1
00:00:04,500 --> 00:00:07,710
In this lecture, we'll be
using analytical models

2
00:00:07,710 --> 00:00:09,900
to prevent heart disease.

3
00:00:09,900 --> 00:00:13,050
The first step is to
identify risk factors,

4
00:00:13,050 --> 00:00:17,350
or the independent variables,
that we will use in our model.

5
00:00:17,350 --> 00:00:21,490
Then, using data, we'll create
a logistic regression model

6
00:00:21,490 --> 00:00:23,800
to predict heart disease.

7
00:00:23,800 --> 00:00:26,570
Using more data, we'll
validate our model

8
00:00:26,570 --> 00:00:29,520
to make sure it performs
well out of sample

9
00:00:29,520 --> 00:00:32,340
and on different populations
than the training set

10
00:00:32,340 --> 00:00:34,310
population.

11
00:00:34,310 --> 00:00:37,260
Lastly, we'll discuss
how medical interventions

12
00:00:37,260 --> 00:00:39,350
can be defined using the model.

13
00:00:41,930 --> 00:00:44,040
We'll be predicting
the 10-year risk

14
00:00:44,040 --> 00:00:47,670
of coronary heart
disease or CHD.

15
00:00:47,670 --> 00:00:51,600
This was the subject of
an important 1998 paper

16
00:00:51,600 --> 00:00:55,420
introducing what is known as
the Framingham Risk Score.

17
00:00:55,420 --> 00:00:58,210
This is one of the most
influential applications

18
00:00:58,210 --> 00:01:00,890
of the Framingham
Heart Study data.

19
00:01:00,890 --> 00:01:05,630
We'll use logistic regression
to create a similar model.

20
00:01:05,630 --> 00:01:10,170
CHD is a disease of the blood
vessels supplying the heart.

21
00:01:10,170 --> 00:01:12,320
This is one type of
heart disease, which

22
00:01:12,320 --> 00:01:17,510
has been the leading cause of
death worldwide since 1921.

23
00:01:17,510 --> 00:01:23,230
In 2008, $7.3 million
people died from CHD.

24
00:01:23,230 --> 00:01:27,500
Even though the number of deaths
due to CHD is still very high,

25
00:01:27,500 --> 00:01:29,480
age-adjusted death
rates have actually

26
00:01:29,480 --> 00:01:33,860
declined 60% since 1950.

27
00:01:33,860 --> 00:01:38,140
This is in part due to earlier
detection and monitoring partly

28
00:01:38,140 --> 00:01:40,210
because of the
Framingham Heart Study.

29
00:01:43,050 --> 00:01:45,920
Before building a
logistic regression model,

30
00:01:45,920 --> 00:01:48,570
we need to identify the
independent variables

31
00:01:48,570 --> 00:01:50,509
we want to use.

32
00:01:50,509 --> 00:01:52,530
When predicting the
risk of a disease,

33
00:01:52,530 --> 00:01:57,070
we want to identify what
are known as risk factors.

34
00:01:57,070 --> 00:01:59,020
These are the
variables that increase

35
00:01:59,020 --> 00:02:02,340
the chances of
developing a disease.

36
00:02:02,340 --> 00:02:04,480
The term risk
factors was actually

37
00:02:04,480 --> 00:02:07,140
coined by William
Kannell and Roy Dawber

38
00:02:07,140 --> 00:02:10,020
from the Framingham Heart Study.

39
00:02:10,020 --> 00:02:12,050
Identifying these
risk factors is

40
00:02:12,050 --> 00:02:14,450
the key to successful
prediction of CHD.

41
00:02:17,220 --> 00:02:20,140
In this lecture, we'll
focus on the risk factors

42
00:02:20,140 --> 00:02:23,060
that they collected data
for in the original data

43
00:02:23,060 --> 00:02:26,100
collection for the
Framingham Heart Study.

44
00:02:26,100 --> 00:02:28,320
We'll be using an
anonymized version

45
00:02:28,320 --> 00:02:31,690
of the original data
that was collected.

46
00:02:31,690 --> 00:02:35,490
This data set includes several
demographic risk factors--

47
00:02:35,490 --> 00:02:38,690
the sex of the patient,
male or female;

48
00:02:38,690 --> 00:02:43,200
the age of the patient in
years; the education level coded

49
00:02:43,200 --> 00:02:45,590
as either 1 for
some high school,

50
00:02:45,590 --> 00:02:48,900
2 for a high school
diploma or GED,

51
00:02:48,900 --> 00:02:51,920
3 for some college
or vocational school,

52
00:02:51,920 --> 00:02:55,700
and 4 for a college degree.

53
00:02:55,700 --> 00:02:58,680
The data set also includes
behavioral risk factors

54
00:02:58,680 --> 00:03:02,060
associated with
smoking-- whether or not

55
00:03:02,060 --> 00:03:06,120
the patient is a current smoker
and the number of cigarettes

56
00:03:06,120 --> 00:03:09,510
that the person smoked
on average in one day.

57
00:03:09,510 --> 00:03:12,930
While it is now widely
known that smoking increases

58
00:03:12,930 --> 00:03:14,980
the risk of heart
disease, the idea

59
00:03:14,980 --> 00:03:20,579
of smoking being bad for you
was a novel idea in the 1940s.

60
00:03:20,579 --> 00:03:24,230
Medical history risk
factors were also included.

61
00:03:24,230 --> 00:03:25,940
These were whether or
not the patient was

62
00:03:25,940 --> 00:03:29,660
on blood pressure medication,
whether or not the patient had

63
00:03:29,660 --> 00:03:33,260
previously had a stroke,
whether or not the patient was

64
00:03:33,260 --> 00:03:36,720
hypertensive, and whether or
not the patient had diabetes.

65
00:03:39,650 --> 00:03:42,220
Lastly, the data set
includes risk factors

66
00:03:42,220 --> 00:03:45,740
from the first physical
examination of the patient.

67
00:03:45,740 --> 00:03:49,720
The total cholesterol level,
systolic blood pressure,

68
00:03:49,720 --> 00:03:55,370
diastolic blood pressure, Body
Mass Index, or BMI, heart rate,

69
00:03:55,370 --> 00:03:59,260
and blood glucose level of
the patient were measured.

70
00:03:59,260 --> 00:04:02,480
In the next video, we'll
use these risk factors

71
00:04:02,480 --> 00:04:06,450
to see if we can predict
the 10-year risk CHD.