AI in health care initiatives
See how to make artificial intelligence work for providers.
WEBVTT
1
00:00:00.270 --> 00:00:03.590 size:90%
What is the first piece of advice that you would
2
00:00:03.590 --> 00:00:06.680 size:90%
give for leaders and provider organizations
3
00:00:06.680 --> 00:00:10.250 size:90%
looking to bring AI technologies into their into
4
00:00:10.250 --> 00:00:14.090 size:90%
their stack? Absolutely so, if you think about AI
5
00:00:14.090 --> 00:00:16.640 size:90%
this is fundamentally a business problem you're
6
00:00:16.640 --> 00:00:18.830 size:90%
looking to solve. So if you're, a health system or
7
00:00:18.830 --> 00:00:21.680 size:90%
a provider, this is about how do you use your
8
00:00:21.680 --> 00:00:25.130 size:90%
business talent to solve an analytics problem?
9
00:00:25.130 --> 00:00:27.350 size:90%
This isn't necessarily a technology problem that
10
00:00:27.350 --> 00:00:28.550 size:90%
you start working.
11
00:00:28.550 --> 00:00:31.940 size:90%
So my advice and my suggestion is focus on what
12
00:00:31.940 --> 00:00:34.370 size:90%
you want to do in terms of the use case. What's
13
00:00:34.370 --> 00:00:35.840 size:90%
the use case you wanna solve whether it's
14
00:00:35.840 --> 00:00:38.210 size:90%
operational or clinical
15
00:00:38.210 --> 00:00:40.070 size:90%
and determine the talent that you need to
16
00:00:40.070 --> 00:00:42.650 size:90%
actually go solve it that technology will be the
17
00:00:42.650 --> 00:00:44.600 size:90%
underlying enabler of that.
18
00:00:44.600 --> 00:00:47.660 size:90%
In fact, we launched our inaugural Optum IQ
19
00:00:47.660 --> 00:00:50.660 size:90%
survey in on artificial intelligence in health
20
00:00:50.660 --> 00:00:53.330 size:90%
care in 2018 and the results were
21
00:00:53.330 --> 00:00:56.240 size:90%
fascinating. 91 percent of those that were
22
00:00:56.240 --> 00:00:57.650 size:90%
interviewed
23
00:00:57.650 --> 00:01:01.070 size:90%
identified AI is the right path to improvement in
24
00:01:01.070 --> 00:01:02.480 size:90%
healthcare.
25
00:01:02.480 --> 00:01:06.500 size:90%
75 percent actually are embarking on a strategy
26
00:01:06.500 --> 00:01:09.020 size:90%
to embed AI into their organization.
27
00:01:09.020 --> 00:01:09.830 size:90%
And if you look at it
28
00:01:09.830 --> 00:01:13.460 size:90%
by organization over the next five years,
29
00:01:13.460 --> 00:01:15.170 size:90%
what we heard was that there's gonna be over
30
00:01:15.170 --> 00:01:17.390 size:90%
30 million dollars per organization that's got
31
00:01:17.390 --> 00:01:20.420 size:90%
to be invested in the AI space.
32
00:01:20.420 --> 00:01:22.940 size:90%
Interesting. Can you tell us a little bit more
33
00:01:22.940 --> 00:01:26.630 size:90%
about Optum's view of practical application of AI
34
00:01:26.630 --> 00:01:30.140 size:90%
versus kind of general strategy for innovation
35
00:01:30.140 --> 00:01:32.720 size:90%
and approaching other projects like what makes an
36
00:01:32.720 --> 00:01:34.910 size:90%
AI initiative unique? What are the unique
37
00:01:34.910 --> 00:01:37.760 size:90%
factors there? Yeah so if you look at the
38
00:01:37.760 --> 00:01:39.860 size:90%
key components of AI that you need to have in
39
00:01:39.860 --> 00:01:42.560 size:90%
place so there's three dimension of that one is
40
00:01:42.560 --> 00:01:45.400 size:90%
your data right so Machine learning and
41
00:01:45.400 --> 00:01:47.920 size:90%
artificial intelligence requires large, large
42
00:01:47.920 --> 00:01:50.860 size:90%
data sets. So it's access to internal and external
43
00:01:50.860 --> 00:01:53.710 size:90%
data sets that you are able to stitch together,
44
00:01:53.710 --> 00:01:57.250 size:90%
clean, enrich, and be able to do things on. Their
45
00:01:57.250 --> 00:01:59.920 size:90%
artificial intelligence models are only going to
46
00:01:59.920 --> 00:02:02.830 size:90%
be as powerful as the data that you have actually
47
00:02:02.830 --> 00:02:05.680 size:90%
building them around. That's number one. Number
48
00:02:05.680 --> 00:02:08.860 size:90%
two is around the methodologies and the algorithms
49
00:02:08.860 --> 00:02:11.560 size:90%
that you actually build in your organization that
50
00:02:11.560 --> 00:02:14.500 size:90%
you're able to actually apply on that data so
51
00:02:14.500 --> 00:02:16.930 size:90%
whether these are basic regression models or deep
52
00:02:16.930 --> 00:02:18.160 size:90%
learning models.
53
00:02:18.160 --> 00:02:21.400 size:90%
Having the talent to be able to better understand
54
00:02:21.400 --> 00:02:24.370 size:90%
what you can go do with that data is one of the
55
00:02:24.370 --> 00:02:27.100 size:90%
first and most important pieces that we adopt in
56
00:02:27.100 --> 00:02:29.380 size:90%
what say you start with. And then third, which
57
00:02:29.380 --> 00:02:31.360 size:90%
would be quite obvious, is around healthcare
58
00:02:31.360 --> 00:02:32.500 size:90%
experience.
59
00:02:32.500 --> 00:02:35.380 size:90%
Oftentimes you get a data scientist that is very
60
00:02:35.380 --> 00:02:37.450 size:90%
deep in the data but may not have the health
61
00:02:37.450 --> 00:02:40.630 size:90%
care background. Keeping very focused in terms of
62
00:02:40.630 --> 00:02:42.910 size:90%
understanding what the data scientists should be
63
00:02:42.910 --> 00:02:45.790 size:90%
focused on and linking it to a business or
64
00:02:45.790 --> 00:02:49.730 size:90%
clinical problem. This is important.
65
00:02:49.730 --> 00:02:53.180 size:90%
Describe a little bit how you would work with
66
00:02:53.180 --> 00:02:57.110 size:90%
senior leadership to build a team to
67
00:02:57.110 --> 00:03:00.110 size:90%
implement an AI project? What kind of talent are
68
00:03:00.110 --> 00:03:02.900 size:90%
you looking for? Are you looking for talent
69
00:03:02.900 --> 00:03:05.630 size:90%
inside the organization? Are you going need to recruit as
70
00:03:05.630 --> 00:03:09.410 size:90%
a combination of both? Yeah talent is a top of
71
00:03:09.410 --> 00:03:12.350 size:90%
mind question for most organizations and this
72
00:03:12.350 --> 00:03:15.620 size:90%
isn't just about how do you hire data scientists
73
00:03:15.620 --> 00:03:18.050 size:90%
this is about the full breadth of analytics
74
00:03:18.050 --> 00:03:20.120 size:90%
talent you need to embark on a broad AI
75
00:03:20.120 --> 00:03:23.330 size:90%
transformation. So this starts at the top getting
76
00:03:23.330 --> 00:03:24.300 size:90%
executive alignment
77
00:03:24.300 --> 00:03:26.880 size:90%
in terms of what you wanna do with AI, hiring
78
00:03:26.880 --> 00:03:30.540 size:90%
ahead of analytics, that oversees your AI program
79
00:03:30.540 --> 00:03:32.640 size:90%
and then there's multiple dimensions. There's the
80
00:03:32.640 --> 00:03:35.880 size:90%
data scientist who's in who's deep in the data
81
00:03:35.880 --> 00:03:38.220 size:90%
there's a translator that actually is able to
82
00:03:38.220 --> 00:03:40.170 size:90%
understand the business problem and translate it
83
00:03:40.170 --> 00:03:43.770 size:90%
for the data scientist, their solution architects,
84
00:03:43.770 --> 00:03:46.500 size:90%
there's data architects so there's a plethora
85
00:03:46.500 --> 00:03:49.590 size:90%
of talent that you need in your organization to do this
86
00:03:49.590 --> 00:03:50.880 size:90%
successfully.
87
00:03:50.880 --> 00:03:54.240 size:90%
This is something that you don't solve overnight.
88
00:03:54.240 --> 00:03:55.420 size:90%
Laying out a very
89
00:03:55.420 --> 00:03:58.090 size:90%
talent plan in terms of what you want to be in
90
00:03:58.090 --> 00:04:00.820 size:90%
the organization, where you focus and whether you
91
00:04:00.820 --> 00:04:04.510 size:90%
hire or acquire this talent via partnership is
92
00:04:04.510 --> 00:04:06.130 size:90%
actually very important to get started with.
93
00:04:06.130 --> 00:04:08.590 size:90%
What kind of timeline should leaders be thinking
94
00:04:08.590 --> 00:04:10.720 size:90%
about when they're planning this kind of you
95
00:04:10.720 --> 00:04:13.360 size:90%
know, work force development as it were well with
96
00:04:13.360 --> 00:04:16.120 size:90%
artificial intelligence? So what I typically
97
00:04:16.120 --> 00:04:20.020 size:90%
advise is this isn't something that you should
98
00:04:20.020 --> 00:04:22.360 size:90%
look to solve overnight. Have a journey, have a
99
00:04:22.360 --> 00:04:25.690 size:90%
plan of multi multi year road map of what you
100
00:04:25.690 --> 00:04:28.100 size:90%
wanna go do. Having said that
101
00:04:28.100 --> 00:04:29.870 size:90%
the best way to succeed with us is through
102
00:04:29.870 --> 00:04:33.710 size:90%
experimentation trial and pilots, so I'm fully
103
00:04:33.710 --> 00:04:36.200 size:90%
supportive of hiring targeted or partnering with
104
00:04:36.200 --> 00:04:39.920 size:90%
companies to target focus use cases in the near
105
00:04:39.920 --> 00:04:43.490 size:90%
term, build on the success of that, show the or why of that
106
00:04:43.490 --> 00:04:45.890 size:90%
and then pursue your broader talent plan
107
00:04:45.890 --> 00:04:49.850 size:90%
over a multi multi month journey. Great let's
108
00:04:49.850 --> 00:04:52.280 size:90%
pivot a little bit to use cases right we're
109
00:04:52.280 --> 00:04:54.260 size:90%
hearing a lot about artificial intelligence and
110
00:04:54.260 --> 00:04:57.890 size:90%
data here at HIMSS'19. What are some practical use
111
00:04:57.890 --> 00:04:59.840 size:90%
cases for artificial intelligence that are
112
00:04:59.840 --> 00:05:02.360 size:90%
beginning to emerge right? Um what are some
113
00:05:02.360 --> 00:05:04.780 size:90%
things maybe that their viewers here on
HIMSS TV
114
00:05:04.780 --> 00:05:07.840 size:90%
haven't thought of yet? Yeah so it's a great
115
00:05:07.840 --> 00:05:10.840 size:90%
question and one of the one of the insights we
116
00:05:10.840 --> 00:05:13.900 size:90%
learned from our Optum IQ survey last year was
117
00:05:13.900 --> 00:05:16.270 size:90%
over 40 percent of those surveyed believe
118
00:05:16.270 --> 00:05:18.760 size:90%
starting with administrative or operational use
119
00:05:18.760 --> 00:05:22.600 size:90%
cases is the best way to get started and so that
120
00:05:22.600 --> 00:05:25.840 size:90%
includes how do you automate processes and tackle
121
00:05:25.840 --> 00:05:27.580 size:90%
the low hanging fruit in your organization of
122
00:05:27.580 --> 00:05:30.310 size:90%
where you have an efficiency.
123
00:05:30.310 --> 00:05:32.020 size:90%
Why is that important? If you think about
124
00:05:32.020 --> 00:05:34.420 size:90%
technology around machine learning and deep
125
00:05:34.420 --> 00:05:36.850 size:90%
learning, that's very much linked to how you can
126
00:05:36.850 --> 00:05:38.560 size:90%
automate processes and how you use that
127
00:05:38.560 --> 00:05:41.330 size:90%
technology to drive value. It's not the only
128
00:05:41.330 --> 00:05:45.440 size:90%
use case so if you think about clinical outcomes and
129
00:05:45.440 --> 00:05:47.840 size:90%
better being able to predict disease, that's
130
00:05:47.840 --> 00:05:50.780 size:90%
another big use case for artificial intelligence
131
00:05:50.780 --> 00:05:53.690 size:90%
and then there's a whole extra exposure around
132
00:05:53.690 --> 00:05:55.490 size:90%
customer experience which is also critically
133
00:05:55.490 --> 00:05:56.330 size:90%
important.
134
00:05:56.330 --> 00:05:59.780 size:90%
If you're a health plan, one of the areas that
135
00:05:59.780 --> 00:06:01.580 size:90%
health plans can focus on using artificial
136
00:06:01.580 --> 00:06:04.910 size:90%
intelligence is how do you better detect claims
137
00:06:04.910 --> 00:06:08.270 size:90%
that will be denied and automate that process? If
138
00:06:08.270 --> 00:06:11.360 size:90%
you're a health system, how do you automate the physician
139
00:06:11.360 --> 00:06:12.590 size:90%
work flows
140
00:06:12.590 --> 00:06:14.480 size:90%
and potentially lower the number of prior
141
00:06:14.480 --> 00:06:16.430 size:90%
authorization is required and if you're a
142
00:06:16.430 --> 00:06:19.130 size:90%
consumer, how do you actually better at here to
143
00:06:19.130 --> 00:06:21.170 size:90%
medication and how do you automate the process of
144
00:06:21.170 --> 00:06:22.910 size:90%
notifying doctors that you need to be treated?
145
00:06:23.960 --> 00:06:27.290 size:90%
Got it I mean these are emerging use cases right
146
00:06:27.290 --> 00:06:32.720 size:90%
so why is adopting AI now versus 24 months from
147
00:06:32.720 --> 00:06:34.850 size:90%
now or even 18 months now? Why is adopting
148
00:06:34.850 --> 00:06:37.340 size:90%
now or beginning now so important for so many
149
00:06:37.340 --> 00:06:39.680 size:90%
different types of provider organizations?Yeah
150
00:06:39.680 --> 00:06:42.140 size:90%
It's a great question and we're
151
00:06:42.140 --> 00:06:44.540 size:90%
at a turning point in the industry in terms of
152
00:06:44.540 --> 00:06:48.360 size:90%
how we use data and analytics. Other industries
153
00:06:48.360 --> 00:06:52.710 size:90%
such as financial services, retail and tech have
154
00:06:52.710 --> 00:06:54.570 size:90%
already matured in terms of how they use
155
00:06:54.570 --> 00:06:56.640 size:90%
artificial intelligence.
156
00:06:56.640 --> 00:06:59.910 size:90%
It's no longer a nice to have for a provider to
157
00:06:59.910 --> 00:07:03.540 size:90%
be able to have artificial intelligence as a core
158
00:07:03.540 --> 00:07:06.210 size:90%
capability. This is a core competency they're
159
00:07:06.210 --> 00:07:08.580 size:90%
gonna need as they move forward and actually
160
00:07:08.580 --> 00:07:11.100 size:90%
build their analytics strategy and drive value
161
00:07:11.100 --> 00:07:15.260 size:90%
and take better care of patients.
Why providers need AI as a core capability
Tushar Mehrotra is senior vice president of analytics at Optum. He sat down with HIMSSTV to discuss artificial intelligence initiatives in health care and AI talent in care provider organizations. The bottom line? Focus on what you want to do in terms of a use case. Watch to learn more.
Learn more about our expert, Tushar Mehrotra.