Early Detection of Malaria Epidemics

1
Early Detection of Malaria Epidemics

• Patricia M Graves PhD
• June 4 2008

2
Epidemiology vs MeteorologyTheoretical vs
Empirical
3
Eritrea Subzone stratification based on patterns
of reported monthly clinical cases, 1996-2003 -
Disease climatology
4
An Abuja target for epidemics

• Malaria epidemics should be detected within 2
  weeks of onset, and properly controlled within 2
  weeks of detection.
  
• WHO Indicator category Disease surveillance,
  epidemic preparedness and response
  
• of malaria epidemics detected within two weeks
  of onset.
  
• of malaria epidemics properly controlled within
  two weeks of onset.
  

5
Malaria epidemics indicator (de Savigny et al)

• Numerator number of epidemics detected in a
  specific geographical area (country, district,
  etc.) within 2 weeks during the last 12 months
  and for which appropriate control measures have
  been initiated within the following 2 weeks.
• Denominator number of malaria epidemics
  recorded during the last 12 months within a
  specific geographical area. Appropriate
  control measures taken means action based on
  preparedness plan of action where such control
  measures are defined or, in case they are not
  locally defined, based on global guidelines
  prepared by WHO.

6
???

• What is an epidemic?
• When did it start? (and how is that different
  from when it was detected?)
  
• When did it end?
• How do we know whether it was properly
  controlled?
  
• Assumes there will be epidemics (as denominator)
• Why not count the epidemics directly as an
  indicator? (still need to define epidemics)

7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
Epidemic definition
11
Epidemic definition

• The occurrence in a community or region of cases
  of an illness, clearly in excess of normal
  expectancy.
  

12
Epidemic definition

• The occurrence in a community or region of cases
  of an illness, clearly in excess of normal
  expectancy.
  
• Need to know

13
Epidemic definition

• The occurrence in a community or region of cases
  of an illness, clearly in excess of normal
  expectancy.
  
• In order to define, need to know
• Geographic area
• Time period
• What is normal for that area / time period
• What is regarded as excess

14
Epidemic definition

• Useful if it triggers action
• Calling something an epidemic is more of a
  political and pragmatic decision, rather than a
  scientific one
  
• Can be useful retrospectively for reporting on
  number of epidemics as measure of control
  
• Identifying abnormal events and their causes may
  help to predict/prevent them in future

15
Unit of analysis

• Based on availability of data in space and time
• Should correspond geographically with planning
  units
  
• Smaller (area) and shorter (time) usually better
• Results will differ depending on units used

16
Eritrea zones and subzones
17
Eritrea reported clinical malaria cases by
month, 1996 to 2003
18
Eritrea reported clinical malaria cases by
month, 1996 to 2003
19
Eritrea reported clinical malaria cases by
month, 1996 to 2003
20
(No Transcript)
21
(No Transcript)
22
(No Transcript)
23
Thresholds To decide what is abnormal

• Third Quartile
• Mean plus 1 or 2 SD (Cullen)
• Moving average plus 1 or 2 SD
• CSUM (cumulative sum)
• Should all be based on at least 5 years data.
• What years do we use for normal?

24
Third Quartile method

• Calculates the threshold as the third quartile
  value of the number of cases per month or week
  for the last five years.
  
• This means that roughly three quarters of the
  time, the case numbers will be below the
  threshold
  
• Recommended by WHO as easy to calculate manually

25
Mean 1 or 2 SDCullen method

• Mean 2 standard deviations of the five or more
  previous years numbers of cases for that month
  or week.
  
• This means that roughly 97.5 of the time, the
  number of cases will be below the threshold.
  
• Modification uses mean plus 1 standard deviation

26
Moving average

• Based on a moving average of current month plus
  the preceding and following months
  
• Avoids false alerts due to variation in onset
  month of the malaria season.
  
• May cause delay in recognising an abnormal rise
  in cases
  
• Used with either 1 or 2 standard deviations
• More difficult to calculate

27
C-SUM

• Cumulative sum of cases for that year
• Start of year can be set at any point
  (conventionally Jan)
  

28
Epidemic characteristics

• Sometimes the start and end of an epidemic is
  clear by all thresholds
  
• Sometimes it is not
• What is defined as an epidemic depends on the
  base years you choose as well as on what
  threshold you choose.
  
• There is no objective standard

29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
Practical Task

• The Eritrea MOH claims that it has controlled
  malaria very well in recent years. To see if you
  agree
  
• Estimate how many malaria epidemics occurred in
  Eritrea during 2003, using the baseline data by
  subzoba/month for 1996-2002, under 3 scenarios
  
• 1) you have no power supply
• 2) you have power and a computer with Excel but
  no internet
  
• 3) you have power, a computer and can access the
  IRI data library.
  
• You will be assigned subzobas at random and we
  will compare results at the end

35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
(No Transcript)
41
(No Transcript)
42
(No Transcript)
43
(No Transcript)