1
00:00:00,330 --> 00:00:02,280
I'd like to welcome Karina Jones.

2
00:00:02,940 --> 00:00:11,880
Professor Karina Jones is here to talk to us about some of her work relevant to the big state epidemiologist motto, which just talks part of.

3
00:00:13,350 --> 00:00:20,940
She's an associate professor at the University of Swansea and she's worked on a number of big data initiatives which she will tell us about.

4
00:00:21,480 --> 00:00:27,150
She's also really involved in information governance and public engagement, which I believe this talk is really focusing.

5
00:00:28,590 --> 00:00:36,750
Her role is to ensure data protection and maximise the social accessibility of the data used across these various sources.

6
00:00:37,480 --> 00:00:40,730
Okay, thank you very much. Thanks very much for the invitation.

7
00:00:40,740 --> 00:00:45,569
It's great to be doing something on this really prestigious course and such a lovely summer as well,

8
00:00:45,570 --> 00:00:50,000
which is I know it's wonderful, but you know, we've got this air con, which is great, but isn't it good?

9
00:00:50,010 --> 00:00:58,580
We've got a good summer at last. And so okay, I'm going to talk about I like intriguing titles, the jugglers and the Black Cat,

10
00:00:58,590 --> 00:01:03,580
just to think, well, you've had a long day, you've been being taught now and people have been teaching you.

11
00:01:03,600 --> 00:01:07,950
So I'm just thinking I'll give you a title so that you think, Well, what's that going to be about then?

12
00:01:08,550 --> 00:01:16,320
But before I talk about that, I'll just tell you a little tiny bit about some of the big data initiatives I work with at the very beginning there.

13
00:01:16,320 --> 00:01:21,480
That's our building, that's the data science building where we house all the work that we do.

14
00:01:22,140 --> 00:01:25,020
So we have the same data bank which you may have heard of,

15
00:01:25,230 --> 00:01:33,180
which is an anonymous data safe haven of records about the population of Wales, so that about 3 million people in Wales.

16
00:01:33,180 --> 00:01:40,650
But this because it's been running now for ten years, but it's gotten retrospective data as I've got data on about 4 million people.

17
00:01:40,950 --> 00:01:47,100
So it can be GP records, hospital records, all sorts of things, screen in cancer registries.

18
00:01:47,310 --> 00:01:51,540
We've got awareness, births and deaths, but we've also got some of the data.

19
00:01:51,540 --> 00:01:56,760
That's not health. We've got education, housing, fire and rescue, that sort of thing.

20
00:01:57,120 --> 00:02:02,880
And the way it works is the data are anonymized by an NHS body and then they come to us,

21
00:02:03,030 --> 00:02:08,070
but they didn't link form because it goes via a trusted third party that in link up form.

22
00:02:08,250 --> 00:02:16,830
So we can have a lot of records, a lot of details about people, but not in a way that has any personal information attached to it.

23
00:02:16,980 --> 00:02:21,750
We only have an anonymous linking field that has no currency outside the system,

24
00:02:22,230 --> 00:02:27,120
and there's a lot tons and tons of research going on with the same databank, if you want to look.

25
00:02:27,120 --> 00:02:30,389
I mean, I think they will distribute the slides for you if you want to.

26
00:02:30,390 --> 00:02:34,650
And you can you can look up the same data bank if you're interested then as well

27
00:02:34,650 --> 00:02:39,690
as what we've just become one of the health data research UK substantial centres.

28
00:02:39,990 --> 00:02:45,330
That's the Amnesty's new investment now in place of the FA Institute though it is a little bit different.

29
00:02:46,110 --> 00:02:53,399
That is set up a number of consortia across the UK to apply cutting edge techniques and to do research.

30
00:02:53,400 --> 00:03:03,360
It's really important for our public benefit and we're also an administrative data research centre funded by the ISC that are four in the UK,

31
00:03:03,360 --> 00:03:09,210
that also rejig in that investment as well. But that's to take forward the use of administrative data.

32
00:03:09,450 --> 00:03:15,000
It doesn't exclude health data, but it's more focussed on government data, local authorities,

33
00:03:15,180 --> 00:03:23,520
economic and social data, benefits, work and pensions and also things like education and on suchlike as well.

34
00:03:24,270 --> 00:03:30,360
And as as Emily mentioned, I'm the academic lead for data governance and public engagement.

35
00:03:30,510 --> 00:03:39,780
So that's about looking at how to keep the data as useful as possible, but also to be safe and to be able to be used in a socially acceptable way.

36
00:03:40,110 --> 00:03:45,569
Because if you think you can lock, date it up and make it aggregated and it's incredibly, incredibly safe,

37
00:03:45,570 --> 00:03:51,000
but it may not actually be useful for answering research questions or you can make it very open,

38
00:03:51,090 --> 00:03:54,090
but then perhaps you haven't got sufficient safeguards.

39
00:03:54,330 --> 00:04:04,080
So it's about finding that proportionate balance so that the data are useful but also can be done within a robust governance framework.

40
00:04:05,520 --> 00:04:09,690
So let's come to the jugglers. So we have a little bit of context.

41
00:04:10,410 --> 00:04:15,510
It's been suggested it can be easier to donate blood and even organs than to genetic data.

42
00:04:15,870 --> 00:04:20,490
This is something that came my way in another meeting I was at in Oxford, actually, in April.

43
00:04:20,760 --> 00:04:28,409
And that's an interesting sort. And if you think about donate, it comes from the Latin for donor, which means a gift.

44
00:04:28,410 --> 00:04:32,100
And the French donor is a very, you know, familiar derivative.

45
00:04:32,820 --> 00:04:36,389
But do you think then donating data is about gifting data?

46
00:04:36,390 --> 00:04:42,690
Donating organs is about gifting them? It's an interesting concept if you think about it in terms of a gift.

47
00:04:43,320 --> 00:04:50,100
So I'm just going to pause a few questions. Are we moving in the right direction with legislation and regulations around this?

48
00:04:50,820 --> 00:04:59,550
Is it possible to find a bioethical balance between individual autonomy, personal exploitation and concepts of social responsibility?

49
00:05:00,000 --> 00:05:05,280
And can we, as individuals and as decision makers, make properly informed choices?

50
00:05:06,990 --> 00:05:10,800
Now, the thing with data is there's a tangibility issue.

51
00:05:11,190 --> 00:05:21,000
Debt is not like a thing. And when you think about ownership and rights, they usually about the right or the act or state of possessing a thing.

52
00:05:21,300 --> 00:05:23,700
And data is not tangible. It's not a thing.

53
00:05:24,270 --> 00:05:33,660
So the laws around data tend to deal with data protection, safeguarding privacy and confidentiality and take and just as a rule of thumb.

54
00:05:33,930 --> 00:05:37,980
Privacy would be about a person confidentiality about the data.

55
00:05:38,250 --> 00:05:45,030
But it's not about ownership of data as much as it is about safeguarding privacy and confidentiality.

56
00:05:45,600 --> 00:05:49,140
Because if you think about it, how can you own what you can't control?

57
00:05:49,380 --> 00:05:54,540
Once someone else is aware of it? If I tell you two things about myself that you don't know.

58
00:05:54,780 --> 00:05:58,920
One is I really, really don't like butternut squash. I think it's awful.

59
00:05:59,400 --> 00:06:04,260
But I want to say, Lambert, you know, now you now know two things about me that you didn't know.

60
00:06:04,380 --> 00:06:06,300
And I really don't mind, you know, in those things.

61
00:06:06,600 --> 00:06:12,420
But there are lots of other things about ourselves that perhaps we don't want people to know or we only want some people to know.

62
00:06:12,750 --> 00:06:16,200
An analogy I heard, which I quite like, is you can think about data.

63
00:06:16,200 --> 00:06:21,149
It's a bit like music. You can hear it, you can't it's not it's not so easy to own it.

64
00:06:21,150 --> 00:06:22,460
You can copy that music. Okay.

65
00:06:22,470 --> 00:06:29,940
But it's quite similar, isn't it, in that, you know, when somebody is aware of data, aware of information, how can you own it?

66
00:06:30,180 --> 00:06:38,490
So the laws and the regulations are more about safeguarding your privacy and the confidentiality of the data rather than an ownership concept,

67
00:06:38,730 --> 00:06:42,900
though we do tend to say, my data, your data, but it is a bit tricky.

68
00:06:43,710 --> 00:06:48,660
And of course we haven't even discovered every type of data that could be considered personal.

69
00:06:49,080 --> 00:06:52,770
So that's a challenge in itself. And there have been recent changes.

70
00:06:52,890 --> 00:06:55,080
If we go back however many years we need to,

71
00:06:55,200 --> 00:07:01,200
when there was no social media that couldn't be considered in data protection legislation because it didn't exist.

72
00:07:01,440 --> 00:07:09,730
And so legislation gets changed to add the concepts of these different kinds of data as we progress and they become available.

73
00:07:11,340 --> 00:07:14,130
But let's think about what it is we could be donating.

74
00:07:14,880 --> 00:07:21,240
So if we think about our little data subject, you know, I've called him Schrodinger's pot and I guess you're going to work out why.

75
00:07:22,230 --> 00:07:27,870
So Pat could donate all sorts of things that pertain to himself.

76
00:07:28,410 --> 00:07:35,040
It could be paired organ, could be a vital organ, could be some tissue, some blood, some stem cells,

77
00:07:35,040 --> 00:07:43,320
perhaps when he was very small, or perhaps his needs to decide about that for a baby or a change of his his DNA,

78
00:07:43,920 --> 00:07:56,640
his health records, social media data, data collected as he uses his phone, something collected via an app like Mapmyrun or store loyalty card data.

79
00:07:57,600 --> 00:08:05,280
Or he could donate his whole body after he dies, which is why he should take part, because, of course, if you don't, it's all about your liver.

80
00:08:05,490 --> 00:08:09,060
Then he must be dead before all of the others he can do while he's alive.

81
00:08:09,750 --> 00:08:15,600
But whatever he donates, even if it's tissue or organs, data are being generated.

82
00:08:15,900 --> 00:08:20,190
So when we say Is it harder or is it easier to donate data or tissue,

83
00:08:20,370 --> 00:08:27,330
we've got to think that even when it's tissue, it's donating data as well, which is another interesting twist.

84
00:08:28,380 --> 00:08:33,450
So if we think about the different sort of data to whom and how we donate it,

85
00:08:33,900 --> 00:08:41,520
we have a store loyalty cards which we go into Sainsbury's or Tesco's and we swipe the card and we think, Great, we're going to get some points.

86
00:08:41,910 --> 00:08:49,320
But the truth is that data, this is of much more value to the shops and the little points we get back and then we get £10 to spend in the shop.

87
00:08:49,590 --> 00:08:52,620
But really it's for the market advantage.

88
00:08:52,950 --> 00:08:57,450
They do it something nice for the customer, but the primary reason is to gain an advantage.

89
00:08:58,140 --> 00:09:02,250
Similarly with social social media, we can share things with friends and family,

90
00:09:02,520 --> 00:09:07,320
but they are gaining information on us which they can use for their own purposes.

91
00:09:07,560 --> 00:09:09,810
And of course we know that they do that.

92
00:09:10,560 --> 00:09:17,280
We'll all be aware of the recent Cambridge Analytica debacle, as it tends to be called debacle after Cambridge Analytica.

93
00:09:17,280 --> 00:09:21,050
That wasn't that. It's a bit like the conciliatory service, ACOSS.

94
00:09:21,170 --> 00:09:26,220
You know, it's one of those things that goes together where they basically Facebook allowed them

95
00:09:26,460 --> 00:09:32,550
to put on a personality test where people filled in details about themselves and then

96
00:09:32,550 --> 00:09:37,410
the company profile people according to the things they clicked like on and then use

97
00:09:37,410 --> 00:09:43,080
that information to target adverts to try and influence the US presidential elections.

98
00:09:43,670 --> 00:09:52,950
And so there's this, things like that. And then just by reason of using our mobile phones or smartphones, every time we make a call or send a text,

99
00:09:53,160 --> 00:09:58,860
the company collects a call detail record which gives the location where.

100
00:09:58,960 --> 00:10:03,670
We are where the recipient of the call is, whether it's a text message or a call.

101
00:10:03,970 --> 00:10:08,590
The duration of the call. So there's some sorts of information just being collected there.

102
00:10:08,830 --> 00:10:11,830
They can also profile the URLs you've visited as well.

103
00:10:12,730 --> 00:10:19,270
And we use all sorts of apps, you know, just like Mapmyrun for GPS locations or all sorts of other reasons.

104
00:10:20,080 --> 00:10:24,970
And then we have medical data, which does tend to be better regulated, of course,

105
00:10:25,510 --> 00:10:32,680
where we have our health status shared, perhaps with a practitioner or a researcher, a genetic data similarly.

106
00:10:32,680 --> 00:10:38,740
But we may also choose to buy some sort of genomic testing via a direct to consumer company.

107
00:10:39,280 --> 00:10:46,540
So that's a different sort of thing because that isn't really the same as someone who's sitting down with us and explaining things to us.

108
00:10:46,780 --> 00:10:54,280
That's a matter of pay. In $200 engaging a company, they send you a little thing to do, a cheek swab, and they send you back a result.

109
00:10:54,670 --> 00:11:03,010
So we provide our data to these companies in different ways, and we may also have apps to monitor our health a health condition.

110
00:11:03,310 --> 00:11:07,510
And depending who owns that app, the data will go to a different place.

111
00:11:07,510 --> 00:11:10,360
Perhaps if it's owned by the health provider, that's one thing.

112
00:11:10,660 --> 00:11:15,670
But if it's owned by a third party, then they will benefit from the data being collected from that app.

113
00:11:16,540 --> 00:11:21,999
And we have everything from properly informed consent to scrolling like crazy and

114
00:11:22,000 --> 00:11:26,950
agreeing to terms and conditions which we I know I'm guilty of just tending to do that.

115
00:11:28,490 --> 00:11:33,070
And there was this story, it was in the BBC News just on I think it was Friday.

116
00:11:35,300 --> 00:11:40,550
What you might have seen. And there are lots of stories like this about what the tech giants really do with your data.

117
00:11:41,510 --> 00:11:46,010
What they did there was they talked to some children up to about age 13, I think they were.

118
00:11:46,280 --> 00:11:53,030
And they got them to read the terms and conditions and they found it very difficult to understand them to be fair, which I think is fair enough.

119
00:11:53,600 --> 00:11:59,510
And the researchers concluded that you probably needed a university degree in order to be able to understand the terms and conditions,

120
00:11:59,690 --> 00:12:08,960
to give a meaningful decision on the sort of things that happened is your location can be tracked even if you don't specifically allow it.

121
00:12:09,200 --> 00:12:12,439
So when you're carrying your phone around, you're just being tracked.

122
00:12:12,440 --> 00:12:20,390
Basically, data passed to affiliates. They're not just capped with the company that collects them and were then bound by the third party terms.

123
00:12:20,510 --> 00:12:24,410
Who's now got the data? And as I mentioned, Cambridge Analytica.

124
00:12:25,190 --> 00:12:31,460
Facebook tracks you if the app is on your device, even if you haven't got an account, that's what they found there.

125
00:12:31,730 --> 00:12:36,680
Yes, I know it's not just if you're not logged in, but even if you haven't got an account,

126
00:12:36,680 --> 00:12:39,860
if the app is on the phone, Facebook is follow in that phone.

127
00:12:41,060 --> 00:12:46,550
And LinkedIn tracks your scans with private messages.

128
00:12:46,880 --> 00:12:49,910
There's often good reasons for these things.

129
00:12:49,910 --> 00:12:55,000
You can see the inverted commas on the podcast where, you know, it may be done for privacy,

130
00:12:55,000 --> 00:12:58,460
it may be done for security, making sure that nothing inappropriate is done.

131
00:12:59,060 --> 00:13:06,770
But the fact is it's still going on. And what we need to think about is anything you put in those situations don't just assume it's private.

132
00:13:06,890 --> 00:13:16,070
You know, it may not be. It may be, but there are cases where it isn't because these are platforms of largely unaccountable power very often.

133
00:13:18,020 --> 00:13:23,120
But are we, at least with Socrates, do we know what we don't know? No, we don't know what we don't know.

134
00:13:23,930 --> 00:13:25,159
I recently did a study.

135
00:13:25,160 --> 00:13:34,880
I don't know if anybody remembers Moth and if it did remember Moth yet I managed to mangle the title of a project into a word acronym MOFT,

136
00:13:35,000 --> 00:13:42,470
which I won't spell out for you now, but it was on looking at how should data governance be done for using mobile phone data for research?

137
00:13:43,160 --> 00:13:49,100
And one of the things we asked people in a series of workshops which I could repeat here is,

138
00:13:49,100 --> 00:13:54,140
okay, has anyone read the terms and conditions for their mobile phone contract?

139
00:13:55,130 --> 00:14:01,100
Anyone know? Okay, you're just the same as everybody else I've ever asked in the study.

140
00:14:01,250 --> 00:14:07,100
Apart from I was doing a presentation in a law conference and it wasn't part of the study,

141
00:14:07,280 --> 00:14:10,969
but I just asked the question and in the whole room there were about 30 people and

142
00:14:10,970 --> 00:14:15,890
three in a law conference full of legal experts had read the terms and conditions.

143
00:14:16,160 --> 00:14:18,620
So it's something we don't do. Okay.

144
00:14:19,140 --> 00:14:26,300
And then there have been some interesting social experiments and over 500 students signed up to a fictitious social media channel.

145
00:14:26,930 --> 00:14:32,420
None of them had read the terms and conditions well enough to notice they'd agreed to hand over their first born child.

146
00:14:33,290 --> 00:14:38,210
There was another example where the terms of a different study, the terms and conditions, said,

147
00:14:38,540 --> 00:14:44,720
if you have read these terms and conditions, you can claim and it was a decent prize and it ran for three months.

148
00:14:44,810 --> 00:14:46,879
And then one lady emailed and then she said,

149
00:14:46,880 --> 00:14:51,500
I've read the terms and conditions and they gave it the place because nobody else had read them, you know.

150
00:14:52,220 --> 00:15:00,680
And even with quite sensitive information like these, direct to consumer genomic companies with 550 people who used the company,

151
00:15:01,070 --> 00:15:07,400
most of them thought they were aware of privacy issues and thought the risk of trouble and repercussions were negligible.

152
00:15:07,850 --> 00:15:13,520
But also 50% of men and about 30% of the women said they hadn't read the terms and conditions anyway,

153
00:15:13,820 --> 00:15:17,690
so they didn't really know whether there were privacy issues to be concerned about.

154
00:15:17,900 --> 00:15:21,170
They were just taking it as I want this thing, therefore I am going to do it.

155
00:15:22,100 --> 00:15:25,550
But what can we think about in terms of perceiving public opinion?

156
00:15:25,730 --> 00:15:29,120
And we can ask ourselves the question as well. It's not just the public out there.

157
00:15:29,270 --> 00:15:37,190
We are the public in this context as well. Do we really engage with information or are there very good reasons why we don't or can't?

158
00:15:38,990 --> 00:15:48,979
So if we look just at the legal position for a minute, Wales has adopted an organ opt out system, organ donation, opt out after death.

159
00:15:48,980 --> 00:15:53,480
We've done this since 2015. England is doing the same. I'm not quite sure when it's coming in.

160
00:15:54,680 --> 00:16:02,300
And the donation of human tissue while you're alive is under the Human Tissue Act, as you know, and the Human Tissue Authority.

161
00:16:02,810 --> 00:16:08,030
There was a study done by the Wales Cancer Bank and they found that over 95% of patients

162
00:16:08,390 --> 00:16:13,310
were willing for surplus tissue to be used by the Wales Cancer Bank for further studies.

163
00:16:13,550 --> 00:16:18,350
That's the sort of thing you often see with people who are close to an issue like that.

164
00:16:19,460 --> 00:16:23,870
On the GDPR, as we now know, has now come in and the new UK Data Protection Act,

165
00:16:24,230 --> 00:16:33,770
which has renewed provisions and some new ones for process in general and special category data as part of this opt out consent.

166
00:16:34,100 --> 00:16:41,090
Is no longer acceptable. So using it in research, let's say you can't just say this is going to happen unless you say no.

167
00:16:41,270 --> 00:16:43,850
And the same with using identifiable data.

168
00:16:44,690 --> 00:16:50,750
There's more clarity on data processing and the need to be more upfront about what the data controller is going to do.

169
00:16:50,900 --> 00:16:58,400
It's not just about having the information. If somebody wants it, you need to be more upfront, more more prospective about it.

170
00:16:59,120 --> 00:17:04,640
And there's an improvement on legacy tiny word escrow. Nobody reads it and sees as well.

171
00:17:05,060 --> 00:17:10,010
And a bit of a a no no on this automated profiling that was going on.

172
00:17:10,160 --> 00:17:13,820
Now, whether that'll have an effect on these massive, powerful platforms.

173
00:17:14,030 --> 00:17:20,420
But that's the idea. And clear provisions for using research in the public interest, which is a good thing,

174
00:17:20,720 --> 00:17:25,820
because the guidance now is it may not be necessary to rely on consent for some research,

175
00:17:26,060 --> 00:17:30,680
provided you can show and defend your position that it's in the public interest.

176
00:17:30,770 --> 00:17:33,020
I mean, it wouldn't be something like a clinical trial, let's say,

177
00:17:33,020 --> 00:17:38,360
but if it's research that's using data, there may be cases where that can be relied upon.

178
00:17:39,650 --> 00:17:46,790
But coming back to the organs, the organs of deceased people can generate DNA and other health related data about living individuals.

179
00:17:47,210 --> 00:17:51,920
So we've got an opt out consent for organ donation after death.

180
00:17:52,220 --> 00:17:59,030
But that organ donation can generate DNA profiles about people that are alive that are related to this person.

181
00:17:59,480 --> 00:18:11,090
So there are some incongruities there, at least it appears so far on some ethicists actually argued against that opt out organ donation after death.

182
00:18:11,750 --> 00:18:17,540
And another query that's come up is, does a donation bid passively lose the spirit of a gift?

183
00:18:17,750 --> 00:18:22,850
If we don't know we're giving something? I don't mean the organ donation after death, because that's got to be specified.

184
00:18:23,180 --> 00:18:30,590
But in terms of passive data collection, if we're doing it passively and we haven't really engaged or been able to engage,

185
00:18:31,220 --> 00:18:38,120
maybe we shouldn't really be thinking about that as a donation, but need to be thinking about do we really want to do that or should it be considered

186
00:18:38,300 --> 00:18:44,510
in another way because we don't have opt out donation for posthumous data?

187
00:18:44,810 --> 00:18:49,040
You know, which is interesting. If we can do it for organs, we can't do it for data.

188
00:18:51,110 --> 00:18:54,320
But of course, Pat isn't an island entire of himself.

189
00:18:54,770 --> 00:19:01,550
And when people make decisions for all of us, we need to think about the other people that we implicate.

190
00:19:02,090 --> 00:19:08,600
And we generally take these as four bioethical principles in making decisions based on bioethics.

191
00:19:09,050 --> 00:19:13,850
So autonomy, the right to make informed decisions free of coercion.

192
00:19:14,510 --> 00:19:20,120
Not maleficent's. Not intentionally harming someone through acts of commission or omission.

193
00:19:20,780 --> 00:19:25,730
Beneficence. The duty to benefit patients, individuals and society and justice.

194
00:19:25,910 --> 00:19:28,430
Equity in provision of care and resources.

195
00:19:29,120 --> 00:19:35,420
But coming back to that question from the beginning, where's that bioethical balance between individual autonomy,

196
00:19:35,570 --> 00:19:44,000
personal exploitation and concepts of social responsibility in these different data contexts where you've got data provided to practitioner,

197
00:19:44,210 --> 00:19:49,190
that's okay. Data provided to a researcher with all the correct procedures in place,

198
00:19:49,460 --> 00:19:55,670
but data provided to some where else, where we don't actually know where it's going.

199
00:19:56,060 --> 00:20:01,070
You know, we could be thinking, well, there are different sorts of ethical philosophies going on here.

200
00:20:01,250 --> 00:20:07,190
We could think, well, autonomy is that individual autonomy, or if that's in relation to a company,

201
00:20:07,400 --> 00:20:12,530
is it more related to we want this regardless of what effect it has on other people?

202
00:20:12,860 --> 00:20:16,490
Because we know that some of the companies I pick on Facebook, again,

203
00:20:17,030 --> 00:20:22,160
some people who work for them have come upfront and said, we've actually got psychologists working.

204
00:20:22,340 --> 00:20:26,060
We know what you know how to manipulate people.

205
00:20:26,240 --> 00:20:29,840
And an engineer, they're thinking in order to get them to do what we want.

206
00:20:29,930 --> 00:20:33,230
And they've they've come up and said that that's what is being done.

207
00:20:33,950 --> 00:20:34,609
Altruism,

208
00:20:34,610 --> 00:20:42,710
we think of as something the benefits are the people without necessarily benefiting ourselves and sometimes with a bit of a sacrifice attached to it,

209
00:20:43,220 --> 00:20:47,840
such as perhaps someone would be prepared to provide data about themselves.

210
00:20:47,840 --> 00:20:53,659
It may take a lengthy interview. They may need to go through some procedures, but it's not going to help them.

211
00:20:53,660 --> 00:21:01,790
But it may help research in the future, or utilitarianism where it may have benefits for a large number of people, including ourselves.

212
00:21:02,660 --> 00:21:10,219
But even so, to what extent can it be said decisions and practices are actually guided by bioethics when

213
00:21:10,220 --> 00:21:15,500
we've got this huge interplay of actors and where the might or right comes to the front.

214
00:21:17,120 --> 00:21:25,220
So if we think of ourselves and everybody, really, we are bombarded by information data from all sorts of places.

215
00:21:25,400 --> 00:21:33,170
We've got this little kind of data splat over here where all sorts of information about health comes to us books,

216
00:21:33,170 --> 00:21:39,980
news media, magazines, social media, health professionals of friends, all sorts of places.

217
00:21:40,070 --> 00:21:45,500
Information comes our way now for health care practitioners. And I worked in the NHS and I work in a university.

218
00:21:45,740 --> 00:21:53,810
We've got an advantage position because we're aware that there are good sources of information where to go to find them.

219
00:21:54,140 --> 00:22:00,560
But for the general public they may not be aware of that and they're more subject to this with these people being quite small,

220
00:22:00,650 --> 00:22:05,450
what is for us perhaps that a bigger influence and research may be a bigger influence?

221
00:22:06,410 --> 00:22:13,940
So when individuals make decisions to donate or not donate their data, we've got this multiplicity of data contexts.

222
00:22:14,360 --> 00:22:20,810
We've got active and passive data donation, we've got the getting something versus the giving something.

223
00:22:21,110 --> 00:22:30,440
If I want a mobile phone or if I need to Skype in the next 10 minutes and I don't have Skype got installed on, I must go through this long thing.

224
00:22:30,740 --> 00:22:35,809
I'm just going to press accept because I need the thing or I want the thing compared to

225
00:22:35,810 --> 00:22:42,500
giving where we see ourselves as data subjects in a study or a clinical trial where okay,

226
00:22:42,530 --> 00:22:49,820
I may be able to benefit someone in the future. There's a getting giving concept going on across these data domains.

227
00:22:50,540 --> 00:22:57,470
We also have the effect of distance from the issues where the whales, cancer bank patients, they were close to the issue.

228
00:22:57,710 --> 00:23:03,080
They could see, well, I've been helped by this. It's going to benefit others if I allow my data to be used.

229
00:23:03,380 --> 00:23:07,340
But if we're in a situation, let's take blood donation. Maybe we donate blood too.

230
00:23:07,340 --> 00:23:11,120
Maybe we don't. But if we think we could, yes, we will do it.

231
00:23:11,270 --> 00:23:17,840
But on the other hand, if there was a local need where there's a big problem near us or to people related to us,

232
00:23:18,050 --> 00:23:22,250
it may be much more foremost in our minds and we think we really need to do this.

233
00:23:22,550 --> 00:23:26,330
But some issues are important, whether or not they're important in our perception.

234
00:23:26,330 --> 00:23:31,370
And they the other thing, of course, is data and things like this.

235
00:23:31,370 --> 00:23:35,390
And research is not the only show in town. And I spent five I was on the train today.

236
00:23:35,720 --> 00:23:44,720
And sometimes I just think to myself, if I was to ask any of these people on this train, have you donated data today and who did you donate it to?

237
00:23:45,260 --> 00:23:48,260
I would imagine people say, no, no, I do that.

238
00:23:48,500 --> 00:23:54,110
No, I haven't donated any data. But the truth is, it's very unlikely that we haven't you know,

239
00:23:54,110 --> 00:23:58,910
if we're carrying around a mobile phone or even travelling in the car with cameras and such like,

240
00:23:59,150 --> 00:24:01,460
there's different sorts of information being collected.

241
00:24:01,700 --> 00:24:06,830
If we've got a Fitbit and all sorts of things like that, there will be information being collected.

242
00:24:08,150 --> 00:24:15,080
But then sometimes it gets personal and it hits that threshold and we have to make some sort of a decision,

243
00:24:15,230 --> 00:24:20,720
whether for ourselves or for somebody else. And then it's about finding a way through moral measures.

244
00:24:20,900 --> 00:24:23,910
And we come back to this situation now.

245
00:24:23,990 --> 00:24:30,500
How does Paterno, which is the bit of information he can trust, how does he know how to make his decision?

246
00:24:30,800 --> 00:24:34,160
It's it's quite a difficult thing when there's so much information.

247
00:24:34,340 --> 00:24:40,790
It's not about getting information anymore, is it, when, you know, maybe 50 years ago you needed to look hard for things and find things in books.

248
00:24:40,970 --> 00:24:46,520
Now it's easy to get information, but it's hard to be judicious and to know what we can trust.

249
00:24:47,810 --> 00:24:51,080
So we find ourselves juggling the ones and zeroes.

250
00:24:51,830 --> 00:24:58,470
We are, wittingly or unwittingly, we are the jugglers because we've got all sorts of things coming our way.

251
00:24:58,550 --> 00:25:03,260
Decisions we have to make, decisions we don't know. We've made all sorts of things going on.

252
00:25:04,190 --> 00:25:08,210
We can donate the same or similar data via different processes.

253
00:25:08,540 --> 00:25:14,450
We can sit down face to face with a researcher or a practitioner and talk about sensitive health data.

254
00:25:14,990 --> 00:25:18,500
Or we can put it on an app and we're not quite sure where it goes,

255
00:25:18,890 --> 00:25:24,800
or we can contact a direct to consumer genomic sequencing company and provide it to them.

256
00:25:25,640 --> 00:25:33,860
Or we can provide different data using the same or different processes, general data or very sensitive data, again, via an app, let's say.

257
00:25:34,220 --> 00:25:37,820
So different things in the same way, similar things in different ways.

258
00:25:38,630 --> 00:25:45,230
And there are limitations on individual choice because we're not always aware of who we are donating data to.

259
00:25:45,590 --> 00:25:49,820
What data do we mean? What will they do with it? Will they pass it on and so on.

260
00:25:50,480 --> 00:25:53,510
So sometimes I think it's too easy to donate data.

261
00:25:53,630 --> 00:25:56,690
It may be quite easy to donate tissue. Yes. Okay.

262
00:25:57,110 --> 00:26:03,920
But sometimes it's too easy to donate dated as well when we're subject to this information overload and distortion.

263
00:26:05,420 --> 00:26:13,910
So we just look at this as a little illustration here. If we think of ourselves as this thinking capital,

264
00:26:14,910 --> 00:26:22,920
we have to make some sort of decision either for ourselves or for the people we're caring for as practitioners or working with us researchers,

265
00:26:23,250 --> 00:26:29,910
whether it's about research and big data, all about this is trying to be packed again with an unusual skin condition this time.

266
00:26:31,050 --> 00:26:40,170
If we think about information coming from primary sources and we just allow us, us to call this objective, depending how it gets to us,

267
00:26:40,560 --> 00:26:50,530
it'll go through different channels, different conduits, and depending how it's interpreted, it may be magnified, diminished or fragmented.

268
00:26:50,550 --> 00:26:53,550
By the time it gets to us and we may not know that.

269
00:26:53,940 --> 00:26:57,030
So all we can really do is to think to ourselves, well,

270
00:26:57,240 --> 00:27:04,860
we need to get as close as possible to the primary source and think about, you know, how does data come to us?

271
00:27:04,980 --> 00:27:08,070
Where can we go to be as close to the primary source as possible?

272
00:27:08,400 --> 00:27:14,850
Because I'm sure we've all been in a situation where someone has said something and it could be something quite ordinary,

273
00:27:14,850 --> 00:27:17,850
but you think that really doesn't work like that.

274
00:27:18,030 --> 00:27:25,110
However, they thought that, well, obviously somewhere down the line they've got some information from somewhere and it's got a bit mangled on the way.

275
00:27:25,500 --> 00:27:31,440
So I think this sort of thing is just with thinking about when you read information as part of your course,

276
00:27:31,440 --> 00:27:38,080
as part of your work, as part of just being a person, you know, think about where does that come from?

277
00:27:38,100 --> 00:27:42,390
And can I get back to as primary a source as possible?

278
00:27:44,410 --> 00:27:47,920
So having got this far, I'm sure there's got to be time for a bit of alliteration, though.

279
00:27:48,010 --> 00:27:53,649
Everybody brings a bit of a little rationing. So minding our P's and Q's, we've got a series of unknowns.

280
00:27:53,650 --> 00:27:57,760
When we deal with data, we've got the unknown package, the data content.

281
00:27:58,000 --> 00:28:05,079
We may not know what data someone holds about us, or we may not know, for example, with, let's say, genomic data,

282
00:28:05,080 --> 00:28:11,140
what the data actually mean, because only some of it is known, you know, it's not as if it's all fully known as yet.

283
00:28:12,040 --> 00:28:20,410
We've got unknowns in who might be using the data. We've got unknowns in how they might use it, on where that data might sit.

284
00:28:20,680 --> 00:28:26,979
Is it somewhere in a data safe haven where it's under secure governance or

285
00:28:26,980 --> 00:28:31,000
is it sitting on somebody's desktop which could be shared with someone else?

286
00:28:31,150 --> 00:28:37,209
We may not know, but I think we need to stop asking questions that can't be answered properly.

287
00:28:37,210 --> 00:28:41,110
Now, what I mean by that is it is something that is changing naturally,

288
00:28:41,350 --> 00:28:46,060
but there's been a lot of work asking the public, do you think your data should be used for research?

289
00:28:46,360 --> 00:28:54,040
Well, let's face it, data are being used for research, and asking that question isn't really going to help anybody anymore,

290
00:28:54,190 --> 00:28:57,760
because if people say no, well, what are we going to do about it anyway?

291
00:28:58,030 --> 00:29:00,770
Because data are being used on a lot of data,

292
00:29:00,790 --> 00:29:09,519
they're being used by large companies who may not being may not be being as caring towards the data subject,

293
00:29:09,520 --> 00:29:15,370
perhaps as others are, as as as generally we should be for research integrity.

294
00:29:16,390 --> 00:29:21,640
So I think it is moving more towards how do you think data should be used?

295
00:29:22,420 --> 00:29:27,729
Who do you are you happy for to use your data or under what conditions are you happy

296
00:29:27,730 --> 00:29:31,660
for your data to be accessed rather than should it be used in the first place?

297
00:29:31,800 --> 00:29:33,910
And I think we are moving on from that.

298
00:29:34,930 --> 00:29:43,149
But we need to be transparent about information and custodians need to be able to show trustworthiness to us all,

299
00:29:43,150 --> 00:29:47,590
us as people, to show that is risk mitigation controls in place.

300
00:29:48,040 --> 00:29:52,179
Be clear, but also be clear about the limits and admit, yes,

301
00:29:52,180 --> 00:30:01,270
there are some unknowns and perhaps we can't totally protect your data against every possible attack that could come from anywhere.

302
00:30:01,570 --> 00:30:08,350
For example, the Personal Genome Project, where people are able to share genomic and health data about themselves actually

303
00:30:08,350 --> 00:30:13,270
tells people We cannot totally guarantee the confidentiality of your information,

304
00:30:13,480 --> 00:30:16,809
so at least people can go in with your eyes open if they want to.

305
00:30:16,810 --> 00:30:23,080
They can. If they don't want to, then that's fine. So can we make informed choices?

306
00:30:23,170 --> 00:30:29,229
Well, I think only if we are asked and we ask the right questions and we're able to engage

307
00:30:29,230 --> 00:30:34,840
and we do engage because otherwise it's quite difficult if we keep that distance.

308
00:30:34,990 --> 00:30:41,049
There's a lot of work with public engagement, but I do think there needs to be a recognition.

309
00:30:41,050 --> 00:30:45,190
We've all got that responsibility to think about it more for ourselves.

310
00:30:46,000 --> 00:30:53,200
And there are now, as I say, some alternatives to consent and we need ways, interpretations of legislation and regulations,

311
00:30:53,710 --> 00:31:00,580
because the GDPR provisions for use in research in the public interest is very helpful for researchers just about data,

312
00:31:00,940 --> 00:31:09,550
because data can be anonymized by a trusted third party as long as that authorised to hold that data, and then the data can be used in anonymous form.

313
00:31:10,390 --> 00:31:14,980
The Digital Economy Act has provisions for data processing, for anonymization,

314
00:31:15,280 --> 00:31:21,580
for non-healthcare organisations, so for government and administrative data organisations.

315
00:31:21,910 --> 00:31:28,900
The Digital Economy Act allows the data to be anonymous by a trusted third party, whereas there wasn't a provision for that before.

316
00:31:29,650 --> 00:31:37,059
Now, as I mentioned, we have the opt out organ donation which is moving in England and we now have in England the National Data Guardian,

317
00:31:37,060 --> 00:31:43,660
opt out where people can object to their data being used for anything beyond the direct care.

318
00:31:45,100 --> 00:31:51,190
Some of the thoughts that are flying around are about licensing data uses compared to actually donating data.

319
00:31:51,700 --> 00:31:56,350
Now if we think back to our full piece, the what, the who, the how and the where,

320
00:31:56,860 --> 00:32:00,830
I think it would be challenging in terms of the practicalities and it would need some really clever,

321
00:32:00,880 --> 00:32:04,630
clever technical solutions and governance frameworks around that.

322
00:32:04,840 --> 00:32:08,020
So I'm not quite sure, to be honest, how that would work.

323
00:32:08,290 --> 00:32:11,379
If we think of donating and forgetting, is it licensing and remembering,

324
00:32:11,380 --> 00:32:19,660
but how would you then engage with that every time you need to be able to engage every possible time your data were going to be used.

325
00:32:20,530 --> 00:32:24,580
But we want to promote data sharing, whether it's donated or licensed.

326
00:32:24,790 --> 00:32:27,040
We know the Data Saves Lives campaign.

327
00:32:27,220 --> 00:32:34,210
This is really, really important and we need to be able to follow the rules and the regulations, but without being rigid.

328
00:32:34,490 --> 00:32:43,360
And we need to be able to make decisions that have consequences for people without being partial so that we have socially acceptable ways.

329
00:32:43,560 --> 00:32:50,850
To use data. So we need wise interpretations of sorry.

330
00:32:51,090 --> 00:33:00,060
And the it's really important to think about what can we lead, what can we shape and what do we need to follow in.

331
00:33:00,900 --> 00:33:10,540
So we come to the black cat. Now, as context for this, I was asked to carry out a study commissioned by the Nuffield Trust,

332
00:33:10,540 --> 00:33:16,570
along with a couple of other colleagues to review the harms for health and biomedical data.

333
00:33:17,800 --> 00:33:25,690
And we did that. And what we found was the largest problem was actually the maladministration of information governance regimes,

334
00:33:26,350 --> 00:33:34,930
not about not having the policies in place, but having them, but there being problems in actually how they were interpreted and implemented.

335
00:33:36,130 --> 00:33:43,450
I won't talk about that now, but what I want to talk about is the alternative, the harms that occur when data are not used.

336
00:33:44,050 --> 00:33:52,450
Now this can be difficult to pin down and there are lots and lots of domain specific efforts going on, like the All Trials campaign,

337
00:33:52,450 --> 00:34:00,580
where a lot of effort is going on to make sure trial data is registered, information is provided properly so people know what's going on.

338
00:34:01,240 --> 00:34:08,500
But the non-use of data, that's quite a tricky concept, and there's a lack of publications in this problem area.

339
00:34:10,090 --> 00:34:18,940
So what we did was an international case study using the published literature looked at why is health related data non-use difficult to ascertain?

340
00:34:19,810 --> 00:34:23,530
What are the sources of data non-use? What are some of the reasons?

341
00:34:23,890 --> 00:34:31,930
What are the implications for citizens in society? And I looked across clinical records, research domains and governance frameworks.

342
00:34:32,260 --> 00:34:35,140
It was focussed on health, but it is applicable more widely.

343
00:34:35,650 --> 00:34:44,230
And it's really important to point out that the harm due to the non-use of data is more than not gaining the benefits of data use.

344
00:34:44,260 --> 00:34:49,690
It's not just one or the other. It's not if you don't use the data, you will get those benefits.

345
00:34:49,840 --> 00:34:53,260
Yes, you will get those benefits, but other things will happen as well.

346
00:34:54,310 --> 00:35:00,580
So if we have a look, first of all, at the clinical regimes, first of all, with sources,

347
00:35:01,870 --> 00:35:07,430
there were a variety of different sources of why data are not use or sources of data non-use.

348
00:35:07,450 --> 00:35:10,810
One of them was the way in which case notes are managed.

349
00:35:10,960 --> 00:35:13,360
Now, this is something that you'll be all very familiar with.

350
00:35:13,660 --> 00:35:21,220
Often a patient has a number of conditions, and so the case notes maybe in one clinic or the other clinic or another clinic.

351
00:35:21,730 --> 00:35:28,390
But that is something that if if everything was beautifully joined up with electronic records, then that obviously wouldn't be a problem.

352
00:35:28,630 --> 00:35:34,840
But while we depend on Steele a lot on on paper records, that can be a problem,

353
00:35:34,840 --> 00:35:38,140
because when the records are looked for, they're not where they meant to be.

354
00:35:38,290 --> 00:35:44,410
Because possibly the interesting case is where someone has got a number of conditions that could be in a number of places,

355
00:35:45,310 --> 00:35:49,240
that could be data entry errors that can happen to anybody.

356
00:35:49,420 --> 00:35:54,410
I've got two really unusual examples for you.

357
00:35:54,760 --> 00:35:59,530
I was working in the Clinical Audit Office at one time and we had a clinical audit assistant

358
00:35:59,770 --> 00:36:04,150
and she was given a spreadsheet with all sorts of information from different departments,

359
00:36:04,660 --> 00:36:10,450
numbers of procedures, bed length, time stay, stays in hospital and suchlike.

360
00:36:10,840 --> 00:36:15,220
And she was adding and she was also needed to sort by date and by department.

361
00:36:15,970 --> 00:36:22,209
But when she added and then sorted, she sorted by date and sorted that column and didn't start the whole spreadsheet,

362
00:36:22,210 --> 00:36:30,610
the dates went in the column, the row stayed the same. Then she sorted on department and the column changed and the roles stayed the same.

363
00:36:31,750 --> 00:36:38,350
So basically as she was adding and sorting and adding and sorting, it was well, it was just completely useless.

364
00:36:38,350 --> 00:36:39,160
It was a shambles.

365
00:36:39,310 --> 00:36:47,590
It was in reverse, it was irreversible because it was randomised basically, you know, and it was found because of some guy, any man.

366
00:36:47,770 --> 00:36:52,450
No, no, that's not right. You know, on this, this obviously became quite a big problem.

367
00:36:52,900 --> 00:37:00,850
Now, I can make this one step worse. After she was found not to be able to work in clinical audit, she was given a job in medical records.

368
00:37:01,180 --> 00:37:09,669
And this is true. I've got another example for you where this was a researcher and he had 4000 records,

369
00:37:09,670 --> 00:37:15,460
quite wide database records, and he had them in SPSS and he wanted to put them in Excel.

370
00:37:16,150 --> 00:37:22,390
Okay. And he took a whole weekend, he printed the whole lot out and he literally hand copied them all in.

371
00:37:23,080 --> 00:37:28,389
And the only reason we found out was he produced a bar chart and he produced a batch and showed it to me.

372
00:37:28,390 --> 00:37:34,980
And I looked at the one it produced the previous week and it bore very little resemblance that it just wasn't even the same shape.

373
00:37:35,320 --> 00:37:40,060
I saw what happened there, and he had actually copied it in row by row.

374
00:37:40,240 --> 00:37:47,080
So some strange things do happen. Unfortunately, sometimes it's about limited data recording.

375
00:37:47,080 --> 00:37:52,180
We don't we don't record all data. We know some information is just not for recording.

376
00:37:52,180 --> 00:38:00,010
That could perhaps be useful. Sometimes there are no clinical systems that are suitable or they don't fit and data are in silos.

377
00:38:00,190 --> 00:38:03,400
We all know that people go to hospital and they're asked.

378
00:38:03,840 --> 00:38:09,719
Medication are you on? And people find it odd because they don't know why the GP has given them certain

379
00:38:09,720 --> 00:38:12,960
medication and the hospital doesn't know that they're taking that medication.

380
00:38:13,050 --> 00:38:16,890
But of course, as we know, it's not joined up, but the public often think it is.

381
00:38:17,040 --> 00:38:23,930
And find it odd when they go to outpatients and they're asked about things and think, well, well, I told that to, you know, and they just don't know.

382
00:38:24,130 --> 00:38:29,550
You really should be more joined up than that. So some of the problems that have happened.

383
00:38:29,760 --> 00:38:35,040
The National Audit Office did a huge review of case notes across 120 hospitals.

384
00:38:35,490 --> 00:38:40,170
And as an example, they did they looked at around two need to audit.

385
00:38:40,500 --> 00:38:48,059
And what happened was when they investigated the results of the audit were better than they should have been because the problem cases,

386
00:38:48,060 --> 00:38:53,350
as I mentioned with the case notes, the case those but in other places and they hadn't been included in the audit.

387
00:38:53,790 --> 00:39:00,410
So where a pregnant lady had a culpability or diabetes or something on the case notes weren't available.

388
00:39:00,420 --> 00:39:09,990
So what about 3% of the case notes that weren't available? But it tended to be a bit systematic in that they were the ones who had the conditions.

389
00:39:11,100 --> 00:39:20,040
We did a survey when we set up the UK EMS register and we asked all the neurology clinics in the UK how they collect their data.

390
00:39:20,340 --> 00:39:29,219
There were about 85 at the time and about 40 something replied to the survey and about six of them still only used paper or a word processor,

391
00:39:29,220 --> 00:39:30,780
no clinical system at all.

392
00:39:31,170 --> 00:39:39,059
And for many specialities I guess that's still the case where it may be people are using an Excel sheet or or an access database.

393
00:39:39,060 --> 00:39:45,240
I don't know if there's any nods going on. Or is everything where you work all beautifully put into a clinical system or is it

394
00:39:45,240 --> 00:39:52,410
still in some cases anybody know large amounts of paper with large amounts of free text?

395
00:39:52,560 --> 00:39:56,400
Yes. Free text which isn't coded or. Yeah, searchable.

396
00:39:56,580 --> 00:39:59,580
Yes, yes, indeed. Yes. Thank you. Yes. Yes.

397
00:39:59,580 --> 00:40:06,120
Although this is improving, it's still something that needs a lot of work and of course different coding standards

398
00:40:06,330 --> 00:40:11,820
so that one system doesn't talk to another and systems themselves are incompatible.

399
00:40:12,210 --> 00:40:16,800
The reasons I mean the NHS is wonderful. I totally believe that I've worked in the NHS.

400
00:40:16,950 --> 00:40:22,650
I wasn't clinical myself but totally aware of the work that goes on.

401
00:40:23,340 --> 00:40:29,970
But it's so massively overstretched and it's about the priority, it's about patient care.

402
00:40:30,330 --> 00:40:37,290
The work of the NHS is to try and make sure all the data is done right that needs to be brought in and made possible

403
00:40:37,500 --> 00:40:45,000
rather than expecting it to be done by magic in some countries as performance a payment performance model,

404
00:40:45,870 --> 00:40:51,180
which also influences the way things are done. And it's not by design.

405
00:40:51,390 --> 00:40:57,510
These things are done not from the beginning where the systems are created in order to be interoperable and suitable.

406
00:40:57,870 --> 00:40:59,250
It's a bit bolt on, isn't it?

407
00:40:59,340 --> 00:41:08,550
Well, I think from the top and not really looking across the big picture, the implications are hugely this happens misdiagnosis,

408
00:41:09,120 --> 00:41:14,580
unnecessary intrusion, where if information is missing, people may be asked the same questions again.

409
00:41:14,760 --> 00:41:18,839
They may need to go through the same tests again because information is just not

410
00:41:18,840 --> 00:41:24,030
available in order to proceed or procedures that they need to have may be delayed.

411
00:41:24,690 --> 00:41:32,130
There may be poor outcomes, emotional damage, deaths, litigation then against the NHS and of course additional costs.

412
00:41:34,540 --> 00:41:44,380
In research. If we're dealing with information based on clinical records, we may find our study is subject to selection bias.

413
00:41:46,650 --> 00:41:51,450
Because, of course, people that are sicker have more data entry.

414
00:41:51,640 --> 00:41:52,920
You know, they have more data records.

415
00:41:53,160 --> 00:42:00,810
So that in itself can cause a selection bias or they may be improper study design that influences how data are used.

416
00:42:01,230 --> 00:42:05,010
Sometimes results are held back for personal or political reasons.

417
00:42:05,790 --> 00:42:12,570
Sometimes this cherry picking where someone might do a study is X, influenced by, let's say,

418
00:42:12,690 --> 00:42:18,450
and choose a lot of variables on a huge data set so that something will be significant and then report,

419
00:42:18,540 --> 00:42:25,710
Hey, this is significant, but really it's been done in a way that's cherry picking, which calls into question research, integrity.

420
00:42:26,430 --> 00:42:32,340
Some examples. I don't know if you remember, this trial was in 2006.

421
00:42:32,670 --> 00:42:35,820
It was a an immunomodulatory drug.

422
00:42:35,940 --> 00:42:41,400
It was a first in man trial. And they gave the drug to four or five people.

423
00:42:41,580 --> 00:42:45,240
They had an awful immune response. I think a couple of them died.

424
00:42:46,230 --> 00:42:48,810
Sorry, that's the one. Yes, indeed.

425
00:42:49,440 --> 00:42:58,740
It later transpired during the investigation that a close enough drug to be relevant had been tried in one person and it caused a big problem.

426
00:42:58,980 --> 00:43:02,130
But this information hadn't been available at the time.

427
00:43:03,870 --> 00:43:11,820
Clinical trial success rates. Commercial clinical trials are reported to have 85% of them have positive results,

428
00:43:12,030 --> 00:43:16,530
whereas in non-commercial trials, only about 50% of them have positive results.

429
00:43:16,740 --> 00:43:21,810
That's a little quote from Ben Goldacre, and obviously he's thinking about that and that's interesting.

430
00:43:21,960 --> 00:43:29,310
How does that quite work then? This one about the arrhythmia drug that was quite a big disaster.

431
00:43:29,580 --> 00:43:33,540
About 100,000 people died before the problem was realised.

432
00:43:33,930 --> 00:43:39,629
And what happened was it was in the 1980s and a drug was developed for arrhythmia and it was given to

433
00:43:39,630 --> 00:43:45,420
people who'd had a heart attack and it was fine as long as it was given to people who'd had a heart attack,

434
00:43:45,420 --> 00:43:53,999
who had arrhythmia. But when it was given to people who didn't have arrhythmia but had had a heart attack, the died basically on 100,000 people.

435
00:43:54,000 --> 00:43:58,440
Over a hundred thousand people died before that mistake was realised.

436
00:43:59,100 --> 00:44:02,490
Sadly, there had been a small study but it wasn't published.

437
00:44:02,640 --> 00:44:07,860
But in the investigation it came forward and said, look, we we weren't able to publish this,

438
00:44:08,100 --> 00:44:13,230
but if we'd been able to publish it, then we may have been able to to stop this happening.

439
00:44:14,460 --> 00:44:22,860
The new anticoagulants. This is an interesting one and one of the big drug companies has marketed dabigatran.

440
00:44:23,460 --> 00:44:28,830
And the marvellous thing about the new anticoagulants is that they're meant to be this meant to be.

441
00:44:28,830 --> 00:44:35,190
No need for titrating one dose suits everybody and no need for the frequent monitoring.

442
00:44:36,150 --> 00:44:45,660
Well, it then transpired that hundreds of people died or had a severe bleed many, many more, many times more than on the older style warfarin.

443
00:44:46,260 --> 00:44:50,819
And when it was investigated, it was found that the drug company had wilfully withheld the data,

444
00:44:50,820 --> 00:44:54,960
and they knew that it did need to be titrated and it did need to be monitored.

445
00:44:55,470 --> 00:45:00,030
And they were fined something like 5 million by the FDA in America.

446
00:45:00,870 --> 00:45:08,940
But that drug is a blockbuster drug and it has an annual revenue of $1,000,000,000 and they were fined $5 million.

447
00:45:09,660 --> 00:45:14,610
Then after that, they also produced and marketed the antidote and then no able to sell that as well.

448
00:45:14,940 --> 00:45:20,010
So that's some of the problems that can happen with non-use of data.

449
00:45:20,280 --> 00:45:28,769
And of course, we have this problem with journal publications where we know it's so much easier to publish something that's got a positive result,

450
00:45:28,770 --> 00:45:33,570
a really clear finding than a negative result. Or basically this just didn't work.

451
00:45:35,040 --> 00:45:42,270
So why does this sort of thing happen? Well, there are few incentives for data sharing and few penalties for not.

452
00:45:43,110 --> 00:45:49,170
Sometimes it's about market share, intellectual property investment, either in money or effort.

453
00:45:49,890 --> 00:45:55,560
Kudos by data. You know, sometimes we're just under pressure and some people want to hold it up to themselves.

454
00:45:56,070 --> 00:45:59,100
Pressure to publish on publication bias, as I mentioned.

455
00:45:59,790 --> 00:46:00,840
But as a result,

456
00:46:01,290 --> 00:46:09,510
sometimes we're dealing with working with untrustworthy information because unreliable findings are taken forward because they look shiny.

457
00:46:09,810 --> 00:46:16,200
Whereas the second best or third best, which was actually true, isn't seen as so good and it doesn't get taken forward.

458
00:46:17,010 --> 00:46:24,660
True findings may be concealed with multiple, multiple deaths in terms of governance frameworks.

459
00:46:24,900 --> 00:46:30,120
These, of course, there to safeguard individuals, everything to be done properly.

460
00:46:30,720 --> 00:46:35,160
Often we'd be all familiar with lengthy and duplicative processes,

461
00:46:35,820 --> 00:46:45,620
inconsistent advice and subjective interpretation like that life through a lens image and sometimes excessive disclosure control even.

462
00:46:45,680 --> 00:46:51,350
And when data are anonymized, sometimes there's a drive for or we must do more.

463
00:46:51,350 --> 00:46:53,990
We must do more to it in order to make it safer.

464
00:46:55,610 --> 00:47:02,180
Sometimes there's non-acceptance by an organisation after the regulatory and governance approvals have been gained.

465
00:47:02,390 --> 00:47:11,360
Sometimes a data provider will still say no, they won't provide data, even though ethical approval and everything that's necessary is in place.

466
00:47:13,380 --> 00:47:16,170
Sometimes you get consent by us.

467
00:47:16,410 --> 00:47:24,330
There's an article in the UN Malformation Study where there was a difference, systematic difference between those who consented and those that didn't.

468
00:47:24,630 --> 00:47:30,990
And so that can affect the the study that also happens with what people have been up to,

469
00:47:31,000 --> 00:47:37,170
wicked problems such as substance misuse and psychosocial problems in young people.

470
00:47:37,410 --> 00:47:41,670
What is really hard to get hold of the people in order to get consent.

471
00:47:42,090 --> 00:47:49,410
But as a result, if you are if this insistence on consent, then you can't study the issue properly.

472
00:47:50,430 --> 00:47:54,540
There was a particular problem with substance misuse stated in the US where

473
00:47:54,540 --> 00:47:59,610
the data provider just decided with no changes in regulations or legislation,

474
00:47:59,910 --> 00:48:06,600
that they would now withhold any record that contained an instance of substance misuse from research.

475
00:48:07,140 --> 00:48:10,200
But that means when the subject can't be studied properly.

476
00:48:10,980 --> 00:48:14,660
But that was just something that needed probably needed a lot of fighting back.

477
00:48:14,670 --> 00:48:19,020
There was an article published on it. So I would hope that that has now been resolved.

478
00:48:20,640 --> 00:48:23,520
The reasons can be, of course, there may be framework changes.

479
00:48:23,940 --> 00:48:30,870
Often it's because responsibilities are unclear and there is fear of reputation, risk, fear of getting it wrong.

480
00:48:30,990 --> 00:48:37,020
It's easier to say no than it is to think what happens if I say yes and I get this wrong?

481
00:48:38,070 --> 00:48:43,200
There's this little term here, privacy protectionism, which was coined by some Australian researchers.

482
00:48:43,500 --> 00:48:51,060
And what they mean by that is adding extra controls on the data and doing more and more processing to the data to make it safer,

483
00:48:51,210 --> 00:48:53,010
but actually not making it safer at all.

484
00:48:53,160 --> 00:48:59,520
So it doesn't actually add safeguards, it just adds further and further controls and limits the utility of the data.

485
00:49:00,210 --> 00:49:05,070
And, of course, the to the extent public support for the use of all kinds of data isn't known.

486
00:49:06,090 --> 00:49:11,850
As a result, we get an information deficit which can be delayed or abandoned and benefits left and attained.

487
00:49:13,050 --> 00:49:21,600
I don't know about you, but sometimes we maybe I'm find and I'm seeing more of a caution since the GDPR and not necessarily because of the GDPR,

488
00:49:21,750 --> 00:49:23,430
but because of interpretations.

489
00:49:23,700 --> 00:49:29,730
All of a sudden, everybody's a legal expert because it's a new piece of legislation and lots of people have had to engage with it.

490
00:49:30,150 --> 00:49:36,000
I don't know about you, but you know, just the range of emails I have received from companies I do and don't know.

491
00:49:36,420 --> 00:49:42,209
You need to do basically nothing for us to continue to send you this thing or you need to do an awful lot of

492
00:49:42,210 --> 00:49:47,520
things and do lots and lots of clicking and saying yes and reading in order for us just to send you a newsletter.

493
00:49:47,820 --> 00:49:54,090
And I think, you know, it's just, again, an evidence of that, differences in how things are interpreted.

494
00:49:54,780 --> 00:49:59,400
And of course, the result, again, multiple deaths of financial burdens to societies.

495
00:50:00,120 --> 00:50:04,020
But it's not just the one thing at a time.

496
00:50:04,530 --> 00:50:05,820
There are combined effects.

497
00:50:06,210 --> 00:50:11,550
So if you think of this doctor's handwriting here, which we're probably very familiar to doctors, I could probably read that.

498
00:50:11,550 --> 00:50:19,380
And it doctors among us, if we think of how the data collected, there are issues there in data non-use,

499
00:50:20,340 --> 00:50:24,330
how it's managed and stored, what is actually managed and stored,

500
00:50:24,660 --> 00:50:30,990
how it's used for research, how it's taken on to inform policies and guidelines,

501
00:50:31,320 --> 00:50:37,920
and then how it is received back at the bedside or at school and what is collected at that point.

502
00:50:38,100 --> 00:50:43,410
That comes back around to the state treasury. So it's a kind of big synergistic effect.

503
00:50:43,560 --> 00:50:50,010
It's an effect that's that has a knock on effect. It's a trail and it can compound itself.

504
00:50:51,180 --> 00:50:54,900
So how serious is this problem? Well, as we know,

505
00:50:55,500 --> 00:51:03,090
the non-use of health data is implicated in the deaths of hundreds of thousands of people and billions of dollars in financial burdens for societies.

506
00:51:03,390 --> 00:51:08,580
And this is quite a nice quote from the managed health care executive.

507
00:51:08,910 --> 00:51:17,790
And what they said was the greatest threat, the biggest risk to people with diabetes or heart disease, cancer, HIV, etc.,

508
00:51:18,750 --> 00:51:23,310
seems not to be from unauthorised sharing or use of their personal health information,

509
00:51:23,460 --> 00:51:27,750
but rather from a failure to share or inadequate use of that information.

510
00:51:27,930 --> 00:51:32,610
And sometimes even valuing privacy or overprotected an individual's life,

511
00:51:32,790 --> 00:51:36,390
their health and the health of their families, their friends and their neighbours.

512
00:51:37,410 --> 00:51:45,420
I think that's a really, really interesting point because sometimes even it's not even you know what we mean by this, isn't it?

513
00:51:45,600 --> 00:51:49,860
What do we mean by protecting privacy, which I could waffle on about for quite a long time,

514
00:51:50,160 --> 00:51:56,190
because if we think about is something disclosed, if well, what happens if it is disclosed?

515
00:51:56,340 --> 00:52:01,680
Because sometimes the disclosure of information doesn't cause any problems and sometimes it does.

516
00:52:01,950 --> 00:52:06,810
So not everything needs to be private, but obviously some things do.

517
00:52:07,080 --> 00:52:11,790
And it's quite a difficult balance. So the recommendations from the.

518
00:52:12,050 --> 00:52:16,440
They were. There needs to be support for a culture change and data capture.

519
00:52:16,460 --> 00:52:21,680
How do we do this? It needs to be thought about and built in, not a bolt on solution.

520
00:52:22,040 --> 00:52:25,070
Decision makers need to think beyond the specific settings.

521
00:52:25,700 --> 00:52:30,260
Data capture systems need to be fit for settings and need to be interoperable.

522
00:52:31,620 --> 00:52:36,890
Need to promote onward data sharing. But recognising the investments that have been made by people.

523
00:52:37,100 --> 00:52:42,679
People may have worked hard for years to build a big cohort of data for research and

524
00:52:42,680 --> 00:52:46,850
it's understandable that they they need to be recognised for having done that work.

525
00:52:47,360 --> 00:52:56,630
But the need to be greater repercussions for wilful non-use of data, okay, accidental or unintentional or data is just not available.

526
00:52:56,900 --> 00:53:04,070
But where data are withheld because I don't want to publish yet because or I'm going to market this drug and then I'll

527
00:53:04,070 --> 00:53:08,300
deal with the repercussions because I know I'm going to make far more money than the fine them ever going to get.

528
00:53:08,580 --> 00:53:12,710
You know, we need something more than a tiny fine on a massive revenue.

529
00:53:14,210 --> 00:53:16,430
Governance models need to be proportionate.

530
00:53:16,910 --> 00:53:27,380
We don't need the same level of ethical scrutiny on such like for something that's dealing with something non-intrusive as we do with an intervention.

531
00:53:28,250 --> 00:53:34,790
We need to rebalance that tension between data privacy and utility with greater support for trustworthy re-use.

532
00:53:35,300 --> 00:53:44,750
I came across a term which was data obsolescence, and I just think, no, you know, we really need to be thinking about reusing data,

533
00:53:44,900 --> 00:53:49,730
not creating data for one project and then saying, okay, we can't use that again.

534
00:53:50,030 --> 00:53:53,960
That massive investment that's gone into that needs to be taken into account.

535
00:53:54,830 --> 00:53:58,040
We need to present the public with a balanced view of data use.

536
00:53:58,310 --> 00:54:01,790
Yes, there are some risks on the S that are huge benefits,

537
00:54:01,970 --> 00:54:08,840
but also there are problems when data are not used and if people choose not to allow their data to be used,

538
00:54:09,050 --> 00:54:18,680
even to say, okay, I don't know my data to be anonymized and used, well then there needs to be an understanding of what happens when data aren't used.

539
00:54:18,980 --> 00:54:22,700
And that's so that social responsibility can work both ways.

540
00:54:24,380 --> 00:54:27,140
Just some brief thoughts on the recent developments.

541
00:54:27,470 --> 00:54:35,570
It's a great thing that there's no the public interest provision for data use for research rather than needing to rely on consent all the time.

542
00:54:35,750 --> 00:54:41,900
Because obviously with the GDPR, the bar for consent has been raised and it is much more stringent.

543
00:54:42,650 --> 00:54:47,270
It's good that the Digital Economy Act allows data sharing for anonymization by the

544
00:54:47,270 --> 00:54:52,820
government departments that have been some comments on the National Data Guardian opt out,

545
00:54:53,420 --> 00:54:58,399
the opting out of personal data being used beyond direct care understanding patient data.

546
00:54:58,400 --> 00:55:02,930
I've written a couple of blogs on this because there is some concern that if people opt out,

547
00:55:03,080 --> 00:55:08,000
it could bias research and audit and service development because of,

548
00:55:08,000 --> 00:55:14,720
let's say, rare conditions, rare events, diversity, ethnicity and possibly differences in geography.

549
00:55:14,990 --> 00:55:19,970
So that, I think, needs to be carefully looked after.

550
00:55:20,210 --> 00:55:28,970
But it's obviously understandable that it needs to be done properly and not done in any way that the public wouldn't be finding acceptable.

551
00:55:30,650 --> 00:55:33,620
So there's a need for innovation in data governance.

552
00:55:33,920 --> 00:55:44,210
However, whatever data is generated from where Schrodinger's pot produces his his donations so that in order to get from data to information,

553
00:55:44,240 --> 00:55:46,190
there needs to be innovation.

554
00:55:47,000 --> 00:55:56,090
It is about but it's not just about following the legislation and regulations and the ethical, legal and social implications and due diligence.

555
00:55:56,090 --> 00:55:58,670
It is about them, but there's more to it than that.

556
00:55:59,090 --> 00:56:05,840
We need to take into account how these things are interpreted, how they're perceived, what sort of reservations are there,

557
00:56:06,050 --> 00:56:13,790
what are the aspirations of people present in the different perspectives that are stakeholders to consider the literature,

558
00:56:13,790 --> 00:56:22,370
the debate, the media, that information splat we looked at with that sort of shifting body of knowledge, it floats around shaping social realities.

559
00:56:22,640 --> 00:56:26,959
So if we're going to do this successfully, we need to be innovative.

560
00:56:26,960 --> 00:56:32,720
You need to do it properly because otherwise it risks becoming unethical through not using data.

561
00:56:32,850 --> 00:56:41,360
It would swing too far. So in terms of final words, we need to hone our juggling skills, all of us.

562
00:56:41,360 --> 00:56:47,480
And and as researchers thought responsible, trustworthy and socially acceptable use of data,

563
00:56:47,930 --> 00:56:55,790
because the non-use of data is like a black cat and it's like a large, agile, polymorphic, lethal black cat.

564
00:56:56,000 --> 00:57:00,079
And he certainly is out there and a better understanding of its nature.

565
00:57:00,080 --> 00:57:07,970
It needs to be understood because this is a global problem and he is definitely not truly Schrodinger's pet because he really is there.

566
00:57:08,830 --> 00:57:12,540
And this is the publication which you can have the. Tails off of.

567
00:57:12,780 --> 00:57:17,550
It's the first known study to address perspectives of health, debt and non-use.

568
00:57:18,030 --> 00:57:25,530
It's informing the parliamentary inquiry on research, integrity and the Wellcome Trust Initiative on understanding patient data.

569
00:57:26,070 --> 00:57:31,350
So that's my my cut. And just thank you very much and your questions.

570
00:57:31,350 --> 00:57:34,979
And that's a picture of where I'm from in Swansea, where I work in Swansea.

571
00:57:34,980 --> 00:57:37,560
I don't live there actually, but I'll leave the catch up for you.