Forecasting Malaria Incidence in Botswana Using the DEMETER Data

1
Forecasting Malaria Incidencein Botswana Using
the DEMETER Data

• Simon Mason
• International Research Institute for Climate and
  Society
• The Earth Institute of Columbia University
• ECMWF Users Meeting
• Reading, England, 15 17 June, 2005

2
Malaria in Africa

• Climate controls on
• malaria in Africa
• Temperature highland malaria
• Precipitation desert-fringe malaria

Botswana
3
Malaria in Botswana
Botswana straddles the southern margins of
malaria transmission in sub-Saharan Africa.
The incidence of malaria varies considerably from
district to district showing a general
decreasing north-south pattern from more stable
to less stable malaria. In Botswana the incidence
of malaria varies considerably from year to year
and as such malaria is considered to be
unstable and prone to periodic epidemics.
4
Malaria in Botswana

• Constraints to studying impact of climate
  variability on malaria incidence
• Inadequate surveillance but in Botswana, malaria
  is a notifiable disease.
• Lack of confirmed case data but
  laboratory-confirmed cases are recorded in
  Botswana.
• Short time-series for analysis but Botswana has
  annual records from 1982.
• Many confounding factors dates of changes in
  drug policies in Botswana are known.

5
Malaria in Botswana
The disease is highly seasonal and follows the
rainy season with a lag of about 2 months
6
Malaria in Botswana
Trends in malaria incidence may result from
trends in climate but mostly indicate changes in
vulnerability, e.g. drug or insecticide
resistance, declining control services, etc. The
long term increasing trend 19821996 ends when
revisions to national control policy and practice
occurred in 1997 (new drugs, new insecticide,
revitalized programme.
7
Malaria in Africa
Recent increases in incidence have been
attributed to global warming, but they are much
more likely a result of increases in drug
resistance, and declines in control
activities. Chloroquine resistance was first
reported in East Africa in 1979 since spread
throughout Africa
8
Malaria in Botswana

• Other factors driving trend and/or interannual
  variability
• Intrinsic population dynamics
• Access to health facilities/reporting
• Drug sensitivity
• Insecticide sensitivity
• Seasonal and long term migration
• HIV

9
Evidence for Efficacy of Policy Change
The ratio of confirmed to unconfirmed malaria
cases increases markedly after 1996.
10
Detrended Malaria Index
The skewness of the log-incidence is small (-0.3)
compared to that for the raw incidence
(1.5). Detrending accounts for the policy
change. Climate-related trends are not
removed. High and low years are defined by the
upper and lower quartiles.
11
Relationship to Observed Rainfall
Malaria incidence in Botswana is strongly related
to rainfall variability during the peak rainfall
season December February. The relationship is
non-linear incidence peaks at about 4 mm per day.
12
Relationship to Observed Rainfall
Correlations with the quadratic rainfall index
are strong
13
Relationship to Observed Rainfall
ROC or low incidence years
14
DEMETER Forecasts
Observations
Forecasts
High malaria years
Low malaria years
15
DEMETER Forecasts
16
DEMETER Forecasts
Ensemble-mean forecasts compared to incidence.
17
DEMETER Forecasts
ROC
At shorter lead-times the forecasts improve.
18
Malaria Early Warning System
19
Using Seasonal Climate Forecasts for Malaria
Control Planning
The First Malaria Epidemic Outlook Forum for SADC
was held in September 2004, and an updated
rainfall forecast was provided in December 2004.
20
Summary

• The use of seasonal climate forecasts for malaria
  in Southern Africa is demand led.
• But seasonal forecasts form only part of the
  inputs to a malaria early warning system.
• Institutions are already organised and policies
  in place for the use of seasonal forecasts.
• Causal relationship between climate and malaria
  known.
• There is a strong influence of seasonal rainfall
  on detrended malaria incidence.
• There is high predictability of detrended malaria
  incidence using DEMETER forecasts.