1
00:00:04,500 --> 00:00:08,950
Let us see what we learn about
the patterns that emerge.

2
00:00:08,950 --> 00:00:12,180
We will show that the
clusters are interpretable

3
00:00:12,180 --> 00:00:14,990
and reveal unique patterns
of diagnostic history

4
00:00:14,990 --> 00:00:16,670
among the population.

5
00:00:16,670 --> 00:00:22,350
We selected six patterns to
present in this lecture--

6
00:00:22,350 --> 00:00:25,860
Cluster 1, 6, and
7, in Cost Bucket 2,

7
00:00:25,860 --> 00:00:28,930
and Clusters 4, 5, and
10, in Cost Bucket 3.

8
00:00:32,330 --> 00:00:35,310
The first pattern shows the
occurrence of chest pain

9
00:00:35,310 --> 00:00:37,890
three months before
the heart attack.

10
00:00:37,890 --> 00:00:42,420
Note that the red dots depict
the visits per diagnosis

11
00:00:42,420 --> 00:00:44,720
for patients in Cluster
1-- this is, we think,

12
00:00:44,720 --> 00:00:48,350
Bucket 2-- and the blue
dots depict the visits

13
00:00:48,350 --> 00:00:52,260
per diagnosis for patients
in Bucket 2 throughout.

14
00:00:52,260 --> 00:00:54,430
Note the very
significant increase

15
00:00:54,430 --> 00:00:56,850
for visits related
to chest pains

16
00:00:56,850 --> 00:00:59,290
three months before the event.

17
00:00:59,290 --> 00:01:04,910
About 17, three months
before for the red patients,

18
00:01:04,910 --> 00:01:08,460
and about 1 and 1/2 visits
for the blue patients.

19
00:01:12,710 --> 00:01:15,650
The next pattern reveals
an increasing occurrence

20
00:01:15,650 --> 00:01:21,570
of chronic obstructive pulmonary
disease, COPD, for short.

21
00:01:21,570 --> 00:01:24,240
Patients from
Cluster 7 in Bucket 2

22
00:01:24,240 --> 00:01:27,640
have regular doctor
visits for COPD.

23
00:01:27,640 --> 00:01:33,030
Note that nine months
before, we have

24
00:01:33,030 --> 00:01:38,160
4 and 1/2 visits
versus 0.5 visits.

25
00:01:38,160 --> 00:01:43,110
Six months before, we have
almost 7 visits versus 1/2

26
00:01:43,110 --> 00:01:45,680
a visit, and three
months before, we

27
00:01:45,680 --> 00:01:50,509
have 9 visits versus
1/2 a visit for COPD,

28
00:01:50,509 --> 00:01:54,190
so a clear increasing pattern.

29
00:01:54,190 --> 00:01:57,910
The next pattern shows
gradually increasing occurrence

30
00:01:57,910 --> 00:01:59,350
of anemia.

31
00:01:59,350 --> 00:02:02,500
The red line shows the
patient's in Cluster 4

32
00:02:02,500 --> 00:02:06,530
increasingly visit the doctor
for anemia from nine months

33
00:02:06,530 --> 00:02:08,389
on before the event.

34
00:02:08,389 --> 00:02:11,600
Nine months before, members
have an average of 9 visits

35
00:02:11,600 --> 00:02:12,740
to the doctor for anemia.

36
00:02:17,430 --> 00:02:23,620
This increases to an average
of 11 visits six months

37
00:02:23,620 --> 00:02:25,970
before the event, and
then an average of 15

38
00:02:25,970 --> 00:02:29,600
visits three months before
the event, a clear increasing

39
00:02:29,600 --> 00:02:30,100
pattern.

40
00:02:36,430 --> 00:02:39,370
The final pattern shows
the occurrence of diabetes

41
00:02:39,370 --> 00:02:41,210
as a pattern for heart attacks.

42
00:02:41,210 --> 00:02:45,320
It is well known that
both types 1 and 2

43
00:02:45,320 --> 00:02:49,170
diabetes are associated with
accelerated atherosclerosis,

44
00:02:49,170 --> 00:02:53,250
one of the main causes of
myocardial infarction-- heart

45
00:02:53,250 --> 00:02:54,800
attacks, that is.

46
00:02:54,800 --> 00:02:57,180
Well known diagnoses
associated with heart attacks,

47
00:02:57,180 --> 00:03:00,430
such as diabetes, hypertension,
and hyperlipidemia,

48
00:03:00,430 --> 00:03:03,700
characterize many of the
patterns of the consistency

49
00:03:03,700 --> 00:03:08,730
of care throughout all of the
cost buckets and clustering

50
00:03:08,730 --> 00:03:09,230
models.

51
00:03:13,630 --> 00:03:15,970
You observe a difference,
here, of the number

52
00:03:15,970 --> 00:03:19,480
of visits for diabetes
for the population

53
00:03:19,480 --> 00:03:23,090
that had the event versus
the other population.