Chapter 2.

Forecasting

Chapter 2: Quantitatve Yasar A. Ozcan 1

Methods in Health Care
 Outline
Outline

 Why
Why Forecast?
Forecast?

 Steps
Steps in
in the
the Forecasting
Forecasting Process
Process

 Forecasting
Forecasting Approaches
Approaches
– Judgmental
Judgmental
– Time
Time Series--
Series-- Historical
Historical Data
Data
– Techniques
Techniques for
for Averaging
Averaging
– Techniques
Techniques for
for Trend
Trend
– Techniques
Techniques for
for Seasonality
Seasonality
– Associative
Associative

 Accuracy
Accuracy and
and Control
Control of
of Forecasts
Forecasts

 Choosing
Choosing aa Forecasting
Forecasting Technique
Technique

Chapter 2: Quantitatve Yasar A. Ozcan 2

Methods in Health Care
 Why
Why forecasting
forecasting is
is important?
important?

 Forecasts
Forecasts serve
serve as
as aa basis
basis for
for planning
planning

 Enable
Enable health
health care
care managers
managers to to anticipate
anticipate the
the
future
future toto plan
plan the
the system
system andand plan
plan the
the use
use of
of
that
that system
system

 Forecasting
Forecasting is is more
more than
than predicting
predicting demand
demand

 It
It is
is not
not an
an exact
exact science;
science; oneone must
must blend
blend
experience,
experience, judgment,
judgment, andand technical
technical expertise
expertise

Chapter 2: Quantitatve Yasar A. Ozcan 3

Methods in Health Care
 All
All forecasts
forecasts have
have common
common elements
elements
 Assumption
Assumption that
that past
past continues
continues into
into future
future
 Errors
 Errors occur--
occur-- actual
actual differs
differs from
from
predicted;
predicted; presence
presence ofof randomness
randomness
 Forecasts
 Forecasts ofof group
group of
of items
items (aggregate)
(aggregate)
tends
tends to
to be
be more
more accurate
accurate than
than individual
individual
items
items (i.e.,
(i.e., departmental
departmental vs. vs. whole
whole
hospital)
hospital)
 Forecast
 Forecast accuracy
accuracy decreases
decreases as as time
time
horizon
horizon increases
increases

Chapter 2: Quantitatve Yasar A. Ozcan 4

Methods in Health Care
 Characteristics of a Good Forecast

Timely

Reliable

Accurate

Meaningful units ($$’s, visits, discharges, patient days, etc.)

Easy to use

Chapter 2: Quantitatve Yasar A. Ozcan 5

Methods in Health Care
 Steps in the Forecasting Process

Step 1 Identify the goal of the forecast

Step 2 Establish a time horizon

Step 3 Select a forecasting technique

Step 4 Conduct the forecast (analyze data)

Step 5 Determine its accuracy

Step 6 Monitor the forecast

Chapter 2: Quantitatve Yasar A. Ozcan 6

Methods in Health Care
 What
What approaches
approaches can
can we
we use?
use?

 Judgmental
– Delphi method
– Executive opinions
– Contracts/insurance/HMO/PPO/POS
estimates
– Consumer surveys
– Outside opinions
– Opinions of managers/staff

Chapter 2: Quantitatve Yasar A. Ozcan 7

Methods in Health Care
 The
The Delphi
Delphi Method
Method
 Method
 Method of
of obtaining
obtaining opinions
opinions of
of
managers
managers andand staff
staff
 Involves
 Involves circulating
circulating aa series
series of
of
questionnaires,
questionnaires, each
each developed
developed from
from the
the
previous
previous one,
one, to
to achieve
achieve aa consensus
consensus on
on
an
an issue
issue (in
(in this
this case,
case, aa forecast)
forecast)
 Useful
 Useful for
for forecasting
forecasting technological
technological
changes
changes and
and their
their impacts
impacts

Chapter 2: Quantitatve Yasar A. Ozcan 8

Methods in Health Care
 The
The Delphi
Delphi Approach,
Approach, cont.
cont.

 Advantages
Advantages
–
– More
More individuals
individuals may
may be
be engaged
engaged than
than can
can effectively
effectively
interact
interact face-to-face
face-to-face
–
– It
It is
is important
important to
to avoid
avoid bandwagon
bandwagon effect
effect
–
– Preserves
Preserves anonymity
anonymity ofof participants
participants

 Weaknesses
Weaknesses
–
– Questions
Questions may
may bebe ambiguous
ambiguous leading
leading to
to false
false
consensus
consensus
–
– Panel
Panel members
members may
may change
change
–
– Studies
Studies dodo not
not prove
prove that
that Delphi
Delphi forecasts
forecasts are
are highly
highly
accurate
accurate
–
– Preserving
Preserving anonymity
anonymity removes
removes accountability
accountability

Chapter 2: Quantitatve Yasar A. Ozcan 9

Methods in Health Care
 Forecasting
Forecasting Approaches,
Approaches, cont.
cont.

 Time
Time series--
series-- identify
identify the
the behavior
behavior of
of the
the series
series by
by using
using
factors
factors such
such as
as trend,
trend, seasonality,
seasonality, cycles,
cycles, irregular
irregular
variations,
variations, and
and random
random variations
variations

 Techniques
Techniques for
for averaging
averaging
 Naive
 Naive forecasts
forecasts
 Moving
 Moving averages
averages (MA)
(MA)
 Exponential
 Exponential smoothing
smoothing
–
– Techniques
Techniques for
for trend
trend
 Linear
 Linear equations
equations using
using regression
regression (y
(ytt =
= aa +
+ bx
bxtt))
 Trend
 Trend adjusted
adjusted exponential
exponential smoothing
smoothing
–
– Techniques
Techniques for
for seasonality
seasonality
 Seasonal
 Seasonal Variations
Variations
 Indices
 Indices Technique
Technique

Chapter 2: Quantitatve Yasar A. Ozcan 10

Methods in Health Care
 Forecasting
Forecasting Approaches,
Approaches, cont.
cont.

 Associative
Associative Techniques
Techniques
– Simple
Simple linear
linear regression
regression (y
(y =
= aa +
+ bx)
bx)
– Scatter
Scatter diagram--
diagram-- plot
plot data
data
– Correlations
Correlations

Chapter 2: Quantitatve Yasar A. Ozcan 11

Methods in Health Care
Figure 2.1 Variation Characteristics
Seasonal
Seasonal
Variation
Variation

2005
2004

2003

Jan Mar May Jul Sep Nov

Cycle
Random Variation

Trend
Chapter 2: Quantitatve Yasar A. Ozcan 12
Methods in Health Care
 Averaging Techniques

 Smooth
Smooth out
out fluctuations
fluctuations in
in time
time serious
serious
because
because individual
individual highs
highs and
and lows
lows cancel
cancel
each
each other
other out
out
So,
So, would
would forecasts
forecasts based
based on on averages
averages
exhibit
exhibit more
more oror less
less variability?
variability?

Chapter 2: Quantitatve Yasar A. Ozcan 13

Methods in Health Care
 Naive Forecasts
A
 naive forecast for any period equals
the previous period’s actual value
 Low cost, easy to prepare, easy to

understand, but less accurate forecasts
 Can be applied to seasonal or trend

data
Examples:
If last week’s demand was 50 units, the naive forecast
for the coming week is 50 units.

If seasonal pattern exists, the naive forecast for next January

would equal the actual demand for January of this year.

Chapter 2: Quantitatve Yasar A. Ozcan 14

Methods in Health Care
 Moving
Moving Averages
Averages
Forecast uses a number of the most recent actual
data values in generating a forecast

Ft  MAn 
 A i

where, i = “Age” of data (i=1,2,3. . .)

n = number of periods in moving average
Ai = actual value with age i

Chapter 2: Quantitatve Yasar A. Ozcan 15

Methods in Health Care
 Moving
Moving Averages
Averages
Example 2.1:
An OB/GYN clinic has the following yearly patient
visits, and would like to predict the volume of
business for the next year for budgeting purposes.

Period (t) Age Visits

1 5 15908

2 4 15504

3 3 14272

4 2 13174

5 1 10022

Chapter 2: Quantitatve Yasar A. Ozcan 16

Methods in Health Care
 Moving
Moving Averages,
Averages, cont.
cont.
Solution:
The three-period moving average (MA3) for period 6 is

F6 = MA3 = (14272+13174+10022) ÷ 3 = 12489.3

Period (t) Age Visits Forecast

1 5 15908

2 4 15504

3 3 14272
4 2 13174 15228
5 1 10022 14317

6 12489

Chapter 2: Quantitatve Yasar A. Ozcan 17

Methods in Health Care
 Moving
Moving Averages,
Averages, cont.
cont.
The greater the number of periods in a moving
average, the greater the forecast will lag with changes
in the data

MA3

Data
MA5

1 2 3 4 5 6 7 8 9
Chapter 2: Quantitatve Yasar A. Ozcan 18
Methods in Health Care
 Moving
Moving Averages,
Averages, cont.
cont.

 Easy to compute and understand, but data
storage requirements can be high and all
values are weighted equally (i.e., in a ten year
moving average, each value is given a weight
of 1/10, adding up to 1).


 A weighted average assigns more weight to
recent values

Chapter 2: Quantitatve Yasar A. Ozcan 19

Methods in Health Care
 Using
Using Weighted
Weighted Values
Values
Example:
Example: Continuing
Continuing Period
with
with Example
Example 2.1;2.1; since
since (t) Age Visits Weights
there
there is
is aa downward
downward
trend
trend in
in visits
visits and
and in in
period
period 55 there
there is
is aa sharp
sharp 1 5 15908
decline,
decline, aa weight
weight of of .5
.5 or
or
even
even higher
higher isis justified
justified
by 2 4 15504
by the
the healthcare
healthcare
manager
manager to to calculate
calculate aa
weighted
weighted average
average for for 3 3 14272 0.2
period
period 66

Ft  MAn   wi Ai 4 2 13174 0.3

5 1 10022 0.5

6
Chapter 2: Quantitatve Yasar A. Ozcan 20
Methods in Health Care
 Using
Using Weighted
Weighted Values
Values
Solution:
Solution:
In
In this
this analysis,
analysis, aa weighted
weighted average,
average, using
using
formula
formula [2.2],
[2.2], for
for the
the OB/GYN
OB/GYN clinic
clinic for
for the
the
period
period 6 6 would
would be:
be:
Period
FF66 =
= 14272*.2+13174*.3+10022*.5
14272*.2+13174*.3+10022*.5 (t) Age Visits Weights Forecast
FF66 =
= 11818.
11818.
1 5 15908

2 4 15504

3 3 14272 0.2

4 2 13174 0.3

5 1 10022 0.5

6 11818
Chapter 2: Quantitatve Yasar A. Ozcan 21
Methods in Health Care
 Simple
Simple Exponential
Exponential Smoothing
Smoothing

 Each
Each new
new forecast
forecast is
is based
based on
on the
the previous
previous forecast
forecast plus
plus aa
percentage
percentage ofof the
the difference
difference between
between that
that forecast
forecast the
the actual
actual
value
value of
of the
the series
series at
at that
that point
point


 New
New forecast
forecast =
= Old
Old forecast + αα(Actual-Old
forecast + (Actual-Old forecast),
forecast), where
where
αα is
is aa percentage
percentage or
or

FFtt =
= FFt-1
t-1 +
+ αα(A t-1 -
(At-1 - FFt-1
t-1),
),

where,
where, FFtt =
= Forecast
Forecast for
for period
period tt
FFt-1
t-1 == Forecast
Forecast for
for period
period t-1
t-1
αα= = Smoothing
Smoothing constant
constant
A t-1 =
At-1 = Actual
Actual demand
demand or or sales
sales in
in period
period t-1
t-1

Chapter 2: Quantitatve Yasar A. Ozcan 22

Methods in Health Care
 Exponential
Exponential Smoothing,
Smoothing, cont.
cont.
Example
Example 2.4:
2.4: Using
Using the
the
data
data from
from Example
Example 2.1,
2.1, Smoothing constant α = 0.3 Error

build
build forecasts
forecasts with
with
smoothing
smoothing constant
constant αα==
0.3
0.3 Period (t) Actual (Visits) Forecast (Actual – Forecast)

Solution:
Solution:
Following
Following the
the previous
previous
example
example and
and formula
formula [2.3],
[2.3], 1 15908 --

we
we can
can build
build forecasts
forecasts for
for
periods
periods as
as data
data become
become 2 15504 15908 -404.0

available.
available.
FF33 =
= 15908
15908 + + .30(15504-
.30(15504- 3 14272 15786.8 -1514.8

15908)
15908)
4 13174 15332.4 -2158.4
FF33 =
= 15786.8
15786.8
5 10022 14684.9 -4662.9

Chapter 2: Quantitatve Yasar A. Ozcan 23

Methods in Health Care
 Exponential Smoothing, cont.

Example 2.5: Using

Using the
the Smoothing constant α = 0.5 Error

data
data from
from Example
Example 2.1
2.1,, build
build Period(t) Visits Forecast (Actual – Forecast)
forecasts
forecasts with
with smoothing
smoothing
constant
constant
αα =
= 0.5.
0.5.
1 15908 --

Solution: 2 15504 15908 -404.0

3 14272 15706.0 -1434.0

4 13174 14989.0 -1815.0

5 10022 14081.5 -4059.5

Chapter 2: Quantitatve Yasar A. Ozcan 24

Methods in Health Care
 Exponential Smoothing, cont.
Example
Example 2.6:
2.6: Using
Using the
the data
data
from
from Example
Example 2.1
2.1,, build
build forecasts
forecasts
with smoothing constants
with smoothing constants
αα =
= 0.0
0.0 and
and αα =
= 1.0.
1.0.
Solution:
Solution:

Period α = 0.0 Error α = 1.0 Error

(t)
Visits Forecast (Actual – Forecast) Visits Forecast (Actual – Forecast)

1 15908 -- 15908 --

2 15504 15908 -404.0 15504 15908 -404.0

3 14272 15908.0 -1636.0 14272 15504.0 -1232.0
4 13174 15908.0 -2734.0 13174 14272.0 -1098.0

5 10022 15908.0 -5886.0 10022 13174.0 -3152.0

6   15908.0   10022.0

Chapter 2: Quantitatve Yasar A. Ozcan 25

Methods in Health Care
 Techniques
Techniques for
for Trends
Trends

 Least
Least squares
squares regression
regression-- -- minimizes
minimizes the
the sum
sum
of
of the
the squared
squared errors
errors
Least
Least squares
squares line
line::
y
y==a a++ bx
bx,, yy == predicted
predicted (dependent)
(dependent)
variable
variable
xx =
= predictor
predictor (independent)
(independent) variable
variable
bb =
= slope
slope of
of data
data line
line
aa =
= value
value of
of yy when
when xx ==00

b=
n(xy) - (x)(y) a = y - bx
n(x2) - (x)2 n

Chapter 2: Quantitatve Yasar A. Ozcan 26

Methods in Health Care
Figure
Figure 2.9
2.9 Linear
Linear Regression
Regression

y
y = a + bx
error
error

Δy

Δx b =(Δy/Δx) , where b>0

Chapter 2: Quantitatve Yasar A. Ozcan 27

Methods in Health Care
 Techniques
Techniques for
for Trends
Trends
Example 2.7: A multi-hospital system (MHS) owns 12 hospitals.
Revenues (x, or the independent variable) and profits (y, or the
dependent variable) for each hospital are given below. Obtain a
regression line for the data, and predict profits for a hospital with
$10 million in revenues. All figures are in millions of dollars.
Multi Hospital System Revenues and Profits Data

Hospital Revenue (x) Profit (y) x*y x2

1 7 0.15 1.05 49

2 2 0.10 0.2 4

3 6 0.13 0.78 36

4 4 0.15 0.6 16

5 14 0.25 3.5 196

6 15 0.27 4.05 225

7 16 0.24 3.84 256

8 12 0.20 2.4 144

9 14 0.27 3.78 196

10 20 0.44 8.8 400

11 15 0.34 5.1 225

12 7 0.17 1.19 49

Total 132 2.71 35.29 1796

Chapter 2: Quantitatve Yasar A. Ozcan 28

Methods in Health Care
 Solution:
Solution:
After calculating
   
x , y , xy , x 2

substitute into the equations [2.5] for a and [2.6] for b, respectively.

b
 n( xy)  ( x)( y ) 12(35.29)  132(2.71)
  0.01593.
n (  x )  ( x )
2 2
12(1796)  132(132)

a
 y  b x 2.71  0.01593(132)
  0.0506.
n 12
Hence, the regression line is:
yx = 0.0506 + 0.01593x.
To predict the profits for a hospital with $10 million in revenue,
simply plug 10 in as the value of x in the regression equation:
Profit = 0.0506 + 0.01593(10) = .209903
Multiplying this value by one million,
the profit level with $10 million in revenue is found to be $209,903.

Chapter 2: Quantitatve Yasar A. Ozcan 29

Methods in Health Care
 Techniques
Techniques for
for Trends
Trends
Linear
Linear Regression
Regression as
as a
a Trend
Trend Line
Line

yy==a a++ b*t

b*t
yy =
= predicted
predicted (dependent)
(dependent) variable
tt =
= predictor
predictor (time)
(time) variable
variable
bb== slope
slope of
of data
data line
line
aa = value
value of
of yy when
when xx =
=00

n(ty) - (t)(y) a= y - bt

b= n(t2) - (t)2 n

Chapter 2: Quantitatve Yasar A. Ozcan 30

Methods in Health Care
 Example
Example 2.8:
2.8: Referring
Referring back
back to
to the
the OB/GYN
OB/GYN example,
example, the
the health
health care
care manager
manager can
can
estimate
estimate the
the trend
trend line
line using
using regression
regression analysis.
analysis.
Solution:
Solution:

Chapter 2: Quantitatve Yasar A. Ozcan 31

Methods in Health Care
 Techniques
Techniques for
for Seasonality
Seasonality
Seasonal
Seasonal variations
variations in
in aa data
data set
set consistently
consistently repeat
repeat upward
upward oror
downward
downward movements
movements ofof the
the data
data values
values that
that can
can be
be traced
traced to
to
recurrent
recurrent events.
events.

In
In the
the additive
additive model
model,, seasonality
seasonality is
is expressed
expressed as
as aa quantity
quantity
(example:
(example: 55 units),
units), which
which is
is added
added or
or subtracted
subtracted from
from the
the
series average in order to incorporate seasonality.
series average in order to incorporate seasonality.

In
In the
the multiplicative
multiplicative model
model,, seasonality
seasonality is
is expressed
expressed as
as aa
percentage
percentage of
of the
the average
average amount
amount (example:
(example: 1.15)
1.15)

Quarterly,
Quarterly, Monthly,
Monthly, Daily
Daily Indices
Indices Technique
Technique

Chapter 2: Quantitatve Yasar A. Ozcan 32

Methods in Health Care
 Techniques
Techniques for
for Seasonality
Seasonality
Employing
Employing Seasonal
Seasonal Indices
Indices in
in Forecasts
Forecasts
Example
Example 2.10:2.10: AA forecast
forecast based
based on
on linear
linear regression
regression yields
yields the
the following
following
trend equation
trend equation
Demand
Demand (Y
(Ytt)) =
= 511.06
511.06 ++ 1.259
1.259 t.
t.

The
The forecast
forecast of
of demand
demand for
for periods
periods 29
29 through
through 31
31 would
would be:
be:
Y29
Y29 == 511.06
511.06 ++ 1.259
1.259 (29)
(29) =
= 547.6.
547.6.
Y30
Y30 == 511.06
511.06 +1.259
+1.259 (30)
(30) =
= 548.8.
548.8.
Y31
Y31 == 511.06
511.06 ++ 1.259
1.259 (31)
(31) =
= 550.1.
550.1.

Having
Having forecast
forecast the
the next
next three
three months,
months, the
the healthcare
healthcare manager
manager needs
needs toto
incorporate
incorporate seasonality
seasonality back
back into
into those
those forecasts.
forecasts. The
The periods
periods tt =
= 29,
29, 30
30
and 31 represent the months of November, December and
and 31 represent the months of November, December and January, January,
respectively,
respectively, with
with corresponding
corresponding monthly
monthly indices
indices 0.984,
0.984, 0.973,
0.973, and
and 1.036.
1.036.
Monthly
Monthly adjustments
adjustments to to those
those forecasts
forecasts are
are calculated
calculated

Monthly
Monthly Adjusted
Adjusted Forecast
Forecast (t):
(t): Forecast
Forecast *
* Monthly
Monthly Index
Index
Period
Period 29
29 (November):
(November): 547.6
547.6 (0.984)
(0.984) == 538.8.
538.8.
Period
Period 30
30 (December):
(December): 548.9
548.9 (0.973)
(0.973) == 534.0.
534.0.
Period
Period 31
31 (January)
(January) :: 550.1
550.1 (1.036)
(1.036) =
= 569.9.
569.9.

Chapter 2: Quantitatve Yasar A. Ozcan 33

Methods in Health Care
 Techniques
Techniques for
for Seasonality
Seasonality
Employing
Employing Seasonal
Seasonal Indices
Indices in in Forecasts
Forecasts
The
The next
next step
step in
in adjustment
adjustment of of the
the forecasted
forecasted
demand
demand would
would bebe for
for daily
daily fluctuations.
fluctuations. AsAs was
was
shown
shown inin Table
Table 2.4,
2.4, Heal
Heal Me
Me Hospital
Hospital experiences
experiences
daily
daily variation
variation in
in demand.
demand. Thus,
Thus, the
the monthly
monthly
index
index adjusted
adjusted forecasts
forecasts should
should bebe further
further adjusted
adjusted
for
for daily
daily variations.
variations.

Daily
Daily Adjusted
Adjusted Forecast
Forecast =
= Monthly
Monthly Adjusted
Adjusted Forecast
Forecast (t)
(t) *
*
Daily
Daily Index
Index

For
For example,
example, for
for November
November (period
(period 29),
29), the
the
adjusted
adjusted forecasts
forecasts for
for Monday
Monday and
and Tuesday
Tuesday are:
are:
Monday,
Monday, November:
November: 538.8
538.8 ** (0.972)
(0.972) == 523.7.
523.7.
Tuesday,
Tuesday, November:
November: 538.8
538.8 ** (1.023)
(1.023) =
= 551.2.
551.2.

Chapter 2: Quantitatve Yasar A. Ozcan 34

Methods in Health Care
 How
How accurate
accurate are
are we?
we?
Forecast Error equals the actual value minus the forecasted value.
Error = Actual – Forecast
 Errors
Errors may
may be
be caused
caused by:
by:
– An
An inadequate
inadequate forecasting
forecasting model
model
– Irregular
Irregular variations
variations due
due to
to severe
severe weather,
weather,
shortages
shortages or
or breakdowns,
breakdowns, catastrophes,
catastrophes, etc.
etc.
– Forecasting
Forecasting technique
technique may
may be be used
used
improperly
improperly
– There
There may
may be
be random
random variations
variations in
in the
the data
data

Chapter 2: Quantitatve Yasar A. Ozcan 35

Methods in Health Care
 Is
Is your
your forecast
forecast accurate?
accurate?
• Mean Absolute Deviation (MAD)
MAD weights all
MAD 
 | Actual  Forecast |
errors evenly.
n
• Mean Absolute Percent Error (MAPE)
MAPE avoids the

MAPE 
 | Actual  Forecast | problem of
interpreting the
 Actual measure of accuracy
Seek lowest of MAD or MAPE for given set of data; relative to the
also examine historical performance versus magnitudes of the
responsiveness to current situation.
actual and the
forecast values.
Chapter 2: Quantitatve Yasar A. Ozcan 36
Methods in Health Care
 Is your forecast accurate?
Using data from Example 2.4, SES with α = 0.3, we
observe the necessary error calculations in Table below.
Period Smoothing constant α = .3 Error Absolute Error
t
Actual Forecast (Actual – Forecast) |Actual – Forecast|

1 15908 --
2 15504 15908 -404 404
3 14272 15786.8 -1514.8 1515
4 13174 15332.4 -2158.4 2158
5 10022 14684.9 -4662.9 4662.9

6 13286

Sum Σ 52972 8740.1

Hence,
MAD = 8740.1 ÷ 4 = 2185.03, and
MAPE = 8740.1 ÷ 52972 = 0.165 or 16.5%.

Chapter 2: Quantitatve Yasar A. Ozcan 37

Methods in Health Care
 Is
Is your
your forecast
forecast accurate?
accurate?

 Controlling
Controlling forecasts--
forecasts-- set
set predetermined
predetermined upper/lower
upper/lower
limits
limits for
for forecast
forecast errors
errors

 Forecasts
Forecasts cancan be
be monitored
monitored using
using either
either aa tracking
tracking
signal
signal oror control
control chart
chart..
– Tracking
Tracking signals
signals show
show cumulative
cumulative errors
errors
-- Control
Control Charts--
Charts-- set
set upper
upper and
and lower
lower limits
limits for
for
individual
individual forecast
forecast errors
errors

Tracking signal 
 ( Actual  Forecast )
MAD

Chapter 2: Quantitatve Yasar A. Ozcan 38

Methods in Health Care
 Control
Control Chart
Chart for
for Tracking
Tracking Signal
Signal
6

Range of
Tracking signal

0 Acceptable
-2
1 3 5 7 9 11 13 15 17 19 21 23 25 27
Variation
-4

-6 Need for
Corrective Action
-During periods 12 through 15 the tracking signal went beyond the
acceptable control limits (down to -5.51), but recovered at period 16 and
stayed within acceptable limits after that.
-Until period 8 the predicted values were below the actual. That changed
from period 9 to period 20, when forecasts were higher than actual data.
-At the period 21 a return to under-forecast occurred.

Chapter 2: Quantitatve Yasar A. Ozcan 39

Methods in Health Care
 So
So what
what technique
technique should
should we
we use?
use?
 Factors
Factors of
of importance:
importance:
– Frequency
Frequency
– Level
Level of
of aggregation
aggregation
– Type
Type ofof Model-
Model- Errors
Errors [MAD,
[MAD, MAPE]
MAPE]
– Degree
Degree ofof managerial
managerial involvement
involvement
– Cost
Cost per
per forecast
forecast
 Time
Time horizon
horizon considerations--
considerations-- short,
short,
intermediate,
intermediate, or
or long
long

Chapter 2: Quantitatve Yasar A. Ozcan 40

Methods in Health Care
 What
What makes
makes aa forecast
forecast aa good
good one?
one?

 Timeliness
Timeliness
 Accuracy
Accuracy
 Meaningful
Meaningful Units
Units ($$’s,
($$’s, visits,
visits, etc.)
etc.)
 In
Inwriting
writing
 Simple
Simple to
to understand
understand and
and use
use

Chapter 2: Quantitatve Yasar A. Ozcan 41

Methods in Health Care
 The End

Chapter 2: Quantitatve Yasar A. Ozcan 42

Methods in Health Care