El Nio and communicable disease: Examples from Oceania

1
El Niño and communicable disease Examples from
Oceania

• Simon Hales
• Wednesday 1st October 2003

2
Examples

• Ciguatera
• Dengue
• Diarrhoeal illness

3
Ciguatera

• Reports of fish poisoning in Pacific Islands
  (1973-1994) were obtained from the Secretariat of
  the Pacific Community (SPC)
• SST time series were estimated using the
  International Research Institute for Climate
  Prediction (IRI) data library INGRID
• http//ingrid.ldgo.columbia.edu/

4
(No Transcript)
5
Analysis

• Pearson correlations were calculated using
  January-December annual averages of SOI, SST and
  fish poisoning reports
• Average fish poisoning reporting rates were
  calculated based on 1994 populations

6
Correlation coefficients SOI vs fish poisoning
7
Correlation coefficients SOI vs local SST
8
(No Transcript)
9
(No Transcript)
10
Summary

• Annual reports of fish poisoning were found to be
  correlated with El Niño events
• Positive correlations between the annual
  incidence of fish poisoning and local SST
  anomalies were confined to an Easterly group of
  islands which experience warming during El Niño
  conditions

11
Fish poisoning reports per 1000
12
Dengue fever in the Pacific Islands regional
scale

• Correlate annual number of epidemics of dengue
  in the Pacific Islands (from literature
  searches) with SOI (1970-1998 data)
• Find that epidemics tend to be initiated in La
  Niña years...

13
(No Transcript)
14
(No Transcript)
15
Dengue fever in the Pacific Islands national
scale

• Using monthly data from SPC (1973-1994),
  calculate correlations between annual averages of
  SOI, dengue reports, and local temperature and
  rainfall
• Find positive correlations between SOI and dengue
  in 10 countries

16

• In half of these islands, there were positive
  correlations between SOI and estimates of local
  temperature and rainfall
• Also temporal correlations between monthly case
  numbers in two groups of islands

17
Diarrhoeal illness

• Monthly reports of diarrhoea in infants for Fiji
  (1978-1989, SPC data)
• National climate estimates as before
• Using Poisson regression models, investigate
  effect of 1-month lag variables (after Checkley
  et al, 2000)

18
Regression analysis

• Parameterise rainfall using two variables
  indicating departure from the mean (5 units),
  rationale as follows
• Remove seasonal effects from rainfall
• Plot deseasonalised residuals against rainfall

19
(No Transcript)
20
(No Transcript)
21
Regression results

• Find increased diarrhoea at extremes of rainfall
  in the same month (as in the figure) and with
  increasing temperature in the same month
• Similar effects at 1-month lag for temperature
  and low rainfall (but no effect of increased
  rainfall)
• Using the model to make out-of-sample predictions
  for 1993-1998, a period which includes several
  ENSO events, the model predicts these data
  quite well (r20.5)

22
Weaknesses

• Crude estimates of climate
• Sparse data
• The communicable disease data is of uncertain
  quality, (may not be disease- specific)
• (But note that biological mechanisms not be
  disease-specific either)