intervention, because all the data collected in the platform can be used to understand the outcome and the physical activity patterns. We implemented the i-Engage program in the small community of We in Lifou. I just wanted to show you here what it looks like. So, there is the area in which we have the school on the right side, and you can see the housing dispersed along different streets, and there's facilitated access to the beach. So, they have several possibilities to go swimming and do some aquatic activities, and there's a stadium as well that is quite accessible. So the context is kind of in favour of physical activity. I. 
 
Professor Corrine Caillaud [00:35:03] That was a small cohort, we had 35 kids involved. But the way their days were made here we have some indication about sleep and the time using screen. I'm not going to spend too much time because we are not directly addressing those data. But overall, the kids really had not enough sleep every day. Around 35% of them do at least 60 minutes of moderate to physical activity and 50% have good physical fitness, but we still have 38% of those children and adolescents living with overweight and obesity. We thought it was a great opportunity to deliver a digital health education program in that context, because the school context was quite in favour of physical activity with quite a lot of space in the school that the children could use for formal and informal physical activity, good communication with teachers and community leaders with which we have engaged, and the students as well as in previous years were engaged to try to understand physical activity. The school and the community are well served. with an Internet connection.
 
Professor Corrine Caillaud [00:36:39] Here, just a little bit of description of the program. The program was made of ten modules that were designed into an app. Each module lasted 60 minutes. They were self-directed so each kid could access the module and do all the different activity at their own pace. There was a little bit of reading. You can see the different activities included quizzes about the learning and the knowledge they had about physical activity, some assessing knowledge. They had goal setting activities, goal achievement activities as well, once they have data into the platform to try to understand if they had achieved their goals. Lots of peer learning activity within the classroom. Maybe I didn't mention it was in the school setting, so the platform was delivered in classes so the whole class would do the program at the same time. We had a little bit of a physical activity session, but it was very small, it was never more than 5 minutes during the module because it was experiential learning for the kids using the data from the sensor to understand the different levels of physical activity and what it means to be physically active. And on the right side, we had the platform mapped with the different domains related to the Australian physical literacy:  physical domain, psychological, social, and creative. 
 
Professor Corrine Caillaud [00:38:12] The design was, we believe, quite inclusive. We didn't want to identify any profile or any kids in any way, so we took the option to use mascots and animals to drive the kids and to help the kids to engage with the different activities. It was an inclusive program because each kid had a tablet, a booklet. So, it was a mixture of getting the intervention from the tablet but being able to take notes and take the notes home so they could connect with the parents and the community. And the activity tracker, which was quite small, they had to keep this one for the ten modules, which means over the five weeks of the program. And so, this is how the information was communicated and how the data was communicated to the kids in the platform. 
 
Professor Corrine Caillaud [00:39:15] I Just wanted to share a few pictures. I really want to make the point that it's completely possible and it's absolutely amazing to work with and to bring technology into remote communities, even if sometimes there's issues in terms of finding the right room to set up all of the activity sensors, having to work late at night, not sometimes having the shop for repair, you can see the industrial zones, which is where you repair in your technology at all times. But you have the people around you helping. And we had in this program when we did it in New Caledonia, we had videos within the program of community leaders communicating very important health messages to the children in the local language, engaging in the dual language. And it was absolutely amazing how it empowered the children to engage with the program. Really, this mix between the technology coming from us, but also the community and the local culture and the local language engaging as well in the program. Here we had videos demonstrating some activities and again this was where local leaders were involved in that and all the kids really engaged in different ways with the platform engaging with the different activities. 
 
Professor Corrine Caillaud [00:41:11] So what is the kind of data we get? And I really wanted to share that because it's about delivering the program, but it's also about the type of data we collect within the platform that really help us understand how the participants responded to the different modules. And what you can see on the left slide, it's about the goals the children set for themselves. And we can see in the blue line that progressively they set goals of increasing challenge and the goals were in steps.  85% of the kids would try to achieve their 11,000 steps a day. And the colourful bars at the bottom going from blue to green show how progressively the cohorts moved from lower daily steps into daily steps to higher daily steps. 
 
Professor Corrine Caillaud [00:42:18] On the right side, we have not only the goals on the dotted line, but we have as well achievement and we can see that progressively the kids kind of increase the number of steps that they’re doing. The graph at the bottom with the green and the yellow bars shows the percentage of days on which the kids were meeting the goals. Okay so at the beginning the goals were easier so most of them would meet the goals, but we can see that in the middle when it becomes a little bit more challenging, a smaller proportion of the kids would meet the goals. And again, they kind of toward the end get to achieving the goals. 
 
Professor Corrine Caillaud [00:43:06] In this paper published with my colleagues involved in this work, A/Prof Olivier Galy from the University of New Caledonia, and A/Prof Kalina Yacef from Computer Science at the University of Sydney, what we have here through this activity tracker providing steps during the program, we have on a daily basis how the consistency of the days above 11,000 steps, so the percentage of days across the program in which the kids performed more than 11,000 steps. What we can see is there are changes. So we can see that if we compare the beginning to the end, there is an increase in terms of the percentage of days showing you did 11,000 steps. But we can see it's quite variable and we need to understand what is it that potentially triggers the kids to do more, but what is it potentially as well within the programs that at some point made the response not what we would expect here. So, the only thing we could find is that they had quite a lot of rain at that time, which could be preventing physical activity. But at the end they were still increasing their physical activity. We can see the dotted line, the kids that were at the beginning the less active. So they are exactly on the same kind of pattern and they benefit as well from the program. 
 
Professor Corrine Caillaud [00:44:42] What we want to show on that slide are again using the days during the program recorded by the activity tracker, we used our technique of clustering the different days and here we clustered across three clusters. So the less active, the moderately active and the very active kids and the figure on the right side shows the average consistency of days in the most active cluster. So, we can see there's quite a lot of variation and the trends increase toward the end. But the dotted line showing the participants who were less active at the beginning, it seems like they benefit a little bit more from the program. And you can see in the middle of the program we have a steep increase here and this is something that is consistent across our data coming from the study in Australia and in New Caledonia. And we are investigating which modules here in the middle of the program were that effective. 
 
Professor Corrine Caillaud 00:46:07] The technique I talked about a bit earlier when we cluster the kids in days, taking into consideration both the sedentary time and moderate to vigorous physical activity. Here using the research grade activity trackers that the kids were at the beginning and after the i-Engage program we measured physical activity across several days and then we clustered the days and then we clustered the participants into those different clusters of different sedentary time and physical activity. And so, the less active is the number 1 and the more active with less sedentary time is the number 6. And everything you can see here in green, all the participants that have shifted after the program into a cluster that was either number 4, so already met the World Health Organization activity guidelines or that's transitioned into a more active cluster. It's a technique that is able to identify how the activity and the sedentary time changed and helped to understand as well at the individual as well as the cohort level the impact of the intervention. So, we can see the shift in terms of behaviour when we compare before versus after. 
 
Professor Corrine Caillaud [00:47:50] This is the last thing I wanted to show you today. I talked about clustering by days, but what we've done as well is clustering the hours during the day because kids attend school and at school a lot of the time is not their choice in terms of being physically active or not. So, we identified all the clusters taking the different hours of the day. And we wanted to see if when we do the i-Engage intervention, there is a change not only during the day, but if there are particles of time during the day where the change is more obvious or significant. So this is what we did. And when we analysed the hours, we identified four different clusters that would classify all the different hours, and we found that it was after school and most of the time after 4pm that there was a shift in terms of kids getting into more physical activity and spending more time in those clusters of physical activity. And if I wanted to show you something that maybe speaks a little bit more to that, here we have a graph in which we have on the left side the hours of the day, and we have separated sedentary time, MVPA and light. And you have the two curves showing the physical activity pre and post. So, if we look at the green in the middle, and the bright green are the data post intervention, we can see that the kids take opportunities at different times during the day, but mostly during recess at school. And very importantly, after school, they found opportunities here to increase both the light activity and the vigorous activity and to do less sedentary time, which is, I think, very interesting because it shows that the kids probably in connection with the parents via the booklets they can connect with the family and the communities, they were empowered, it was not restricted activity. They were actually empowered to find the time for more physical activity and particularly for more moderate to vigorous physical activity. So this is looking at the data this way. 
 
Professor Corrine Caillaud [00:50:26] So what's next for us? What we would like to do is to take better advantage of the step data we collect during the program to provide automated feedback. So, develop algorithms that would help us to identify which days or which hour the person has been more active to identify that time and say something around, oh, this is amazing, you know, yesterday or the day before or the two previous days, you've been able to do more physical activity at 10am, maybe you can try to reproduce that. In terms of the messaging, this is where we really would like to get into co-design with young people to understand what type of messaging would be more effective for them. 
 
Professor Corrine Caillaud [00:51:23] And the last thing is, I've only talked about physical activity in this talk. In the i-Engage program we have modules about food, well It's more about sugar in the food and soft drinks, but it's very difficult to, as you know, not think about food when we talk about energy expenditure, about the energy intake and the quality of the food. So, we have developed for the Pacific region a digital 24-hour recall app in which we have a database with the food of the Pacific region. And it's very particular because in this region, the traditional food and the family farming and the raw food is still very present, but the processed food is kind of extremely present as well. We want to merge that information coming from the sensors, the i-Engage platform and the 24-hour digital recall app to really understand the patterns. When do people eat , what they eat, how often, when do people engage in physical activity and how often? Because we know that the different patterns and how food intake and the quality of intake combines with physical activity has impacts as well on metabolic health. That's it for me. And I just wanted to thank you for your attention and just mentioned that this work wouldn't be possible without partners, particularly industry partners. It's connected to the work we do in the Pacific and it's been funded as well by different groups. Thank you so much. 
 
Professor Kate Steinbeck [00:53:15] Thank you Corrine for a fantastic talk. We've got time for some questions. And I'm going to focus on the ones that are quite specific to measurements because I think they’re possibly going to be the most useful. Hopefully we can get all of them done. So, the first question is, "thanks for a super interesting and useful talk A questionvabout accelerometer. If we're working in a setting where we use the same accelerometer in both adults and children, would the data algorithm analysis need to be tailored to each age group?" 
 
Professor Corrine Caillaud [00:53:58] I would say yes, our work would show that, but there's already different papers, for example, with GENEActiv accelerator, which is the one we've used, you can find in the literature cut-off values for children and adolescents, toddlers as well, I believe, but as well for adults and then older adults. And I guess when we go toward older age, the bout duration is not that important because people tend to engage and stay longer in a typical activity. But when we talk about kids, it's really bursts of activity. The short answer is yes, you would need to use different threshold values at least. 
 
Professor Kate Steinbeck [00:54:43] And just following on "what are the implications for sleep tracking with the same device, we use 30 second bouts for sleep, but if we collected the data at higher resolution that means the physical activity data could be used" I think you can see the question there. 
 
Professor Corrine Caillaud [00:55:05] Yes, I can see the question. Yes, it's actually a very good question because there are some studies in which we wanted to capture both sleep and physical activity. And I guess it's always a tradeoff between what's possible. I know with sleep there's some trackers that are better than others. Algorithms have already been developed like for GENEActiv. I'm not sure all of the algorithms have been fully developed for assessing sleep patterns in young people. So it's always a tradeoff. If you have 30 seconds, depending on the age of the population, I guess you would have to look a little bit in the literature what people say about the 30 seconds, but you have lots of studies using the 30 minutes average, so I think you are still within collecting a very good data set. There are a few things you may not be able to capture, but it doesn't mean that you won't be able to capture some of the physical activity and the MVPA and if you're using the same methods before, after or within your different groups, I guess you still have a good a good dataset and a good way to analyse and to have good outcomes. 
 
Professor Kate Steinbeck [00:56:38] I'm going to try taking a couple more questions before we finish. The next one is from Ian asking given reliance on accelerometer data to infer physical activity intensity, can you comment on how research in this space can account for the likelihood of false positive records eg young people waving their arms about to artificially increase recorded steps? 
 
Professor Corrine Caillaud [00:57:12] Yes, very good question. I mean, this is always the difficulty with wrist accelerometry, how do you make the difference between agitation, you know false positive. So, one way, for example, we did was looking at our data set we wouldn't use the 1 second bout because within that 1 there's quite a lot of false positives. So we had to look at the dataset and we tried to identify what would be for our population the ideal bouts to avoid losing too much of that moderate to vigorous physical activity without taking into account too many of the false positives. This is correct. It may also relate to the question you are asking in your research. If you really want to identify the patterns, when it happens, which time of the day in a more precise way. And the thing with young people I would like to say, and I haven't said is that amongst all the interventions that have done, the results are often very disappointing. So when after several weeks or months you increase MVPA by 5 minutes, you're happy. So the question is, is it really that disappointing or is it because we are missing some of that information? I guess  that's the question we would like to ask, but definitely want to avoid having bouts of very short duration that would increase the false positives. There's always a bit of a tradeoff. 
 
Professor Kate Steinbeck [00:58:58] I'm going to ask one last question, which is, “was ethics approval covered by the Australian ethics or did your team have local ethics and another one about how communities were consulted.” 
 
Professor Corrine Caillaud [00:59:31] And we have two different ethics applications, two different platforms. One was in French; one was in English. So, it's the same platform, two different languages that went through the ethical frameworks in each country and the way we engage with the community, the work we've been doing in that community in Lifou spanned across several years and it really started with trying to understand the physical activity on sites, trying to do surveys, realising surveys didn't provide the right information because the questions didn’t make sense for the kids. So, trying to use the technology and understand how the technology could work and we have engaged with the community a little bit on the goal with that study because i-Engage is a very innovative program and it was very difficult to try to explain before delivering. So, we started with the framework and the way we collaborated as well with the industry partner was that on the go, we could add a few things just before delivering the program. But what we wanted to do is absolutely to engage now that we know what we can do, how we can deliver, we've learned so much from those initial studies we actually want to engage with the communities and with the children in a more structured way. Absolutely. This is for us something we will look at. 
 
Professor Kate Steinbeck [01:01:18] Well, I think we've run out of time. Corrine, thank you for a fascinating talk. I'm sure you'd be happy to answer more questions if they came through on the email.
 
Professor Corrine Caillaud [01:01:31] Absolutely, yes. 
 
Professor Kate Steinbeck [01:01:32] And thank you, everyone, for attending today.