Difference between revisions of "Health impact of radon in Europe"

Revision as of 09:40, 2 February 2011

Scope

Purpose

What are the health impacts of radon in the indoor air of residential buildings in Europe?

Boundaries etc

Boundaries, scenarios, intended users, and participants are the same as in the Mega case study.

Definition

Decision variables

• Building policies in Europe

Other variables

• Population of Europe by Country
• WHO mortality data
• heande:HI:Air exchange rate for European residences
• Radon concentrations in European residences
• ERF for long-term indoor exposure to radon and lung cancer

Indicators

• Lung cancer cases due to radon in Europe

Analyses

Indoor radon case study

• Estimates health impacts of radon in indoor air of residences.
• Nation-wide radon concentration distributions were obtained from EnVIE project and UNSCEAR 2000 report.
• United Kingdom, Czech Republic, and Slovenia were rejected because of lack of data.
• Exposure-response function was 16 % increase of lung cancer incidence due to 100 Bq/m³ increase of radon (Darby 2004 and 2005).
• The same lung cancer background rate was assumed for the whole Europe: 58.2 cases/100000 person-years (Globocan 2008).
• Linear no-threshold ERF assumed for the whole population in each country.
• The model development, data storage, and model runs were all performed in Opasnet using R software and Opasnet Base.
• The main page of the sub-assessment is http://en.opasnet.org/w/Health_impact_of_radon_in_Europe

Outcomes of interest of radon sub-assessment

• Main health impact: number of lung cancer cases attributable to radon.
• Preliminary estimate of DALYs attributable to radon (only draft estimates for duration of disease).
  • Disability weight 0.146 (from WHO) used for disease duration.
  • Duration of disease estimated at 2-36 months; life-years lost due to lung cancer death estimated at 1-15 years.
• Preliminary estimate of monetary impact was obtained indirectly by converting DALYs into euros; other cost types were ignored.
  • One DALY estimated to be worth 30-60 k€.
• Methodological outcome: proof of concept for running assessment models via open internet interface.

R code

• This code features R functions described on Opasnet Base Connection for R and Operating intelligently with multidimensional arrays in R.

library(ggplot2)
cancer <- op_baseGetData("opasnet_base", "Op_en4715", exclude = 48823)
array <- DataframeToArray(cancer)
array <- array[,,,c(2,1,3,4),,]
##### Cases #####
means <- apply(array, c(2,3,4), mean, na.rm=TRUE)
means <- apply(means, c(2,3), sum, na.rm=TRUE)
plot1 <- as.data.frame(as.table(means))
plot1 <- ggplot(plot1[plot1[,"Freq"]!=0,], aes(Year, weight=Freq, fill=policy)) + geom_bar(position="dodge") + 
scale_x_discrete("Year") + scale_y_continuous("Cases")
plot1
ci <- apply(apply(array, c(1,3,4), sum, na.rm=TRUE), c(2,3), quantile, probs=c(0.025,0.975))
final1 <- means
final1[,] <- paste(round(means), " (", round(ci[1,,]), "-", round(ci[2,,]), ")", sep="")
final1[c(2:4,7:8,11:12)] <- NA
final1
##### DALYs #####
DALY <- array(NA, dim = c(dim(array), 3), dimnames = dimnames(array))
DALY[,,,,1] <- array
DALY[,,,,2] <- runif(dim(array)[1],2,36)/12
DALY[,,,,3] <- runif(dim(array)[1],1,15)
DALY <- DALY[,,,,1]*0.146*DALY[,,,,2]+DALY[,,,,1]*DALY[,,,,3]
means <- apply(DALY, c(2,3,4), mean, na.rm=TRUE)
means <- apply(means, c(2,3), sum, na.rm=TRUE)
plot2 <- as.data.frame(as.table(means))
plot2 <- ggplot(plot2[plot2[,"Freq"]!=0,], aes(Year, weight=Freq, fill=policy)) + geom_bar(position="dodge") + 
scale_x_discrete("Year") + scale_y_continuous("DALYs")
plot2
ci <- apply(apply(DALY, c(1,3,4), sum, na.rm=TRUE), c(2,3), quantile, probs=c(0.025,0.975))
final2 <- means
final2[,] <- paste(round(means), " (", round(ci[1,,]), "-", round(ci[2,,]), ")", sep="")
final2[c(2:4,7:8,11:12)] <- NA
final2
##### Cost #####
mpdaly <- op_baseGetData("opasnet_base", "Op_en4858")
cost <- IntArray(mpdaly, DALY, "DALYs")
cost <- data.frame(cost[,c("obs","Country","policy","Year")], Result=cost[,"Result"]*cost[,"DALYs"])
cost <- DataframeToArray(cost)
cost <- cost[,,c(2,1,3,4),]
means <- apply(cost, c(2,3,4), mean, na.rm=TRUE)
means <- apply(means, c(2,3), sum, na.rm=TRUE)/10^6
plot3 <- as.data.frame(as.table(means))
plot3 <- ggplot(plot3[plot3[,"Freq"]!=0,], aes(Year, weight=Freq, fill=policy)) + geom_bar(position="dodge") + 
scale_x_discrete("Year") + scale_y_continuous("Cost (M€)")
plot3
ci <- apply(apply(cost, c(1,3,4), sum, na.rm=TRUE), c(2,3), quantile, probs=c(0.025,0.975))/10^6
final3 <- means
final3[,] <- paste(round(means), " (", round(ci[1,,]), "-", round(ci[2,,]), ")", sep="")
final3[c(2:4,7:8,11:12)] <- NA
final3
##### Probability density plot #####
costdf <- as.data.frame(as.table(apply(cost, c(1,3,4), sum)/1e9))
costdf <- costdf[is.na(costdf[,"Freq"])==FALSE,]
plot4 <- ggplot(costdf, aes(x=Freq, y=..density.., fill=policy)) + geom_density(alpha=0.2, adjust=4) + 
scale_x_continuous(expression("Cost ("*10^9*"€)"), limits = c(-50,300)) + scale_y_continuous("Density") + facet_wrap(~Year)
plot4
##### Expected Value of Perfect Information #####
evpi <- (apply(apply(cost, c(2,3,4), mean, na.rm=TRUE), c(1,3), min, na.rm=TRUE) - apply(apply(cost, c(1,2,4), min, 
na.rm=TRUE), c(2,3), mean, na.rm=TRUE))/1e6
plot5 <- as.data.frame(as.table(apply(evpi, 2, sum)))
plot5 <- ggplot(plot5, aes(Var1, weight=Freq)) + geom_bar(position="dodge") + 
scale_x_discrete("Year") + scale_y_continuous("Value of perfect information (M€)")
plot5
##### Expected Value of Partial Perfect Information #####
#Same as that of perfect information, because of only one decision variable
ae <- op_baseGetData("opasnet_base", "Erac2499")
aer <- DataframeToArray(ae)
aer <- aer[,,c(2,1,4,5),]
dropnonmax <- function(x) {
	x[x<max(x, na.rm = TRUE)] <- NA
	return(x)
}
aer <- apply(aer, c(1,2,4), dropnonmax)
aer <- as.data.frame(as.table(aer))
aer <- aer[,c(2,3,1,4,5)]
colnames(aer)[3] <- "policy"
aer <- aer[is.na(aer[,"Freq"])==FALSE,]
aer <- IntArray(aer, cost, "Cost")
aer <- DataframeToArray(aer[,c("obs","Country","Year","Cost")],"Cost")
test2 <- (apply(apply(cost, c(2,3,4), mean, na.rm=TRUE), c(1,3), min, na.rm=TRUE) - apply(aer, c(2,3), mean))/1e6
plot6 <- as.data.frame(as.table(apply(test2, 2, sum)))
plot6 <- ggplot(plot6, aes(Var1, weight=Freq)) + geom_bar(position="dodge") + 
scale_x_discrete("Year") + scale_y_continuous("Value of perfect information (M€)")
plot6
test2==evpi #Test if the values are the same

Result

Results

Lung cancer cases due to radon in Europe: Show results

• Results for the Biomass scenario are wrong and the scenario is perhaps irrelevant because biomass usage does not affect air exchange rates which this assessment is concerned with, so it should be ignored.

Conclusions

See also

• Radon
• ERF for long-term indoor exposure to radon and lung cancer
• An Overview of Radon Surveys in Europe. Joint Research Centre, 2005. ISBN 92-79-01066-2
• In Heande (password-protected)
• heande:HI:Radon Indoor Air Case
  • Indoor air quality & its impact on man
  • Radon
  • Indoor air
  • Radon sisäilma annos-vaste
  • Radon sisäilma altistus Suomi
  • Radon ja pitkäikäiset nuklidit porakaivo
  • Radon ja pitkäikäiset nuklidit porakaivo
  • Radon ja pitkäikäiset nuklidit porakaivovedessä

Keywords

Radon, indoor air, lung cancer, Europe

References

Related files

<mfanonymousfilelist></mfanonymousfilelist>