1
00:00:04,490 --> 00:00:08,109
Our last example involves
a company called D2Hawkeye,

2
00:00:08,109 --> 00:00:12,430
a medical software
company founded in 2001.

3
00:00:12,430 --> 00:00:15,420
The company combined
data with analytics

4
00:00:15,420 --> 00:00:18,800
to improve quality and cost
management in health care.

5
00:00:18,800 --> 00:00:23,570
In 2009, the company was
analyzing 20 million people

6
00:00:23,570 --> 00:00:25,690
monthly.

7
00:00:25,690 --> 00:00:29,310
It is impossible for humans to
sift through patient records

8
00:00:29,310 --> 00:00:32,210
and assess quality and
cost without algorithms.

9
00:00:35,060 --> 00:00:40,590
This motivated the approach
that D2Hawkeye took.

10
00:00:40,590 --> 00:00:44,740
Let us discuss the data
that D2Hawkeye utilized.

11
00:00:44,740 --> 00:00:47,100
Health care industry
is data rich,

12
00:00:47,100 --> 00:00:49,530
but data may be hard to assess.

13
00:00:49,530 --> 00:00:52,320
It is often unstructured
and unavailable.

14
00:00:52,320 --> 00:00:55,790
The company used insurance
data regarding procedures,

15
00:00:55,790 --> 00:00:57,790
prescriptions, and diagnosis.

16
00:00:57,790 --> 00:01:00,470
It further defined
new risk factors

17
00:01:00,470 --> 00:01:02,030
based on doctor's insights.

18
00:01:02,030 --> 00:01:05,349
For example, obesity
and depression.

19
00:01:05,349 --> 00:01:07,870
Finally, it used
demographic information,

20
00:01:07,870 --> 00:01:09,330
like gender and age.

21
00:01:11,940 --> 00:01:13,990
What were the analytics used?

22
00:01:13,990 --> 00:01:17,280
The goal was to predict
future heath care costs,

23
00:01:17,280 --> 00:01:19,700
and identify high-risk
patients to be

24
00:01:19,700 --> 00:01:21,940
prioritized for intervention.

25
00:01:21,940 --> 00:01:24,510
The company created
interpretable models

26
00:01:24,510 --> 00:01:26,780
for doctors to
analyze and verify.

27
00:01:26,780 --> 00:01:29,260
This led to significant
improvements

28
00:01:29,260 --> 00:01:32,259
over just using
historical costs.

29
00:01:35,000 --> 00:01:36,110
So what is the edge?

30
00:01:38,960 --> 00:01:42,289
The analytics used led to
substantial improvement

31
00:01:42,289 --> 00:01:44,350
in D2Hawkeye's
ability to identify

32
00:01:44,350 --> 00:01:47,300
patients who need
more attention.

33
00:01:47,300 --> 00:01:49,900
The approach allowed
expert knowledge

34
00:01:49,900 --> 00:01:54,830
to identify new variables and
refine existing variables.

35
00:01:54,830 --> 00:01:57,630
Further, it allowed the
ability to make predictions

36
00:01:57,630 --> 00:01:59,729
for millions of patients
without manually

37
00:01:59,729 --> 00:02:02,480
reading patient's files.