1
00:00:09,490 --> 00:00:13,070
This is a story of D2Hawkeye,
a medical data mining

2
00:00:13,070 --> 00:00:18,720
company located in
Waltham, Massachusetts.

3
00:00:18,720 --> 00:00:23,790
The company was founded by Chris
Kryder, a medical doctor who

4
00:00:23,790 --> 00:00:26,940
was an MIT physician
in the 1990s.

5
00:00:26,940 --> 00:00:29,820
The company was founded in 2001.

6
00:00:29,820 --> 00:00:35,530
I have been in the board
of this company since 2002.

7
00:00:35,530 --> 00:00:38,730
The company combines
expert knowledge

8
00:00:38,730 --> 00:00:43,110
and databases with analytics
to improve quality and course

9
00:00:43,110 --> 00:00:46,120
management in health care.

10
00:00:46,120 --> 00:00:48,670
The company was located
in Waltham, Massachusetts.

11
00:00:48,670 --> 00:00:55,950
It grew very fast and was sold
to Verisk Analytics in 2009.

12
00:00:55,950 --> 00:01:01,240
The overall process that
D2Hawkeye uses is as follows.

13
00:01:01,240 --> 00:01:03,700
It starts with
medical claims that

14
00:01:03,700 --> 00:01:11,630
consist of diagnoses,
procedures, and drugs.

15
00:01:11,630 --> 00:01:16,350
These medical claims
are then processed

16
00:01:16,350 --> 00:01:22,150
via process of aggregation,
cleaning, and normalization.

17
00:01:22,150 --> 00:01:28,340
This data then enters
secure databases

18
00:01:28,340 --> 00:01:32,450
on which predictive
models are applied.

19
00:01:32,450 --> 00:01:36,600
The output of predictive
models are specific reports

20
00:01:36,600 --> 00:01:40,940
that give insight to
the various questions

21
00:01:40,940 --> 00:01:43,770
that D2Hawkeye
aspires to answer.

22
00:01:49,789 --> 00:01:54,330
The company tries to improve
health care case management.

23
00:01:54,330 --> 00:02:00,440
Specifically, it tries to
identify high-risk patients,

24
00:02:00,440 --> 00:02:04,410
work with patients to manage
treatment and associated costs,

25
00:02:04,410 --> 00:02:06,180
and arrange specialist care.

26
00:02:09,320 --> 00:02:12,780
Medical costs, of course,
is a serious matter

27
00:02:12,780 --> 00:02:17,300
both for the patient as
well as the provider.

28
00:02:17,300 --> 00:02:21,180
Being able to predict this
cost is an important problem

29
00:02:21,180 --> 00:02:26,170
that interests both the patients
as well as the providers.

30
00:02:26,170 --> 00:02:30,520
The overall goal of
D2Hawkeye is to improve

31
00:02:30,520 --> 00:02:35,150
the quality of cost predictions.

32
00:02:35,150 --> 00:02:39,230
D2Hawkeye had many
different types of clients.

33
00:02:39,230 --> 00:02:42,400
The most important were
third party administrators

34
00:02:42,400 --> 00:02:43,880
of medical claims.

35
00:02:43,880 --> 00:02:46,000
Third party administrators
are companies

36
00:02:46,000 --> 00:02:51,240
hired by the employer who manage
the claims of the employees.

37
00:02:51,240 --> 00:02:54,320
Now the type of clients were
case management companies,

38
00:02:54,320 --> 00:02:57,120
benefits consultants,
and health plans.

39
00:02:57,120 --> 00:03:03,450
The company grew and by 2009,
it analyzed monthly millions

40
00:03:03,450 --> 00:03:09,040
of people through the
analytic platform it built.

41
00:03:09,040 --> 00:03:11,830
This corresponded to
thousands of employers

42
00:03:11,830 --> 00:03:13,050
that were processed monthly.

43
00:03:18,630 --> 00:03:21,860
To analyze the data,
the company used

44
00:03:21,860 --> 00:03:25,440
what we call a
pre-analytics approach.

45
00:03:25,440 --> 00:03:29,200
This was based on the human
judgment of physicians

46
00:03:29,200 --> 00:03:33,670
who manually analyze patient
histories and developed

47
00:03:33,670 --> 00:03:36,620
medical rules.

48
00:03:36,620 --> 00:03:41,970
Of course, this
involved human judgment,

49
00:03:41,970 --> 00:03:44,850
utilized a limited
set of data, it

50
00:03:44,850 --> 00:03:49,400
was often costly, and
somewhat inefficient.

51
00:03:49,400 --> 00:03:51,840
The key question we
analyze in this lecture

52
00:03:51,840 --> 00:03:54,260
is "Can we use
analytics instead?"