1
00:00:11,310 --> 00:00:16,680
Hello. Welcome to the How Epidemics End project, which is based at the University of Oxford.

2
00:00:16,680 --> 00:00:23,280
My name is Erica Charters and in these videos I discuss with experts how they research disease in a variety of ways,

3
00:00:23,280 --> 00:00:28,680
as well as how they research how epidemics end. Today, I'm here with Crystal Donnelly,

4
00:00:28,680 --> 00:00:37,800
who's professor of applied statistics at the University of Oxford and also professor of statistical epidemiology at Imperial College London.

5
00:00:37,800 --> 00:00:40,860
Bristol, thanks for being here. Thanks for having me.

6
00:00:40,860 --> 00:00:50,040
I know that you studied a range of diseases you've studied, I think SA's tuberculosis and BSE amongst livestock and you've also studied Ebola.

7
00:00:50,040 --> 00:00:54,810
Can you talk a little bit about how you study these diseases and what you look at?

8
00:00:54,810 --> 00:00:59,790
Sure, so as you said, I'm a statistical epidemiologist,

9
00:00:59,790 --> 00:01:05,940
and many more people are familiar with the term epidemiologist now, so I study patterns of disease.

10
00:01:05,940 --> 00:01:12,000
But infectious disease is in particular, and that's important because with infectious diseases,

11
00:01:12,000 --> 00:01:17,040
the risk to you is not just dependent on your own behaviour, but the behaviour of people around you.

12
00:01:17,040 --> 00:01:22,860
And so that's true of animals as well, so we can see transmission between different animal species.

13
00:01:22,860 --> 00:01:30,150
So I've worked across the range of statistical and mathematical approaches to epidemiology,

14
00:01:30,150 --> 00:01:37,350
so measuring risk and also looking at patterns to see if we can quantify the dynamics of the disease.

15
00:01:37,350 --> 00:01:42,630
So anticipating when it might go from an increasing trend to a decreasing trend,

16
00:01:42,630 --> 00:01:46,650
and that can happen because of changes in behaviour, as we've seen with COVID.

17
00:01:46,650 --> 00:01:50,250
If we contact many fewer people, there's less transmission.

18
00:01:50,250 --> 00:01:55,890
Or it could be because a disease is actually becoming so common that there aren't as many susceptible

19
00:01:55,890 --> 00:02:01,470
individuals around anymore that we obviously don't want to be the case with a very serious disease,

20
00:02:01,470 --> 00:02:09,030
whether it's of animals or of humans. But by quantifying the patterns of who's getting infected, how it's transmitted from one group to another,

21
00:02:09,030 --> 00:02:14,990
we can understand more about how it's transmitting, but also what the opportunities are for control.

22
00:02:14,990 --> 00:02:21,140
So can you give us a few examples of how this works in practise, because I think probably a lot of people just know, oh,

23
00:02:21,140 --> 00:02:27,230
there's something called data, there's some models and some numbers, and then some kind of projection comes out the other end.

24
00:02:27,230 --> 00:02:35,960
So say for BSE or for something one of those, I think epidemic's that was very much in the news and in Britain.

25
00:02:35,960 --> 00:02:39,890
What exactly do you do and what kind of data do you rely on?

26
00:02:39,890 --> 00:02:48,770
OK, so BSE, which is bovine spongiform encephalopathy, which was also casually known as Mad Cow Disease that was first identified in the late

27
00:02:48,770 --> 00:02:55,580
80s and it was a disease that caused neurological deterioration and death in cattle.

28
00:02:55,580 --> 00:03:00,710
And so initially, there were studies to understand how it was transmitted,

29
00:03:00,710 --> 00:03:06,190
and it was transmitted actually from cattle being fed protein from infected cattle.

30
00:03:06,190 --> 00:03:11,450
And obviously, the farmers didn't know that when they were giving them the feed, but it was included cattle protein.

31
00:03:11,450 --> 00:03:20,120
And so that set up a mechanism for transmission within the species. So that had all happened before we got involved at Oxford.

32
00:03:20,120 --> 00:03:26,630
And when we became involved was when there was an announcement in the House of Commons, which was in 1996.

33
00:03:26,630 --> 00:03:35,270
That disease had been identified in humans, which was then called new variant of Yakub disease that was associated with BSE in cattle.

34
00:03:35,270 --> 00:03:39,110
So we then got access to surveillance data.

35
00:03:39,110 --> 00:03:45,890
So surveillance is how we watch for diseases as a collective population.

36
00:03:45,890 --> 00:03:54,050
And so then they recorded for individual cattle, which yeah, when they had the disease, what age the animal was,

37
00:03:54,050 --> 00:04:00,350
when it was onset, which form it was in and to some extent, what forms it had been on before that was known.

38
00:04:00,350 --> 00:04:06,470
And so we were able to look at the patterns of how many cases there were in each year and

39
00:04:06,470 --> 00:04:12,680
when those animals were born and see how there had been an increasing risk initially.

40
00:04:12,680 --> 00:04:16,610
But then measures that were brought in in 1988,

41
00:04:16,610 --> 00:04:23,510
which banned the feeding of ruminant protein to ruminants, to ruminants or things like cattle and sheep.

42
00:04:23,510 --> 00:04:32,480
And so they said no more feeding protein from those animals back to themselves, and that stopped the cycle of transmission.

43
00:04:32,480 --> 00:04:37,480
Predominantly, there may have been a little bit of cow to calf transmission.

44
00:04:37,480 --> 00:04:43,150
But it wasn't immediate, and that was what was difficult was the measure was brought in in the late 80s.

45
00:04:43,150 --> 00:04:48,310
But we didn't see the peak until years later and the reason for that is the incubation period.

46
00:04:48,310 --> 00:04:56,230
And that's the time from when an animal was infected to when it actually showed the clinical signs of disease, and that was five years on average.

47
00:04:56,230 --> 00:05:05,590
So that was quite a long time between when there had been a dramatic reduction in infections and when we saw the subsequent reduction in cases.

48
00:05:05,590 --> 00:05:09,550
But we were able to analyse the patterns and also additional data where they had done

49
00:05:09,550 --> 00:05:14,560
deliberate infections and saw how long those take to show clinical signs of disease.

50
00:05:14,560 --> 00:05:20,560
So it's bringing all those different pieces of data together, both to get estimates and particularly as a statistician.

51
00:05:20,560 --> 00:05:29,410
It's important to me to quantify how uncertain those estimates are because uncertainty can be key if you're deciding on a particular control measure.

52
00:05:29,410 --> 00:05:37,720
Well, there's a good estimate of how what the impact will be, how impactful a particular measure will be, but how uncertain are we?

53
00:05:37,720 --> 00:05:42,220
Are we very sure? Or do we think that that's our best estimate?

54
00:05:42,220 --> 00:05:48,880
But there's quite a lot of uncertainty and that can make a real difference, particularly if it's expensive to do.

55
00:05:48,880 --> 00:05:57,490
On what decision is made, I think that it's really fascinating to think about trying to to kind of pin down uncertainty in situations which

56
00:05:57,490 --> 00:06:04,810
obviously must be very fluid and in which obviously I'm sure politicians and the public really want certainty instead.

57
00:06:04,810 --> 00:06:09,370
And you're actually under pressure to kind of explain the limits of what you can

58
00:06:09,370 --> 00:06:14,050
provide detail on and what certain knowledge and what might be somewhat uncertain.

59
00:06:14,050 --> 00:06:18,340
And I was thinking, I know you've also worked on Ebola, which of course, also had, I think,

60
00:06:18,340 --> 00:06:23,290
a long development thinking about what it might mean to have projections and also

61
00:06:23,290 --> 00:06:28,180
thinking about what the end of Ebola would look like and how we might determine that.

62
00:06:28,180 --> 00:06:31,900
So can you explain a little bit about your research on that? Sure.

63
00:06:31,900 --> 00:06:36,460
So I initially started working on Ebola with colleagues at Imperial College,

64
00:06:36,460 --> 00:06:43,480
and when we got involved with helping the W.H.O. respond to the outbreak in West Africa.

65
00:06:43,480 --> 00:06:46,900
And so that was the first outbreak of Ebola that had happened in West Africa,

66
00:06:46,900 --> 00:06:54,460
although there had been previous outbreaks before much smaller and ended up costing thousands of lives in West Africa.

67
00:06:54,460 --> 00:07:01,270
It's a highly infectious disease. It kills the majority of people, even with medical support who get infected,

68
00:07:01,270 --> 00:07:08,590
so it's just devastating to communities and to the individuals who are affected.

69
00:07:08,590 --> 00:07:14,810
But it can respond well to non-pharmaceutical interventions, which we now hear more about.

70
00:07:14,810 --> 00:07:20,470
So those are the isolating cases, isolating people who may have been exposed,

71
00:07:20,470 --> 00:07:29,920
monitoring them very quickly so that if they have have become infected and potentially becoming infectious, that they don't go on to infect others.

72
00:07:29,920 --> 00:07:36,340
And so that's a lot of on the ground work with people who are in a situation where there is this deadly disease spreading,

73
00:07:36,340 --> 00:07:44,230
going and finding individuals finding out from themselves or they're the people who live with them,

74
00:07:44,230 --> 00:07:47,410
who they came into contact with and tracing all those individuals.

75
00:07:47,410 --> 00:07:55,030
Some people would have public health officials show up and take their temperature every day while they're being monitored for a couple of weeks.

76
00:07:55,030 --> 00:08:01,150
It is a disease where there are infections out in an animal reservoir.

77
00:08:01,150 --> 00:08:07,510
And so that is believed to mainly be bats so that there can be potential for transmission through other species,

78
00:08:07,510 --> 00:08:12,790
for example, through great apes can get it as well. And so people could get it that way.

79
00:08:12,790 --> 00:08:20,710
And in that case, there will always be this opportunity for another transmission into the human population,

80
00:08:20,710 --> 00:08:23,800
and then it can very easily spread from person to person.

81
00:08:23,800 --> 00:08:32,620
So what we look for at the end of an epidemic, if that's the end of that particular Ebola epidemic rather than all of them in particular.

82
00:08:32,620 --> 00:08:41,080
And so usually there is a sort of time frame at the end where there are no additional cases identified.

83
00:08:41,080 --> 00:08:45,370
But then you don't just say, Well, OK, we haven't seen one today. We're off.

84
00:08:45,370 --> 00:08:51,460
It's a matter of how long you wait until you're confident that there won't be any more cases.

85
00:08:51,460 --> 00:08:56,710
And we have to allow for the fact that not just how long it takes that incubation period,

86
00:08:56,710 --> 00:08:59,770
the time from becoming infected to showing the signs of disease,

87
00:08:59,770 --> 00:09:05,560
but also the possibility that the surveillance system may miss, occasionally some cases.

88
00:09:05,560 --> 00:09:10,300
And so it's important to make sure that there was time for not just that you saw no cases,

89
00:09:10,300 --> 00:09:15,490
but in case you missed one or two that time, did they infect any others?

90
00:09:15,490 --> 00:09:19,690
And it can be made more difficult because it has been shown in a couple of cases

91
00:09:19,690 --> 00:09:24,340
that people have become infectious later after having recovered from the disease.

92
00:09:24,340 --> 00:09:30,370
And so that can lead to flare ups even at the end and particularly in West Africa,

93
00:09:30,370 --> 00:09:34,900
where there were multiple countries involved and so therefore a big geographic area.

94
00:09:34,900 --> 00:09:40,600
There were times when a country declared. There a particular component of the epidemic over,

95
00:09:40,600 --> 00:09:45,910
and then they had additional cases now sometimes that can happen for reintroduction from a nearby country,

96
00:09:45,910 --> 00:09:55,450
or it could be this case where you have a flare up and somebody suddenly apparently they did have virus in them and it became infectious again.

97
00:09:55,450 --> 00:09:59,980
So both of those things mean that you have to be very cautious at the end of an epidemic.

98
00:09:59,980 --> 00:10:08,980
But right now, we don't have any Ebola epidemics ongoing, but we're very certain that there will be additional ones.

99
00:10:08,980 --> 00:10:14,170
And it's just, you know, as a as a community, we know how to deal with them.

100
00:10:14,170 --> 00:10:17,830
But there still can be devastating in that time between when they're identified

101
00:10:17,830 --> 00:10:22,490
and when they finally are brought under control and reduced to no cases.

102
00:10:22,490 --> 00:10:30,380
I think one point that's very striking is is how, as you described with Ebola, there you're discussing outbreaks,

103
00:10:30,380 --> 00:10:36,800
so repeated outbreaks, and you can talk about a specific outbreak ending, but then also thinking that another thing might continue.

104
00:10:36,800 --> 00:10:44,630
But also the, I guess, the political and the economic significance of being able to state that something that outbreak is over because of course,

105
00:10:44,630 --> 00:10:49,520
that has policy implications for people travelling for people returning to work.

106
00:10:49,520 --> 00:10:52,490
You've also mentioned, of course, how disease continues on.

107
00:10:52,490 --> 00:10:58,790
And so I wondered if you could talk about other types of diseases where it's not so much that the outbreak ends,

108
00:10:58,790 --> 00:11:04,520
but that perhaps cases go down to a particular level or they become more predictable?

109
00:11:04,520 --> 00:11:13,760
And I know you've worked on influenza, and that might help us also to think about what we mean when we talk about the end of, say, COVID.

110
00:11:13,760 --> 00:11:22,400
Yeah, certainly. So before any of us were born, there was influenza and there will be influenza long after us.

111
00:11:22,400 --> 00:11:32,900
And what makes it tricky is it's a viral disease that has multiple animal hosts, so there can be swine flus, there can be avian flus.

112
00:11:32,900 --> 00:11:35,840
And of course, we have transmission from person to person.

113
00:11:35,840 --> 00:11:43,970
So the fact that we have animal reservoirs that can maintain influenza means that we can't really think of ever eradicating it.

114
00:11:43,970 --> 00:11:52,280
But on the other hand, you know, we do function as a society quite well, despite there being influenza around every year.

115
00:11:52,280 --> 00:11:55,760
Now, influenza does kill people every year,

116
00:11:55,760 --> 00:12:04,550
but what we have to reduce those impacts are vaccination and also within particular settings, for example, care home.

117
00:12:04,550 --> 00:12:10,100
There can be measures that are brought in if there appears to be influenza within a care home

118
00:12:10,100 --> 00:12:15,590
setting to try and reduce the risk of transmission between residents so that can come in.

119
00:12:15,590 --> 00:12:22,130
We had a large pandemic of what was a new.

120
00:12:22,130 --> 00:12:26,420
That's what causes the pandemic is there's a new strain of influenza that shows up.

121
00:12:26,420 --> 00:12:32,930
This was in 2009. Probably the estimate is about a third of us got infected over the course of that epidemic,

122
00:12:32,930 --> 00:12:39,830
which is very large, although of course we're seeing very large spread of another disease right now.

123
00:12:39,830 --> 00:12:46,130
On the other hand, we didn't see that then disappear. So it goes down to lower levels.

124
00:12:46,130 --> 00:12:53,060
It becomes more predictable in the sense that we see now seasonal flus in the way we'd expected before.

125
00:12:53,060 --> 00:13:00,800
We have seen less flus over these last couple of winters because of social distancing, because if we stayed further away from other people,

126
00:13:00,800 --> 00:13:08,120
it reduces the risk of spreading COVID between each other, but also spreading other respiratory diseases, including influenza.

127
00:13:08,120 --> 00:13:13,640
So there have been the last two years have been unusual for the transmission of influenza,

128
00:13:13,640 --> 00:13:24,440
but it's still the case that we have multiple strains of influenza out there and grouped into big groups of influenza or influenza B.

129
00:13:24,440 --> 00:13:30,020
But it's certainly the case that people still get this sort of influenza that caused the pandemic.

130
00:13:30,020 --> 00:13:35,540
But it's just because there's so much immunity that's built up partly from natural infections,

131
00:13:35,540 --> 00:13:41,810
partly from vaccination, that it's not causing the huge problems that it did before.

132
00:13:41,810 --> 00:13:48,920
But these are still out there, and I'm very confident in predicting there will be another influenza pandemic,

133
00:13:48,920 --> 00:13:55,130
but the severity of it will very much determine its overall impact on populations.

134
00:13:55,130 --> 00:14:03,380
Lots of people will get infected, but because we can't immediately produce vaccines straight away for worldwide production.

135
00:14:03,380 --> 00:14:09,380
But as you've seen, vaccines that are developed more quickly and can be delivered more quickly than ever before,

136
00:14:09,380 --> 00:14:15,800
and that helps mitigate the impacts of these epidemics. But when does it end?

137
00:14:15,800 --> 00:14:22,070
There's not really one point at which you can say, OK, it was a pandemic.

138
00:14:22,070 --> 00:14:28,130
Now the you know, there is a declaration of the pandemic being over, but that's sort of a a doubly W.H.O.

139
00:14:28,130 --> 00:14:37,310
That's World Health Organisation decision based on various characteristics when it when the epidemic itself is actually over.

140
00:14:37,310 --> 00:14:43,520
There's no one point at which you say, OK, well, it's now changed from an epidemic to being endemic disease,

141
00:14:43,520 --> 00:14:52,550
which means that it's it's sort of always with us at a sort of more predictable level that can still be higher in the winter than in the summer.

142
00:14:52,550 --> 00:14:58,380
But that's sort of a gradual process of it sort of settling down. So there isn't one.

143
00:14:58,380 --> 00:15:08,070
You know, it's not like a shutting of a door, but it's it's sort of gradually moves from being more like, you know, a large outbreak at.

144
00:15:08,070 --> 00:15:14,020
So. From being less like a large outbreak and more into the subtle pattern.

145
00:15:14,020 --> 00:15:23,480
So would I be right in suggesting that partly what we're thinking about is how you study patterns in data and especially quantitative data.

146
00:15:23,480 --> 00:15:31,150
So thinking about the patterns of numbers and trying to make sense of them and in some ways never suggesting that we can be certain say

147
00:15:31,150 --> 00:15:37,240
about predictions because my sense is a lot of people think that you might be able to predict what's going to happen in terms of numbers.

148
00:15:37,240 --> 00:15:42,520
And you probably say that you can create models and projections to think about this,

149
00:15:42,520 --> 00:15:49,630
but especially when it comes to people asking about COVID that we're probably not thinking about, say, when there will be no more COVID.

150
00:15:49,630 --> 00:15:57,520
But what kinds of patterns that we might be able to establish by looking at numbers of target cases, for example?

151
00:15:57,520 --> 00:16:02,680
Indeed. So there are different scales of predict prediction or projection.

152
00:16:02,680 --> 00:16:08,140
You know, if you're saying, well, you know, how many cases might we see over the next week?

153
00:16:08,140 --> 00:16:13,780
That's a much easier thing to try to do accurately than to predict over the, you know,

154
00:16:13,780 --> 00:16:20,020
over a month or over three months, particularly because we have the possibility of these new variants showing up.

155
00:16:20,020 --> 00:16:24,190
But we try to analyse data from several different studies.

156
00:16:24,190 --> 00:16:32,830
So I'm involved with the study called React, which is Real-Time Assessment of transmission in the community of COVID 19.

157
00:16:32,830 --> 00:16:41,260
And so what happens is essentially no swab kids get sent out to people about 100000 people each round,

158
00:16:41,260 --> 00:16:45,730
reply randomly chosen, send theirs back and then it gets sent for PCR,

159
00:16:45,730 --> 00:16:55,000
which is a very sensitive diagnostic test that allows us to get an idea of how much SARS-CoV-2 infection there is in the community.

160
00:16:55,000 --> 00:17:02,890
And that's really important because then it's separated from what the tendency is to seek a diagnosis.

161
00:17:02,890 --> 00:17:04,120
So we're trying to get a sense of that.

162
00:17:04,120 --> 00:17:14,110
And so if a lot of those infections are asymptomatic, which may be more the case now that we have more vaccination and also depends on the variant,

163
00:17:14,110 --> 00:17:19,360
some might have more asymptomatic infection than others. Looking at those patterns is really crucial.

164
00:17:19,360 --> 00:17:23,980
We combine that with data on numbers of individuals who have severe disease,

165
00:17:23,980 --> 00:17:29,860
so they go to hospital numbers of people who sadly die infected with SARS-CoV-2,

166
00:17:29,860 --> 00:17:36,160
and those pieces all come together to form as complete as we can a picture of the situation now.

167
00:17:36,160 --> 00:17:43,750
But predicting what's going to happen in future is fraught, and particularly we can say that new variants will arise.

168
00:17:43,750 --> 00:17:48,430
I'm very confident saying that this isn't the last important variant,

169
00:17:48,430 --> 00:17:54,880
but what it will look like where it would show up first, and nobody can say with any certainty.

170
00:17:54,880 --> 00:18:04,240
So there are some inherent uncertainties in the system, but we can be prepared and know what data to analyse when they do arise.

171
00:18:04,240 --> 00:18:11,540
That's a great summary of what I know is a very complex practise and methodology and all sorts of different kinds of data analysis ongoing.

172
00:18:11,540 --> 00:18:16,960
So thank you very much, Crystal, for sharing your research and your expertise with us.

173
00:18:16,960 --> 00:18:23,410
And thank you also for watching your videos. I hope that you will watch some of the other videos in the project series and please also do fill out the

174
00:18:23,410 --> 00:18:29,320
feedback forms so that we can help to further inform and shape research at the University of Oxford.

175
00:18:29,320 --> 00:18:39,060
Thank you very much.