METHODOLOGY FOR REGIONAL AVERAGES OF HEALTH INDICATORS

1
METHODOLOGY FOR REGIONAL AVERAGES OF HEALTH
INDICATORS

• Abhaya Indrayan
• University College of Medical Sciences Delhi

2
NEED OF REGIONAL AVERAGES

• Inter-Regional comparison
• Assess health needs of the Region as a whole
• Assess health performance of the Region
• How far is one country of the Region from the
  average
  

3

• Some countries do publish averages
• Usual method Weighted mean
• Statistically correct method but may not provide
  socially correct perspective
  
• Same averaging method may not be appropriate to
  achieve all the objectives

4
DIFFICULTIES WITH WEIGHTED MEAN

• A country with large population (consequently
  large number of births, child population, etc.)
  becomes primary determinant of the average
  
• If this country does well for an indicator, the
  entire Region on average would seem doing well.
  If this country does not do well for another
  indicator, the entire Region on average would
  seem to be doing poorly

5
In SEARO, if India does poorly for, say, female
U5MR, the weighted mean (weight is the number of
female live births), is high

• WHO 2002
• Bangladesh 73
• Bhutan 92
• DPR Korea 54
• India 95
• Indonesia 36
• Maldives 43
• Myanmar 78
• Nepal 87
• Sri Lanka 16
• Thailand 26

• Weighted mean 82
• Close to Indias value
• So, investigate alternative methods
• They may or may not work

6
WHO HQ once proposed

• Weighting by log-square adjusted population for
  resource allocation (? health needs)
  

7
Log-square adjustment is uneven
8
AVERAGES INVESTIGATED

• SIMPLE MEAN
• Easy but ignores population of the country

9

• WEIGHTED MEAN
• Statistically correct but biased for large
  country
  
• Denominator (w) needed (ELB?)

10

• SIMPLE MEDIAN
• if n is odd
• Average of and if n
  is even
  
• Ignores values

11

• HARMONIC MEAN
• Not good for many health indicators

12

• MEAN ADJUSTED FOR POPULATION RANK
• Seems good but lacks continuity

13

• MEAN ADJUSTED BY LOG-SQUARE
• Too complex and uneven
• Not smooth for small number of countries

14
IMR BY DIFFERENT METHODS

• Actual IMR (WHO 2002)
• Bangladesh 51.0
• Bhutan 60.5
• DPR Korea 21.8
• India 68.0
• Indonesia 41.4
• Maldives 21.0
• Myanmar 59.8
• Nepal 64.2
• Sri Lanka 15.4
• Thailand 21.5
• Timor Leste 82.5
• (Av. of 70 95)

• Simple mean 46.1
• Weighted mean
• (for 10 countries) 60.4
• Simple median 51.0
• Harmonic mean 33.9
• Population rank
• adjusted mean 47.8
• Population log-square
• adjusted mean 51.3

15
RECOMMENDATION OF STATISTICAL EXPERTS IN SEARO
MEETING

• No need of Regional averages, except selected
  few.
  
• HQ may be requested to supply Regional averages
  of indicators like expectation of life and U5MR
  for which HQ makes global estimates.
  

16
RECOMMENDATION OF STATISTICAL EXPERTS IN SEARO
MEETING (contd.)

• For indicators such as incidence and prevalence
  of diseases, respective technical units may be
  approached for Regional averages.
  
• If there are any more, weighted mean method
  should be used (even if it gives large weightage
  to a big country). In this case, range (Min and
  Max) and coefficient of variation should also be
  stated.