TEDGlobal 2012
John Wilbanks: Let's pool our medical data
Filmed:
Readability: 3.7
581,818 views
When you're getting medical treatment, or taking part in medical testing, privacy is important; strict laws limit what researchers can see and know about you. But what if your medical data could be used -- anonymously -- by anyone seeking to test a hypothesis? John Wilbanks wonders if the desire to protect our privacy is slowing research, and if opening up medical data could lead to a wave of health care innovation.
John Wilbanks - Data Commons Advocate
Imagine the discoveries that could result from a giant pool of freely available health and genomic data. John Wilbanks is working to build it. Full bio
Imagine the discoveries that could result from a giant pool of freely available health and genomic data. John Wilbanks is working to build it. Full bio
Double-click the English transcript below to play the video.
00:15
So I have bad news, I have good news,
0
98
3061
00:19
and I have a task.
1
3159
1865
00:20
So the bad news is that we all get sick.
2
5024
2943
00:23
I get sick. You get sick.
3
7967
2272
00:26
And every one of us gets sick, and the question really is,
4
10239
2542
00:28
how sick do we get? Is it something that kills us?
5
12781
2877
00:31
Is it something that we survive?
6
15658
1345
00:32
Is it something that we can treat?
7
17003
1928
00:34
And we've gotten sick as long as we've been people.
8
18931
3256
00:38
And so we've always looked for reasons to explain why we get sick.
9
22187
3486
00:41
And for a long time, it was the gods, right?
10
25673
1957
00:43
The gods are angry with me, or the gods are testing me,
11
27630
3154
00:46
right? Or God, singular, more recently,
12
30784
2416
00:49
is punishing me or judging me.
13
33200
2664
00:51
And as long as we've looked for explanations,
14
35864
2680
00:54
we've wound up with something that gets closer and closer to science,
15
38544
3711
00:58
which is hypotheses as to why we get sick,
16
42255
2489
01:00
and as long as we've had hypotheses about why we get sick, we've tried to treat it as well.
17
44744
4740
01:05
So this is Avicenna. He wrote a book over a thousand years ago called "The Canon of Medicine,"
18
49484
4033
01:09
and the rules he laid out for testing medicines
19
53517
2406
01:11
are actually really similar to the rules we have today,
20
55923
1789
01:13
that the disease and the medicine must be the same strength,
21
57712
2945
01:16
the medicine needs to be pure, and in the end we need
22
60657
2397
01:18
to test it in people. And so if you put together these themes
23
63054
3141
01:22
of a narrative or a hypothesis in human testing,
24
66195
4465
01:26
right, you get some beautiful results,
25
70660
2656
01:29
even when we didn't have very good technologies.
26
73316
1442
01:30
This is a guy named Carlos Finlay. He had a hypothesis
27
74758
3062
01:33
that was way outside the box for his time, in the late 1800s.
28
77820
2905
01:36
He thought yellow fever was not transmitted by dirty clothing.
29
80725
2848
01:39
He thought it was transmitted by mosquitos.
30
83573
2426
01:41
And they laughed at him. For 20 years, they called this guy
31
85999
2362
01:44
"the mosquito man." But he ran an experiment in people,
32
88361
3489
01:47
right? He had this hypothesis, and he tested it in people.
33
91850
3103
01:50
So he got volunteers to go move to Cuba and live in tents
34
94953
4642
01:55
and be voluntarily infected with yellow fever.
35
99595
3035
01:58
So some of the people in some of the tents had dirty clothes
36
102630
3022
02:01
and some of the people were in tents that were full
37
105652
1219
02:02
of mosquitos that had been exposed to yellow fever.
38
106871
2256
02:05
And it definitively proved that it wasn't this magic dust
39
109127
3401
02:08
called fomites in your clothes that caused yellow fever.
40
112528
3422
02:11
But it wasn't until we tested it in people that we actually knew.
41
115950
3376
02:15
And this is what those people signed up for.
42
119326
1959
02:17
This is what it looked like to have yellow fever in Cuba
43
121285
3090
02:20
at that time. You suffered in a tent, in the heat, alone,
44
124375
4534
02:24
and you probably died.
45
128909
2696
02:27
But people volunteered for this.
46
131605
3217
02:30
And it's not just a cool example of a scientific design
47
134822
3311
02:34
of experiment in theory. They also did this beautiful thing.
48
138133
2913
02:36
They signed this document, and it's called an informed consent document.
49
141046
3919
02:40
And informed consent is an idea that we should be
50
144965
2513
02:43
very proud of as a society, right? It's something that
51
147478
2226
02:45
separates us from the Nazis at Nuremberg,
52
149704
2766
02:48
enforced medical experimentation. It's the idea
53
152470
2875
02:51
that agreement to join a study without understanding isn't agreement.
54
155345
3788
02:55
It's something that protects us from harm, from hucksters,
55
159133
4109
02:59
from people that would try to hoodwink us into a clinical
56
163242
2853
03:01
study that we don't understand, or that we don't agree to.
57
166095
3752
03:05
And so you put together the thread of narrative hypothesis,
58
169847
4329
03:10
experimentation in humans, and informed consent,
59
174176
2597
03:12
and you get what we call clinical study, and it's how we do
60
176773
2665
03:15
the vast majority of medical work. It doesn't really matter
61
179438
3015
03:18
if you're in the north, the south, the east, the west.
62
182453
2342
03:20
Clinical studies form the basis of how we investigate,
63
184795
4113
03:24
so if we're going to look at a new drug, right,
64
188908
1859
03:26
we test it in people, we draw blood, we do experiments,
65
190767
2998
03:29
and we gain consent for that study, to make sure
66
193765
2329
03:31
that we're not screwing people over as part of it.
67
196094
2649
03:34
But the world is changing around the clinical study,
68
198743
3664
03:38
which has been fairly well established for tens of years
69
202407
3366
03:41
if not 50 to 100 years.
70
205773
1900
03:43
So now we're able to gather data about our genomes,
71
207673
3051
03:46
but, as we saw earlier, our genomes aren't dispositive.
72
210724
2860
03:49
We're able to gather information about our environment.
73
213584
2766
03:52
And more importantly, we're able to gather information
74
216350
1910
03:54
about our choices, because it turns out that what we think of
75
218260
2840
03:57
as our health is more like the interaction of our bodies,
76
221100
2720
03:59
our genomes, our choices and our environment.
77
223820
3649
04:03
And the clinical methods that we've got aren't very good
78
227469
2744
04:06
at studying that because they are based on the idea
79
230213
2632
04:08
of person-to-person interaction. You interact
80
232845
1914
04:10
with your doctor and you get enrolled in the study.
81
234759
2095
04:12
So this is my grandfather. I actually never met him,
82
236854
2615
04:15
but he's holding my mom, and his genes are in me, right?
83
239469
3795
04:19
His choices ran through to me. He was a smoker,
84
243264
2891
04:22
like most people were. This is my son.
85
246155
2584
04:24
So my grandfather's genes go all the way through to him,
86
248739
3442
04:28
and my choices are going to affect his health.
87
252181
2552
04:30
The technology between these two pictures
88
254733
2694
04:33
cannot be more different, but the methodology
89
257427
3673
04:37
for clinical studies has not radically changed over that time period.
90
261100
4124
04:41
We just have better statistics.
91
265224
2668
04:43
The way we gain informed consent was formed in large part
92
267892
3452
04:47
after World War II, around the time that picture was taken.
93
271344
2591
04:49
That was 70 years ago, and the way we gain informed consent,
94
273935
3934
04:53
this tool that was created to protect us from harm,
95
277869
2877
04:56
now creates silos. So the data that we collect
96
280746
3666
05:00
for prostate cancer or for Alzheimer's trials
97
284412
2726
05:03
goes into silos where it can only be used
98
287138
2615
05:05
for prostate cancer or for Alzheimer's research.
99
289753
3224
05:08
Right? It can't be networked. It can't be integrated.
100
292977
2894
05:11
It cannot be used by people who aren't credentialed.
101
295871
3533
05:15
So a physicist can't get access to it without filing paperwork.
102
299404
2953
05:18
A computer scientist can't get access to it without filing paperwork.
103
302357
3068
05:21
Computer scientists aren't patient. They don't file paperwork.
104
305425
4143
05:25
And this is an accident. These are tools that we created
105
309568
3986
05:29
to protect us from harm, but what they're doing
106
313554
3267
05:32
is protecting us from innovation now.
107
316821
2530
05:35
And that wasn't the goal. It wasn't the point. Right?
108
319351
3265
05:38
It's a side effect, if you will, of a power we created
109
322616
2699
05:41
to take us for good.
110
325315
2359
05:43
And so if you think about it, the depressing thing is that
111
327674
3144
05:46
Facebook would never make a change to something
112
330818
2133
05:48
as important as an advertising algorithm
113
332951
2571
05:51
with a sample size as small as a Phase III clinical trial.
114
335522
4411
05:55
We cannot take the information from past trials
115
339933
3662
05:59
and put them together to form statistically significant samples.
116
343595
4154
06:03
And that sucks, right? So 45 percent of men develop
117
347749
3484
06:07
cancer. Thirty-eight percent of women develop cancer.
118
351233
3097
06:10
One in four men dies of cancer.
119
354330
2344
06:12
One in five women dies of cancer, at least in the United States.
120
356674
3556
06:16
And three out of the four drugs we give you
121
360230
2228
06:18
if you get cancer fail. And this is personal to me.
122
362458
3513
06:21
My sister is a cancer survivor.
123
365971
1963
06:23
My mother-in-law is a cancer survivor. Cancer sucks.
124
367934
3589
06:27
And when you have it, you don't have a lot of privacy
125
371523
2190
06:29
in the hospital. You're naked the vast majority of the time.
126
373713
3487
06:33
People you don't know come in and look at you and poke you and prod you,
127
377200
3695
06:36
and when I tell cancer survivors that this tool we created
128
380895
3441
06:40
to protect them is actually preventing their data from being used,
129
384336
3098
06:43
especially when only three to four percent of people
130
387434
2050
06:45
who have cancer ever even sign up for a clinical study,
131
389484
2798
06:48
their reaction is not, "Thank you, God, for protecting my privacy."
132
392282
3558
06:51
It's outrage
133
395840
2697
06:54
that we have this information and we can't use it.
134
398537
2125
06:56
And it's an accident.
135
400662
2476
06:59
So the cost in blood and treasure of this is enormous.
136
403138
3055
07:02
Two hundred and twenty-six billion a year is spent on cancer in the United States.
137
406193
3655
07:05
Fifteen hundred people a day die in the United States.
138
409848
3219
07:08
And it's getting worse.
139
413067
2573
07:11
So the good news is that some things have changed,
140
415640
2982
07:14
and the most important thing that's changed
141
418622
1553
07:16
is that we can now measure ourselves in ways
142
420175
2338
07:18
that used to be the dominion of the health system.
143
422513
3058
07:21
So a lot of people talk about it as digital exhaust.
144
425571
2158
07:23
I like to think of it as the dust that runs along behind my kid.
145
427729
3042
07:26
We can reach back and grab that dust,
146
430771
2376
07:29
and we can learn a lot about health from it, so if our choices
147
433147
2414
07:31
are part of our health, what we eat is a really important
148
435561
2680
07:34
aspect of our health. So you can do something very simple
149
438241
2689
07:36
and basic and take a picture of your food,
150
440930
1957
07:38
and if enough people do that, we can learn a lot about
151
442887
2884
07:41
how our food affects our health.
152
445771
1425
07:43
One interesting thing that came out of this — this is an app for iPhones called The Eatery —
153
447196
4516
07:47
is that we think our pizza is significantly healthier
154
451712
2490
07:50
than other people's pizza is. Okay? (Laughter)
155
454202
3438
07:53
And it seems like a trivial result, but this is the sort of research
156
457640
3608
07:57
that used to take the health system years
157
461248
2314
07:59
and hundreds of thousands of dollars to accomplish.
158
463562
2293
08:01
It was done in five months by a startup company of a couple of people.
159
465855
3724
08:05
I don't have any financial interest in it.
160
469579
2624
08:08
But more nontrivially, we can get our genotypes done,
161
472203
2696
08:10
and although our genotypes aren't dispositive, they give us clues.
162
474899
2818
08:13
So I could show you mine. It's just A's, T's, C's and G's.
163
477717
2806
08:16
This is the interpretation of it. As you can see,
164
480523
2232
08:18
I carry a 32 percent risk of prostate cancer,
165
482755
2600
08:21
22 percent risk of psoriasis and a 14 percent risk of Alzheimer's disease.
166
485355
4223
08:25
So that means, if you're a geneticist, you're freaking out,
167
489578
2607
08:28
going, "Oh my God, you told everyone you carry the ApoE E4 allele. What's wrong with you?"
168
492185
4034
08:32
Right? When I got these results, I started talking to doctors,
169
496219
3688
08:35
and they told me not to tell anyone, and my reaction is,
170
499907
2409
08:38
"Is that going to help anyone cure me when I get the disease?"
171
502316
3288
08:41
And no one could tell me yes.
172
505604
3064
08:44
And I live in a web world where, when you share things,
173
508668
2806
08:47
beautiful stuff happens, not bad stuff.
174
511474
2710
08:50
So I started putting this in my slide decks,
175
514184
1900
08:51
and I got even more obnoxious, and I went to my doctor,
176
516084
2461
08:54
and I said, "I'd like to actually get my bloodwork.
177
518545
1982
08:56
Please give me back my data." So this is my most recent bloodwork.
178
520527
2790
08:59
As you can see, I have high cholesterol.
179
523317
2369
09:01
I have particularly high bad cholesterol, and I have some
180
525686
2751
09:04
bad liver numbers, but those are because we had a dinner party with a lot of good wine
181
528437
3003
09:07
the night before we ran the test. (Laughter)
182
531440
2709
09:10
Right. But look at how non-computable this information is.
183
534149
4413
09:14
This is like the photograph of my granddad holding my mom
184
538562
2974
09:17
from a data perspective, and I had to go into the system
185
541536
3599
09:21
and get it out.
186
545135
2162
09:23
So the thing that I'm proposing we do here
187
547297
3282
09:26
is that we reach behind us and we grab the dust,
188
550579
2416
09:28
that we reach into our bodies and we grab the genotype,
189
552995
2978
09:31
and we reach into the medical system and we grab our records,
190
555973
2701
09:34
and we use it to build something together, which is a commons.
191
558674
3440
09:38
And there's been a lot of talk about commonses, right,
192
562114
3144
09:41
here, there, everywhere, right. A commons is nothing more
193
565258
2948
09:44
than a public good that we build out of private goods.
194
568206
2928
09:47
We do it voluntarily, and we do it through standardized
195
571134
2769
09:49
legal tools. We do it through standardized technologies.
196
573903
2800
09:52
Right. That's all a commons is. It's something that we build
197
576703
3271
09:55
together because we think it's important.
198
579974
2520
09:58
And a commons of data is something that's really unique,
199
582494
2632
10:01
because we make it from our own data. And although
200
585126
2868
10:03
a lot of people like privacy as their methodology of control
201
587994
2287
10:06
around data, and obsess around privacy, at least
202
590281
2255
10:08
some of us really like to share as a form of control,
203
592536
3048
10:11
and what's remarkable about digital commonses
204
595584
2353
10:13
is you don't need a big percentage if your sample size is big enough
205
597937
3532
10:17
to generate something massive and beautiful.
206
601469
2511
10:19
So not that many programmers write free software,
207
603980
2558
10:22
but we have the Apache web server.
208
606538
2335
10:24
Not that many people who read Wikipedia edit,
209
608873
2697
10:27
but it works. So as long as some people like to share
210
611570
4009
10:31
as their form of control, we can build a commons, as long as we can get the information out.
211
615579
3744
10:35
And in biology, the numbers are even better.
212
619323
2376
10:37
So Vanderbilt ran a study asking people, we'd like to take
213
621699
2552
10:40
your biosamples, your blood, and share them in a biobank,
214
624251
3322
10:43
and only five percent of the people opted out.
215
627573
2372
10:45
I'm from Tennessee. It's not the most science-positive state
216
629945
3092
10:48
in the United States of America. (Laughter)
217
633037
3039
10:51
But only five percent of the people wanted out.
218
636076
2378
10:54
So people like to share, if you give them the opportunity and the choice.
219
638454
4023
10:58
And the reason that I got obsessed with this, besides the obvious family aspects,
220
642477
4483
11:02
is that I spend a lot of time around mathematicians,
221
646960
3273
11:06
and mathematicians are drawn to places where there's a lot of data
222
650233
2914
11:09
because they can use it to tease signals out of noise.
223
653147
2943
11:11
And those correlations that they can tease out, they're not
224
656090
2968
11:14
necessarily causal agents, but math, in this day and age,
225
659058
3872
11:18
is like a giant set of power tools
226
662930
2360
11:21
that we're leaving on the floor, not plugged in in health,
227
665290
3875
11:25
while we use hand saws.
228
669165
2312
11:27
If we have a lot of shared genotypes, and a lot of shared
229
671477
4438
11:31
outcomes, and a lot of shared lifestyle choices,
230
675915
2748
11:34
and a lot of shared environmental information, we can start
231
678663
2776
11:37
to tease out the correlations between subtle variations
232
681439
2896
11:40
in people, the choices they make and the health that they create as a result of those choices,
233
684335
5311
11:45
and there's open-source infrastructure to do all of this.
234
689646
2486
11:48
Sage Bionetworks is a nonprofit that's built a giant math system
235
692132
3094
11:51
that's waiting for data, but there isn't any.
236
695226
4572
11:55
So that's what I do. I've actually started what we think is
237
699798
3888
11:59
the world's first fully digital, fully self-contributed,
238
703686
3938
12:03
unlimited in scope, global in participation, ethically approved
239
707624
5035
12:08
clinical research study where you contribute the data.
240
712659
3655
12:12
So if you reach behind yourself and you grab the dust,
241
716314
2206
12:14
if you reach into your body and grab your genome,
242
718520
2626
12:17
if you reach into the medical system and somehow extract your medical record,
243
721146
3047
12:20
you can actually go through an online informed consent process --
244
724193
3323
12:23
because the donation to the commons must be voluntary
245
727516
2646
12:26
and it must be informed -- and you can actually upload
246
730162
2793
12:28
your information and have it syndicated to the
247
732955
2592
12:31
mathematicians who will do this sort of big data research,
248
735547
3096
12:34
and the goal is to get 100,000 in the first year
249
738643
2856
12:37
and a million in the first five years so that we have
250
741499
2358
12:39
a statistically significant cohort that you can use to take
251
743857
3834
12:43
smaller sample sizes from traditional research
252
747691
2422
12:46
and map it against,
253
750113
1599
12:47
so that you can use it to tease out those subtle correlations
254
751712
2922
12:50
between the variations that make us unique
255
754634
2529
12:53
and the kinds of health that we need to move forward as a society.
256
757163
4024
12:57
And I've spent a lot of time around other commons.
257
761187
3024
13:00
I've been around the early web. I've been around
258
764211
2680
13:02
the early creative commons world, and there's four things
259
766891
2608
13:05
that all of these share, which is, they're all really simple.
260
769499
3354
13:08
And so if you were to go to the website and enroll in this study,
261
772853
2727
13:11
you're not going to see something complicated.
262
775580
2255
13:13
But it's not simplistic. These things are weak intentionally,
263
777835
5049
13:18
right, because you can always add power and control to a system,
264
782884
3023
13:21
but it's very difficult to remove those things if you put them in at the beginning,
265
785907
3964
13:25
and so being simple doesn't mean being simplistic,
266
789871
2545
13:28
and being weak doesn't mean weakness.
267
792416
2184
13:30
Those are strengths in the system.
268
794600
2351
13:32
And open doesn't mean that there's no money.
269
796951
2665
13:35
Closed systems, corporations, make a lot of money
270
799616
3020
13:38
on the open web, and they're one of the reasons why the open web lives
271
802636
3539
13:42
is that corporations have a vested interest in the openness
272
806175
2827
13:44
of the system.
273
809002
2334
13:47
And so all of these things are part of the clinical study that we've created,
274
811336
3794
13:51
so you can actually come in, all you have to be is 14 years old,
275
815130
3429
13:54
willing to sign a contract that says I'm not going to be a jerk,
276
818559
2027
13:56
basically, and you're in.
277
820586
2665
13:59
You can start analyzing the data.
278
823251
1573
14:00
You do have to solve a CAPTCHA as well. (Laughter)
279
824824
4159
14:04
And if you'd like to build corporate structures on top of it,
280
828983
3581
14:08
that's okay too. That's all in the consent,
281
832564
3146
14:11
so if you don't like those terms, you don't come in.
282
835710
2564
14:14
It's very much the design principles of a commons
283
838274
3092
14:17
that we're trying to bring to health data.
284
841366
2594
14:19
And the other thing about these systems is that it only takes
285
843960
2979
14:22
a small number of really unreasonable people working together
286
846939
3179
14:26
to create them. It didn't take that many people
287
850118
3182
14:29
to make Wikipedia Wikipedia, or to keep it Wikipedia.
288
853300
3472
14:32
And we're not supposed to be unreasonable in health,
289
856772
2068
14:34
and so I hate this word "patient."
290
858840
2276
14:37
I don't like being patient when systems are broken,
291
861116
3167
14:40
and health care is broken.
292
864283
2627
14:42
I'm not talking about the politics of health care, I'm talking about the way we scientifically approach health care.
293
866910
4164
14:46
So I don't want to be patient. And the task I'm giving to you
294
871074
3270
14:50
is to not be patient. So I'd like you to actually try,
295
874344
3046
14:53
when you go home, to get your data.
296
877390
2717
14:56
You'll be shocked and offended and, I would bet, outraged,
297
880107
2717
14:58
at how hard it is to get it.
298
882824
2876
15:01
But it's a challenge that I hope you'll take,
299
885700
2619
15:04
and maybe you'll share it. Maybe you won't.
300
888319
2461
15:06
If you don't have anyone in your family who's sick,
301
890780
1444
15:08
maybe you wouldn't be unreasonable. But if you do,
302
892224
2993
15:11
or if you've been sick, then maybe you would.
303
895217
2207
15:13
And we're going to be able to do an experiment in the next several months
304
897424
3088
15:16
that lets us know exactly how many unreasonable people are out there.
305
900512
3157
15:19
So this is the Athena Breast Health Network. It's a study
306
903669
2122
15:21
of 150,000 women in California, and they're going to
307
905791
3818
15:25
return all the data to the participants of the study
308
909609
2718
15:28
in a computable form, with one-clickability to load it into
309
912327
3146
15:31
the study that I've put together. So we'll know exactly
310
915473
2616
15:33
how many people are willing to be unreasonable.
311
918089
2304
15:36
So what I'd end [with] is,
312
920393
2384
15:38
the most beautiful thing I've learned since I quit my job
313
922777
3320
15:41
almost a year ago to do this, is that it really doesn't take
314
926097
3383
15:45
very many of us to achieve spectacular results.
315
929480
3808
15:49
You just have to be willing to be unreasonable,
316
933288
2712
15:51
and the risk we're running is not the risk those 14 men
317
936000
2331
15:54
who got yellow fever ran. Right?
318
938331
1868
15:56
It's to be naked, digitally, in public. So you know more
319
940199
2861
15:58
about me and my health than I know about you. It's asymmetric now.
320
943060
3433
16:02
And being naked and alone can be terrifying.
321
946493
3630
16:06
But to be naked in a group, voluntarily, can be quite beautiful.
322
950123
4467
16:10
And so it doesn't take all of us.
323
954590
1888
16:12
It just takes all of some of us. Thank you.
324
956478
3006
16:15
(Applause)
325
959484
5590
ABOUT THE SPEAKER
John Wilbanks - Data Commons AdvocateImagine the discoveries that could result from a giant pool of freely available health and genomic data. John Wilbanks is working to build it.
Why you should listen
Performing a medical or genomic experiment on a human requires informed consent and careful boundaries around privacy. But what if the data that results, once scrubbed of identifying marks, was released into the wild? At WeConsent.us, John Wilbanks thinks through the ethical and procedural steps to create an open, massive, mine-able database of data about health and genomics from many sources. One step: the Portable Legal Consent for Common Genomics Research (PLC-CGR), an experimental bioethics protocol that would allow any test subject to say, "Yes, once this experiment is over, you can use my data, anonymously, to answer any other questions you can think of." Compiling piles of test results in one place, Wilbanks suggests, would turn genetic info into big data--giving researchers the potential to spot patterns that simply aren't viewable up close.
A campaigner for the wide adoption of data sharing in science, Wilbanks is also a Senior Fellow with the Kauffman Foundation, a Research Fellow at Lybba and supported by Sage Bionetworks.
In February 2013, the US government responded to a We the People petition spearheaded by Wilbanks and signed by 65,000 people, and announced a plan to open up taxpayer-funded research data and make it available for free.
John Wilbanks | Speaker | TED.com