1. Previous Host Immunity Affected Clustering of Influenza in Sado island, Japan 2. Prediction of Onset Timing of Seasonal Influenza Epidemic, Japan Yugo Shobugawa, MD, PhD yugo@ med.niigata-u.ac.jp Post Doctoral Fellow, School of Public Health, Loma Linda University, CA Assistant Professor, Department of Public Health, Niigata University, Japan
46
Embed
Prediction of Onset Timing of Seasonal Influenza Epidemic ... · PDF filePrediction of Onset Timing of Seasonal Influenza Epidemic, Japan Yugo Shobugawa, MD, PhD yugo@ med.niigata-u.ac.jp
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1. Previous Host Immunity Affected Clustering of Influenza in Sado island, Japan
2. Prediction of Onset Timing of Seasonal Influenza Epidemic, Japan
Yugo Shobugawa, MD, PhD yugo@ med.niigata-u.ac.jp
Post Doctoral Fellow, School of Public Health, Loma Linda University, CA
Assistant Professor, Department of Public Health, Niigata University, Japan
Background
During seasonal influenza epidemics, 5-15% of the population are affected with infections.
Worldwide epidemics result in three to five million cases of severe illness and even death in each year.
Reducing morbidity and mortality with influenza is major issue in every country from the public health perspective.
Case 1.
General information, Sado island, Japan
Total population: 63,313 in May, 2011 • 0-14: 11.8% •15-64: 53.0% •65<: 35.3%
Total area: 855 square kilometers. Rural and mountain forest area covers a large part of the island. One hour by high-speed boat, 150 minutes by ferry from the main land.
Rural and mountain forest area covers a large part of the island.
Landscape in Sado island
Coast area
Sight seeing spot especially in summer “Tarai-bune” (tub-turned boat) which is a traditional
Japanese fishing boat used for catching Abalone and other mollusks.
Study flow
Step 1: Basic statistics of Flu outbreaks
Step 2: Cluster detection
Step 3: Exploring cause of clustering
Step 1: Basic statistics of Flu outbreaks Data: individual patients data (age, sex, date
of onset, zip code of resident area, and diagnosis: Flu A or B by commercial rapid test kit) were collected by physicians and pediatricians in hospitals or clinics in winter seasons, 2005-2009.
More than 80% of all the hospitals/clinics in the island cooperated with the study.
Outbreak description
Outbreaks of Flu A, no. of cases (median age, range) 2005/2006 656 (28, 0-95 y/o)
2006/2007 1,884 (16, 0-97 y/o)
2007/2008 140 (16, 0-62 y/o)
2008/2009 2,070 (14, 0-95 y/o)
Outbreaks of Flu B , no. of cases (median age, range) 2007/2008 299 (7, 0-92 y/o)
2008/2009 796 (7, 0-77 y/o)
Flu epidemic, 2005-2009 (weekly no. of cases)
2006 2007 2008 2009
0
100
200
300
w0552 w0626 w0652 w0726 w0752 w0826 w0852 w0926
Influenza A cases
Influenza B cases
Wee
kly
no. o
f Flu
cas
es
2006 2007 2008 2009
Step 2: cluster detection
Geocoding: by zip code of patients’ resident (Sado island was divided into 218 zip code areas).
Pure space clusters in each winter were detected by Poisson Model on SaTScan software. Population and age adjusted by six age subsets:
Age 0-4, 5-14, 15-19, 20-39, 40-64, >65 Maximum cluster size was set at 10% of the
population.
Flu cluster areas in each season
Locations of clusters varied by season and type.
Why?
What factor(s) affect for clustering of Flu?
>> next step
Step 3: Exploring cause(s) of clustering Binary logistic regression analysis
Dependent variable: Cluster or not? Explanatory variables:
○ Population density ○ No. of family members in each household ○ Flu vaccination rate in school aged children ○ Incident rate of Flu in former 2 seasons in school aged children
Flu vaccination rate and incident rate in school aged children were calculated by using another data set collected from all the elementary school in the island.
Calculation has been done by using SPSS software.
Significant factors for Flu clustering 2005-2006 season, Flu A
Discussions High population density and greater number of
people in a household can get more frequent chance to be infected because influenza virus can be transmitted from human to human by droplet.
Previous infection with influenza may reduce chance to be infected with influenza which reason is possibly explained by pre-existing host immunity.
For areas where Flu outbreak did not occur largely former years, preventive activity should be enforced from the public health perspective (cough etiquette, hand hygiene, monitoring and so on).
Limitations
Only data of school aged children in vaccination and incident rate is available.
No laboratory confirmation has been done, but Flu was diagnosed by commercial rapid test kit (which has still high sensitivity and high specificity).
Conclusions
Locations of Flu cluster varied by season and type of Flu.
High density of population and more family member/household promoted Flu clustering.
High incident rate of Flu in school aged children in former seasons suppressed Flu clustering.
From these results, we can plan any effective interventions for outbreak prevention before epidemic.
Spread manner of Flu has still not been elucidated from this study.
(http://idsc.nih.go.jp/index.html) ILI cases were reported by clinician in sentinel
hospital/clinic. Average ILI cases in each prefecture were used for analysis.
ILI is defined by sudden onset of fever >38°C, respiratory symptoms, and myalgia.
Total of 11.1 million ILI cases from 1999 to 2009 for 46 prefectures were enrolled.
How to count ILI cases e.g. in certain prefecture
How to count ILI cases
Sum of these ILI cases from the all sentinels divided by no of sentinels was reported weekly.
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
®v
6
8
03
5
795
1
11
2115
12
H Sentinel hospital/clinic
In this prefecture, Sum of the ILI = 153 No. of sentinels = 13 Av. no. of ILI cases = 153/13 = 11.8
Japan ILI
>> movie
Spatial indicator for representing of compactness of distribution
Weighted Standard Distance (WSD)
(X, Y ) indicates the mean center for all the features in the study area. It takes the squared difference in coordinate values between each point and the weighted mean center and multiplies it by the weight, sums the weighted differences, then divides the summed values by the sum of the weights. X and Y are the x- and y- coordinates of the weighted mean center.
WSD: Compactness of distributions
2nd week, 2002 19th week, 2002
Clustered Dispersed
Japan ILI with WSD
>> movie
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
45.0
50.0
0
100
200
300
400
500
600
700
800
Apr-
99
Jul-9
9
Oct
-99
Jan-
00
Apr-
00
Jul-0
0
Oct
-00
Jan-
01
Apr-
01
Jul-0
1
Oct
-01
Jan-
02
Apr-
02
Jul-0
2
Oct
-02
Jan-
03
Apr-
03
Jul-0
3
Oct
-03
Jan-
04
Apr-
04
Jul-0
4
Oct
-04
Jan -
05
Apr-
05
Jul-0
5
Oct
-05
Jan-
06
Apr-
06
Jul-0
6
Oct
-06
Jan-
07
Apr-
07
Jul-0
7
Oct
-07
Jan-
08
Apr-
08
Jul-0
8
Oct
-08
Jan-
09
Apr-
09
Aver
age
wee
kly
repo
rted
ILI c
ases
per
sen
tinel
Wei
ghte
d st
anda
rd d
ista
nce,
km
Trend of WSD and ILI cases, 1999-2009 Weekly reported ILI cases Standard distance
Gap between minimum standard distance week and initiation week for influenza outbreak
Influenza outbreak warning is stated by IDSC (Infectious Diseases Surveillance Center) on the week, ILI cases reached 10.0 or greater in Japan.
WSD trend
ILI trend
Table 1. Relationship between gap duration and the prevelance of influenza virus type/subtype
* The week when ILI cases reached 10.0 or greater for the first time in each season through all Japan.
Prevalence of type or subtype of influenza virusSeasons Lowest WSD week ILI>10.0 week* Gap duration
(weeks)
Relationship between gap duration and the prevalence of influenza virus type/subtype
0%
20%
40%
60%
80%
100%
0 5 10 15
A/H3N2
A/H1N1
B
Linear (A/H3N2)
Linear (A/H1N1)
Linear (B)
R = -0.69
Gap duration, week
Prop
ortio
n of
circ
ulat
ing
influ
enza
type
/sub
type
Result summary
Standard distance weighted by average no. of ILI cases from sentinels decreased to a minimal value before the each peak.
Duration gap between the lowest WSD week and week ILI reached 10.0 or greater varied by season. However, the gap showed significant negative correlation with proportion of prevalence of A/H3N2 virus.
Discussion and Conclusion
Weighted standard distance will be a possible indicator to predict timing of onset of influenza epidemic.
A/H3N2 virus showed faster spreading pattern rather than in A/H1N1 and B.
Limitations
Only Japanese data has been analyzed. Application this method to continental countries is needed.