Survival - Notes (Lecture 2)

Lecture 2: Survival Models and Analysis
Survival and Hazard Functions
Ms. Beryl Ang’iro
September 7, 2020
Ms. Beryl Ang’iro September 7, 2020 1 / 18

Survival and Hazard functions
Survival function
Survival function
Let T be a non-negative random variable representing the survival times of

individuals in some population.(time-to-event of our interest)
T is a positive random variable which has to be unambiguously defined;

that is, we must be very specific about the start and end with the length
of the time period in-between corresponding to T.
Examples
• Survival time (in general): measured from birth to death for an
individual. This is the survival time we need to investigate in a life
expectancy study.
• Survival time of a treatment for a population with certain disease:
measured from the time of treatment initiation until death.

Survival function
Survival function


Examples
expectancy study.

Survival function
Survival function


Examples
expectancy study.

Survival function
Survival function


Examples
expectancy study.

Survival function
• Survival time due to heart disease: (the event is death from heart
disease): measured from birth (or other time point such as treatment
initiation for heart disease patients) to death caused by heart disease.
(This may be a bit tricky if individuals die from other causes).
This is competing risk problem. That is, other risks are competing with
heart disease to produce an event ( death.)
The distribution of the random variable T can be described in a number of
equivalent ways. There is of course the usual (cumulative) distribution
function;
F (t) = P[T ≤ t]; t ≥ 0.

which is right continuous, i.e., limu→t + F(u)=F(t). When T is a survival
time, F(t) is the probability that a randomly selected subject from the
population will die before time t.

Survival function
function;
F (t) = P[T ≤ t]; t ≥ 0.


Survival function
function;
F (t) = P[T ≤ t]; t ≥ 0.


Survival function
function;
F (t) = P[T ≤ t]; t ≥ 0.


Survival function
If T is a continuous random variable, then it has a density function f(t),

which is related to F(t) through following equations;
Z t
dF (t)
f (t) = ; F (t) = f (u)du
dt 0
In biomedical applications, it is often common to use the survival function
S(t) = P(T ≥ t) = 1 − F (t − ).
where F (t − ) = limu→t − F (u). When T is a survival time, S(t) is the

probability that a randomly selected individual will survive to time t or
beyond. (So S(t) has the name of survival function.)
Note: Some authors use the following definition of a survival function
S(t) = P(T > t) = 1 − F (t).
This definition will be identical to the above one if T is a continuous

random variable, which is the case we will focus on in this course.
Survival function

Z t
dF (t)
f (t) = ; F (t) = f (u)du
dt 0
S(t) = P(T ≥ t) = 1 − F (t − ).

S(t) = P(T > t) = 1 − F (t).

Survival function

Z t
dF (t)
f (t) = ; F (t) = f (u)du
dt 0
S(t) = P(T ≥ t) = 1 − F (t − ).

S(t) = P(T > t) = 1 − F (t).

Survival function

Z t
dF (t)
f (t) = ; F (t) = f (u)du
dt 0
S(t) = P(T ≥ t) = 1 − F (t − ).

S(t) = P(T > t) = 1 − F (t).

Survival function

Z t
dF (t)
f (t) = ; F (t) = f (u)du
dt 0
S(t) = P(T ≥ t) = 1 − F (t − ).

S(t) = P(T > t) = 1 − F (t).

Survival function
All functions, unless stated otherwise, are defined over the interval [0,1).
The survival function S(t) is a non-increasing function over time taking

on the value 1 at t = 0, i.e., S(0)=1.
For a proper random variable T, S(1) = 0, which means that everyone

will eventually experience the event.
If T is a continuous r.v., we have;

Z ∞
dS(t)
S(t) = f (u)du, f (t) = −
t dt
That is, there is a one-to-one correspondence between f(t) and S(t).

Survival function



Z ∞
dS(t)
S(t) = f (u)du, f (t) = −
t dt

Survival function



Z ∞
dS(t)
S(t) = f (u)du, f (t) = −
t dt

Survival function



Z ∞
dS(t)
S(t) = f (u)du, f (t) = −
t dt

Survival function
One use of the survivor function is to predict quantiles of the survival time.
Mean Survival Time: Due to censoring, sample mean of observed
survival times is no longer an unbiased estimate of µ= E(T). If we can
estimate S(t) well, then we can estimate µ= E(T) using the following
fact: Z ∞
E (t) = S(t)dt
0
Median Survival Time: The median survival time m is defined as the
quantity m satisfying S(m)= 0.5. Sometimes denoted by t0.5 .
If S(t) is not strictly decreasing, m is the smallest one such that S(m)≤
0.5. (The median may be preferable to the mean as a measure of
centrality if the data are highly skewed.)
p th quantile of Survival Time (100p th percentile): tp such that S(tp )
= 1-p (0 < p < 1). If S(t) is not strictly decreasing, tp is the smallest
one such that S(tp ) ≤ 1 − p.
Survival function
fact: Z ∞
E (t) = S(t)dt
0
Survival function
fact: Z ∞
E (t) = S(t)dt
0
Survival function
fact: Z ∞
E (t) = S(t)dt
0
Survival function
fact: Z ∞
E (t) = S(t)dt
0
Survival function
Mean Residual life Time (mrl):
mrl(t0 ) = E [T − t0 |T ≥ t0 ];
That is mrl(t0 ) is the average remaining survival time given the population
has survived beyond t0 . It can be shown that
R∞
S(t)dt
mrl(t0 ) = t0
S(t0 )
For example, in the hypothetical population shown in Figure 1, we have a

population where 70% of the individuals will survive 2 years (i.e t0.3 = 2)
and the median survival time is 2.8 years (i.e., 50% of the population will
survive at least 2.8 years).

Survival function
mrl(t0 ) = E [T − t0 |T ≥ t0 ];
R∞
S(t)dt
mrl(t0 ) = t0
S(t0 )


Survival function
mrl(t0 ) = E [T − t0 |T ≥ t0 ];
R∞
S(t)dt
mrl(t0 ) = t0
S(t0 )


Survival function
Figure 1: The survival function for a hypothetical population

Survival function
Hazard function
The hazard rate is a useful way of describing the distribution of ”time to

event” because it has a natural interpretation that relates to the aging of a
population.
We motivate the definition of hazard rate by first defining mortality rate

which is a discrete version of the hazard rate.
The mortality rate at time t, where t is generally taken to be an integer in

terms of some unit of time (e.g., years, months, days, etc), is the
proportion of the population who fail (die) between times t and t + 1
among individuals alive at time t, i.e.,
m(t) = P[t ≤ T < t + 1|T ≥ t]

Survival function
Hazard function

population.


m(t) = P[t ≤ T < t + 1|T ≥ t]

Survival function
Hazard function

population.


m(t) = P[t ≤ T < t + 1|T ≥ t]

Survival function
In a human population, the mortality rate has the typical pattern shown in
Figure 2
Figure 2: A typical mortality pattern for human

Survival function
The hazard rate λ(t) is the limit of a mortality rate if the interval of time
is taken to be small (rather than one unit).
The hazard rate is the instantaneous rate of failure (experiencing the
event) at time t given that an individual is alive at time t.
Specifically, hazard rate λ(t) is defined by the following equation
P[t ≤ T < t + h|T ≥ t]

λ(t) = lim
h→0 h
Therefore if h is very small
P[t ≤ T < t + h|T ≥ t] ≈ λ(t)h

The definition of the hazard function implies that
limh→0 P[t≤Th<t+h] f (t)

λ(t) = =
P[T ≥ t] S(t)

Survival function
P[t ≤ T < t + h|T ≥ t]

λ(t) = lim
h→0 h
P[t ≤ T < t + h|T ≥ t] ≈ λ(t)h


λ(t) = =
P[T ≥ t] S(t)

Survival function
P[t ≤ T < t + h|T ≥ t]

λ(t) = lim
h→0 h
P[t ≤ T < t + h|T ≥ t] ≈ λ(t)h


λ(t) = =
P[T ≥ t] S(t)

Survival function
P[t ≤ T < t + h|T ≥ t]

λ(t) = lim
h→0 h
P[t ≤ T < t + h|T ≥ t] ≈ λ(t)h


λ(t) = =
P[T ≥ t] S(t)

Survival function
P[t ≤ T < t + h|T ≥ t]

λ(t) = lim
h→0 h
P[t ≤ T < t + h|T ≥ t] ≈ λ(t)h


λ(t) = =
P[T ≥ t] S(t)

Survival function
0
S (t) dlog {S(t)}
=− =−
S(t) dt
From this, we can integrate both sides to get
Z t
Λ(t) = λ(u)du = −log {S(t)}
0
where Λ(t) is referred to as the cumulative hazard function. Here we used

the fact that S(0) = 1. Hence,
Rt
S(t) = e −Λ(t) = e − 0 λ(u)du
Since S(∞) = 0 it follows that Λ(∞) = limt→∞ Λ(t) = ∞. Thus the

hazard function has the property
Z ∞
λ(t) ≥ 0 and λ(t)dt = ∞
0

Survival function
0
S (t) dlog {S(t)}
=− =−
S(t) dt
Z t
0

Rt
S(t) = e −Λ(t) = e − 0 λ(u)du

Z ∞
λ(t) ≥ 0 and λ(t)dt = ∞
0

Survival function
0
S (t) dlog {S(t)}
=− =−
S(t) dt
Z t
0

Rt
S(t) = e −Λ(t) = e − 0 λ(u)du

Z ∞
λ(t) ≥ 0 and λ(t)dt = ∞
0

Survival function
Also
0 0
F (t) = 1 − S(t) =⇒ f (t) = F (t) = −S (t) = λ(t)exp{−Λ(t)}
Note:
1 There is a one-to-one relationship between hazard rate λ(t); t ≥ 0
and survival function S(t), namely,
Rt
S(t) = e − 0 λ(u)du
and
λ(t) = − dlog {S(t)}
dt
2 The hazard rate is NOT a probability, it is a probability rate.
Therefore it is possible that a hazard rate can exceed one in the same
fashion as a density function f(t) may exceed one.

Survival function
Also
0 0
Note:
Rt
S(t) = e − 0 λ(u)du
and
dt

Survival function
Also
0 0
Note:
Rt
S(t) = e − 0 λ(u)du
and
dt

Survival function
Also
0 0
Note:
Rt
S(t) = e − 0 λ(u)du
and
dt

Survival function
Common Parametric Models:

Example 1: Exponential distribution
Suppose T has pdf
(
λexp(−λt), for t > 0, λ > 0
f (t) =
0, otherwise
The survival function: S(t) = e −λt

f (t)
The hazard function: λ(t) = S(t) =λ
In other words, the hazard function is constant when the survival time is
exponentially distributed.
The cumulative hazard function: Λ(t) = λt
The constant hazard function is a consequence of the memory less property
of the exponential distribution: the distribution of the subject’s remaining
survival time given that s/he has survived till time t does not depend on t.
Survival function

Suppose T has pdf
(
f (t) =
0, otherwise

f (t)
Survival function

Suppose T has pdf
(
f (t) =
0, otherwise

f (t)
Survival function

Suppose T has pdf
(
f (t) =
0, otherwise

f (t)
Survival function

Suppose T has pdf
(
f (t) =
0, otherwise

f (t)
Survival function

Suppose T has pdf
(
f (t) =
0, otherwise

f (t)
Survival function
In other words, the probability of death in a time interval [t, t + h] does

not depend on the starting point, t.
Mean survival time (lifetime):
Z ∞ Z ∞ Z ∞
1
µ = E (T ) = tf (t)dt = S(t)dt = e −λt dt =
0 0 0 λ
Median survival time:
Letting
S(t0.5 ) = e −λt0.5 = 0.5

ln 0.5 ln 2
t0.5 = − =
λ λ
The mean residual life time after t0
R∞ R ∞ −λt
t0 S(t)dt e 1
mrl(t0 ) = = t0 −λt = = E (T )
S(t0 ) e 0 λ
Which is the same as the mean lifetime.
Survival function

Z ∞ Z ∞ Z ∞
1
0 0 0 λ
Letting
S(t0.5 ) = e −λt0.5 = 0.5

ln 0.5 ln 2
t0.5 = − =
λ λ
R∞ R ∞ −λt
t0 S(t)dt e 1
mrl(t0 ) = = t0 −λt = = E (T )
S(t0 ) e 0 λ
Survival function

Z ∞ Z ∞ Z ∞
1
0 0 0 λ
Letting
S(t0.5 ) = e −λt0.5 = 0.5

ln 0.5 ln 2
t0.5 = − =
λ λ
R∞ R ∞ −λt
t0 S(t)dt e 1
mrl(t0 ) = = t0 −λt = = E (T )
S(t0 ) e 0 λ
Survival function

Z ∞ Z ∞ Z ∞
1
0 0 0 λ
Letting
S(t0.5 ) = e −λt0.5 = 0.5

ln 0.5 ln 2
t0.5 = − =
λ λ
R∞ R ∞ −λt
t0 S(t)dt e 1
mrl(t0 ) = = t0 −λt = = E (T )
S(t0 ) e 0 λ
Survival function
Example 2: Weibull Distribution
Suppose T has pdf

(
αλt α−1 exp(−λt α ), for t > 0, α > 0, λ > 0
f (t) =
0, otherwise
This is weibull distribution with shape parameter α and scale parameter λ
The survival function of T is S(t) = exp(−λt α )

f (t)
The hazard function: λ(t) = S(t) = λαt α−1
The cummulative hazard function: Λ(t) = λt α

Survival function
Suppose T has pdf

(
f (t) =
0, otherwise

f (t)

Survival function
Suppose T has pdf

(
f (t) =
0, otherwise

f (t)

Survival function
Suppose T has pdf

(
f (t) =
0, otherwise

f (t)

Survival function
Suppose T has pdf

(
f (t) =
0, otherwise

f (t)

Survival function
For the Weibull distribution, the hazard function depends on t.
We can see that, depending on whether α is greater than or less than 1,

the hazard can increase or decrease with increasing t.
This is often more realistic than the assumption of a constant hazard

function (as in the exponential case).
Since the exponential distribution is a special case of the Weibull with

α = 1, one way of analyzing the hazard rate is to fit the (more general)
Weibull model and then test whether α = 1.
So λ(t) is monotonically increasing when α > 1 and monotonically

decreasing when 0 < α < 1.

Survival function





Survival function





Survival function





Survival function





Survival function
Note that this model allows:

• Constant hazard: α = 1
• Increasing hazard: α > 1
• Decreasing hazard: α < 1
and has the hazard patterns shown in Figure 3
Figure 3: Three hazard patterns

Survival function


Survival function


Survival function


Survival function



Survival - Notes (Lecture 2)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Survival - Notes (Lecture 2)

Uploaded by

Copyright:

Available Formats

Lecture 2: Survival Models and Analysis

Survival and Hazard Functions

Ms. Beryl Ang’iro

Ms. Beryl Ang’iro September 7, 2020 1 / 18

Let T be a non-negative random variable representing the survival times of

T is a positive random variable which has to be unambiguously defined;

Ms. Beryl Ang’iro September 7, 2020 2 / 18

Let T be a non-negative random variable representing the survival times of

T is a positive random variable which has to be unambiguously defined;

Ms. Beryl Ang’iro September 7, 2020 2 / 18

Let T be a non-negative random variable representing the survival times of

T is a positive random variable which has to be unambiguously defined;

Ms. Beryl Ang’iro September 7, 2020 2 / 18

Let T be a non-negative random variable representing the survival times of

T is a positive random variable which has to be unambiguously defined;

Ms. Beryl Ang’iro September 7, 2020 2 / 18

F (t) = P[T ≤ t]; t ≥ 0.

Ms. Beryl Ang’iro September 7, 2020 3 / 18

F (t) = P[T ≤ t]; t ≥ 0.

Ms. Beryl Ang’iro September 7, 2020 3 / 18

F (t) = P[T ≤ t]; t ≥ 0.

Ms. Beryl Ang’iro September 7, 2020 3 / 18

F (t) = P[T ≤ t]; t ≥ 0.

Ms. Beryl Ang’iro September 7, 2020 3 / 18

If T is a continuous random variable, then it has a density function f(t),

In biomedical applications, it is often common to use the survival function

where F (t − ) = limu→t − F (u). When T is a survival time, S(t) is the

S(t) = P(T > t) = 1 − F (t).

This definition will be identical to the above one if T is a continuous

If T is a continuous random variable, then it has a density function f(t),

In biomedical applications, it is often common to use the survival function

where F (t − ) = limu→t − F (u). When T is a survival time, S(t) is the

S(t) = P(T > t) = 1 − F (t).

This definition will be identical to the above one if T is a continuous

If T is a continuous random variable, then it has a density function f(t),

In biomedical applications, it is often common to use the survival function

where F (t − ) = limu→t − F (u). When T is a survival time, S(t) is the

S(t) = P(T > t) = 1 − F (t).

This definition will be identical to the above one if T is a continuous

If T is a continuous random variable, then it has a density function f(t),

In biomedical applications, it is often common to use the survival function

where F (t − ) = limu→t − F (u). When T is a survival time, S(t) is the

S(t) = P(T > t) = 1 − F (t).

This definition will be identical to the above one if T is a continuous

If T is a continuous random variable, then it has a density function f(t),

In biomedical applications, it is often common to use the survival function

where F (t − ) = limu→t − F (u). When T is a survival time, S(t) is the

S(t) = P(T > t) = 1 − F (t).

This definition will be identical to the above one if T is a continuous

The survival function S(t) is a non-increasing function over time taking

For a proper random variable T, S(1) = 0, which means that everyone

If T is a continuous r.v., we have;

Ms. Beryl Ang’iro September 7, 2020 5 / 18

The survival function S(t) is a non-increasing function over time taking

For a proper random variable T, S(1) = 0, which means that everyone

If T is a continuous r.v., we have;

Ms. Beryl Ang’iro September 7, 2020 5 / 18

The survival function S(t) is a non-increasing function over time taking

For a proper random variable T, S(1) = 0, which means that everyone

If T is a continuous r.v., we have;

Ms. Beryl Ang’iro September 7, 2020 5 / 18

The survival function S(t) is a non-increasing function over time taking

For a proper random variable T, S(1) = 0, which means that everyone