Data Processing

1
Data Processing
Health Management Information Systems

• Topic 4

João Carlos de Timóteo Mavimbe Oslo, April 2007
2
(No Transcript)
3
Learning objectives (1)

• Explain the use of the data handling process as a
  strategy to provide good data quality
• Explain the techniques for providing and ensuring
  good quality data
• Describe mechanisms for assessing data quality

4
Learning objectives (2)

• Examine the importance of good data quality
• Appreciate the importance of accuracy in health
  data
• Understand why errors occur
• Acquire the skills required to detect, correct
  and prevent future errors
• Establish and apply the logistics of error
  checking

5
Processing data in the information cycle
6
Ensuring data accuracy

• Once data has been collected, it should be
  checked for any inaccuracies and obvious errors.
• Ideally this should be done as close to the point
  of data collection as possible.
• But also at all steps of the information cycle.

7
Why checking data is vital?

• Use of inaccurate data is DANGEROUS
• Producing data is EXPENSIVE
• Inaccurate data are USELESS data
• Producing inaccurate data is a WASTE of money and
  time

8
Why checking data is vital?
Better to have NO data, than to have inaccurate
data!!!
9
Common problems with data

• large gaps
• unusual month to month variations
• duplication
• Promoted by different vertical programs
• inconsistencies
• unlikely values
• data is present where it should not be
• typing errors
• maths problems poor calculation
• data entered in wrong boxes

10
Good quality data

• WHAT?
• data that are complete, correct and consistent
  (and timely)
• WHY?
• facilitates
• good decision-making
• appropriate planning
• ongoing Monitoring Evaluation
• improvement of coverage and quality of care
• HOW?
• provides an accurate picture of health
  programmes and services

11
Visual scanning (eyeballing)? checking for 3 Cs

• Completeness
• Correctness
• Consistency

12
Are data complete?

• submission by all (most) reporting facilities
• physical events observed events registered
  (how?)
• registered data collated data (how?)
• all data elements registered

13
Are data correct?

• data within normal ranges
• logical data
• existing standardised definitions used adequately
• legible handwriting
• are there any preferential end digits used?

14
Are data consistent?

• data in the similar range as this time last year
  (last reporting period)
• no large gaps
• is the correct target population being used?

15
Accuracy enhancing principles

• Training
• User-friendly collection/collation tools
• Feedback on data errors
• Feedback of analysed Information
• Use of information (and prove it!)

16
How do you detect Errors?

• general accuracy checking measures
• specific accuracy checking measures

17
General accuracy checks

• Completeness
• Proper place
• Friendly tools
• Arithmetic

18
Specific Accuracy Checks

• Time-trend consistency
• Time-trend variation
• Minimum/maximum
• Realism
• Comparison
• Parts vs whole
• Preferential end-digits

19
(No Transcript)
20
PREFERENTIAL END-DIGITS
Other examples ?
21
Practical error checking procedures

• Check completeness of the data forms
• Set minimum and maximum values
• Examine a printout of data for errors using
  general and specific error checks
• Hold an error feedback session

22
What to do if you find errors?

• Find the cause
• Correct the error
• Prevent future errors

23
Good data quality 10 steps to achieve it
24

• small, essential dataset - EDS
• clear definitions - standardized
• careful collection and collation of data good
  tools
• local analysis of data using relevant indicators
• presentation of information to all collectors
• regular feedback on both data and information
• supportive supervision - at all levels
• ongoing training and support
• discussion of information at facility team
  meetings
• monitoring use of information

25
Please remember!
26
Data, in order to be locally useful, should be

• AVAILABLE ON TIME fix dates for reporting
• AVAILABLE AT ALL LEVELS who reports to whom?
  - feedback mechanisms
• RELIABLE ACCURATE check that all data is
  correct, complete, consistent
• COMPREHENSIVE collected from all possible data
  sources
• USABLE if no action, throw data away
• COMPARABLE same numerator and denominator
  definitions used by all

27
Controlling quality with DHIS

• Maximum / minimum values
• 13-month retrospective
• Regression line
• Validation rules absolute
  statistical
• Validation reminders