WEBVTT

00:00:14.890 --> 00:00:17.830
PETER SZOLOVITS:
As David said, I've

00:00:17.830 --> 00:00:20.900
been at this for a long time.

00:00:20.900 --> 00:00:22.450
I am not a medical doctor.

00:00:22.450 --> 00:00:24.400
But I've probably
learned enough medicine

00:00:24.400 --> 00:00:29.440
to be able to play one on
television over the years.

00:00:29.440 --> 00:00:32.140
And actually, that's
relevant to today's lecture

00:00:32.140 --> 00:00:34.450
because today's
lecture is really

00:00:34.450 --> 00:00:38.210
trying to set the scene
for you to say, well,

00:00:38.210 --> 00:00:40.360
what are the kinds of
problems that doctors

00:00:40.360 --> 00:00:45.050
are interested in by looking
at what is it that they do.

00:00:45.050 --> 00:00:45.820
OK.

00:00:45.820 --> 00:00:49.090
So that's our goal for today.

00:00:49.090 --> 00:00:50.590
So what we're going
to do today is

00:00:50.590 --> 00:00:53.740
to talk about, in a
very general way, what

00:00:53.740 --> 00:00:55.640
are the goals of health care.

00:00:55.640 --> 00:00:59.510
So how many of you are doctors?

00:00:59.510 --> 00:01:00.370
A couple.

00:01:00.370 --> 00:01:01.070
Great.

00:01:01.070 --> 00:01:01.600
OK.

00:01:01.600 --> 00:01:03.910
So fix me when I blow it.

00:01:03.910 --> 00:01:04.870
All right.

00:01:04.870 --> 00:01:09.220
Please feel free to interrupt.

00:01:09.220 --> 00:01:11.470
So that's going to
be my first task.

00:01:11.470 --> 00:01:15.010
And then the second
one is going to be

00:01:15.010 --> 00:01:18.940
what are the things that
people actually do in order

00:01:18.940 --> 00:01:21.040
to try to achieve these goals.

00:01:21.040 --> 00:01:24.400
What is the practice
of medicine like?

00:01:24.400 --> 00:01:27.430
What is the process
that generates the data

00:01:27.430 --> 00:01:30.820
that we're going to be
using to learn from?

00:01:30.820 --> 00:01:33.100
And then I can't resist
talking a little bit

00:01:33.100 --> 00:01:36.850
about paying for health care
at the end of the lecture

00:01:36.850 --> 00:01:40.030
because a lot of the
problems that come up

00:01:40.030 --> 00:01:43.000
and a lot of the
interest that people show

00:01:43.000 --> 00:01:46.360
in doing the kind of analysis
we're talking about in fact

00:01:46.360 --> 00:01:48.260
is motivated by money.

00:01:48.260 --> 00:01:50.920
They want to be able to
save money, or spend less

00:01:50.920 --> 00:01:52.700
money, or something like that.

00:01:52.700 --> 00:01:55.200
So it's important to know that.

00:01:55.200 --> 00:01:56.310
OK.

00:01:56.310 --> 00:01:58.610
Medicine's been around
for a long time.

00:01:58.610 --> 00:02:03.290
I think from probably the
earliest of recorded history,

00:02:03.290 --> 00:02:06.970
there are discussions
of people wondering

00:02:06.970 --> 00:02:11.530
what the cause of disease
is, how to cure it.

00:02:11.530 --> 00:02:15.100
They came up with some
fairly cockamamie theories

00:02:15.100 --> 00:02:18.940
because they didn't have a
lot of scientific, modern

00:02:18.940 --> 00:02:20.620
approaches to it.

00:02:20.620 --> 00:02:24.010
But for example, this
is a photo on the left

00:02:24.010 --> 00:02:29.290
of a shaman I think from a
Canadian Indian tribe who's

00:02:29.290 --> 00:02:33.100
working on the boy
lying there who's sick.

00:02:33.100 --> 00:02:38.290
And this shaman would use
his knowledge of experience

00:02:38.290 --> 00:02:40.720
that he'd had with
other patients.

00:02:40.720 --> 00:02:45.040
They did know a lot
about medicinal plants.

00:02:45.040 --> 00:02:49.120
They knew something about how
to care for injuries and things

00:02:49.120 --> 00:02:50.150
like that.

00:02:50.150 --> 00:02:53.510
And so this was an effective
form of health care.

00:02:53.510 --> 00:02:54.780
Not much record keeping.

00:02:54.780 --> 00:03:00.940
You don't see electronic medical
records system in that scene.

00:03:00.940 --> 00:03:05.170
On the right is a modern
shaman practicing at one

00:03:05.170 --> 00:03:07.870
of the New York area hospitals.

00:03:07.870 --> 00:03:10.330
And so there are
traditional cultures

00:03:10.330 --> 00:03:14.920
in which that sort of hands
on interaction with the healer

00:03:14.920 --> 00:03:17.350
is considered a
very important part

00:03:17.350 --> 00:03:19.550
of the practice of medicine.

00:03:19.550 --> 00:03:23.440
And if you listen to futurist
doctors talking about what

00:03:23.440 --> 00:03:27.850
medicine is likely to be
like, they emphasize the fact

00:03:27.850 --> 00:03:31.000
that the role of a
healer is not just

00:03:31.000 --> 00:03:36.250
to be a good automaton who
figures out the right things

00:03:36.250 --> 00:03:40.630
but is to persuade
a patient to trust

00:03:40.630 --> 00:03:44.230
them to do the things that
he or she is suggesting

00:03:44.230 --> 00:03:46.040
to the patient.

00:03:46.040 --> 00:03:47.890
And there are a lot
of placebo effects

00:03:47.890 --> 00:03:50.860
that we know from lots
and lots of experiments

00:03:50.860 --> 00:03:53.350
that say that if you think
you're going to get better,

00:03:53.350 --> 00:03:57.190
you are going to get
better, on average.

00:03:57.190 --> 00:03:59.140
No guarantees.

00:03:59.140 --> 00:04:00.340
OK.

00:04:00.340 --> 00:04:04.460
Now modern medicine actually
looks more like this.

00:04:04.460 --> 00:04:06.820
So this is an
intensive care unit.

00:04:06.820 --> 00:04:09.250
And what you see
is a patient who

00:04:09.250 --> 00:04:14.260
has got all kinds of
electrical leads, and tubes,

00:04:14.260 --> 00:04:17.019
and things going into
them and is surrounded

00:04:17.019 --> 00:04:19.990
by tons and tons
of equipment which

00:04:19.990 --> 00:04:25.900
is monitoring him and
perhaps keeping him alive.

00:04:25.900 --> 00:04:29.050
And so this is the
high tech medicine

00:04:29.050 --> 00:04:31.690
that we think of as the
contemporary version

00:04:31.690 --> 00:04:32.740
of clinical care.

00:04:35.250 --> 00:04:41.340
Well, you might say, OK, what
does it mean to be healthy?

00:04:41.340 --> 00:04:41.840
Right.

00:04:41.840 --> 00:04:46.070
If the goal of medicine
is to make people or keep

00:04:46.070 --> 00:04:48.350
people healthy, what is health?

00:04:48.350 --> 00:04:51.170
So we turn to the World
Health Organization.

00:04:51.170 --> 00:04:54.290
And they have this lovely,
very comprehensive notion

00:04:54.290 --> 00:04:56.240
of the definition of health.

00:04:56.240 --> 00:05:00.050
A state of complete physical,
mental, and social well-being,

00:05:00.050 --> 00:05:03.420
not merely the absence
of disease or infirmity.

00:05:03.420 --> 00:05:04.550
And then they categorize.

00:05:04.550 --> 00:05:07.190
This and they say, well,
there's physical health,

00:05:07.190 --> 00:05:10.350
there's mental health,
and there's social health.

00:05:10.350 --> 00:05:12.680
Social health is
especially hard to measure.

00:05:12.680 --> 00:05:15.750
And I'll come back to
that in a little while.

00:05:15.750 --> 00:05:20.280
So what's easiest to measure
is how long people live.

00:05:20.280 --> 00:05:24.750
And so we've had data
on survival analysis

00:05:24.750 --> 00:05:26.370
for a long time.

00:05:26.370 --> 00:05:28.950
And this is kind of shocking.

00:05:28.950 --> 00:05:35.700
If you look here, this lower
curve is from around 1800.

00:05:35.700 --> 00:05:40.710
And what it shows you is that if
you lived in India around 1800,

00:05:40.710 --> 00:05:45.560
your life expectancy
was about 25 years.

00:05:45.560 --> 00:05:47.980
It's not very good.

00:05:47.980 --> 00:05:52.360
And if you lived in the richest
countries, which in those days

00:05:52.360 --> 00:05:55.060
were typically
European, in Belgium,

00:05:55.060 --> 00:05:59.850
your life expectancy was
way up there at 40 years.

00:05:59.850 --> 00:06:00.440
Right.

00:06:00.440 --> 00:06:03.400
How many of you knew that?

00:06:03.400 --> 00:06:03.900
OK.

00:06:03.900 --> 00:06:04.590
Good.

00:06:04.590 --> 00:06:07.660
I didn't until I
started looking at this.

00:06:07.660 --> 00:06:11.130
Now by 1950, which is
not that long ago, it

00:06:11.130 --> 00:06:18.030
was like 69 years ago, in
Norway your expectation

00:06:18.030 --> 00:06:20.730
was that you'd live
into your early 70s,

00:06:20.730 --> 00:06:25.710
in the US that you would live
into your late 60s on average.

00:06:25.710 --> 00:06:28.440
There was still a
huge cliff where,

00:06:28.440 --> 00:06:33.120
if you lived in Bhutan,
or Somalia, or something,

00:06:33.120 --> 00:06:34.830
you were still down around 30.

00:06:37.740 --> 00:06:42.330
Today, well in 2012,
we're doing a lot better.

00:06:42.330 --> 00:06:45.630
And the thing that's
striking is not only

00:06:45.630 --> 00:06:49.900
that the people who were
doing well have gotten better

00:06:49.900 --> 00:06:52.650
but that a lot of the people
who were doing very poorly

00:06:52.650 --> 00:06:54.330
have also gotten better.

00:06:54.330 --> 00:06:56.550
And so we're now up--

00:06:56.550 --> 00:07:00.360
India, remember, was at 25
years of life expectancy

00:07:00.360 --> 00:07:02.760
and now it's in the high 60s.

00:07:02.760 --> 00:07:04.410
Now of course,
these are averages.

00:07:04.410 --> 00:07:07.560
And so individuals vary a lot.

00:07:07.560 --> 00:07:09.250
But it's kind of interesting.

00:07:09.250 --> 00:07:12.180
So if you look at
the numbers, you

00:07:12.180 --> 00:07:16.930
see that even on a shorter
term, there are big changes.

00:07:16.930 --> 00:07:20.580
So for example, if
you're a male living

00:07:20.580 --> 00:07:26.190
in Rwanda, which is among the
worst places in terms of life

00:07:26.190 --> 00:07:29.120
expectancy, your
life expectancy,

00:07:29.120 --> 00:07:33.400
if you were born today,
is about 62 and 1/2 years.

00:07:33.400 --> 00:07:34.430
Right.

00:07:34.430 --> 00:07:38.970
If you were born in 2001,
it was only 38 years.

00:07:38.970 --> 00:07:45.360
Now what was going
on in Rwanda in 2001?

00:07:45.360 --> 00:07:45.990
Genocide.

00:07:45.990 --> 00:07:46.490
Yeah.

00:07:46.490 --> 00:07:47.760
They were killing each other.

00:07:47.760 --> 00:07:52.110
So that's sort of an
exceptional situation.

00:07:52.110 --> 00:07:54.960
And that's gotten much
better because they've

00:07:54.960 --> 00:07:58.260
stopped killing each
other as genocidal attacks

00:07:58.260 --> 00:08:03.930
between the Hutu and Tutsi,
I think, if I remember right.

00:08:03.930 --> 00:08:05.160
What about South Africa?

00:08:05.160 --> 00:08:15.640
What was going on in
South Africa in 2001.

00:08:15.640 --> 00:08:16.910
I'm not sure I heard you.

00:08:16.910 --> 00:08:19.025
AUDIENCE: Failure to
address the HIV crisis.

00:08:19.025 --> 00:08:19.900
PETER SZOLOVITS: Yes.

00:08:19.900 --> 00:08:23.530
The government at
the time was claiming

00:08:23.530 --> 00:08:26.980
that HIV was not
the cause of AIDS

00:08:26.980 --> 00:08:30.610
and therefore there was no point
in controlling HIV infections

00:08:30.610 --> 00:08:34.070
because AIDS was caused
by something else.

00:08:34.070 --> 00:08:36.260
Pretty crazy.

00:08:36.260 --> 00:08:37.390
So that was terrible.

00:08:37.390 --> 00:08:39.850
And they've gotten
much better at it.

00:08:39.850 --> 00:08:44.810
So that's what you tend to see
in a lot of African countries.

00:08:44.810 --> 00:08:48.130
And what you also see is
that there has really been

00:08:48.130 --> 00:08:50.000
improvement everywhere.

00:08:50.000 --> 00:08:53.800
So in the US, we went
from males expected

00:08:53.800 --> 00:08:59.500
to live 74 years to almost 78
years, so about a four year

00:08:59.500 --> 00:09:05.330
increase in life expectancy
over a period of just 17 years.

00:09:05.330 --> 00:09:12.720
You women, by the way, are going
to outlive us men, on average.

00:09:12.720 --> 00:09:16.156
There's some biological thing
that seems to work that way.

00:09:16.156 --> 00:09:18.180
OK.

00:09:18.180 --> 00:09:21.870
So a typical way
that people look

00:09:21.870 --> 00:09:25.800
at the survival
of a population is

00:09:25.800 --> 00:09:28.920
to say, well, given
a cohort of people

00:09:28.920 --> 00:09:32.820
born at some instant zero,
what fraction of them

00:09:32.820 --> 00:09:37.090
are still alive after a
certain period of time?

00:09:37.090 --> 00:09:39.620
And what you see
is that, of course,

00:09:39.620 --> 00:09:42.240
2031 we haven't reached yet.

00:09:42.240 --> 00:09:44.820
And so these are
projections based

00:09:44.820 --> 00:09:48.630
on sort of theoretical
extrapolations of actual data.

00:09:48.630 --> 00:09:51.300
But the older data is real.

00:09:51.300 --> 00:09:57.150
And what you see is that
from 1851 to, you know,

00:09:57.150 --> 00:10:02.370
2011 let's say, these
numbers have gone way up.

00:10:02.370 --> 00:10:04.650
Now where have they
gone up the most?

00:10:04.650 --> 00:10:09.540
Well, it used to be that
childhood mortality was

00:10:09.540 --> 00:10:11.350
enormous.

00:10:11.350 --> 00:10:17.830
And so if you look at 1851, by
age 10 about 30% of children

00:10:17.830 --> 00:10:19.930
had died.

00:10:19.930 --> 00:10:23.490
And so we've gotten a lot
better at stopping that

00:10:23.490 --> 00:10:25.790
from happening.

00:10:25.790 --> 00:10:30.250
People also look at
curves like this.

00:10:30.250 --> 00:10:34.440
So this is a distribution
of death rates by age.

00:10:34.440 --> 00:10:37.620
This happens to be for
Japan a few years ago.

00:10:37.620 --> 00:10:40.620
And again, females
do better than males.

00:10:40.620 --> 00:10:45.850
The gold curve in the middle
is the average of the two.

00:10:45.850 --> 00:10:49.770
And this is very typical
of almost any country

00:10:49.770 --> 00:10:50.760
that you look at.

00:10:50.760 --> 00:10:55.290
The shape of this curve
is pretty universal.

00:10:55.290 --> 00:10:56.970
So what does this say?

00:10:56.970 --> 00:11:02.390
It says that when you are born,
there is a relatively high risk

00:11:02.390 --> 00:11:03.780
that you're going to die.

00:11:03.780 --> 00:11:07.820
So these are kids who have
congenital abnormalities,

00:11:07.820 --> 00:11:13.460
have prenatal problems, have
all kinds of difficulties.

00:11:13.460 --> 00:11:14.840
And they don't make it.

00:11:14.840 --> 00:11:19.370
So there's a fairly high
death rate at birth.

00:11:19.370 --> 00:11:24.260
But once you make it to, I
think, about two years old,

00:11:24.260 --> 00:11:31.920
the death rate is down to
about one in 10,000 per year.

00:11:31.920 --> 00:11:32.760
Right.

00:11:32.760 --> 00:11:38.860
And then it stays quite low
until you become a teenager.

00:11:38.860 --> 00:11:41.860
Now why might the death rate go
up when you become a teenager?

00:11:45.300 --> 00:11:48.390
Well, suicide is the
extreme example of that.

00:11:48.390 --> 00:11:51.900
But teenagers tend
to be risk seeking

00:11:51.900 --> 00:11:54.270
rather than risk averse.

00:11:54.270 --> 00:11:56.070
You know, they
start driving cars.

00:11:56.070 --> 00:11:59.610
They go skiing,
skydiving, whatever

00:11:59.610 --> 00:12:01.960
it is that they're doing.

00:12:01.960 --> 00:12:04.050
And they start dying.

00:12:04.050 --> 00:12:07.890
But then if you
make it to about 20,

00:12:07.890 --> 00:12:11.490
then there is a relatively
flat region where by then

00:12:11.490 --> 00:12:14.730
you've developed enough sense
to know what risks are worth

00:12:14.730 --> 00:12:17.350
taking and which ones aren't.

00:12:17.350 --> 00:12:23.190
And so it's relatively flat
until about age 35 or 40

00:12:23.190 --> 00:12:28.230
at which point it starts
inexorably rising.

00:12:28.230 --> 00:12:30.690
And of course, as you
get older and older,

00:12:30.690 --> 00:12:33.450
the probability that you're
going to die in the next year

00:12:33.450 --> 00:12:35.410
becomes higher and higher.

00:12:35.410 --> 00:12:35.910
Right.

00:12:35.910 --> 00:12:37.860
This is uncomfortable
for somebody

00:12:37.860 --> 00:12:41.250
with my amount of gray hair.

00:12:41.250 --> 00:12:44.940
Now there is a peculiarity
in Japan which people

00:12:44.940 --> 00:12:46.540
puzzled over for a while.

00:12:46.540 --> 00:12:52.210
And that is that weird
dip up at age 106.

00:12:52.210 --> 00:12:56.030
So first of all, that's a
very small number of people

00:12:56.030 --> 00:12:57.990
that that represents.

00:12:57.990 --> 00:13:01.240
And it turned out
that it was fraud.

00:13:01.240 --> 00:13:04.740
So there were
families who failed

00:13:04.740 --> 00:13:08.730
to report the death of
their ancient grandmother

00:13:08.730 --> 00:13:11.400
or great-grandmother
because they wanted

00:13:11.400 --> 00:13:13.500
to continue collecting
social security

00:13:13.500 --> 00:13:16.050
payments from the government.

00:13:16.050 --> 00:13:17.175
So that's an artifact.

00:13:19.680 --> 00:13:20.180
OK.

00:13:25.200 --> 00:13:27.420
Now this is a
serious problem which

00:13:27.420 --> 00:13:30.390
we're going to return to in
a more technical way later

00:13:30.390 --> 00:13:34.420
in the semester, which is
this problem of disparities.

00:13:34.420 --> 00:13:37.530
So if you look at, for
example, the difference

00:13:37.530 --> 00:13:42.000
between white and black
female life expectancy

00:13:42.000 --> 00:13:45.360
in the United States, you
see that everybody's life

00:13:45.360 --> 00:13:48.720
expectancy, as we've shown,
is going up gradually

00:13:48.720 --> 00:13:53.190
in this case from 1975 to 2015.

00:13:53.190 --> 00:13:57.840
But there continues to be a gap
between black and white females

00:13:57.840 --> 00:14:00.600
and between black
and white males

00:14:00.600 --> 00:14:05.910
where black patients
are more likely to die

00:14:05.910 --> 00:14:12.120
or less likely to survive longer
given the disparities that

00:14:12.120 --> 00:14:14.760
exist socioeconomically.

00:14:14.760 --> 00:14:16.560
Maybe medically.

00:14:16.560 --> 00:14:18.570
We don't know exactly.

00:14:18.570 --> 00:14:21.000
And then if you
look at Hispanics,

00:14:21.000 --> 00:14:23.610
however, they do pretty well.

00:14:23.610 --> 00:14:26.940
So in 2015, you're
actually a little bit

00:14:26.940 --> 00:14:30.000
better off to be Hispanic,
either male or female,

00:14:30.000 --> 00:14:32.640
than you are to be
either white or black.

00:14:32.640 --> 00:14:36.800
But it's still worse to be black
than to be white or Hispanic.

00:14:36.800 --> 00:14:37.500
Right.

00:14:37.500 --> 00:14:40.680
So these are the
kinds of facts that

00:14:40.680 --> 00:14:46.150
drive some of the issues in
what we do in medical care.

00:14:46.150 --> 00:14:48.180
Now what do people die of?

00:14:48.180 --> 00:14:52.020
Well, about a quarter of
them die of heart disease.

00:14:52.020 --> 00:14:54.510
And a little over a fifth
of them die of cancer.

00:14:54.510 --> 00:14:59.100
This is USA data from 2014 so
it's not completely up to date

00:14:59.100 --> 00:15:01.920
but hasn't changed that much.

00:15:01.920 --> 00:15:04.950
And then there's a
decreasing number

00:15:04.950 --> 00:15:08.290
of deaths from
various other causes.

00:15:08.290 --> 00:15:10.110
So heart disease,
cancer, or chronic

00:15:10.110 --> 00:15:11.790
lower respiratory disease.

00:15:11.790 --> 00:15:14.640
So this is like
COPD that's caused

00:15:14.640 --> 00:15:18.150
by smoking, stuff like that.

00:15:18.150 --> 00:15:22.110
Accidents account for
about 5% of deaths.

00:15:22.110 --> 00:15:27.240
Stroke, cerebral vascular
events, Alzheimer's disease,

00:15:27.240 --> 00:15:31.740
diabetes, influenza, pneumonia,
kidney disease, suicide,

00:15:31.740 --> 00:15:34.920
and then everything else
is about another quarter.

00:15:34.920 --> 00:15:36.340
OK.

00:15:36.340 --> 00:15:38.440
Now take a look at those.

00:15:38.440 --> 00:15:42.035
What kind of diseases
are these, the biggies?

00:15:47.400 --> 00:15:47.900
Well.

00:15:47.900 --> 00:15:48.830
They're chronic.

00:15:48.830 --> 00:15:51.950
Most of them are chronic.

00:15:51.950 --> 00:15:54.910
They're also not infectious.

00:15:54.910 --> 00:15:58.010
But except for
influenza and pneumonia,

00:15:58.010 --> 00:16:02.310
nothing else there is
infectious as far as we know.

00:16:02.310 --> 00:16:05.840
I mean, yeah.

00:16:05.840 --> 00:16:10.460
That asterisk you should
put after every statement

00:16:10.460 --> 00:16:13.040
about current medicine.

00:16:13.040 --> 00:16:16.820
So that's interesting, because
if you wrote the same table

00:16:16.820 --> 00:16:20.730
back in 1850, you would
find that a lot of people

00:16:20.730 --> 00:16:22.860
were dying of infections.

00:16:22.860 --> 00:16:24.810
And they weren't
typically living

00:16:24.810 --> 00:16:28.170
long enough to develop these
lovely chronic diseases

00:16:28.170 --> 00:16:30.540
of the aging.

00:16:30.540 --> 00:16:35.040
So there have been
big changes there.

00:16:35.040 --> 00:16:37.860
Now the other thing
that's worth looking at

00:16:37.860 --> 00:16:42.120
is, in addition to the
reasons that people die,

00:16:42.120 --> 00:16:43.860
they start getting sicker.

00:16:43.860 --> 00:16:48.780
And they are getting sort
of less value out of life

00:16:48.780 --> 00:16:52.670
because they're developing
all these other conditions.

00:16:52.670 --> 00:16:58.310
So if you look at people
over 65, about half of them

00:16:58.310 --> 00:17:02.970
have some form of arthritis.

00:17:02.970 --> 00:17:05.220
And about 40%-- yeah.

00:17:05.220 --> 00:17:09.329
About 40% have hypertension.

00:17:09.329 --> 00:17:12.750
By the way, if you
have trouble with any

00:17:12.750 --> 00:17:15.839
of the medicalese
words, just interrupt.

00:17:15.839 --> 00:17:18.569
Hypertension is
high blood pressure.

00:17:18.569 --> 00:17:21.240
Hearing impairment.

00:17:21.240 --> 00:17:25.079
Me, I'm wearing a
hearing aid on one side

00:17:25.079 --> 00:17:28.470
because my ears are going bad.

00:17:28.470 --> 00:17:30.640
Heart disease, about a quarter.

00:17:30.640 --> 00:17:33.840
Orthostatic impairment--
that means people who wobble

00:17:33.840 --> 00:17:37.440
because their sense of
balance is not so good--

00:17:37.440 --> 00:17:38.580
16%.

00:17:38.580 --> 00:17:43.530
Cataracts, chronic
sinusitis, visual impairment,

00:17:43.530 --> 00:17:48.430
genitourinary problems,
diabetes, et cetera.

00:17:48.430 --> 00:17:50.500
So these are all growing.

00:17:50.500 --> 00:17:53.400
Here's the list of the next 10.

00:17:53.400 --> 00:17:58.890
And varicose veins, hernia,
hemorrhoids, psoriasis,

00:17:58.890 --> 00:18:02.430
hardening of the arteries,
tinnitus, corns, calluses,

00:18:02.430 --> 00:18:06.450
constipation, hay fever, and
cerebral vascular problems.

00:18:06.450 --> 00:18:07.410
All right.

00:18:07.410 --> 00:18:13.500
So people develop these by
the time they're over 65.

00:18:13.500 --> 00:18:16.230
So one question we
might ask is, well,

00:18:16.230 --> 00:18:18.670
what is the quality of life?

00:18:18.670 --> 00:18:22.740
So for example, a
lot of the doctors

00:18:22.740 --> 00:18:26.280
that I started working
with in the 1970s

00:18:26.280 --> 00:18:29.580
were great advocates
of the application

00:18:29.580 --> 00:18:32.280
of decision analysis
decision theory

00:18:32.280 --> 00:18:34.870
to making medical decisions.

00:18:34.870 --> 00:18:39.240
And so the problem is how
do you evaluate an outcome?

00:18:39.240 --> 00:18:42.810
And they said, well, the way
we evaluate an outcome is

00:18:42.810 --> 00:18:45.330
we look at your longevity.

00:18:45.330 --> 00:18:49.350
Obviously, the longer you
live, the better typically.

00:18:49.350 --> 00:18:53.400
But we also look at your quality
of life during that time.

00:18:53.400 --> 00:18:58.710
And we say that if you're
confined to a wheelchair,

00:18:58.710 --> 00:19:01.320
let's say, your quality
of life might not

00:19:01.320 --> 00:19:05.520
be as good as if you
were able to run around,

00:19:05.520 --> 00:19:08.370
or if you're suffering
from chronic pain,

00:19:08.370 --> 00:19:10.890
your quality of life
might not be as good as

00:19:10.890 --> 00:19:13.930
if you were pain free.

00:19:13.930 --> 00:19:16.750
And so we came up with
this model that says,

00:19:16.750 --> 00:19:23.730
well, the value of your life
is essentially an integral

00:19:23.730 --> 00:19:26.760
from time zero to
however long you're

00:19:26.760 --> 00:19:30.510
going to live of
a function Q that

00:19:30.510 --> 00:19:34.920
says this is a measure of
how good your quality of life

00:19:34.920 --> 00:19:39.420
is at that particular point
in time and then some discount

00:19:39.420 --> 00:19:40.600
factor.

00:19:40.600 --> 00:19:41.100
Right.

00:19:41.100 --> 00:19:43.770
So what's the role of
the discount factor?

00:19:43.770 --> 00:19:45.810
Well, it's just
like in economics.

00:19:45.810 --> 00:19:50.790
If I offer you some
horribly painful thing

00:19:50.790 --> 00:19:56.400
today versus 10 years from now,
which are you going to choose?

00:19:56.400 --> 00:20:00.340
Most of us will say later.

00:20:00.340 --> 00:20:02.940
So that's what the
discount factor does.

00:20:02.940 --> 00:20:06.780
Now who knows what the
right discount rate is?

00:20:06.780 --> 00:20:09.600
So in some of their work,
they were doing crazy things

00:20:09.600 --> 00:20:14.340
like taking the financial
discount factors about bank

00:20:14.340 --> 00:20:16.470
interest rates and
things like that

00:20:16.470 --> 00:20:18.810
and applying them to
these health things

00:20:18.810 --> 00:20:21.690
just because they didn't have
any better numbers to do.

00:20:21.690 --> 00:20:23.370
That seems a little suspicious.

00:20:23.370 --> 00:20:25.440
But nevertheless,
methodologically, it's

00:20:25.440 --> 00:20:27.490
a way of doing it.

00:20:27.490 --> 00:20:28.200
OK.

00:20:28.200 --> 00:20:30.930
So how do you measure
the quality of life?

00:20:30.930 --> 00:20:35.680
Well, there is this notion of
the activities of daily living.

00:20:35.680 --> 00:20:38.000
So can you bathe and shower?

00:20:38.000 --> 00:20:41.940
Can you brush your teeth
and groom your hair?

00:20:41.940 --> 00:20:42.900
Can you get dressed?

00:20:42.900 --> 00:20:46.170
Can you go to the toilet,
clean yourself up?

00:20:46.170 --> 00:20:49.260
Are you able to walk,
get in and out of bed,

00:20:49.260 --> 00:20:51.300
get in and out of a chair?

00:20:51.300 --> 00:20:53.530
Can you feed yourself?

00:20:53.530 --> 00:20:56.850
And then there are a bunch of
instrumental factors, things

00:20:56.850 --> 00:20:59.700
like are you able
to clean your house

00:20:59.700 --> 00:21:02.890
and can you manage
your money and so on.

00:21:02.890 --> 00:21:06.150
So these are typically
for older people.

00:21:06.150 --> 00:21:10.440
But they are ways of trying to
quantify that quality of life

00:21:10.440 --> 00:21:15.070
by saying how many of these
things are you able to do.

00:21:15.070 --> 00:21:17.040
And there are a lot of
federal regulations,

00:21:17.040 --> 00:21:22.180
for example, that take advantage
of quantification like this.

00:21:22.180 --> 00:21:26.760
So if you're asking to be put
on some sort of disability

00:21:26.760 --> 00:21:30.900
where the government sends
you a check to keep you alive,

00:21:30.900 --> 00:21:33.390
you have to demonstrate
that you are

00:21:33.390 --> 00:21:35.550
at a certain point
on the scale that's

00:21:35.550 --> 00:21:38.640
derived from these
capabilities in order

00:21:38.640 --> 00:21:41.860
to justifiably get that.

00:21:41.860 --> 00:21:45.220
So occupational therapy
is one of the things

00:21:45.220 --> 00:21:49.300
that people try to
teach the elderly.

00:21:49.300 --> 00:21:54.250
My parents died in their
late 80s and my dad was 90.

00:21:54.250 --> 00:21:59.270
And I remember when he would
have some medical problem,

00:21:59.270 --> 00:22:01.360
then he would be put
into the clutches

00:22:01.360 --> 00:22:05.230
of an occupational therapist
who would try to make sure

00:22:05.230 --> 00:22:13.240
that he was able to communicate,
and get around, and not

00:22:13.240 --> 00:22:17.620
fall for tricks where people
wanted to get him to send all

00:22:17.620 --> 00:22:21.790
his money to them,
or meal preparation

00:22:21.790 --> 00:22:22.940
and stuff like that.

00:22:22.940 --> 00:22:24.580
So these are sort of the--

00:22:27.090 --> 00:22:29.160
occupational is a
funny term for it,

00:22:29.160 --> 00:22:31.710
because this typically
applies to people

00:22:31.710 --> 00:22:35.190
who are retired so it's
not really an occupation.

00:22:35.190 --> 00:22:36.870
But it's the sort
of things that you

00:22:36.870 --> 00:22:40.960
need to do in order to be
able to have a decent life.

00:22:40.960 --> 00:22:44.450
Well, now there's an
interesting valuation question.

00:22:44.450 --> 00:22:48.750
So if you look at
the top right model,

00:22:48.750 --> 00:22:53.160
we actually don't have very
good data on anything other

00:22:53.160 --> 00:22:54.240
than mortality.

00:22:54.240 --> 00:22:56.490
So mortality is who dies.

00:22:56.490 --> 00:22:59.850
So the blue curve
there is the curve

00:22:59.850 --> 00:23:02.910
that you've seen before,
which is, of a cohort,

00:23:02.910 --> 00:23:05.550
how many people are still
alive after a certain number

00:23:05.550 --> 00:23:07.360
of years?

00:23:07.360 --> 00:23:11.730
The red curve is
a morbidity curve

00:23:11.730 --> 00:23:17.700
which says how many of those
people are still alive and have

00:23:17.700 --> 00:23:22.830
no sort of problematic
chronic diseases.

00:23:22.830 --> 00:23:24.600
So they're not in constant pain.

00:23:24.600 --> 00:23:27.090
And they're not immobilized.

00:23:27.090 --> 00:23:31.710
And they're not unable to
do the things that I just

00:23:31.710 --> 00:23:35.520
listed on the previous slides.

00:23:35.520 --> 00:23:39.870
And then disability
is when you really

00:23:39.870 --> 00:23:43.500
become incapable of
taking care of yourself.

00:23:43.500 --> 00:23:47.160
And it typically involves
moving into an assisted living

00:23:47.160 --> 00:23:50.520
facility, or a nursing home,
or something like that, which

00:23:50.520 --> 00:23:53.760
is kind of a nightmare
for many people

00:23:53.760 --> 00:23:56.770
and also very
expensive for society.

00:23:56.770 --> 00:24:00.270
So as I said, the
blue curve there

00:24:00.270 --> 00:24:05.940
is based on actual data for
American females in 1980.

00:24:05.940 --> 00:24:09.360
The red curve is a
hypothetical curve

00:24:09.360 --> 00:24:14.760
where I just assumed that the
rate of developing a morbidity

00:24:14.760 --> 00:24:18.220
is about twice
the rate of dying.

00:24:18.220 --> 00:24:22.060
And the green curve I
assumed that the rate

00:24:22.060 --> 00:24:27.040
of developing a disability is--

00:24:27.040 --> 00:24:30.430
I can't remember-- I think
three times as high as the rate

00:24:30.430 --> 00:24:32.180
of dying, something like that.

00:24:32.180 --> 00:24:35.290
So that's why those
curves are lower.

00:24:35.290 --> 00:24:37.210
And they look
approximately right.

00:24:37.210 --> 00:24:40.270
But we don't have
good data on those.

00:24:40.270 --> 00:24:43.360
Now the question
that you have to ask

00:24:43.360 --> 00:24:46.060
is, how do you want
to change this?

00:24:46.060 --> 00:24:50.980
So for example, suppose that
we kept the same situation.

00:24:50.980 --> 00:24:56.280
We reduced mortality to
20% of its actual rate.

00:24:56.280 --> 00:25:00.810
But we kept the disability
and morbidity rates the same

00:25:00.810 --> 00:25:03.460
as they are on the top left.

00:25:03.460 --> 00:25:04.690
So what would that do?

00:25:04.690 --> 00:25:07.980
Well, what that would do
is create a huge number

00:25:07.980 --> 00:25:18.000
of people who are
disabled because they're

00:25:18.000 --> 00:25:21.000
going to live longer
beyond the point

00:25:21.000 --> 00:25:24.620
where they're able
to function fully.

00:25:24.620 --> 00:25:26.180
So this is-- yeah.

00:25:26.180 --> 00:25:28.450
AUDIENCE: Can I just ask,
why does green not just

00:25:28.450 --> 00:25:29.223
mean healthy?

00:25:29.223 --> 00:25:30.140
To me, it just seems--

00:25:30.140 --> 00:25:31.830
PETER SZOLOVITS: Green
just means healthy.

00:25:31.830 --> 00:25:32.372
AUDIENCE: OK.

00:25:32.372 --> 00:25:33.183
It does.

00:25:33.183 --> 00:25:34.100
PETER SZOLOVITS: Yeah.

00:25:34.100 --> 00:25:37.910
So beyond green is what I
meant is the morbidity curve.

00:25:37.910 --> 00:25:40.850
And beyond red is
the disability curve.

00:25:40.850 --> 00:25:41.870
OK.

00:25:41.870 --> 00:25:44.630
I may have misspoken.

00:25:44.630 --> 00:25:47.520
So green is healthy.

00:25:47.520 --> 00:25:52.160
Red means suffering
from some morbidity.

00:25:52.160 --> 00:25:55.680
And blue means disabled.

00:25:55.680 --> 00:26:00.000
So if we just extend life but
we don't make it any better,

00:26:00.000 --> 00:26:03.470
then that's not a very
attractive picture.

00:26:03.470 --> 00:26:06.710
So other possibilities are
compression of morbidity.

00:26:06.710 --> 00:26:15.080
So for example, if we reduce the
rate at which people get sick

00:26:15.080 --> 00:26:20.300
and die but we increase it--

00:26:20.300 --> 00:26:23.600
we decrease it initially
and then increase it

00:26:23.600 --> 00:26:27.260
so that, on average, people live
about the same length of time

00:26:27.260 --> 00:26:31.580
that they now do, then we're
going to have fewer people who

00:26:31.580 --> 00:26:33.830
are suffering from
morbidity or who

00:26:33.830 --> 00:26:41.490
are disabled because you last
doing well and then you die.

00:26:41.490 --> 00:26:44.960
So this is the
wonderful one horse shay

00:26:44.960 --> 00:26:50.360
where everything falls apart at
once, which, you know, frankly,

00:26:50.360 --> 00:26:52.790
as somebody who's a
little closer to the end

00:26:52.790 --> 00:26:56.640
than you guys are, I wouldn't
mind that kind of exit.

00:26:56.640 --> 00:26:57.830
Right.

00:26:57.830 --> 00:27:01.610
I don't want to be
disabled for 20 years.

00:27:01.610 --> 00:27:03.560
I'd rather live a
long time healthy

00:27:03.560 --> 00:27:05.710
and then drop dead
at some point.

00:27:05.710 --> 00:27:06.210
OK.

00:27:09.650 --> 00:27:12.440
My dad used to say
that he wanted to die

00:27:12.440 --> 00:27:15.080
by being hit by a meteor.

00:27:15.080 --> 00:27:15.800
Right.

00:27:15.800 --> 00:27:17.570
He wouldn't know it's coming.

00:27:17.570 --> 00:27:18.770
It's instant.

00:27:18.770 --> 00:27:20.420
No suffering, no pain.

00:27:20.420 --> 00:27:21.590
Perfect.

00:27:21.590 --> 00:27:25.760
He almost got it but not quite.

00:27:25.760 --> 00:27:26.620
OK.

00:27:26.620 --> 00:27:30.700
And then the final story
is lifespan extension,

00:27:30.700 --> 00:27:38.640
which is where we simply lower
the mortality rate and all

00:27:38.640 --> 00:27:40.870
the other rates in proportion.

00:27:40.870 --> 00:27:42.630
And what happens
is that you start

00:27:42.630 --> 00:27:48.780
having healthy 107-year-olds
and not so healthy

00:27:48.780 --> 00:27:53.580
120-year-olds in the
population in larger numbers

00:27:53.580 --> 00:27:55.790
than we do now.

00:27:55.790 --> 00:27:57.400
OK.

00:27:57.400 --> 00:27:59.200
Social quality of life.

00:27:59.200 --> 00:28:01.102
That's a tough one.

00:28:01.102 --> 00:28:02.320
All right.

00:28:02.320 --> 00:28:05.310
So here's a naive theory.

00:28:05.310 --> 00:28:08.235
We take the quality of life
of every person on the planet

00:28:08.235 --> 00:28:10.500
and we sum them up.

00:28:10.500 --> 00:28:13.240
And we say, OK, that's the
social quality of life.

00:28:16.050 --> 00:28:19.090
Is that a good idea?

00:28:19.090 --> 00:28:19.980
Probably not.

00:28:19.980 --> 00:28:20.670
Right.

00:28:20.670 --> 00:28:22.440
For one thing, it
would say that we

00:28:22.440 --> 00:28:26.730
ought to have as much population
explosion as possible,

00:28:26.730 --> 00:28:30.340
because then we'd have more
people to integrate over,

00:28:30.340 --> 00:28:34.140
which doesn't seem sensible.

00:28:34.140 --> 00:28:34.700
Right.

00:28:34.700 --> 00:28:39.060
Now obviously if we start
packing the world so that it's

00:28:39.060 --> 00:28:41.910
super crowded, then
the quality of life

00:28:41.910 --> 00:28:45.750
will go down eventually
enough that adding more people

00:28:45.750 --> 00:28:48.120
is probably not optimal.

00:28:48.120 --> 00:28:54.240
But nevertheless, that doesn't
seem like a real satisfying

00:28:54.240 --> 00:28:56.800
solution.

00:28:56.800 --> 00:28:58.370
How about less?

00:28:58.370 --> 00:29:00.840
It was popular
about 10 years ago

00:29:00.840 --> 00:29:03.900
for people to write these
sort of speculative books

00:29:03.900 --> 00:29:08.170
about how would the world
look if half the people died.

00:29:08.170 --> 00:29:09.690
Right.

00:29:09.690 --> 00:29:11.250
Other than the
trauma of the half

00:29:11.250 --> 00:29:14.430
the people dying,
you know, they were

00:29:14.430 --> 00:29:20.250
proposing that this would
be some wonderful sylvan

00:29:20.250 --> 00:29:25.440
sort of ideal old
fashioned kind of world.

00:29:25.440 --> 00:29:27.060
I didn't buy that.

00:29:27.060 --> 00:29:29.220
And of course, we don't
know of a good way

00:29:29.220 --> 00:29:32.160
to get there,
although it is true

00:29:32.160 --> 00:29:34.560
that in at least
developed countries,

00:29:34.560 --> 00:29:39.150
birth rates keep falling to
the point where populations are

00:29:39.150 --> 00:29:41.910
worried about underpopulation.

00:29:41.910 --> 00:29:46.020
The Japanese, for example,
have very strict immigration

00:29:46.020 --> 00:29:48.840
policies so you
can't become Japanese

00:29:48.840 --> 00:29:51.270
if you're not born Japanese.

00:29:51.270 --> 00:29:53.820
And Japanese people
aren't having enough kids

00:29:53.820 --> 00:29:55.630
to replace themselves.

00:29:55.630 --> 00:30:01.170
And so the natural population
of Japan is falling.

00:30:01.170 --> 00:30:05.010
Italy is in the same
quandary except that Italy

00:30:05.010 --> 00:30:08.640
has all these immigrants
coming in and trying

00:30:08.640 --> 00:30:09.900
to become Italians.

00:30:09.900 --> 00:30:13.920
And of course, this leads to big
political fights about who gets

00:30:13.920 --> 00:30:18.080
to be an Italian
and who doesn't.

00:30:18.080 --> 00:30:19.880
And then, of course,
there's the question

00:30:19.880 --> 00:30:23.390
of money, which as I say
we'll return to later.

00:30:23.390 --> 00:30:26.830
Now one other important
thing to consider

00:30:26.830 --> 00:30:32.860
is that because of the
increase in life expectancy,

00:30:32.860 --> 00:30:37.180
there has been a big change
in timescale in the way

00:30:37.180 --> 00:30:40.120
people think about medical care.

00:30:40.120 --> 00:30:46.060
So it used to be long,
long ago in the shaman era,

00:30:46.060 --> 00:30:49.600
you wouldn't go to a shaman
to say keep me healthy.

00:30:49.600 --> 00:30:53.170
You would go to a shaman saying,
you know, I broke my arm,

00:30:53.170 --> 00:30:58.400
or I have this pain in my
leg, or fix me somehow.

00:30:58.400 --> 00:31:03.140
And so things were focused
on the notion of cure.

00:31:03.140 --> 00:31:06.960
And that was applicable
to acute illnesses.

00:31:06.960 --> 00:31:09.750
But as we've gotten
better at treating

00:31:09.750 --> 00:31:11.610
acute illnesses--
which, by the way,

00:31:11.610 --> 00:31:13.950
didn't happen all that long ago.

00:31:13.950 --> 00:31:16.410
I mean antibiotics
were only invented

00:31:16.410 --> 00:31:18.480
in the early 20th century.

00:31:18.480 --> 00:31:22.050
And that made a huge
difference in stopping people

00:31:22.050 --> 00:31:24.780
from dying of infections.

00:31:24.780 --> 00:31:28.140
So then it became more a
matter of managing long term

00:31:28.140 --> 00:31:29.970
chronic illnesses.

00:31:29.970 --> 00:31:32.070
And that's pretty
much where we are now.

00:31:32.070 --> 00:31:36.360
The medical world at the
moment, most of the action

00:31:36.360 --> 00:31:39.210
is in trying to understand
things like diabetes,

00:31:39.210 --> 00:31:41.850
and heart disease,
and cancer, and things

00:31:41.850 --> 00:31:44.700
like that that develop
over a long time.

00:31:44.700 --> 00:31:46.500
And they don't
kill you instantly

00:31:46.500 --> 00:31:49.720
like infectious diseases did.

00:31:49.720 --> 00:31:53.370
But they produce a real burden.

00:31:53.370 --> 00:31:56.820
And then, of course, the next
step that everybody expects

00:31:56.820 --> 00:32:00.210
is, well, how do
we prevent disease?

00:32:00.210 --> 00:32:03.720
So how can we change
your exposures?

00:32:03.720 --> 00:32:05.590
How can we change
your motivation?

00:32:05.590 --> 00:32:06.930
How can we change your diet?

00:32:06.930 --> 00:32:10.650
How can we change whatever
it is that we need to change?

00:32:10.650 --> 00:32:13.650
How can we change your
genes to prevent you

00:32:13.650 --> 00:32:16.930
from developing these
diseases in the first place?

00:32:16.930 --> 00:32:18.435
And that's sort of the future.

00:32:19.500 --> 00:32:21.090
OK.

00:32:21.090 --> 00:32:25.440
So that's about what
medicine tries to do.

00:32:25.440 --> 00:32:26.820
But how does it do it?

00:32:26.820 --> 00:32:28.830
And so we're going
to talk a little bit

00:32:28.830 --> 00:32:32.940
about the traditional tasks that
are attributed to health care

00:32:32.940 --> 00:32:34.050
practice.

00:32:34.050 --> 00:32:38.320
So traditionally, people talk
about diagnosis, prognosis,

00:32:38.320 --> 00:32:39.660
and therapy.

00:32:39.660 --> 00:32:42.390
So diagnosis-- I
go to my doctor.

00:32:42.390 --> 00:32:45.210
I say, doc, I've got
this horrible headache.

00:32:45.210 --> 00:32:47.460
I've had it for two weeks.

00:32:47.460 --> 00:32:50.340
What's wrong with me?

00:32:50.340 --> 00:32:56.446
And his job-- in my case,
it happens to be a "he."

00:32:56.446 --> 00:33:00.545
His job is to come
up with an answer.

00:33:00.545 --> 00:33:01.420
What's wrong with me?

00:33:04.340 --> 00:33:08.330
Prognosis is he's
supposed to predict

00:33:08.330 --> 00:33:11.660
what's going to happen to
me, at least if he doesn't

00:33:11.660 --> 00:33:14.130
do anything.

00:33:14.130 --> 00:33:16.770
So is this headache
going to go away

00:33:16.770 --> 00:33:20.400
or is it going to turn out
to be a brain tumor that

00:33:20.400 --> 00:33:25.350
will kill me, or is it going
to be some amoeba that's living

00:33:25.350 --> 00:33:30.270
in my brain and eat my neurons?

00:33:30.270 --> 00:33:34.630
All kinds of horrible
things are possible.

00:33:34.630 --> 00:33:37.470
So it's the prospect of
recovery as anticipated

00:33:37.470 --> 00:33:42.450
from the usual course of disease
or peculiarities of the case.

00:33:42.450 --> 00:33:46.570
And then therapy, of course,
is what do you do about it.

00:33:46.570 --> 00:33:53.420
And prognosis is definitely
informed by diagnosis,

00:33:53.420 --> 00:33:56.570
because if you don't know
what's wrong with me,

00:33:56.570 --> 00:34:00.500
then it's much harder to predict
what's going to happen to me.

00:34:00.500 --> 00:34:03.650
And if you can't predict
what's going to happen to me,

00:34:03.650 --> 00:34:05.690
then it's much
harder to figure out

00:34:05.690 --> 00:34:08.449
what to do to prevent
that from happening

00:34:08.449 --> 00:34:11.670
or to encourage that
to happen, right?

00:34:11.670 --> 00:34:15.570
So this is kind of
a serial process.

00:34:15.570 --> 00:34:18.540
And the way I look
at it is that there's

00:34:18.540 --> 00:34:23.159
a kind of cyclic
process of care.

00:34:23.159 --> 00:34:27.360
And the process starts with
an initial presentation.

00:34:27.360 --> 00:34:29.520
So I show up at
my doctor's office

00:34:29.520 --> 00:34:31.949
and I complain about something.

00:34:31.949 --> 00:34:36.989
And if you ever listen
to a doctor interacting

00:34:36.989 --> 00:34:39.940
with a patient, the first
time the patient comes in,

00:34:39.940 --> 00:34:43.530
the first question is always,
what brought you here?

00:34:43.530 --> 00:34:44.340
Right?

00:34:44.340 --> 00:34:47.400
And that's called the
presenting complaint.

00:34:47.400 --> 00:34:51.659
So if I say, you
know, my ankle hurts

00:34:51.659 --> 00:34:55.170
like hell, that's very
different than if I say,

00:34:55.170 --> 00:34:58.350
I stopped being able to
hear in my right ear,

00:34:58.350 --> 00:35:02.410
or I have this horrible skin
rash on my arm, or whatever.

00:35:02.410 --> 00:35:05.990
And that's going to take me
in very different directions.

00:35:05.990 --> 00:35:10.030
So then what happens is
that the doctor examines you

00:35:10.030 --> 00:35:12.590
and generates a bunch of data.

00:35:12.590 --> 00:35:13.750
So these are measurements.

00:35:13.750 --> 00:35:16.810
And of course, it used to be
that those measurements were

00:35:16.810 --> 00:35:18.850
based on observation.

00:35:18.850 --> 00:35:22.480
So there are very famous
doctors from 100 years ago

00:35:22.480 --> 00:35:26.980
who were spectacularly good at
being able to look at a patient

00:35:26.980 --> 00:35:28.910
and figure out what's
wrong with them

00:35:28.910 --> 00:35:32.050
by being very astute observers--

00:35:32.050 --> 00:35:36.070
the Sherlock Holmes
kind of subtle,

00:35:36.070 --> 00:35:39.820
oh, I see that you have a cut
on the inside of your shoe,

00:35:39.820 --> 00:35:43.030
which means you must have
been going through brambles,

00:35:43.030 --> 00:35:44.620
and you know, whatever.

00:35:44.620 --> 00:35:48.040
I'm making up a Sherlock
Holmes story here.

00:35:48.040 --> 00:35:50.290
So that generates data.

00:35:50.290 --> 00:35:53.110
And then we interpret
that data to generate

00:35:53.110 --> 00:35:56.740
some kind of information
or interpreted data

00:35:56.740 --> 00:35:58.960
about the patient.

00:35:58.960 --> 00:36:02.810
And based on that, we
determine a diagnosis.

00:36:02.810 --> 00:36:05.020
Now, do we determine
a diagnosis?

00:36:05.020 --> 00:36:05.800
Maybe not.

00:36:05.800 --> 00:36:08.870
Maybe we guess a diagnosis.

00:36:08.870 --> 00:36:12.550
One of the things I learned
early on working in this field

00:36:12.550 --> 00:36:17.010
is that doctors are actually
quite willing to make guesses,

00:36:17.010 --> 00:36:20.580
because it's so useful to
believe that you understand

00:36:20.580 --> 00:36:22.530
what's going on.

00:36:22.530 --> 00:36:27.300
If you say, well, there's
some probability distribution

00:36:27.300 --> 00:36:30.720
over a vast number of
possible things, that

00:36:30.720 --> 00:36:34.470
doesn't give you very good
guidance on what to do next.

00:36:34.470 --> 00:36:36.660
Whereas if you can
say, oh, I think

00:36:36.660 --> 00:36:40.530
this patient is developing
type 2 diabetes,

00:36:40.530 --> 00:36:44.640
then you're locked
into a set of questions

00:36:44.640 --> 00:36:48.400
and a set of approaches
that you might try.

00:36:48.400 --> 00:36:52.090
Now, when we come back to
looking at machine learning,

00:36:52.090 --> 00:36:56.060
machines don't have the
same limitations as people.

00:36:56.060 --> 00:36:59.590
And so for a machine to
integrate over a vast number of

00:36:59.590 --> 00:37:04.960
possibilities is not difficult.
But for a human cognition to do

00:37:04.960 --> 00:37:06.890
that is very hard.

00:37:06.890 --> 00:37:10.120
And so this is actually an
important characteristic

00:37:10.120 --> 00:37:14.660
of the way doctors think
about diagnostic reasoning.

00:37:14.660 --> 00:37:18.340
So then, having made a
diagnosis or made a guess,

00:37:18.340 --> 00:37:20.680
they plan some kind of therapy.

00:37:20.680 --> 00:37:24.220
They apply that therapy,
and then they wait awhile

00:37:24.220 --> 00:37:26.410
and they see what happened.

00:37:26.410 --> 00:37:31.210
So if your diagnosis led
you to a choice of therapy,

00:37:31.210 --> 00:37:33.910
you gave that therapy to
the patient and the patient

00:37:33.910 --> 00:37:36.340
got better, then you
say, well, it must

00:37:36.340 --> 00:37:38.980
have been the right diagnosis.

00:37:38.980 --> 00:37:41.140
If the patient
didn't get better,

00:37:41.140 --> 00:37:44.640
then you say, well, how did
what happened to the patient

00:37:44.640 --> 00:37:49.570
differ from what I expected
to happen to the patient?

00:37:49.570 --> 00:37:54.070
And that drives your revision
of this whole process.

00:37:54.070 --> 00:37:57.090
So we again examine
what happened

00:37:57.090 --> 00:37:58.950
as a result of the therapy.

00:37:58.950 --> 00:38:00.090
We gather more data.

00:38:00.090 --> 00:38:01.020
We interpret it.

00:38:01.020 --> 00:38:04.530
We come up with the revised
diagnostic hypothesis.

00:38:04.530 --> 00:38:07.350
We come up with a
revised therapeutic plan,

00:38:07.350 --> 00:38:11.040
and we keep going
around the cycle.

00:38:11.040 --> 00:38:14.910
Now, that cycle
happens very quickly

00:38:14.910 --> 00:38:18.360
if you're a hospitalized
patient, because you're there

00:38:18.360 --> 00:38:19.110
all the time.

00:38:19.110 --> 00:38:20.070
You're available.

00:38:20.070 --> 00:38:23.170
They're trying to do
things to you constantly.

00:38:23.170 --> 00:38:28.600
And so this cycle happens on
the order of hours or a day,

00:38:28.600 --> 00:38:30.570
whereas if you were
an outpatient, then

00:38:30.570 --> 00:38:32.940
you're not dealing with
some urgent problem.

00:38:32.940 --> 00:38:36.060
It may happen over a
much slower period.

00:38:36.060 --> 00:38:39.190
It may be that your
doctor says, well,

00:38:39.190 --> 00:38:40.860
we're going to
adjust your drug dose

00:38:40.860 --> 00:38:44.400
and see if that helps bring
down your cholesterol,

00:38:44.400 --> 00:38:48.750
or manage your pain, or whatever
it is that he's trying to do.

00:38:48.750 --> 00:38:51.060
Or worse yet, we're
going to try to convince

00:38:51.060 --> 00:38:55.050
you to eat more healthy food,
and six months later we'll

00:38:55.050 --> 00:38:58.500
see if your hemoglobin
A1C came down, that you're

00:38:58.500 --> 00:39:02.310
less close to being diabetic.

00:39:02.310 --> 00:39:04.560
So the time scale
is very different.

00:39:04.560 --> 00:39:11.220
But that process of continually
reinterpreting things

00:39:11.220 --> 00:39:15.280
is a really critical feature, I
think, of all of medical care.

00:39:15.280 --> 00:39:20.640
And if you look back,
Alan Turing actually

00:39:20.640 --> 00:39:25.260
talked, in the early
1950s, about health care

00:39:25.260 --> 00:39:28.440
as being one of the
interesting application areas

00:39:28.440 --> 00:39:31.980
of artificial intelligence.

00:39:31.980 --> 00:39:32.550
Why?

00:39:32.550 --> 00:39:35.130
Well, because it was
an important topic.

00:39:35.130 --> 00:39:37.530
And he had the vision
that says, as we

00:39:37.530 --> 00:39:40.510
start getting more
data about health care,

00:39:40.510 --> 00:39:42.990
we're going to be able to
build the kinds of models

00:39:42.990 --> 00:39:47.670
that we're going to be
talking about in this class.

00:39:47.670 --> 00:39:53.803
But a lot of the early work took
a kind of one-shot approach.

00:39:53.803 --> 00:39:55.470
So they said, well,
we're going to solve

00:39:55.470 --> 00:39:57.520
the diagnostic problem.

00:39:57.520 --> 00:40:00.570
So we're going to take a
snapshot of a patient, all

00:40:00.570 --> 00:40:02.340
their data at a
particular moment.

00:40:02.340 --> 00:40:04.530
We're going to feed
it into an algorithm.

00:40:04.530 --> 00:40:06.660
It'll come up with a diagnosis.

00:40:06.660 --> 00:40:09.490
We're done.

00:40:09.490 --> 00:40:13.060
And that wasn't very
useful, because it

00:40:13.060 --> 00:40:17.650
didn't obey the cyclic nature of
the process of providing health

00:40:17.650 --> 00:40:18.730
care.

00:40:18.730 --> 00:40:22.480
And so this was a revolution
that started, really,

00:40:22.480 --> 00:40:26.440
around the 1980s when people
realized that you have

00:40:26.440 --> 00:40:29.620
to be in it for the
long-run and not for sort

00:40:29.620 --> 00:40:31.550
of single interactions.

00:40:31.550 --> 00:40:37.640
OK, well, this is
just some definitions

00:40:37.640 --> 00:40:40.260
of these care processes.

00:40:40.260 --> 00:40:46.580
So here I've listed some ideas
that came from a 1976 paper

00:40:46.580 --> 00:40:50.780
by several of my
colleagues, who said, well,

00:40:50.780 --> 00:40:53.750
here's a cognitive
theory of diagnosis.

00:40:53.750 --> 00:40:59.240
From the initial complaints,
guess a suitable hypothesis.

00:40:59.240 --> 00:41:01.850
Use the current
active hypotheses

00:41:01.850 --> 00:41:03.710
to guide questioning--

00:41:03.710 --> 00:41:08.430
so to order more tests, to
ask questions of the patient.

00:41:08.430 --> 00:41:11.460
And it's the failure
to satisfy expectations

00:41:11.460 --> 00:41:15.750
that's the strongest clue to how
to develop a better hypothesis.

00:41:18.540 --> 00:41:22.140
And then the hypotheses
could be in an activated,

00:41:22.140 --> 00:41:25.740
deactivated, confirmed,
or rejected state.

00:41:25.740 --> 00:41:27.930
They actually built
a computer program

00:41:27.930 --> 00:41:32.460
that implemented this theory
of diagnostic reasoning.

00:41:32.460 --> 00:41:38.540
And these rules, essentially,
about whether to activate,

00:41:38.540 --> 00:41:41.370
deactivate, confirm,
or reject something

00:41:41.370 --> 00:41:44.340
could be based both
on logical criteria

00:41:44.340 --> 00:41:49.210
and on a kind of very
bad probabilistic model.

00:41:49.210 --> 00:41:52.230
So it was very bad,
because what they really

00:41:52.230 --> 00:41:54.510
needed was Bayesian networks.

00:41:54.510 --> 00:41:59.280
And those were about a decade
in the future at that point.

00:41:59.280 --> 00:42:04.920
So they and every other
system built in the 1970s

00:42:04.920 --> 00:42:07.320
had really horrible
probabilistic models,

00:42:07.320 --> 00:42:10.570
because we didn't understand
how to do it correctly.

00:42:10.570 --> 00:42:13.080
Now, what's interesting
is somebody noticed

00:42:13.080 --> 00:42:16.110
that if you strip away
the medicine from this,

00:42:16.110 --> 00:42:20.760
this is kind of like the
scientific method, right?

00:42:20.760 --> 00:42:23.340
If you're trying to
understand something,

00:42:23.340 --> 00:42:24.810
you form a hypothesis.

00:42:24.810 --> 00:42:26.670
You perform an experiment.

00:42:26.670 --> 00:42:30.330
If the experiment is consistent
with your expectations,

00:42:30.330 --> 00:42:33.420
then you go on and you've gotten
a little bit more confident

00:42:33.420 --> 00:42:35.010
in your hypothesis.

00:42:35.010 --> 00:42:36.960
If your experiment
is inconsistent

00:42:36.960 --> 00:42:40.740
with your expectations, then
you have to change your theory,

00:42:40.740 --> 00:42:42.510
change your hypothesis.

00:42:42.510 --> 00:42:46.140
You go back and gather more
data, and then keep doing this

00:42:46.140 --> 00:42:48.180
until you're satisfied
that you've come up

00:42:48.180 --> 00:42:50.290
with an adequate theory.

00:42:50.290 --> 00:42:52.890
So this was a
surprise to doctors,

00:42:52.890 --> 00:42:55.410
because they thought
of themselves more

00:42:55.410 --> 00:42:58.090
as artists than as scientists.

00:42:58.090 --> 00:43:00.480
But in a way, they
act like scientists,

00:43:00.480 --> 00:43:02.620
which is kind of cool.

00:43:02.620 --> 00:43:06.390
All right, this doesn't
stop with caring

00:43:06.390 --> 00:43:08.310
for a single patient.

00:43:08.310 --> 00:43:12.510
So we have all these meta-level
processes about the acquisition

00:43:12.510 --> 00:43:15.840
and the application of
knowledge about education,

00:43:15.840 --> 00:43:20.280
quality control and process
improvement, cost containment,

00:43:20.280 --> 00:43:22.260
and developing references.

00:43:22.260 --> 00:43:25.320
So this is a picture
from David Margulies,

00:43:25.320 --> 00:43:28.950
who was chief information
officer at Children's Hospital

00:43:28.950 --> 00:43:29.820
here.

00:43:29.820 --> 00:43:34.500
And the cycle that I described
is the one right here.

00:43:34.500 --> 00:43:38.160
This is the care team
taking care of a patient.

00:43:38.160 --> 00:43:42.420
But of course at some point,
that patient is discharged.

00:43:42.420 --> 00:43:46.080
And then they're in
community care and self-care,

00:43:46.080 --> 00:43:49.230
and then maybe they're in some
sort of active health status

00:43:49.230 --> 00:43:51.480
management.

00:43:51.480 --> 00:43:54.390
And then if that goes
badly, then there's

00:43:54.390 --> 00:43:58.080
some episode where they
reconnect to the health care

00:43:58.080 --> 00:43:59.170
system.

00:43:59.170 --> 00:44:01.890
They get authorized
to come back.

00:44:01.890 --> 00:44:06.310
They get scheduled for a visit,
and they're back in the cycle.

00:44:06.310 --> 00:44:10.590
And so the process
of care involves

00:44:10.590 --> 00:44:13.350
people going to the
hospital, getting taken

00:44:13.350 --> 00:44:17.040
care of, they get better,
they get discharged,

00:44:17.040 --> 00:44:19.350
they live their
lives for a while.

00:44:19.350 --> 00:44:21.060
Maybe they get sick again.

00:44:21.060 --> 00:44:21.960
They come back.

00:44:21.960 --> 00:44:26.910
And so there's this larger
cycle around that issue.

00:44:26.910 --> 00:44:31.590
And then around that are all
kinds of things about health

00:44:31.590 --> 00:44:36.660
plan design and membership
and what coverage

00:44:36.660 --> 00:44:38.400
you have and so on.

00:44:38.400 --> 00:44:41.880
And then I would add
one more idea, which

00:44:41.880 --> 00:44:45.690
is that if you have
a system like this,

00:44:45.690 --> 00:44:50.340
you actually want to study,
at the next meta-level,

00:44:50.340 --> 00:44:53.980
that system, make
observations about it,

00:44:53.980 --> 00:44:57.360
analyze it, model it,
plan some improvements,

00:44:57.360 --> 00:45:02.310
and then intervene in the system
and observe how it's working

00:45:02.310 --> 00:45:05.340
and try to make it better.

00:45:05.340 --> 00:45:11.340
So in terms of tasks that are
important for us in this class,

00:45:11.340 --> 00:45:17.490
this class of tasks is
central, because one

00:45:17.490 --> 00:45:20.100
of the things we're
trying to do is

00:45:20.100 --> 00:45:22.020
to look at the way
health care works

00:45:22.020 --> 00:45:24.240
and to figure out
how to make it better

00:45:24.240 --> 00:45:26.520
by examining its operation.

00:45:26.520 --> 00:45:29.310
And that can be done at
any of these three levels.

00:45:29.310 --> 00:45:32.940
It can be done in the
more acute phase, where

00:45:32.940 --> 00:45:34.440
we're dealing with
somebody who's

00:45:34.440 --> 00:45:36.090
in the middle of an illness.

00:45:36.090 --> 00:45:39.840
It can be done at the larger
phase of somebody who's

00:45:39.840 --> 00:45:42.630
going through the cycle
of being well for a while

00:45:42.630 --> 00:45:46.080
and then being sick, and then
being well again and being sick

00:45:46.080 --> 00:45:46.950
again.

00:45:46.950 --> 00:45:50.980
And it can be in terms
of the system itself,

00:45:50.980 --> 00:45:54.570
of how do you design a health
care system that works better

00:45:54.570 --> 00:45:56.290
for the population?

00:45:56.290 --> 00:45:59.100
So this notion of a
learning health care system

00:45:59.100 --> 00:46:01.290
is now a journal.

00:46:01.290 --> 00:46:05.790
So in 2017, Chuck Friedman
at the University of Michigan

00:46:05.790 --> 00:46:08.130
started this new journal.

00:46:08.130 --> 00:46:13.350
And it's full of articles
about this third level

00:46:13.350 --> 00:46:17.790
of external cycle.

00:46:17.790 --> 00:46:20.960
So how does the health
care system learn?

00:46:20.960 --> 00:46:23.900
Well, I'll tell you an anecdote.

00:46:23.900 --> 00:46:30.900
In the mid-1980s, I was teaching
an AI expert systems course.

00:46:30.900 --> 00:46:33.440
And I had just come
back from a conference

00:46:33.440 --> 00:46:35.960
of medical informatics
people, where

00:46:35.960 --> 00:46:39.200
they were talking about
this great new idea called

00:46:39.200 --> 00:46:42.090
evidence-based medicine.

00:46:42.090 --> 00:46:45.600
And I remember mentioning this
to a bunch of MIT engineering

00:46:45.600 --> 00:46:46.260
students.

00:46:46.260 --> 00:46:50.900
And one of them raised his hand
and said, as opposed to what?

00:46:50.900 --> 00:46:52.160
Right?

00:46:52.160 --> 00:46:54.540
I mean, to an
engineer, it's obvious

00:46:54.540 --> 00:46:58.260
that evidence is the
basis on which you analyze

00:46:58.260 --> 00:47:00.460
things and make things better.

00:47:00.460 --> 00:47:03.120
But it wasn't
obvious to doctors.

00:47:03.120 --> 00:47:06.820
And so this was almost
a revolutionary change.

00:47:06.820 --> 00:47:09.120
And the idea that
they fostered was

00:47:09.120 --> 00:47:13.510
the idea of the randomized
controlled clinical trial.

00:47:13.510 --> 00:47:17.070
So I'm going to sketch
what that looks like.

00:47:17.070 --> 00:47:18.750
Of course, there
are many variations.

00:47:18.750 --> 00:47:22.530
But suppose that I'm one of the
drug companies around Kendall

00:47:22.530 --> 00:47:25.830
Square, and I come
up with a new drug

00:47:25.830 --> 00:47:28.770
and I want to prove that it's
more effective for condition

00:47:28.770 --> 00:47:32.540
X than some existing drug B.

00:47:32.540 --> 00:47:34.130
So what do I do?

00:47:34.130 --> 00:47:38.060
I find some patients
who are suffering from X

00:47:38.060 --> 00:47:41.000
and I try very hard to
find patients who are not

00:47:41.000 --> 00:47:43.220
suffering from anything else.

00:47:43.220 --> 00:47:46.290
I want the purest case possible.

00:47:46.290 --> 00:47:49.190
I then go to my
statisticians and I

00:47:49.190 --> 00:47:52.310
say, let's design
an experiment where

00:47:52.310 --> 00:47:55.550
we're going to collect
a standard set of data

00:47:55.550 --> 00:47:57.880
about all these patients.

00:47:57.880 --> 00:48:02.500
And then we're going to give
some of Drug A and some of them

00:48:02.500 --> 00:48:06.280
Drug B, and we'll see
which of them do better.

00:48:06.280 --> 00:48:10.180
And we'll pre-define what
we mean by do better.

00:48:10.180 --> 00:48:13.210
So like, not dying is
considered doing better,

00:48:13.210 --> 00:48:17.970
or not suffering some bad thing
that people are suffering from

00:48:17.970 --> 00:48:21.830
is considered doing better.

00:48:21.830 --> 00:48:25.240
And then the
statisticians also will

00:48:25.240 --> 00:48:30.370
tell me, given that you
expect Drug A to be,

00:48:30.370 --> 00:48:33.820
let's say, 10%
better than Drug B,

00:48:33.820 --> 00:48:37.480
how many patients do you
have to enroll in this trial

00:48:37.480 --> 00:48:40.990
in order to be likely to get
a statistically significant

00:48:40.990 --> 00:48:43.690
answer to that question?

00:48:43.690 --> 00:48:45.370
And then they do it.

00:48:45.370 --> 00:48:48.250
The statisticians
analyze the data.

00:48:48.250 --> 00:48:51.610
Hopefully you've gotten
p less than 0.05.

00:48:51.610 --> 00:48:53.680
You go to the Food and
Drug Administration

00:48:53.680 --> 00:48:57.670
and say, give me permission to
market this drug as the hottest

00:48:57.670 --> 00:49:01.150
new cure for something
or other, and then you

00:49:01.150 --> 00:49:02.990
make billions of dollars.

00:49:02.990 --> 00:49:03.490
Right?

00:49:03.490 --> 00:49:07.490
This is the standard
way that pharma works.

00:49:07.490 --> 00:49:09.770
Now, there are some problems.

00:49:09.770 --> 00:49:16.080
So most of the cases
to which the results

00:49:16.080 --> 00:49:19.620
of a trial like this are
applied wouldn't have

00:49:19.620 --> 00:49:23.530
qualified to be in the trial.

00:49:23.530 --> 00:49:29.150
For example, we talked
about morbidities, about

00:49:29.150 --> 00:49:31.860
the chronic problems
that people have.

00:49:31.860 --> 00:49:34.140
Well, if you're dealing
with one disease,

00:49:34.140 --> 00:49:37.620
you want to make sure that those
populations you're dealing with

00:49:37.620 --> 00:49:39.880
don't have any of
these other diseases.

00:49:39.880 --> 00:49:42.430
But in the real
world, people do.

00:49:42.430 --> 00:49:44.430
And so we've never
actually measured

00:49:44.430 --> 00:49:47.820
what happens to those people
if you give them this drug

00:49:47.820 --> 00:49:52.020
and they have these
comorbidities.

00:49:52.020 --> 00:49:54.420
The other problem is
that the drug company

00:49:54.420 --> 00:49:57.450
wants to start making
these billions of dollars

00:49:57.450 --> 00:49:59.650
as soon as possible.

00:49:59.650 --> 00:50:03.780
And so they want the trial
to be as short as possible.

00:50:03.780 --> 00:50:06.900
And they want it to be
as small a sample as they

00:50:06.900 --> 00:50:11.320
need to get that 0.05
statistical significance.

00:50:11.320 --> 00:50:17.410
So these are all problematic,
and they lead to real problems.

00:50:17.410 --> 00:50:19.560
So I didn't bring any
examples, but there

00:50:19.560 --> 00:50:23.400
are plenty of stories
where the FDA has approved

00:50:23.400 --> 00:50:25.770
some drug on this
basis, and then

00:50:25.770 --> 00:50:30.450
later they discovered that
although the drug works well

00:50:30.450 --> 00:50:32.910
in the short-term,
it has horrible side

00:50:32.910 --> 00:50:36.600
effects in the long-term,
or that it has interactions

00:50:36.600 --> 00:50:39.960
with other diseases where
it doesn't work effectively

00:50:39.960 --> 00:50:45.600
for people except for these pure
cases that it was tested on.

00:50:45.600 --> 00:50:50.520
So the other idea, the
competing idea is to say, let's

00:50:50.520 --> 00:50:54.120
build this learning health
care system in which progress

00:50:54.120 --> 00:50:55.650
in science, informatics--

00:50:55.650 --> 00:50:58.680
whatever-- generates
new knowledge

00:50:58.680 --> 00:51:02.820
as an ongoing natural byproduct
of the care experience

00:51:02.820 --> 00:51:07.200
and we seamlessly refine
and deliver best practices

00:51:07.200 --> 00:51:10.350
for continuous improvement
in health and health care.

00:51:10.350 --> 00:51:13.170
Wonderful words from the
Institute of Medicine,

00:51:13.170 --> 00:51:17.760
now called the National
Academy of Medicine.

00:51:17.760 --> 00:51:19.870
But it's hard to do this.

00:51:19.870 --> 00:51:21.660
And the reason it's
hard to do this

00:51:21.660 --> 00:51:26.880
is mainly for a very
profound underlying reason,

00:51:26.880 --> 00:51:29.700
which is that people
are not treated

00:51:29.700 --> 00:51:32.280
by experimental protocols.

00:51:32.280 --> 00:51:35.520
So it's very important
in that sketch

00:51:35.520 --> 00:51:39.120
that I gave you of the
Drug A versus Drug B

00:51:39.120 --> 00:51:43.950
that there is a randomization
step where I flip a coin

00:51:43.950 --> 00:51:47.430
to decide which drug any
particular individual is going

00:51:47.430 --> 00:51:49.230
to get.

00:51:49.230 --> 00:51:55.140
If I allow that decision to
be biased by my expectations

00:51:55.140 --> 00:51:57.990
or by something else I
know about the patient,

00:51:57.990 --> 00:52:01.320
then I'm no longer
doing a fair trial.

00:52:01.320 --> 00:52:03.600
And of course,
when I collect data

00:52:03.600 --> 00:52:06.840
about how actual patients
are being treated,

00:52:06.840 --> 00:52:10.020
they're being treated according
to what their doctors think

00:52:10.020 --> 00:52:14.360
is best for them, and so
there is no randomization.

00:52:14.360 --> 00:52:17.310
I mean, if I went to Mass
General and said, could you

00:52:17.310 --> 00:52:19.470
guys please treat
everybody randomly

00:52:19.470 --> 00:52:23.010
so that we collect
really good data,

00:52:23.010 --> 00:52:27.310
they would throw
me out properly.

00:52:27.310 --> 00:52:32.520
OK, so we also need a whole lot
of technical infrastructure.

00:52:32.520 --> 00:52:35.910
We need to capture all kinds of
novel data sources, which we'll

00:52:35.910 --> 00:52:38.830
talk about in the next lecture.

00:52:38.830 --> 00:52:40.920
And then we need a
technical infrastructure

00:52:40.920 --> 00:52:42.780
for truly big data.

00:52:42.780 --> 00:52:51.230
So just for example, Dana Farber
started about five years ago--

00:52:51.230 --> 00:52:52.370
it's a cancer hospital.

00:52:52.370 --> 00:52:54.980
And so for every
solid tumor, they

00:52:54.980 --> 00:52:58.670
would take samples of the
tumor and genotype it--

00:52:58.670 --> 00:53:02.940
multiple samples, because
tumors are not uniform.

00:53:02.940 --> 00:53:06.230
So just storing that stuff
is a technical challenge,

00:53:06.230 --> 00:53:08.960
and being able to
come up with it.

00:53:08.960 --> 00:53:16.250
You've got three gigabases, so
about over a gigabyte of data--

00:53:16.250 --> 00:53:19.060
from each sample,
from each tumor,

00:53:19.060 --> 00:53:23.420
times all the people who
come in and have this test.

00:53:23.420 --> 00:53:26.720
So you buy some big
disk drives or you farm

00:53:26.720 --> 00:53:28.620
it out to Google or something.

00:53:28.620 --> 00:53:30.830
But then you need to
organize it in some way

00:53:30.830 --> 00:53:36.170
so that it's usefully
easy to find that data.

00:53:36.170 --> 00:53:40.520
So today's technique,
today's prejudice

00:53:40.520 --> 00:53:44.870
is what I call the meat
grinder story, which

00:53:44.870 --> 00:53:47.220
is you take medical
records, genetic data,

00:53:47.220 --> 00:53:50.180
environmental data,
data from wearables,

00:53:50.180 --> 00:53:53.210
you put them into an
old-fashioned meat grinder,

00:53:53.210 --> 00:53:56.780
and out come bits, which
you store on a disk.

00:53:56.780 --> 00:53:58.880
And then you have all
the data from which

00:53:58.880 --> 00:54:00.570
you can build models.

00:54:00.570 --> 00:54:02.030
And that's what we do.

00:54:02.030 --> 00:54:05.990
You're going to see a lot
of that in this course.

00:54:05.990 --> 00:54:11.070
OK, the other thing that
medicine tries to do

00:54:11.070 --> 00:54:14.430
is not to cure people
but to keep them healthy.

00:54:14.430 --> 00:54:19.200
And this has been pretty much
the domain of the public health

00:54:19.200 --> 00:54:20.730
infrastructure.

00:54:20.730 --> 00:54:24.980
So if you go across the river
to the Harvard Medical area,

00:54:24.980 --> 00:54:26.730
there are a couple of
big buildings, which

00:54:26.730 --> 00:54:28.560
is the Harvard School
of Public Health,

00:54:28.560 --> 00:54:31.720
and this is what
they're all about.

00:54:31.720 --> 00:54:35.490
So they do things like
tracking disease prevalence

00:54:35.490 --> 00:54:37.860
and tracking
infections, and worrying

00:54:37.860 --> 00:54:40.110
about quarantining people.

00:54:40.110 --> 00:54:42.330
They also do a lot
of the kind of work

00:54:42.330 --> 00:54:45.570
we're going to talk about in
this class, which is modeling

00:54:45.570 --> 00:54:50.130
in order to try to understand
what's going on in individual's

00:54:50.130 --> 00:54:52.620
health, in the health
of a population,

00:54:52.620 --> 00:54:55.620
in the operations of
a health care system.

00:54:55.620 --> 00:54:57.810
So they're very much into this.

00:54:57.810 --> 00:55:03.320
Now historically, I looked
back and it turns out

00:55:03.320 --> 00:55:07.160
there's something called the
London Bills of Mortality

00:55:07.160 --> 00:55:12.890
in the 17th century, started by
a gentleman named John Graunt.

00:55:12.890 --> 00:55:17.210
And he was interested
just in figuring out,

00:55:17.210 --> 00:55:20.090
how long do people live?

00:55:20.090 --> 00:55:24.020
And so he came up with
these bills of mortality,

00:55:24.020 --> 00:55:26.930
where he went around to
different parts of London,

00:55:26.930 --> 00:55:29.780
talked to undertakers
and hospitals

00:55:29.780 --> 00:55:34.130
and whatever health care
providers existed at the time,

00:55:34.130 --> 00:55:38.210
and collected data
on what people died

00:55:38.210 --> 00:55:40.890
and how many people were
living in that area.

00:55:40.890 --> 00:55:46.550
And so for example, he estimated
that the mortality before age

00:55:46.550 --> 00:55:52.220
six in the 17th century-- this
was a long time ago, 1600s--

00:55:52.220 --> 00:55:54.750
was about 36%.

00:55:54.750 --> 00:56:00.060
So if you were a kid, your
chances of making it to age six

00:56:00.060 --> 00:56:06.720
were only about 64%,
so less than 2/3.

00:56:06.720 --> 00:56:08.850
Kind of shocking.

00:56:08.850 --> 00:56:12.030
In the 18th century,
people you've never heard

00:56:12.030 --> 00:56:14.850
of-- and Linnaeus, who
you probably have heard

00:56:14.850 --> 00:56:17.790
of, because he was one
of the early taxonomists

00:56:17.790 --> 00:56:22.710
for biological and animal
species and so on--

00:56:22.710 --> 00:56:26.580
made the first attempts at
systemic classification.

00:56:26.580 --> 00:56:31.110
In the mid-1850s,
mid-1800s, there

00:56:31.110 --> 00:56:34.890
was a Congress of the First
International Statistical

00:56:34.890 --> 00:56:36.270
Congress.

00:56:36.270 --> 00:56:39.990
And a gentleman named
William Farr came up

00:56:39.990 --> 00:56:44.110
with an interesting
categorization that said, well,

00:56:44.110 --> 00:56:46.980
if we're going to
taxonomize the diseases,

00:56:46.980 --> 00:56:50.010
we should separate
epidemic diseases

00:56:50.010 --> 00:56:54.480
from constitutional diseases
from local diseases,

00:56:54.480 --> 00:56:57.060
whereby "local"
he means affecting

00:56:57.060 --> 00:57:02.810
a particular part of the body
from developmental diseases.

00:57:02.810 --> 00:57:06.070
So this is things
like stunted growth

00:57:06.070 --> 00:57:10.930
or failure of mental development
or speech development,

00:57:10.930 --> 00:57:14.000
and then diseases that are
the direct result of violence.

00:57:14.000 --> 00:57:18.880
So this is things like broken
bones and bar fight results

00:57:18.880 --> 00:57:21.110
and stuff like that.

00:57:21.110 --> 00:57:24.400
So that was the
first classification

00:57:24.400 --> 00:57:28.780
of disease in about 1853.

00:57:28.780 --> 00:57:32.170
Note, by the way, that this
is before Louis Pasteur

00:57:32.170 --> 00:57:36.260
and his theory of the
germ cause of disease.

00:57:36.260 --> 00:57:38.500
And so this was a
pretty early attempt,

00:57:38.500 --> 00:57:42.400
and obviously could
have benefited from what

00:57:42.400 --> 00:57:45.010
Pasteur later discovered.

00:57:45.010 --> 00:57:49.210
So by the 1890s,
which is post-Pasteur,

00:57:49.210 --> 00:57:51.730
they came up with
a classification

00:57:51.730 --> 00:57:54.730
that was a hierarchic
classification, 44

00:57:54.730 --> 00:58:00.910
top-level hierarchies broken up
into 99 lower-level categories

00:58:00.910 --> 00:58:04.390
and 161 particular titles.

00:58:04.390 --> 00:58:07.990
And they adopted this
as a way of getting,

00:58:07.990 --> 00:58:10.900
typically, mortality
data of what was it

00:58:10.900 --> 00:58:12.820
that people were dying of.

00:58:12.820 --> 00:58:17.740
And by the 1920s,
you've heard of ICD--

00:58:17.740 --> 00:58:20.090
ICD-9, ICD-10.

00:58:20.090 --> 00:58:23.980
So this is currently used as
a way of classifying diseases

00:58:23.980 --> 00:58:26.530
and disorders.

00:58:26.530 --> 00:58:30.370
The International List
of the Causes of Death

00:58:30.370 --> 00:58:35.410
was the first ICD back
in, I think, the 1920s.

00:58:35.410 --> 00:58:39.820
And then it kept developing
through multiple versions.

00:58:39.820 --> 00:58:43.210
In 1975, ICD-9 was adopted.

00:58:43.210 --> 00:58:46.660
In 2015, ICD-10.

00:58:46.660 --> 00:58:48.880
And these are under the
control of the World Health

00:58:48.880 --> 00:58:52.720
Organization, which is now a
UN agency, although I think

00:58:52.720 --> 00:58:56.290
it predates the UN actually.

00:58:56.290 --> 00:59:00.630
And then we have ICD-9
CM and ICD-10 CM,

00:59:00.630 --> 00:59:04.150
are US Clinical
Modifications that

00:59:04.150 --> 00:59:08.390
are an extension of the
ICD-9 and 10 coding.

00:59:08.390 --> 00:59:12.130
And they are primarily
used for billing.

00:59:12.130 --> 00:59:15.490
But they're also used for
epidemiological research.

00:59:15.490 --> 00:59:21.040
And if you look at the Centers
for Disease Control, CDC,

00:59:21.040 --> 00:59:24.280
they collect death
certificates like this

00:59:24.280 --> 00:59:26.180
from all over the country.

00:59:26.180 --> 00:59:30.970
And so this is a person who died
of a cerebral hemorrhage which

00:59:30.970 --> 00:59:35.140
was due to nephritis, which was
due to cirrhosis of the liver.

00:59:35.140 --> 00:59:38.170
And so you can use this
kind of data to say,

00:59:38.170 --> 00:59:40.820
well, here's the
immediate cause of death,

00:59:40.820 --> 00:59:43.240
here's sort of the
proximate cause of death,

00:59:43.240 --> 00:59:47.010
and here's the underlying
cause of death.

00:59:47.010 --> 00:59:50.310
And so this is the sort
of statistical data

00:59:50.310 --> 00:59:52.800
that we now have available.

00:59:52.800 --> 00:59:59.352
Now, do any of you watch
the PBS show Victoria?

00:59:59.352 --> 01:00:01.070
Nobody?

01:00:01.070 --> 01:00:02.720
You're not television watchers.

01:00:02.720 --> 01:00:06.260
All right, that's pretty cool.

01:00:06.260 --> 01:00:12.410
I was stunned, because as I
was preparing this lecture

01:00:12.410 --> 01:00:15.500
and I had the next
slide, it turns out

01:00:15.500 --> 01:00:17.840
this plays a role in
one of the episodes

01:00:17.840 --> 01:00:20.150
that was broadcast about
a week and a half ago.

01:00:23.210 --> 01:00:30.050
So in the 1850s, there was a big
outbreak of cholera in London.

01:00:30.050 --> 01:00:34.460
And John Snow was
a doctor who did

01:00:34.460 --> 01:00:38.780
this amazing epidemiological
study to try to figure out

01:00:38.780 --> 01:00:40.820
what was causing cholera.

01:00:40.820 --> 01:00:44.600
The accepted opinion was that
cholera was caused by miasma.

01:00:47.420 --> 01:00:49.652
What's a miasma?

01:00:49.652 --> 01:00:51.110
STUDENT: Bad air.

01:00:51.110 --> 01:00:53.360
PETER SZOLOVITS: Bad air, OK?

01:00:53.360 --> 01:00:54.620
So it's bad air.

01:00:54.620 --> 01:01:00.710
Somehow, bad air was causing
people to get sick and die.

01:01:00.710 --> 01:01:04.040
And several hundred people died.

01:01:04.040 --> 01:01:09.140
Whereas Snow started plotting
on a map of the area in London

01:01:09.140 --> 01:01:13.660
where these were concentrated
where everybody lived.

01:01:13.660 --> 01:01:16.690
And what he discovered,
interestingly enough,

01:01:16.690 --> 01:01:19.660
is that right in the
middle of Broad Street,

01:01:19.660 --> 01:01:23.560
which is pretty much at the
epicenter of all these people

01:01:23.560 --> 01:01:28.530
dying, was a water pump that
everybody in the neighborhood

01:01:28.530 --> 01:01:29.040
used.

01:01:29.040 --> 01:01:35.170
And that water pump, its supply
had become infected by cholera.

01:01:35.170 --> 01:01:37.140
And so people were
pumping water,

01:01:37.140 --> 01:01:40.230
taking it home, drinking
it, and then dying,

01:01:40.230 --> 01:01:42.630
or at least getting very sick.

01:01:42.630 --> 01:01:45.900
And he looked at this
and he said, well,

01:01:45.900 --> 01:01:52.620
if we turn off that pump,
the epidemic will stop.

01:01:52.620 --> 01:01:55.890
And he actually went to
the queen, Queen Victoria--

01:01:55.890 --> 01:01:59.250
hence the tie-in to
the television show--

01:01:59.250 --> 01:02:03.540
and convinced her that this
was worth trying, because they

01:02:03.540 --> 01:02:05.670
didn't have any better ideas.

01:02:05.670 --> 01:02:08.010
And they took the pump
handle off the pump,

01:02:08.010 --> 01:02:12.030
and sure enough the
cholera epidemic abated.

01:02:12.030 --> 01:02:17.320
Now, of course the underlying
problem was sanitation.

01:02:17.320 --> 01:02:18.610
And they didn't fix that.

01:02:18.610 --> 01:02:21.520
That took longer.

01:02:21.520 --> 01:02:24.070
But what's
interesting is here is

01:02:24.070 --> 01:02:30.590
a 2003 study of the
spread of West Nile virus.

01:02:30.590 --> 01:02:35.710
So this is mosquitoes that are
biting people and infecting

01:02:35.710 --> 01:02:38.650
them with this nasty disease.

01:02:38.650 --> 01:02:41.920
And they actually used
very similar techniques

01:02:41.920 --> 01:02:45.070
to figure out that
maybe this was coming in

01:02:45.070 --> 01:02:48.310
on airplanes through JFK.

01:02:48.310 --> 01:02:53.470
So mosquitoes were hitching
a ride on an airplane

01:02:53.470 --> 01:02:55.780
and coming into the US.

01:02:55.780 --> 01:03:01.675
We need to build a border
wall against mosquitoes.

01:03:04.470 --> 01:03:08.120
There is also a very
controversial practice

01:03:08.120 --> 01:03:11.870
that used to be used a lot by
public health officials, which

01:03:11.870 --> 01:03:14.400
was to quarantine people.

01:03:14.400 --> 01:03:18.110
And so there are
lots of examples.

01:03:18.110 --> 01:03:22.200
Anybody's relatives come
through Ellis Island?

01:03:22.200 --> 01:03:23.610
Must be a few.

01:03:23.610 --> 01:03:27.240
OK, so they were
subject to quarantine.

01:03:27.240 --> 01:03:30.480
If you were sick when you
arrived at Ellis Island

01:03:30.480 --> 01:03:33.630
and they didn't know
exactly how sick you were,

01:03:33.630 --> 01:03:36.840
they would put you in a
building and wait a month

01:03:36.840 --> 01:03:39.540
and see if you got better
or if you got worse,

01:03:39.540 --> 01:03:42.870
and then decide whether to
send you back or to let you in.

01:03:42.870 --> 01:03:45.060
So that was a pretty
common practice.

01:03:45.060 --> 01:03:47.940
There's a famous
story of Typhoid Mary,

01:03:47.940 --> 01:03:50.610
who was a carrier
of typhoid fever

01:03:50.610 --> 01:03:53.280
but was not herself affected.

01:03:53.280 --> 01:03:55.460
Unfortunately, she was a cook.

01:03:55.460 --> 01:04:00.210
And wherever she was employed,
people got really sick.

01:04:00.210 --> 01:04:03.690
And eventually, the New
York Department of Health

01:04:03.690 --> 01:04:09.810
forcibly essentially jailed
her in some sanatorium

01:04:09.810 --> 01:04:11.910
in order to keep
her from continuing

01:04:11.910 --> 01:04:15.120
to infect people-- it was a
very controversial case, as you

01:04:15.120 --> 01:04:17.160
might imagine.

01:04:17.160 --> 01:04:18.870
You don't have to
go that far back.

01:04:18.870 --> 01:04:25.170
Here's a 1987 article from
UPI from The Chicago Tribune.

01:04:25.170 --> 01:04:28.200
So Jesse Helms,
who was a senator,

01:04:28.200 --> 01:04:33.550
was calling for everybody who
has AIDS to be quarantined.

01:04:33.550 --> 01:04:36.220
So fortunately,
that didn't happen.

01:04:36.220 --> 01:04:40.780
But the idea is still
around, to say, well,

01:04:40.780 --> 01:04:43.990
we're going to stop this
infection by quarantining

01:04:43.990 --> 01:04:45.280
people.

01:04:45.280 --> 01:04:51.740
And then here's a recent
report about the Ebola response

01:04:51.740 --> 01:04:55.090
in Africa over the
past few years,

01:04:55.090 --> 01:04:59.230
when Ebola was ravaging
parts of that continent.

01:04:59.230 --> 01:05:03.160
And their conclusions
are it's controversial,

01:05:03.160 --> 01:05:07.210
a debated issue, significant
risks related to human rights.

01:05:07.210 --> 01:05:09.970
Quarantine should be
used as a last resort.

01:05:09.970 --> 01:05:13.570
Quarantines in urban
areas are really hard.

01:05:13.570 --> 01:05:15.900
Mobile populations make it hard.

01:05:15.900 --> 01:05:20.920
And this is the most
technical conclusion,

01:05:20.920 --> 01:05:23.320
that if you're going to
quarantine a bunch of people,

01:05:23.320 --> 01:05:26.720
you have a huge waste disposal
problem on your hands.

01:05:26.720 --> 01:05:29.680
Because if you have people
who might have Ebola,

01:05:29.680 --> 01:05:31.900
you can't just
take their garbage

01:05:31.900 --> 01:05:33.730
and throw it out somewhere.

01:05:33.730 --> 01:05:36.700
You have to dispose
of it properly.

01:05:36.700 --> 01:05:39.910
OK, so the last thing I
wanted to talk about--

01:05:39.910 --> 01:05:45.260
hopefully mercifully short--
will be paying for health care.

01:05:45.260 --> 01:05:49.780
So I remember reading
about 20 years ago

01:05:49.780 --> 01:05:53.260
that if you bought a
Chevy from General Motors,

01:05:53.260 --> 01:05:56.740
they spent more money on
health insurance and health

01:05:56.740 --> 01:06:00.580
care for their employees than
they did on the steel that

01:06:00.580 --> 01:06:02.498
went into your car.

01:06:02.498 --> 01:06:03.415
That's pretty amazing.

01:06:06.600 --> 01:06:07.570
So why is this?

01:06:07.570 --> 01:06:10.440
Well, essentially, there's
an insatiable demand

01:06:10.440 --> 01:06:12.780
for health care, right?

01:06:12.780 --> 01:06:17.970
Nobody wants to die, except
for the suicidal people.

01:06:17.970 --> 01:06:22.890
And so if I'm sick, I want
the best care possible,

01:06:22.890 --> 01:06:26.340
and I want as much
of it as possible.

01:06:26.340 --> 01:06:29.940
Because you know, what
is more important in life

01:06:29.940 --> 01:06:33.040
than continuing to live?

01:06:33.040 --> 01:06:39.330
So we also have gotten better at
making drugs and better tests,

01:06:39.330 --> 01:06:40.410
and so on.

01:06:40.410 --> 01:06:43.410
I remember one MRI
machines became

01:06:43.410 --> 01:06:47.040
popular about 30 years ago.

01:06:47.040 --> 01:06:49.120
The state of
Massachusetts, for example,

01:06:49.120 --> 01:06:51.750
had a commission that
you had to convince

01:06:51.750 --> 01:06:57.360
them to be allowed to buy an
MRI machine for your hospital,

01:06:57.360 --> 01:07:00.570
because MRI machines
were very expensive

01:07:00.570 --> 01:07:04.260
and MRIs were, at that
time, hugely expensive.

01:07:04.260 --> 01:07:07.380
And so they wanted to contain
the cost of health care

01:07:07.380 --> 01:07:10.020
by limiting the number
of such machines.

01:07:10.020 --> 01:07:12.070
Well, eventually
the costs came down.

01:07:12.070 --> 01:07:13.600
And so we're doing better.

01:07:13.600 --> 01:07:15.090
But if you read
the newspapers, you

01:07:15.090 --> 01:07:18.550
see that drug therapy
is very expensive.

01:07:18.550 --> 01:07:24.060
We have these wonder drugs
for rare diseases or cancers

01:07:24.060 --> 01:07:29.490
that cost $1 million a year
to pay for your dosage.

01:07:29.490 --> 01:07:33.090
And so there is a
high human motivation

01:07:33.090 --> 01:07:36.450
to do this, and
not much pushback--

01:07:36.450 --> 01:07:38.970
except from insurance companies.

01:07:38.970 --> 01:07:43.410
But they just pass the cost
on in insurance contracts.

01:07:43.410 --> 01:07:45.630
There's also waste.

01:07:45.630 --> 01:07:48.690
So there are lots of
stories about half

01:07:48.690 --> 01:07:51.120
of health care
expenses are spent

01:07:51.120 --> 01:07:54.090
in the last year
of somebody's life.

01:07:54.090 --> 01:07:57.390
Although, Ingelfinger, who was
the editor of The New England

01:07:57.390 --> 01:08:00.820
Journal, gave a talk
here about 25 years ago.

01:08:00.820 --> 01:08:03.420
And he said, you know, when
I was a practicing doctor,

01:08:03.420 --> 01:08:06.210
no patient ever came into
my office saying, doc,

01:08:06.210 --> 01:08:09.120
I'm in the last year of my life.

01:08:09.120 --> 01:08:12.900
And so that was a
difficult criterion

01:08:12.900 --> 01:08:15.660
to try to operationalize.

01:08:15.660 --> 01:08:18.359
There are marginally
useful procedures.

01:08:18.359 --> 01:08:20.609
The IOM estimated
that there are sort

01:08:20.609 --> 01:08:24.750
of 40,000 to 100,000 quote
unquote unnecessary deaths

01:08:24.750 --> 01:08:26.649
per year--

01:08:26.649 --> 01:08:28.410
in other words,
deaths that could

01:08:28.410 --> 01:08:31.140
be avoided by just being
more careful and a little bit

01:08:31.140 --> 01:08:32.490
smarter.

01:08:32.490 --> 01:08:36.899
Well, so the result of this is
that if you look at health care

01:08:36.899 --> 01:08:41.250
spending as a percentage
of gross domestic product

01:08:41.250 --> 01:08:47.069
from 1970 to I think 2017,
I believe, on this graph,

01:08:47.069 --> 01:08:50.279
what you see is
one real outlier.

01:08:50.279 --> 01:08:54.500
And that's the
United States on top.

01:08:54.500 --> 01:08:58.939
So I've selected a few
of these just to look at.

01:08:58.939 --> 01:09:01.000
There's the US on top.

01:09:01.000 --> 01:09:05.109
France, Germany, a lot
of the European countries

01:09:05.109 --> 01:09:11.109
are roughly at that
highest of the crowd level.

01:09:11.109 --> 01:09:12.460
Canada is down there.

01:09:12.460 --> 01:09:14.020
The UK is a little lower.

01:09:14.020 --> 01:09:15.760
Spain is a little lower.

01:09:15.760 --> 01:09:17.140
Israel is a little lower.

01:09:17.140 --> 01:09:20.620
Turkey is the lowest
of the OECD countries

01:09:20.620 --> 01:09:25.000
in terms of percentage of GDP
that they spend on health care.

01:09:25.000 --> 01:09:26.590
So well, that's OK.

01:09:26.590 --> 01:09:30.130
But maybe we're getting
more for our money

01:09:30.130 --> 01:09:32.359
than these other countries.

01:09:32.359 --> 01:09:34.600
And so there are
a lot of analyses

01:09:34.600 --> 01:09:36.160
that look at stuff like this.

01:09:36.160 --> 01:09:39.310
They say, well, if
we spend so much

01:09:39.310 --> 01:09:43.990
money per patient
per year, what do

01:09:43.990 --> 01:09:46.090
we get in terms of
the simplest thing

01:09:46.090 --> 01:09:48.729
to measure, which
is life expectancy?

01:09:48.729 --> 01:09:50.859
How long do people live?

01:09:50.859 --> 01:09:53.950
And what we discover is
that in the United States,

01:09:53.950 --> 01:09:59.440
we're spending $9,000
a year on patients

01:09:59.440 --> 01:10:05.170
and we're getting a life span
of somewhere in the high 70s.

01:10:05.170 --> 01:10:08.290
So this is 2015 data.

01:10:08.290 --> 01:10:11.800
Whereas in Switzerland,
they're spending about $6,000--

01:10:11.800 --> 01:10:17.800
so about 2/3-- and they're
getting 83 years or something

01:10:17.800 --> 01:10:22.150
of life for the
same price, and so

01:10:22.150 --> 01:10:24.610
on for all these
different countries.

01:10:24.610 --> 01:10:27.880
By the way, this is
all from Gapminder,

01:10:27.880 --> 01:10:31.023
which is a wonderful
data visualization tool.

01:10:31.023 --> 01:10:32.440
And I don't have
time to show you,

01:10:32.440 --> 01:10:34.990
but you can click
on individual lines

01:10:34.990 --> 01:10:40.280
there and slide the slider about
what euro you're talking about,

01:10:40.280 --> 01:10:41.800
and the data moves.

01:10:41.800 --> 01:10:44.290
And it's beautiful, and
it's a wonderful way

01:10:44.290 --> 01:10:46.640
of understanding it.

01:10:46.640 --> 01:10:52.230
So there's an important
thing to remember,

01:10:52.230 --> 01:10:54.750
which was taught to
me by my friend Chris

01:10:54.750 --> 01:10:57.420
Dede at the Harvard
Education School.

01:10:57.420 --> 01:11:00.480
And that is that it's not even
good enough to stand still.

01:11:00.480 --> 01:11:04.620
So he suggested the
following scenario.

01:11:04.620 --> 01:11:07.350
If you look at the
growth of productivity

01:11:07.350 --> 01:11:12.240
over this period of
10 years by industry,

01:11:12.240 --> 01:11:14.520
you discover that
productivity went up

01:11:14.520 --> 01:11:18.990
by seven point something
percent for durable goods

01:11:18.990 --> 01:11:23.420
and went down by something
like 2% for mining.

01:11:23.420 --> 01:11:26.240
About 1% for construction.

01:11:26.240 --> 01:11:31.440
Information technologies grew
by 5 and 1/2% or something.

01:11:31.440 --> 01:11:34.430
So if you ask the
question, what happens

01:11:34.430 --> 01:11:37.280
if the demand for
these goods remains

01:11:37.280 --> 01:11:41.030
constant over a period of time?

01:11:41.030 --> 01:11:45.710
And what you discover is that
because the more productive

01:11:45.710 --> 01:11:48.590
things become
cheaper, they wind up

01:11:48.590 --> 01:11:52.610
occupying a smaller fraction
of the total amount of money

01:11:52.610 --> 01:11:55.370
that's being spent.

01:11:55.370 --> 01:11:59.500
So your computer, your
laptop, is a lot cheaper today

01:11:59.500 --> 01:12:02.210
than it was 30 years ago.

01:12:02.210 --> 01:12:04.300
And so that means
that the amount

01:12:04.300 --> 01:12:08.350
of spending that people do
on things like information

01:12:08.350 --> 01:12:13.180
technology, at least
per item, is much lower

01:12:13.180 --> 01:12:14.890
than it used to be.

01:12:14.890 --> 01:12:18.160
And if in the aggregate it's
also lower than it used to be,

01:12:18.160 --> 01:12:22.020
that means that something
else must be higher, right?

01:12:22.020 --> 01:12:24.640
Because it sums to 100%.

01:12:24.640 --> 01:12:27.780
And so what this
shows is that if you

01:12:27.780 --> 01:12:31.830
spend 30 years at the same
rates of productivity growth,

01:12:31.830 --> 01:12:37.080
mining grows from whatever
fraction of the economy

01:12:37.080 --> 01:12:42.650
it was to something that's about
three times as big a fraction.

01:12:42.650 --> 01:12:44.450
Right?

01:12:44.450 --> 01:12:48.290
And if productivity
growth is better,

01:12:48.290 --> 01:12:52.770
then that sector of
the economy shrinks.

01:12:52.770 --> 01:12:55.680
So I think something
like this is

01:12:55.680 --> 01:12:57.930
happening in health
care, which is

01:12:57.930 --> 01:13:00.240
that there's infinite demand.

01:13:00.240 --> 01:13:02.940
And health care is not
terribly productive.

01:13:02.940 --> 01:13:06.210
We're not getting better
as fast as electronics

01:13:06.210 --> 01:13:09.890
is getting better, for example.

01:13:09.890 --> 01:13:13.250
So people have tried
doing various things.

01:13:13.250 --> 01:13:18.038
"Managed care" was the
buzzword of the 1980s and '90s.

01:13:18.038 --> 01:13:19.580
And they said, well,
what we're going

01:13:19.580 --> 01:13:22.490
to do is to prevent
people from overusing

01:13:22.490 --> 01:13:27.410
medical services by requiring
pre-admission review, continued

01:13:27.410 --> 01:13:31.160
stay review, second
surgical opinion.

01:13:31.160 --> 01:13:34.250
We're going to have
post-care management, where

01:13:34.250 --> 01:13:36.140
if you're released
from the hospital

01:13:36.140 --> 01:13:38.840
and you're in that
second of the cycles,

01:13:38.840 --> 01:13:41.960
people will call you at home
and try to help you out and make

01:13:41.960 --> 01:13:45.200
sure that you're doing whatever
is best to keep you out

01:13:45.200 --> 01:13:46.770
of the hospital.

01:13:46.770 --> 01:13:47.270
And

01:13:47.270 --> 01:13:49.880
We're going to try
various experiments,

01:13:49.880 --> 01:13:53.150
like institutional
arrangements, that say, well,

01:13:53.150 --> 01:14:00.620
if I as a doctor agree to refer
all my patients to Mass General

01:14:00.620 --> 01:14:05.360
rather than to the Brigham or
the BI, they'll pay me extra.

01:14:05.360 --> 01:14:10.380
And they'll get some kind of
efficiencies from aggregation.

01:14:10.380 --> 01:14:14.030
And so maybe that's one
way of controlling costs.

01:14:14.030 --> 01:14:17.990
So leakage is the idea of
keeping people in-system.

01:14:17.990 --> 01:14:20.270
And capitation is
an interesting idea,

01:14:20.270 --> 01:14:23.000
which says that
instead of paying

01:14:23.000 --> 01:14:25.760
for what the
hospital does for me

01:14:25.760 --> 01:14:28.400
or what the doctor
does for me, we simply

01:14:28.400 --> 01:14:34.290
pay him a flat fee for the
year to take care of me.

01:14:34.290 --> 01:14:36.480
And that takes away
the incentive for him

01:14:36.480 --> 01:14:39.572
to do more and more and
more to get paid more.

01:14:39.572 --> 01:14:41.280
But of course, it
gives them an incentive

01:14:41.280 --> 01:14:44.940
to do less and less and
less so that he doesn't

01:14:44.940 --> 01:14:46.380
have to spend as much money.

01:14:46.380 --> 01:14:48.450
And so it's sort
of a knife balance

01:14:48.450 --> 01:14:50.250
to figure out how to do this.

01:14:50.250 --> 01:14:53.170
But that's an
important component.

01:14:53.170 --> 01:14:56.130
So if you look at the
evaluation of managed care

01:14:56.130 --> 01:14:59.370
a long time ago,
what they said was it

01:14:59.370 --> 01:15:05.100
helped reduce inpatient costs
by increasing outpatient costs.

01:15:05.100 --> 01:15:07.200
So what it's done is
it's pushed people

01:15:07.200 --> 01:15:11.220
from going to the hospital into
going to their doctor's office.

01:15:11.220 --> 01:15:14.370
But it's been pretty much a wash
in terms of overall spending.

01:15:17.680 --> 01:15:20.320
Doctors also hated managed care.

01:15:20.320 --> 01:15:22.900
I was sitting with
one of my colleagues

01:15:22.900 --> 01:15:26.830
at a Boston hospital, and
an insurance company clerk

01:15:26.830 --> 01:15:32.500
called him to dun him for
having ordered a certain test

01:15:32.500 --> 01:15:34.030
on a heart patient.

01:15:34.030 --> 01:15:36.550
And so he was furious.

01:15:36.550 --> 01:15:39.640
And he turns to her and
says, so which medical school

01:15:39.640 --> 01:15:41.740
do you have your diploma from?

01:15:41.740 --> 01:15:44.340
And of course, she doesn't
have a medical degree.

01:15:44.340 --> 01:15:47.050
She's following some
rules on a sheet

01:15:47.050 --> 01:15:51.730
about how to harass doctors
not to order expensive tests.

01:15:51.730 --> 01:15:55.900
And so we have Edward Annis,
the past president of the AMA,

01:15:55.900 --> 01:15:58.510
who says, well,
in the glory days

01:15:58.510 --> 01:16:01.780
there was no bureaucratic
regimentation, no forms,

01:16:01.780 --> 01:16:03.370
no blah, blah, blah.

01:16:03.370 --> 01:16:05.980
And all my patients were
happy, and I was happy,

01:16:05.980 --> 01:16:08.380
and things were ideal.

01:16:08.380 --> 01:16:10.900
If you actually look
back at those days,

01:16:10.900 --> 01:16:15.170
it wasn't as good as
the cracks it up to be.

01:16:15.170 --> 01:16:17.230
And some of those
issues of disparity

01:16:17.230 --> 01:16:19.720
that we were talking
about were horrible.

01:16:19.720 --> 01:16:22.180
So for his patients
who were rich

01:16:22.180 --> 01:16:26.500
and who could afford to see
him, life was pretty good.

01:16:26.500 --> 01:16:31.840
But not so much for
underserved populations.

01:16:31.840 --> 01:16:37.750
So you've seen ObamaCare come
and partly go, and continues

01:16:37.750 --> 01:16:39.670
to be controversial.

01:16:39.670 --> 01:16:46.450
But it's trying to foster
better information technology

01:16:46.450 --> 01:16:51.080
as a basis for getting doctors
to make better decisions.

01:16:51.080 --> 01:16:54.560
It's trying to foster these
accountable care organizations,

01:16:54.560 --> 01:16:56.800
which is a version
of capitation,

01:16:56.800 --> 01:17:00.790
to put the pressure on to reduce
the amount of health services

01:17:00.790 --> 01:17:03.400
that people are asking for.

01:17:03.400 --> 01:17:07.960
There is a hospital readmission
reduction program now

01:17:07.960 --> 01:17:14.050
that says if you are a
Medicare patient, for example,

01:17:14.050 --> 01:17:16.210
and you're discharged
from the hospital

01:17:16.210 --> 01:17:21.090
and you're readmitted within
30 days of your discharge,

01:17:21.090 --> 01:17:24.120
then they're going to
dun you and not pay you

01:17:24.120 --> 01:17:25.920
for that readmission,
or not pay you

01:17:25.920 --> 01:17:27.870
for part of that readmission.

01:17:27.870 --> 01:17:31.530
But if you look at the
statistics, which I just did,

01:17:31.530 --> 01:17:35.400
that's the distribution of the
payment adjustment factors.

01:17:35.400 --> 01:17:39.066
So it turns out the
lowest number is 97%.

01:17:39.066 --> 01:17:41.760
So a 3% decrease
in reimbursement

01:17:41.760 --> 01:17:44.610
is something that the
CFO of your hospital

01:17:44.610 --> 01:17:45.940
would really care about.

01:17:45.940 --> 01:17:50.310
But it's not like a 25%
reduction in reimbursements.

01:17:50.310 --> 01:17:53.235
So this has had a
fairly minor effect.

01:17:56.780 --> 01:18:01.790
Let me just finish by saying
that money determines much.

01:18:01.790 --> 01:18:05.210
From our point of view,
one of the problems we face

01:18:05.210 --> 01:18:09.410
is that IT traditionally
gets about 1% to 2%

01:18:09.410 --> 01:18:14.720
of spending in medical centers,
whereas it gets about 6% or 7%

01:18:14.720 --> 01:18:19.850
in business overall and
about 10% to 12% for banking.

01:18:19.850 --> 01:18:24.230
And a lot of these systems
are managed by accountants,

01:18:24.230 --> 01:18:26.390
although that is
slowly changing.

01:18:26.390 --> 01:18:31.010
So in the 1990s,
HST with Harvard

01:18:31.010 --> 01:18:35.360
started a training program
for medical doctors

01:18:35.360 --> 01:18:39.260
to become medical
informaticians-- so practicing

01:18:39.260 --> 01:18:40.970
this sort of witchcraft.

01:18:40.970 --> 01:18:43.820
And our first two
graduates, one of them

01:18:43.820 --> 01:18:47.300
is now CIO at the
Beth Israel Deaconess.

01:18:47.300 --> 01:18:50.330
The second one is CIO
at Children's Hospital.

01:18:50.330 --> 01:18:53.540
And so one of my big
successes personally

01:18:53.540 --> 01:18:56.210
has been to displace
some accountants

01:18:56.210 --> 01:18:58.610
by doctors who
actually understand

01:18:58.610 --> 01:19:01.580
this technology to some extent.

01:19:01.580 --> 01:19:05.130
OK, I think that's
all I'm going to say.

01:19:05.130 --> 01:19:08.780
There's a funny last
slide here, which

01:19:08.780 --> 01:19:12.290
has a pointer that I would
want you to remember.

01:19:12.290 --> 01:19:15.530
The slides will be
up on our website.

01:19:15.530 --> 01:19:17.330
You can follow it.

01:19:17.330 --> 01:19:19.460
MIT has a program called GEMS.

01:19:19.460 --> 01:19:23.840
It's the General Education
in Medical Sciences, intended

01:19:23.840 --> 01:19:27.170
as a minor program for
people in the PhD program

01:19:27.170 --> 01:19:29.090
in some other field.

01:19:29.090 --> 01:19:33.650
And if you're serious about
really concentrating on health

01:19:33.650 --> 01:19:37.640
care, developing at least
the kind of understanding

01:19:37.640 --> 01:19:41.250
of the health care process that
I've tried to give you today

01:19:41.250 --> 01:19:44.690
and that would allow you to
play a doctor on television

01:19:44.690 --> 01:19:46.340
is really important.

01:19:46.340 --> 01:19:49.550
And there is a program that
helps you achieve that,

01:19:49.550 --> 01:19:51.170
which I commend to people.

01:19:51.170 --> 01:19:54.130
OK, see you next week.