Data refining of sick listing data for statistics, analysis and forecasting

1
Data refining of sick listing data for
statistics, analysis and forecasting
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
2
Part 1 What do we want to create?
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
3
Administrative data in Sweden
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• Sweden has a history of extensive gathering of
  administrative individual data
• Every person has an individual civic registration
  number which contains the birth date and four
  additional numbers
• Due to the civic registration number it is
  possible to combine administrative information
  from various sources (on a individual level)

4
Sick listing in Sweden
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• Sick listing in Sweden can be very lasting
• Different degrees of partiality(25, 50 or 75
  percent)
• Sickness cash benefit and rehabilitation cash
  benefit
• Employers pay for the first 14 days(has been 21
  and 28)
• Relation between sickness insurance and i.e.
  unemployment insurance and parental insurance
• Seasonal variation

5
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
Sickness cash benefit
Rehabilitation cash benefit
Sick pay
Waiting day
Example A sickness case
Degree of partiality
1
3
6
5
4
7
8
11
10
9
2
Time
Higher income entitling to sickness cash benefit
Episodes with the same benefit, degree of
partiality and daily compensation
6
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
Sickness cash benefit
Rehabilitation cash benefit
Sick pay
Waiting day
Example A sickness case
Degree of partiality
1
3
6
5
4
7
8
11
10
9
2
Time
Higher income entitling to sickness cash benefit
Episodes with the same degree of partiality
7
Share of new sickness cases with part-time
absence at the beginning of the case
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
8
Share of new sickness cases with a history of a
sickness case within preceding 90 days
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
9
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• The common way for analysts when making their
  own project databases
• Is it possible to make this process more
  effective?

Special design project databases
Separate processes to transform raw data
Raw Data
10
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
Yes, by setting up a common framework for the raw
data we can be more time efficient and increase
the general quality when we produce a designed
project database
Special design project databases
Refined data transformed into the lowest common
denominator
Raw Data
Refined database (MiDAS)
11
Part 2How do we create this?
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
12
The complexity in data
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• Three ways of dealing with this
• Leave data untouched and put togetherthe
  information as it is
• Make an effort to understand data andtake into
  account any defect in data
• Exclude observations that dont fit in

13
Data refining for analysis and forecasting
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• Quality assurance of individual data
• Take care of as much of the information as
  possible in the administrative systems
• Make data more accessible to optimize the use of
  data
• Create multidimensional databases foranalysis on
  individual level
• Detailed documentation

14
Qualifications needed
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• Front line staff who know about activities and
  routines mirrored in the administrative systems
• Analysts with some experience in programming who
  knows how to analyze data
• In-house people to ensure that the knowledge
  remains within the organization gt prefer
  employees to extern consultants

15
Example Sickness absence data
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
16
Complexity in sick listing data
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• Correct corrections (by the book)
• Incorrect corrections (not allowed but supported
  by the registration system)
• A registration squeezed into an earlier
  registration
• Incorrect registration of dates
• False 1 January
• Late arrival of observations

17
Correct correction of benefit and degree of
partiality
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
18
Incorrect correction of degree of partiality
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
19
A registration squeezed into an earlier
registration
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
4 days rehabilitation cash benefit
7 days sickness cash benefit
19 days rehabilitation cash benefit
20
A registration squeezed into an earlier
registration (continued)
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
21
False 1 January
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
22
Administrative data is complex
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• If you dont handle the complexity,data can be
  hard to analyze on micro level
• Even small errors in micro data willdecrease
  credibility in the data

23
Part 3The result
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
24
Refined databases
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• Episode data
• Sickness case
• part sickness case
• Panel data
• Month
• Quarter
• Year

25
Example Sick case data
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
26
Example Part sick case data
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
27
Example Year data
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
28
Number of people with occurrence of sick listing
in Sweden
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22
29
With this data structure you have
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• all sick listing data in one place time,
  days, compensation
• always the same information but structured
  differently for different purposes

30
With this data structure you can
Data refining of sick listing data Patric
Tirmén and Niklas Österlund 2006-11-22

• fast and easily create countless aggregated
  statistics
• analyze data on a micro level with high
  flexibility
• easily combine this data with other data
• create data sets suitable for forecasting