1
00:00:04,500 --> 00:00:06,110
What is the impact
of clustering?

2
00:00:06,110 --> 00:00:09,100
Clustering members
within each cost bucket

3
00:00:09,100 --> 00:00:13,080
yielded better predictions for
heart attacks within clusters.

4
00:00:13,080 --> 00:00:14,670
Grouping patients
in clusters exhibits

5
00:00:14,670 --> 00:00:17,610
temporal diagnostic
patterns within nine months

6
00:00:17,610 --> 00:00:18,840
of a heart attack.

7
00:00:18,840 --> 00:00:20,450
These patterns can
be incorporated

8
00:00:20,450 --> 00:00:22,920
in the diagnostic rules
for heart attacks.

9
00:00:22,920 --> 00:00:26,690
The approach shows that using
analytics for early heart

10
00:00:26,690 --> 00:00:29,500
failure detection through
pattern recognition

11
00:00:29,500 --> 00:00:31,240
can lead to interesting
new insights.

12
00:00:33,820 --> 00:00:39,230
The findings here are reinforced
by results from our research.

13
00:00:39,230 --> 00:00:44,650
IBM, Sutter Health, and
Geisinger Health Systems

14
00:00:44,650 --> 00:00:48,490
partnered in 2009 to
research analytics tools

15
00:00:48,490 --> 00:00:50,790
in view of early detection.

16
00:00:50,790 --> 00:00:54,110
Steve Steinhubl, a
cardiologist from Geisinger,

17
00:00:54,110 --> 00:00:56,670
wrote, "our early
research showed

18
00:00:56,670 --> 00:01:00,240
the signs and symptoms of
heart failure in patients

19
00:01:00,240 --> 00:01:03,670
are often documented
years before diagnosis.

20
00:01:03,670 --> 00:01:08,289
The pattern of documentation can
offer clinically useful signals

21
00:01:08,289 --> 00:01:11,550
for early detection of
this deadly disease."