WEBVTT 1 00:00:00.000 --> 00:00:07.020 Okay, I think we can start. Welcome everybody. Welcome to Can Big Data Help You Age Well? 2 00:00:07.420 --> 00:00:13.900 My name is Jane Muzzard. I'm the Chief Executive of COTA South Australia, the older people's 3 00:00:13.900 --> 00:00:23.500 movement in South Australia. I want to frame today's conversation. Our hypothesis is that 4 00:00:23.500 --> 00:00:30.600 big data will help us individually and importantly collectively to age well. Nobel Prize winning 5 00:00:30.600 --> 00:00:38.080 economist Ronald Coase said, torture the data and it will confess to anything. So will it 6 00:00:38.080 --> 00:00:43.860 confess to the secrets of ageing well? I think we all have a rough idea of what big data 7 00:00:43.860 --> 00:00:51.320 is. It refers to extremely large data sets that when analysed are capable of revealing 8 00:00:51.320 --> 00:00:53.460 patterns, trends and associations. 9 00:00:53.500 --> 00:01:00.500 The collection of millions of individual records within a service or system provides 10 00:01:02.020 --> 00:01:08.740 a very unique opportunity for us to pose questions of it that will help us spot and understand 11 00:01:09.600 --> 00:01:16.600 patterns that are not otherwise evident. Surely it makes sense then to harness the power of 12 00:01:16.600 --> 00:01:23.460 that data to understand the wonderful fortune of our time, the gift of our ageing and to 13 00:01:23.460 --> 00:01:34.520 help us all age well. While each of us lives lives uniquely and individually there is a knowledge that accrues from lots of 14 00:01:34.520 --> 00:01:43.800 individuals with similar experiences that helps us to understand changes, trends and patterns that will shape policy 15 00:01:43.800 --> 00:01:49.540 systems and services of the future. There are two facts I think that frames the conversation 16 00:01:49.540 --> 00:01:50.060 today. 17 00:01:50.640 --> 00:01:53.200 The first is that whether we like it or not 18 00:01:53.460 --> 00:01:59.480 Our lives generate lots of data that forms a digital record of our many experiences. 19 00:02:01.140 --> 00:02:05.900 And the second, I think, is that we all share an ambition to age well. 20 00:02:06.760 --> 00:02:12.140 Today's panel will focus on the upside of big data using examples that are making very good use 21 00:02:12.140 --> 00:02:16.080 of all of the information that is already being collected by our governments 22 00:02:16.080 --> 00:02:18.540 when we access health and aged care services. 23 00:02:18.540 --> 00:02:24.860 The panel will give you a chance to have a look inside three large-scale health registries, 24 00:02:25.020 --> 00:02:31.580 two very established, one emerging, including the Registry of Older South Australians 25 00:02:31.580 --> 00:02:33.160 that I'll talk to in a minute. 26 00:02:33.800 --> 00:02:38.580 Each has very strict ethical confidentiality and privacy constraints 27 00:02:38.580 --> 00:02:45.060 and each is under very tight governance arrangements of some of our best academic and stakeholder minds. 28 00:02:45.800 --> 00:02:48.440 So let me start by telling you a bit about ROSE. 29 00:02:48.540 --> 00:02:52.880 ROSE is the affectionate name for the Registry of Older South Australians 30 00:02:52.880 --> 00:02:54.980 and it was established in 2017. 31 00:02:56.160 --> 00:03:00.240 It is a new population-based multidisciplinary data platform. 32 00:03:01.480 --> 00:03:05.900 It connects the information from the cohort of people, older people, 33 00:03:06.000 --> 00:03:11.320 who receive aged care services in Australia with their health records. 34 00:03:11.780 --> 00:03:17.860 It's the first time in Australia and indeed one of only very few examples across the world 35 00:03:17.860 --> 00:03:18.500 with aged care services. 36 00:03:18.500 --> 00:03:18.520 It's the first time in Australia and indeed one of only very few examples across the world with aged care services. 37 00:03:18.520 --> 00:03:25.120 It was funded in 2017 by the South Australian Premier's Research and Industry Fund 38 00:03:25.120 --> 00:03:30.360 and by members of the Healthy Ageing Research Consortium here at SAMRI 39 00:03:30.360 --> 00:03:33.980 under the directorship of Professor Steve Wessling. 40 00:03:34.780 --> 00:03:39.680 It's designed by Associate Professor Maria Inacio, who's in our audience today, 41 00:03:39.900 --> 00:03:42.700 who is a registry and population surveillance expert. 42 00:03:43.180 --> 00:03:48.100 The aim of ROSE is to monitor the health and wellbeing of people accessing 43 00:03:48.100 --> 00:03:55.560 aged care services in Australia and in doing so learn more about the patterns of that aged care. 44 00:03:56.460 --> 00:03:59.760 While their individual identity and information remains private, 45 00:03:59.940 --> 00:04:05.820 their collated data, their big data, will be used to help shape and improve services 46 00:04:05.820 --> 00:04:08.040 and interventions into the future. 47 00:04:08.960 --> 00:04:12.240 The first stage of the project, which is already reported, 48 00:04:12.660 --> 00:04:18.080 has been an examination of the historical records of 2.9 million people 49 00:04:18.080 --> 00:04:23.820 across Australia who received 6 million different types of aged care 50 00:04:23.820 --> 00:04:27.280 between 1997 and 2014. 51 00:04:27.940 --> 00:04:32.740 We've just started a prospective phase, which began in April, 52 00:04:32.940 --> 00:04:38.520 which looks at the big data of 16,000 older South Australians 53 00:04:38.520 --> 00:04:41.180 who undergo an aged care assessment each year. 54 00:04:42.220 --> 00:04:46.740 These people are being asked and almost all are giving consent 55 00:04:46.740 --> 00:04:48.060 to have their aged care assessment done. 56 00:04:48.060 --> 00:04:58.600 After that, if they've said yes, then that's usually said. 57 00:04:58.600 --> 00:04:59.480 After that, if they've said no, then that's usually said. 58 00:05:00.160 --> 00:05:04.540 After that, if they've said yes, then that's usually said. 59 00:05:05.320 --> 00:05:07.100 After that, if they've said no, then that's usually said. 60 00:05:07.100 --> 00:05:07.480 After that, if they've said yes, then that's usually said. 61 00:05:07.480 --> 00:05:07.540 After that, if they've said no, then that's usually said. 62 00:05:07.540 --> 00:05:08.380 After that, if they've said yes, then that's usually said. 63 00:05:08.820 --> 00:05:11.100 After that, if they've said no, then that's usually said. 64 00:05:12.500 --> 00:05:16.960 After that, if they've said no, then that's usually said. 65 00:05:16.960 --> 00:05:23.680 care assessment to their information about ambulance use, about Medicare benefit schedule, 66 00:05:23.900 --> 00:05:29.500 about pharmaceutical benefit scheme, with the Australian death index, with emergency 67 00:05:29.500 --> 00:05:30.820 department use. 68 00:05:31.700 --> 00:05:37.920 And so it allows us to check how well those sectors are aligned in supporting the needs 69 00:05:37.920 --> 00:05:39.660 of older South Australians. 70 00:05:40.080 --> 00:05:45.920 The second great strength, of course, is the volume of data available and indeed the completeness 71 00:05:45.920 --> 00:05:47.860 of the available data. 72 00:05:48.900 --> 00:05:56.700 With 16,000 assessments each year and very few opt-outs, the evaluation of infrequent 73 00:05:56.700 --> 00:06:03.160 events, health assessments, surgically treated fractures, cardiac events, and rare conditions, 74 00:06:03.520 --> 00:06:08.960 liver or kidney disease, which can't normally be tracked through clinical studies because 75 00:06:08.960 --> 00:06:13.880 they take too long or too costly, can be followed. 76 00:06:13.880 --> 00:06:19.020 And we can therefore use ROSA to understand the patterns of resource use. 77 00:06:19.200 --> 00:06:22.420 We can ask it questions about risk profiles for events. 78 00:06:23.460 --> 00:06:28.720 And we can ask about the associations between various sectors that can't be evaluated in 79 00:06:28.720 --> 00:06:31.340 the absence of the availability of that big data. 80 00:06:31.760 --> 00:06:38.580 The third thing that ROSA has done is that it's assembled a pretty, I think, a very clever, 81 00:06:38.660 --> 00:06:43.800 sophisticated team, a cross-disciplinary team of programmers. 82 00:06:43.880 --> 00:06:47.300 Statisticians, epidemiologists, and data scientists. 83 00:06:47.660 --> 00:06:53.100 These are people who specialise in handling large amounts of data and are clever when 84 00:06:53.100 --> 00:06:58.880 it comes to creating ways to look at data, to design systems to collect and evaluate 85 00:06:58.880 --> 00:07:02.920 the data, and to use the best strategies for analysing them. 86 00:07:03.220 --> 00:07:13.040 Finally, from my perspective, ROSA also has very, very important clinical and consumer 87 00:07:13.040 --> 00:07:13.860 or research. 88 00:07:13.860 --> 00:07:22.680 And I think that's a really important part of the governance of ROSA. 89 00:07:22.920 --> 00:07:30.100 Because it means that it is designed to ask the right questions of the data that's available. 90 00:07:30.760 --> 00:07:36.320 It also means that we're able to access the opinions, the views about both clinicians 91 00:07:36.320 --> 00:07:40.660 and older people about what are the most important questions to ask of our data. 92 00:07:40.880 --> 00:07:43.840 If we could prioritise those questions, what would those be? 93 00:07:43.840 --> 00:07:48.980 It also allows us to interpret the findings appropriately. 94 00:07:49.800 --> 00:07:54.940 And it has avenues, and it means that we've got avenues to implement the findings of it. 95 00:07:55.040 --> 00:08:01.220 It means that the people who are very interested in what the outcomes of that data analysis 96 00:08:01.220 --> 00:08:04.540 are, are also the people who are part of the governance structure. 97 00:08:05.100 --> 00:08:12.440 So the project is made, the governance of the project is made up of geriatricians, aged 98 00:08:12.440 --> 00:08:13.340 care providers. 99 00:08:13.840 --> 00:08:19.560 But also consumer groups such as my own, the Council on the Ageing in South Australia 100 00:08:19.560 --> 00:08:20.920 and the Health Consumers Alliance. 101 00:08:21.340 --> 00:08:22.640 So that's ROSA. 102 00:08:22.700 --> 00:08:23.680 That's one example. 103 00:08:24.100 --> 00:08:31.660 I want to now introduce you to our panel members and start with, I think Stephen Graves is 104 00:08:31.660 --> 00:08:31.940 first. 105 00:08:32.240 --> 00:08:36.700 Stephen is an orthopaedic surgeon and he's been the director of the Australian Orthopaedic 106 00:08:36.700 --> 00:08:42.460 Association National Joint Replacement Registry since it was established in 1999. 107 00:08:43.220 --> 00:08:49.920 I'm going to give each of our panel members a bike bell reminder at six minutes because 108 00:08:49.920 --> 00:08:53.200 each have got seven minutes to tell you a lot of stuff. 109 00:08:53.480 --> 00:08:54.260 So, Stephen. 110 00:08:54.560 --> 00:08:57.620 Well, hopefully I can get through things a little bit quicker than that. 111 00:08:57.720 --> 00:09:03.180 And all I really want to do is just give you a quick overview of what the registry is and 112 00:09:03.180 --> 00:09:03.800 what it does. 113 00:09:05.060 --> 00:09:11.900 And hopefully leave plenty of time for discussion at the end when we've completed our presentation. 114 00:09:11.960 --> 00:09:12.440 So, Stephen. 115 00:09:12.440 --> 00:09:15.240 I'm going to give you a quick overview of what the registry is. 116 00:09:15.240 --> 00:09:21.140 So, as was mentioned, the registry was established around 20 years ago and it was actually an 117 00:09:21.140 --> 00:09:26.900 initiative of the Australian Orthopaedic Association, which is a professional body that covers all 118 00:09:26.900 --> 00:09:28.440 orthopaedic surgeons in Australia. 119 00:09:28.640 --> 00:09:33.440 And the reason it was established was that orthopaedic surgeons didn't really know how 120 00:09:33.440 --> 00:09:39.060 many people were getting joint replacements in Australia and what were the devices being 121 00:09:39.060 --> 00:09:42.380 used in that and whether those devices all worked equally. 122 00:09:42.380 --> 00:09:47.620 And what were the techniques that should be used in different clinical situations? 123 00:09:48.000 --> 00:09:52.700 Do you use a different technique with a younger patient compared to an older patient? 124 00:09:54.300 --> 00:09:55.760 And there was... 125 00:09:59.300 --> 00:10:05.120 So it was really to look at the practice of joint replacement surgery in the country because 126 00:10:05.120 --> 00:10:06.980 there was really no information on that. 127 00:10:07.540 --> 00:10:09.040 So we established the registry. 128 00:10:09.200 --> 00:10:10.440 I was asked to do that. 129 00:10:11.100 --> 00:10:11.660 And... 130 00:10:11.660 --> 00:10:15.720 What we did was we rolled it out in all the hospitals in Australia. 131 00:10:15.920 --> 00:10:20.260 So every hospital doing joint replacement contributes data to us on every procedure 132 00:10:20.260 --> 00:10:21.420 that is actually undertaken. 133 00:10:21.940 --> 00:10:26.140 And we're able to actually check that we get all the information because the state governments 134 00:10:26.140 --> 00:10:29.140 also provide us the information on who's had a joint replacement. 135 00:10:29.440 --> 00:10:33.200 And so we're able to match the data that we receive from the hospitals and from the state 136 00:10:33.200 --> 00:10:36.360 governments to identify everyone that's had a joint replacement. 137 00:10:36.680 --> 00:10:41.640 And the way that it works is actually very simple in that all we do is record that data. 138 00:10:41.660 --> 00:10:42.100 We collect that information. 139 00:10:42.220 --> 00:10:46.700 And then if someone has a problem with their joint replacement, they usually require what's 140 00:10:46.700 --> 00:10:48.840 called a revision operation or a redo operation. 141 00:10:49.200 --> 00:10:50.900 And so they get registered again. 142 00:10:51.100 --> 00:10:56.460 And then we link that data so that we know when someone has had to have their first operation 143 00:10:57.000 --> 00:10:57.240 redone. 144 00:10:57.400 --> 00:10:59.380 And so that indicates that there was a problem. 145 00:10:59.500 --> 00:11:01.560 And so we ask what the problem was and so on. 146 00:11:02.320 --> 00:11:10.080 And through just that very simple data collection, which was what we established initially, that 147 00:11:10.080 --> 00:11:11.640 we were able to identify the patients. 148 00:11:11.660 --> 00:11:15.300 And so we were able to identify a lot of variation in practice and a lot of things that were 149 00:11:15.300 --> 00:11:17.720 very good and a lot of things that weren't so good. 150 00:11:17.860 --> 00:11:24.480 You may remember that there was a problem a number of years ago with metal on metal 151 00:11:24.480 --> 00:11:30.320 hips and that it was the Australian registry that actually identified that problem. 152 00:11:30.500 --> 00:11:35.160 It was the first organisation globally that was able to identify that there was that problem. 153 00:11:35.340 --> 00:11:40.920 But as well as metal on metal hips, there's 150 different devices that we've identified 154 00:11:40.920 --> 00:11:41.460 that. 155 00:11:41.660 --> 00:11:42.180 And they're all very good. 156 00:11:42.180 --> 00:11:43.420 But they don't work particularly well. 157 00:11:43.500 --> 00:11:45.460 And they've been removed from the market in Australia. 158 00:11:45.840 --> 00:11:50.240 But it's changed the rules around how new devices get introduced, the evidence that's 159 00:11:50.240 --> 00:11:50.700 required. 160 00:11:51.000 --> 00:11:54.000 And so that's actually significantly improved the outcome. 161 00:11:54.680 --> 00:12:00.200 Identifying different patient groups and what is required for those different patient populations 162 00:12:00.200 --> 00:12:03.580 has also been very important in improving outcomes. 163 00:12:03.880 --> 00:12:09.120 And in fact, over the years, we've actually halved the number of people that are requiring 164 00:12:10.380 --> 00:12:18.140 having redo hip operations and that we've reduced the number of people having redo knee 165 00:12:18.140 --> 00:12:20.260 operations by about 30, 40%. 166 00:12:20.260 --> 00:12:22.840 So they're quite dramatic changes. 167 00:12:22.980 --> 00:12:29.140 If you want to put it in, it's thousands of people each year because there's 110,000 people 168 00:12:29.140 --> 00:12:31.840 who have a hip or knee replacement each year in Australia. 169 00:12:32.060 --> 00:12:37.280 So bigger than an MCG full of people each year have hip or knee replacements in this country. 170 00:12:37.540 --> 00:12:38.940 And that's a big change. 171 00:12:39.020 --> 00:12:39.100 And that's a big change. 172 00:12:39.100 --> 00:12:39.660 And that's a big change. 173 00:12:39.660 --> 00:12:45.600 There was an independent assessment as to the value of that over just a 10-year period. 174 00:12:45.780 --> 00:12:52.240 And the benefit was around $640 million to the Australian community in savings. 175 00:12:52.380 --> 00:12:56.460 So that means that more procedures can be done or that money can be spent elsewhere 176 00:12:56.460 --> 00:12:57.660 within the health system. 177 00:12:57.900 --> 00:13:02.440 So it's been very, very dramatic, the benefit. 178 00:13:02.640 --> 00:13:08.720 But not only has it benefited the joint replacement patients in Australia, it's also benefited 179 00:13:08.720 --> 00:13:09.080 them in Australia. 180 00:13:09.080 --> 00:13:09.660 So that's a big change. 181 00:13:09.720 --> 00:13:15.120 Because one of the things the Australian Registry did was link all the registries that exist 182 00:13:15.120 --> 00:13:15.720 globally. 183 00:13:15.760 --> 00:13:19.160 And there's about 50 national registries globally in joint replacement. 184 00:13:19.340 --> 00:13:21.400 And we actually all work very closely together. 185 00:13:21.520 --> 00:13:22.520 It's a very tight network. 186 00:13:22.720 --> 00:13:26.940 So we're able to compare techniques across countries, results across countries. 187 00:13:27.060 --> 00:13:32.820 And so we can help to optimise healthcare globally with respect to joint replacement. 188 00:13:33.100 --> 00:13:35.700 So that was a very simple data set that we collected. 189 00:13:35.940 --> 00:13:37.700 And since then, we've actually grown. 190 00:13:37.880 --> 00:13:39.060 And there's a couple of areas that we've improved. 191 00:13:39.080 --> 00:13:48.980 One is that the federal government has given us permission to link all of our data to their 192 00:13:48.980 --> 00:13:49.580 data sets. 193 00:13:49.820 --> 00:13:59.000 So it's the MBS data, which is all the hospital admissions and so on, as well as all the pharmaceutical 194 00:13:59.000 --> 00:14:00.280 data, the PBS data. 195 00:14:00.580 --> 00:14:05.240 So that means that we get a much better idea of how healthy people are and whether their 196 00:14:05.240 --> 00:14:08.940 health deteriorates or gets better after joint replacement surgery. 197 00:14:09.080 --> 00:14:11.300 And so we can look at what medications they're taking. 198 00:14:11.740 --> 00:14:16.800 I might mention that all of that information, when it's provided to us, is de-identified. 199 00:14:16.980 --> 00:14:18.580 We don't know who the patients are. 200 00:14:18.820 --> 00:14:21.440 And so we're just looking at the massive data. 201 00:14:23.360 --> 00:14:29.200 But the other thing that we're doing, and I think is a very good thing, is that we have 202 00:14:29.960 --> 00:14:34.680 developed an automated system which begins to be rolled out in the country in about two 203 00:14:34.680 --> 00:14:38.900 weeks' time, which contacts patients if they wish to be contacted. 204 00:14:38.920 --> 00:14:39.060 And that's a very good thing. 205 00:14:39.080 --> 00:14:41.560 We've also got a very good system for the patient to be contacted to get their own thoughts 206 00:14:41.560 --> 00:14:44.580 about how their health care has gone with respect to the joint replacement. 207 00:14:44.880 --> 00:14:51.700 So what we're doing is also getting information directly from the patients as to the success 208 00:14:51.700 --> 00:14:52.580 of the procedure. 209 00:14:52.800 --> 00:14:59.980 So we think that that will add a whole variety of new data that we can use to further enhance 210 00:14:59.980 --> 00:15:02.240 the outcomes of joint replacement surgery. 211 00:15:02.560 --> 00:15:03.160 Thank you. 212 00:15:03.720 --> 00:15:09.060 I'd now like to invite Professor Steven MacDonald, who's the director of the MBS Health and Health 213 00:15:09.060 --> 00:15:09.260 Services Centre, to speak to us. 214 00:15:09.260 --> 00:15:13.440 I'm a director of dialysis and senior staff nephrologist at the Royal Adelaide Hospital. 215 00:15:13.600 --> 00:15:18.420 But he's importantly also the executive officer of the Australia and New Zealand Dialysis 216 00:15:18.420 --> 00:15:20.260 and Transplant Registry. 217 00:15:20.540 --> 00:15:21.540 Thank you, Jane. 218 00:15:21.940 --> 00:15:29.500 So the Dialysis and Transplant Registry is based here, but as the name implies, collects 219 00:15:29.500 --> 00:15:35.320 information about everyone who receives long-term dialysis for kidney failure in Australia or 220 00:15:35.320 --> 00:15:37.340 New Zealand or has a kidney transplant. 221 00:15:37.640 --> 00:15:39.040 And I'm going to talk to you about that. 222 00:15:39.060 --> 00:15:47.000 So, as an entity, that's actually existed since 1977, but that was the point at which 223 00:15:47.000 --> 00:15:50.620 the separate dialysis and transplant registries were combined. 224 00:15:51.000 --> 00:15:55.980 And so that data, in fact, goes back to 1963, when those treatments began here in Australia 225 00:15:55.980 --> 00:16:01.560 and, in fact, began here in South Australia, where the first successful kidney transplant 226 00:16:01.560 --> 00:16:02.220 was performed. 227 00:16:02.720 --> 00:16:07.960 And one of the pleasures I have in my role across the courtyard is I'm the custodian 228 00:16:07.960 --> 00:16:08.880 of an old 229 00:16:08.880 --> 00:16:09.040 entrance. 230 00:16:09.040 --> 00:16:14.200 exercise book which actually records in long hand the first dialysis treatments in South Australia 231 00:16:14.200 --> 00:16:19.820 done down at the Queen Elizabeth Hospital and it was actually remarkable reading. But fast forward 232 00:16:19.820 --> 00:16:26.100 to the present you say well this is not a common problem and it is and it isn't and you say well 233 00:16:26.100 --> 00:16:32.080 why I'm here talking about big data is that kidney disease and dialysis and transplantation in 234 00:16:32.080 --> 00:16:38.540 particular have become major challenges for the health system and for the health of society and 235 00:16:38.540 --> 00:16:43.900 in particular the older segment of society. So if you look around the country at the moment there 236 00:16:43.900 --> 00:16:51.240 are around about 25,000 people who are in so-called end-stage kidney failure and of that about half a 237 00:16:51.240 --> 00:16:57.800 little over half receive long-term dialysis. Most of those three times a week through hemodialysis 238 00:16:57.800 --> 00:17:05.400 and the other group have functioning kidney transplants. Of the dialysis group a half, 50% 239 00:17:05.400 --> 00:17:08.220 are over 65 years in age. 240 00:17:08.540 --> 00:17:15.060 So as you work through the numbers if you look at the 65 to 75 year old group about one in every 340 people 241 00:17:15.060 --> 00:17:21.220 actually receive long-term dialysis. So again that's a lot of people across the country it's not millions 242 00:17:21.220 --> 00:17:26.880 and millions. Where we do get the attention of the health system is that every year somewhere between 243 00:17:26.880 --> 00:17:33.040 that costs the direct costs of providing that treatment are somewhere between 1.2 and 1.5 billion dollars. 244 00:17:33.700 --> 00:17:38.520 To put that into context the total public health expenditure, public health hospital 245 00:17:38.540 --> 00:17:44.720 expenditure in the country is around about 46 billion dollars. So we have a very big profile with the 246 00:17:44.720 --> 00:17:52.040 health service. And so from the health service perspective ANZDATA has a major role and I have frequent 247 00:17:52.040 --> 00:17:59.040 conversations with in fact every state and territory and nationally about understanding the numbers of 248 00:17:59.040 --> 00:18:06.660 people who receive treatment and what the types of treatment they receive are, where those people live 249 00:18:07.220 --> 00:18:08.520 and where they're going. 250 00:18:08.520 --> 00:18:12.920 And so we have a very big role in understanding where you expect those trends to go in the next five and ten years time. 251 00:18:13.180 --> 00:18:20.580 And so that is a direct role of the data collection that we have that informs hopefully good health service planning. 252 00:18:21.040 --> 00:18:29.140 What that data collection looks like is that we collect data not from individual patients but from the hospitals and the 253 00:18:29.680 --> 00:18:36.440 renal units that care for those patients. When patients begin dialysis or have a kidney transplant they are given a 254 00:18:36.440 --> 00:18:38.000 consent form and 255 00:18:38.000 --> 00:18:51.680 can opt out but as you've heard from Jane only a small number actually opt out. And we collect data about people's treatment on a yearly basis on an ongoing basis and so the data accrues that way. 256 00:18:52.180 --> 00:19:01.720 What that does end up providing is a rich data set that allows us to provide information that's not just useful for health services but for individuals. 257 00:19:02.360 --> 00:19:07.980 And one of the important things that's facilitated is providing accurate advice to patients about what they need to know. 258 00:19:08.000 --> 00:19:34.140 And so if you've got a kidney disease, if you've got a disease that's usually 259 00:19:34.140 --> 00:19:36.700 we can give you very exact advice 260 00:19:36.700 --> 00:19:41.360 about not just what the average expectation of survival 261 00:19:41.360 --> 00:19:43.100 and related quality of life is, 262 00:19:43.140 --> 00:19:47.840 but what the range is and what predicts how you will go. 263 00:19:48.100 --> 00:19:51.540 And similarly to what Steve said about the orthopaedic registry, 264 00:19:51.620 --> 00:19:53.920 that information really becomes important 265 00:19:53.920 --> 00:19:56.560 in integrating into decision-making. 266 00:19:56.720 --> 00:20:01.160 So if you're, for example, somewhere between 75 and 84 years of age 267 00:20:01.160 --> 00:20:02.820 and you just look across the board 268 00:20:02.820 --> 00:20:04.500 and you're about to start dialysis, 269 00:20:04.700 --> 00:20:08.340 the median survival, so the average if you like, 270 00:20:08.480 --> 00:20:09.940 is about three and a half years. 271 00:20:10.320 --> 00:20:12.020 Okay, and that's not too bad. 272 00:20:12.280 --> 00:20:15.120 It's not as good as the general population. 273 00:20:16.120 --> 00:20:17.680 Importantly, though, that's the average 274 00:20:17.680 --> 00:20:23.300 and 25% of people live less than 15 months, 275 00:20:23.500 --> 00:20:26.280 whereas 25% of people live over six years. 276 00:20:26.420 --> 00:20:28.980 So there's an enormous spread 277 00:20:28.980 --> 00:20:31.060 and similarly there's an enormous spread 278 00:20:31.060 --> 00:20:32.800 in people's quality of life. 279 00:20:32.820 --> 00:20:36.800 And understanding what predicts how you are going to go, 280 00:20:36.940 --> 00:20:39.920 you can see is therefore very important to understand 281 00:20:39.920 --> 00:20:43.280 and help inform people about the decisions 282 00:20:43.280 --> 00:20:44.580 that they need to make. 283 00:20:45.460 --> 00:20:48.960 And one of the things that that's driven 284 00:20:48.960 --> 00:20:53.500 is a trend in recent years to think very carefully 285 00:20:53.500 --> 00:20:56.520 about whether embarking on dialysis treatment 286 00:20:56.520 --> 00:20:58.900 in older people is, for that individual, 287 00:20:59.140 --> 00:21:00.760 a positive or a negative thing. 288 00:21:01.280 --> 00:21:02.800 And we have... 289 00:21:02.820 --> 00:21:06.040 been able to track this through linking our data set 290 00:21:06.040 --> 00:21:09.580 with the National Death Index and some other data sets 291 00:21:09.580 --> 00:21:12.340 so we can look at, if you like, the people that we don't know 292 00:21:12.340 --> 00:21:15.540 about who don't receive treatment but are hospitalised 293 00:21:15.540 --> 00:21:16.840 or die of kidney disease. 294 00:21:17.020 --> 00:21:19.140 And we can say that in that age group, 295 00:21:19.200 --> 00:21:22.040 about half of people embark on dialysis and half don't. 296 00:21:22.260 --> 00:21:23.860 So that's one example. 297 00:21:24.600 --> 00:21:26.340 There are a number of other examples 298 00:21:26.340 --> 00:21:32.040 where we've done data linkage to answer particular questions. 299 00:21:32.160 --> 00:21:32.800 We've demonstrated... 300 00:21:32.820 --> 00:21:35.440 for example, that people who have kidney transplants 301 00:21:35.440 --> 00:21:37.760 have substantially higher rates of cancer, 302 00:21:37.880 --> 00:21:40.300 in particular some types of cancer but not others, 303 00:21:40.400 --> 00:21:42.360 and that rolled through directly to screening. 304 00:21:42.780 --> 00:21:44.940 So, for example, if you have a kidney transplant, 305 00:21:45.000 --> 00:21:46.920 we know you have a higher rate of bowel cancer 306 00:21:46.920 --> 00:21:47.980 but not of breast cancer. 307 00:21:48.180 --> 00:21:51.280 And so you have screened for one but not the other. 308 00:21:51.420 --> 00:21:53.880 The other area that we've become involved with 309 00:21:53.880 --> 00:21:55.440 and so have the Joint Replacement Registry 310 00:21:55.440 --> 00:21:58.320 is using these big data sets as a means 311 00:21:58.320 --> 00:22:01.840 to run randomised controlled trials of new treatments 312 00:22:01.840 --> 00:22:02.240 that... 313 00:22:02.240 --> 00:22:04.320 are not going to be done by drug companies 314 00:22:04.320 --> 00:22:05.960 because they're not cost-effective 315 00:22:05.960 --> 00:22:07.400 or effectively they're not interested. 316 00:22:07.640 --> 00:22:10.100 As an example of that, we are running a trial 317 00:22:10.100 --> 00:22:13.080 looking at the settings of the sodium level in dialysis fluid, 318 00:22:13.180 --> 00:22:16.220 which is simply a change of a setting on a machine. 319 00:22:16.460 --> 00:22:19.100 It's an important question to us as doctors 320 00:22:19.100 --> 00:22:21.720 because it probably, we think, makes a difference. 321 00:22:21.780 --> 00:22:22.400 Nobody knows. 322 00:22:22.840 --> 00:22:25.340 It's not something that's going to make anybody any money 323 00:22:25.340 --> 00:22:28.860 but it's something that we are able to run at very little cost. 324 00:22:29.020 --> 00:22:31.840 And I think that's one of the things you will see much more 325 00:22:32.240 --> 00:22:34.740 of in the future out of big data sets 326 00:22:34.740 --> 00:22:38.240 and registries in particular is not just using them 327 00:22:38.240 --> 00:22:42.340 as platforms to accumulate data in an ongoing fashion 328 00:22:43.000 --> 00:22:45.360 but to actively, if you like, blur the boundaries 329 00:22:45.360 --> 00:22:48.560 which have existed in the past between a large silo 330 00:22:48.560 --> 00:22:51.720 of an observational data set and the other world 331 00:22:51.720 --> 00:22:55.320 of setting up a small study to attack a particular question. 332 00:22:55.660 --> 00:23:01.700 And finally, can I introduce you to Dr. Sigmund Spach? 333 00:23:02.260 --> 00:23:05.540 Who is a computer scientist and senior research associate 334 00:23:05.540 --> 00:23:08.820 in the School of Computer Science at the University of Adelaide. 335 00:23:08.900 --> 00:23:09.360 Sigmund. 336 00:23:09.460 --> 00:23:10.200 Thank you very much. 337 00:23:10.420 --> 00:23:13.000 So I actually work at the Australian Institute 338 00:23:13.000 --> 00:23:15.460 for Machine Learning which is now a research institute 339 00:23:15.460 --> 00:23:16.840 that's been established quite recently. 340 00:23:17.600 --> 00:23:20.320 It's going to be situated at the old RAH site 341 00:23:21.140 --> 00:23:23.620 and it's also an institute at the University of Adelaide. 342 00:23:23.800 --> 00:23:28.680 And what we do is we try to work on effectively using this data, 343 00:23:29.440 --> 00:23:30.680 this big data that we're speaking about, 344 00:23:30.680 --> 00:23:33.780 to create new algorithms, to find new patterns, 345 00:23:33.880 --> 00:23:35.720 to actually solve new problems. 346 00:23:36.100 --> 00:23:39.100 And I can give you a concrete example that's very pertinent 347 00:23:39.100 --> 00:23:43.620 to old ageing, to ageing and things like that. 348 00:23:44.120 --> 00:23:47.180 For the last three years I've been working with a local company 349 00:23:47.180 --> 00:23:51.400 on a new medical device, a handheld medical device that's meant 350 00:23:51.400 --> 00:23:54.540 to monitor wounds, in particular chronic wounds. 351 00:23:54.900 --> 00:23:57.940 So these are wounds that you develop like bed sores 352 00:23:57.940 --> 00:24:00.080 or they could be diabetic ulcers. 353 00:24:00.080 --> 00:24:02.980 And these types of wounds don't heal for a very, very long time. 354 00:24:03.120 --> 00:24:05.800 Some people can have them for six months or longer. 355 00:24:06.180 --> 00:24:09.400 And the people who are particularly prone to getting these wounds 356 00:24:09.400 --> 00:24:11.040 are the ageing population. 357 00:24:11.880 --> 00:24:15.220 And the issue that currently exists is that, you know, 358 00:24:15.260 --> 00:24:17.680 you might develop this wound, you'll get a community nurse 359 00:24:17.680 --> 00:24:21.620 to come and see you, the community nurse might put dressing on. 360 00:24:22.060 --> 00:24:25.120 But two weeks later another nurse comes. 361 00:24:25.220 --> 00:24:27.360 It's hardly ever the same person that comes to see you. 362 00:24:27.800 --> 00:24:29.920 And they have to make a judgement to see 363 00:24:29.920 --> 00:24:31.960 whether is the wound getting better or not. 364 00:24:32.220 --> 00:24:34.280 They're expected to make a lot of different assessments. 365 00:24:34.460 --> 00:24:36.760 They're supposed to determine how deep is your wound, 366 00:24:36.920 --> 00:24:40.120 how shallow is it, what's the area, what's the volume, 367 00:24:40.820 --> 00:24:44.020 what percentage of the tissue is dying, what percentage of it is healing. 368 00:24:44.500 --> 00:24:46.960 And it's a very subjective thing to assess. 369 00:24:47.220 --> 00:24:50.980 If I draw an arbitrary shape and ask you what percentage area is that, 370 00:24:51.180 --> 00:24:54.000 none of us would know how to do it except if it was a circle or rectangle. 371 00:24:54.840 --> 00:24:58.400 But these are the types of things that people are expected to do. 372 00:24:58.400 --> 00:25:03.060 And what's supposed to happen is if I can determine that your wound has gotten bigger, 373 00:25:03.260 --> 00:25:05.380 well, I know that the treatment is not working. 374 00:25:05.500 --> 00:25:07.980 And so I might try a different dressing, I might try a different interpretation. 375 00:25:08.560 --> 00:25:13.040 What we know is that a lot of wounds could actually heal a lot better 376 00:25:13.040 --> 00:25:14.660 if the world's best practice was applied. 377 00:25:14.980 --> 00:25:18.420 But in practice it's very difficult to apply because of all of this subjectivity. 378 00:25:18.880 --> 00:25:24.160 So the purpose of this device is to automatically assess the wound. 379 00:25:24.320 --> 00:25:27.620 So we develop algorithms that take a picture of the wound 380 00:25:27.620 --> 00:25:28.380 and then assess it. 381 00:25:28.400 --> 00:25:33.200 So we assess the picture and predict or quantify what percentage of the wound is dead, 382 00:25:33.300 --> 00:25:36.160 what percentage of it is healing, automatically determine the volume, 383 00:25:36.340 --> 00:25:38.320 automatically determine the depth and things like that. 384 00:25:38.620 --> 00:25:43.400 And we can only do this, so the key ingredient for this type of work is a lot of examples. 385 00:25:43.780 --> 00:25:47.420 So imagine the way you train a nurse, a wound nurse. 386 00:25:47.620 --> 00:25:51.660 You take this person and you show them lots of examples of wounds 387 00:25:51.660 --> 00:25:55.560 and you say, here, this is an example of necrosis and this is an example of sloth. 388 00:25:55.680 --> 00:25:58.380 And they study for several years and after a while they become proficient. 389 00:25:58.380 --> 00:26:01.240 You need to do something similar with computers. 390 00:26:01.540 --> 00:26:05.600 You need to do something similar if you want to teach a machine to automatically detect things. 391 00:26:05.800 --> 00:26:08.280 You have to show the machine lots of examples. 392 00:26:08.840 --> 00:26:13.260 So what we've been doing is been working with people who have had wounds 393 00:26:13.260 --> 00:26:17.000 who have volunteered their data and said, they've effectively said, 394 00:26:17.880 --> 00:26:23.140 I consent to you using my images of my wound to develop the system. 395 00:26:23.540 --> 00:26:27.880 And the system can only work if a lot of people do that because we need to, 396 00:26:27.880 --> 00:26:29.660 like everyone's wound is slightly different. 397 00:26:29.880 --> 00:26:32.840 You know, you'll have a bigger wound, a shallow wound, different cut, etc. 398 00:26:33.240 --> 00:26:35.680 So you need to have the variety. 399 00:26:36.000 --> 00:26:38.940 You need to see all the different ways in which something can manifest. 400 00:26:39.440 --> 00:26:43.700 And what's incredible about this process is that in some sense, 401 00:26:43.840 --> 00:26:48.280