View source for Domestic fish consumption of the general population in Finland

[[Category:Exposures]]
[[Category:Fish]]
{{variable|moderator=Jouni|stub=Yes}}

==Question==

What is the average fish consumption of individuals in the general Finnish population? Imported fish may be considered. Sea and freshwater species may be considered separately. {{disclink|Imported fish?}} 

==Answer==

== Rationale ==

=== Data ===

==== EFSA food database 2016 ====

The EFSA Comprehensive European Food Consumption Database consists of data from many EU countries, collected between 1997 and 2012. The database was published 2016.<ref name="efsa2016"> EFSA Comprehensive European Food Consumption Database. (2016) [https://data.europa.eu/euodp/en/data/dataset/the-efsa-comprehensive-european-food-consumption-database]</ref> The Finnish data comes from the following studies:
* Diabetes Prediction and Prevention Nutrition Study (DIPP) 2001-2009
* Diabetes Prediction and Prevention Nutrition Study (DIPP) 2003 -2006
* National FINDIET 2012 Survey
* National Findiet Surveys
* Nutrition and wellbeing of secondary school pupils
* Special Turku Coronary Risk Factor Intervention Project

The most useful dataset is about the chronic consumption among consumers only, measured as g/d [http://www.efsa.europa.eu/sites/default/files/chronicgdayconsumers.xlsx]. Another dataset is about the whole population including those who do not consume the product [http://www.efsa.europa.eu/sites/default/files/chronicgdaytotpop.xlsx].

* {{#l:EFSA_fish_consumption_consumers_fi_g_per_day.zip}} (see production code below)
* {{#l:EFSA_fish_consumption_consumers_fi_g_per_day.csv}}

==== 2007 study ====

Fish consumption by the population <ref name="Tero"> Tero Hirvonen. 2007. http://www.pyrkilo.fi/beneris/index.php/Image:Food_consumption_FINLAND.xls </ref> for twelve various fish species commonly consumed in Finland as normal distributions (weighed mean, weighed SD) and additionally indexed by age categories and gender. 

{| {{prettytable}}
|+ '''Fish consumption in the subpopulation of fish users (g/d).
! Fish species || Consumption by 0-2 years old || Consumption by 2-18 years old || Consumption by 18-55 years old
|----
|| Baltic herring (Clupea harengus membras) || 5.717 || 11.55 || 29.61
|----
|| Herring (Clupea harengus) || 0 || 3.333 || 19.84
|----
|| Vendace (Coregonus albula) || 7.858 || 8.884 || 37.08
|----
|| Whitefish (Coregonus lavaretus) || 10.05 || 19.4 || 52.57
|----
|| Pike (Esox lucius) || 3.636 || 3.684 || 26.13
|----
|| Rainbow trout (Onchorhynchus mykiss) || 9.686 || 17.59 || 66.31
|----
|| Shrimp (Pandalus sp.) || 0.663 || 4.328 || 19.94
|----
|| Perch (Perca fluviatilis) || 5.807 || 14.64 || 20.31
|----
|| Saithe (Pollachius virens) || 10.09 || 16.48 || 54.51
|----
|| Atlantic salmon (Salmo salar) || 13.82 || 18.79 || 24.95
|----
|| Pike-perch (Sander lucioperca) || 18.65 || 22.52 || 0
|----
|| Tuna (Thunnus thynnus) || 7.981 || 7.733 || 47.43 
|----
|}

<anacode>5.717	5.897	11.55	12.96	29.61	39.02
0.000e+000	0.000e+000	3.333	2.47	19.84	15.22
7.858	7.6	8.884	10.2	37.08	49.41
10.05	8.227	19.4	23.57	52.57	64.87
3.636	2.402	3.684	3.106	26.13	32.6
9.686	7.075	17.59	14.06	66.31	48.59
0.663	1.339	4.328	4.142	19.94	26.41
5.807	3.15	14.64	29.2	20.31	31.37
10.09	7.458	16.48	14.44	54.51	32.06
13.82	14.23	18.79	19.34	24.95	19.84
18.65	18.83	22.52	17.77	0.000e+000
7.981	6.591	7.733	8.539	47.43	39.26</anacode> 

==== Fisheries statistics 2006 ====

There are two kinds of data. Actual consumption data and fishery catch data.<ref>Kalatalous tilastoina 2006 RKTL, page 23</ref> <ref name="RKTL">Kalatalous tilastoina 2006 RKTL, pages 3 and 9</ref> <ref>RKTL 2006(a) Estimates for filleting factors</ref> <ref>RKTL 2006(b) Estimates for factors of edible proportions</ref> 

For the most important species (herring, salmon) the actual consumption estimates are available <ref>Methyl mercury: Kalatalous tilastoina 2006 RKTL, page 23</ref>, but for the most of the species consumption estimates are calculated from fishery catch. <ref>Methyl mercury: Kalatalous tilastoina 2006 RKTL, pages 3 and 9 </ref> For those species, fishery catch is first transformed to filleted fish weight <ref> Methyl mercury: RKTL 2006(a) Estimates for filleting factors</ref> and further with actual edible proportions <ref> Methyl mercury: RKTL 2006(b) Estimates for factors of edible proportions </ref>to fish consumption. 

RKTL has data on overall fish consumption in Finland. This model yields a very similar estimate. {{disclink|Loss of fish during process?}} 

{| {{prettytable}}
|+ '''Fish consumption as filleted weight (g/d) (or kg/a?)
! Fish species || Fish consumption 
|----
|| Farmed salmon (sea+inland) || 3.83
|----
|| Wild salmon || 0.23
|----
|| Herring(sea) || 1.92
|----
|| White fish(sea) || 0.3
|----
|| Sprat(sea) || 0.11
|----
|| Perch(sea) || 0.44
|----
|| Flounder(sea) || 0.01
|----
|| Pike-perch(sea) || 0.2
|----
|| Bream(sea) || 0.07
|----
|| Pike(sea) || 0.35
|----
|| Vendace(sea) || 0.06
|----
|| Burbot(sea) || 0.02
|----
|| Wild salmon(inland) || 0.45
|----
|| White fish(inland) || 0.48
|----
|| Perch(inland) || 1.25
|----
|| Pike-perch(inland) || 0.27
|----
|| Bream(inland) || 0.31
|----
|| Pike(inland) || 1.53
|----
|| Vendace(inland) || 1.38
|----
|| Burbot(inland) || 0.16
|----
|}

==== Other data ====

*[http://www.norden.org/pub/sk/showpub.asp?pubnr=2008:531&lang=1 TemaNord report on PCDD/F and PCB in Nordic countries]

The aim is to find probability distributions of daily consumption of various sea and freshwater fish species in six subgroups of the general Finnish population. The subgroups of interest are males and females in two age categories 0-2yr and 2-18yr, and adults (males and females combined) aged 18-55yr and 55yr+. The data used to derive these distributions comes from <ref name="Tero" /> where the summary statistics (mean, sd and percentiles: 5th, 25th, 50th, 75th, 95th) of daily consumption of twelve fish species among fish users and information on the percentage of users are given. The data is provided separately for males and females in the following age groups: 1yr, 3yr, 6-9yr, 25-34yr, 35-44yr, 45-54yr and 55-64yr. There is a single data point of daily intake of whitefish by women aged 25-34yr and 35-44yr, and by males aged 45-54yr. This information is not enough to create probability distribution. Thus, it was assumed that in each of these three sex/age groups the intake of whitefish is the same as intake of vendace. First, linear interpolation was used to approximate values of 5th, 25th, 50th, 75th, 95th percentiles of daily fish consumption among consumers in missing age/sex groups. For example, the i-th percentile of daily consumption of species j in a group of 2 year old males was found by linear interpolation between points corresponding to i-th percentiles of daily consumption of species j in groups of 1 year old and 3 year old males (for interpolation the mid age of every subgroup was used). The same technique was applied to age groups 4-5yr, 10-13yr, 14-17yr, 18-24yr. The i-th percentile of daily fish consumption by 0yr old children (users only) was approximated by the arithmetic mean of values of i-th percentiles for months 7 through 11 obtained by linear interpolation between i-th percentiles of fish consumption at age of 6 months and at age of 1yr (mid age of 1.5yr). Since young children don’t eat fish until they are 6 months old the fish consumption and the percentage of fish users at age of 6 months is zero. Moreover, it was assumed that the fish consumption and the percentage of fish consumers in age group 65-74yr and in age group 75yr+ is the same as in group 55-64yr. In the similar way the percentage of users of particular fish species in every missing age/sex group was estimated. Then, the available data (percentiles, mean and variance) was used to reconstruct the probability distributions of daily fish consumption among users in every age/sex group. The following procedure was used for that purpose. First, for a given fish species, age and gender the starting/initial cumulative distribution function (CDF) of daily consumption was constructed by linear interpolation between known percentiles. In general, this distribution will not reproduce the mean and sd as provided by the data. Therefore, the constrained optimization was further used to find the CDF that satisfies the percentiles given and whose mean and variance are as close as possible to the mean and variance from the data. For age groups for which the mean and variance were unknown only the first step of this procedure was applied. After that, information about the percentage of users was used to determine the probability distribution of daily consumption of particular fish species among all individuals in a certain sex/age group (users and non-users). The resulting distributions were further grouped and combined (first by age and then, only in case of adults, by gender; population data can be found here <ref> Population of Finland years 1980 - 2007. http://en.opasnet.org/w/Image:Population_of_Finland_years_1980-2007.xls </ref>) to derive the fish consumption distributions in six target age/sex groups. Finally, since in <ref name="Tero" /> the fish habitat was not differentiated the species specific data on sea and inland intake proportions (assumed to be the same as proportion of sea and inland fishery catch and extracted from <ref name="RKTL" />) was used to estimate consumption of different sea and inland species by target subgroups. 

Unit: g/d 

=== Calculations ===

==== EFSA food database 2016 ====

EFSA data was fetched and preprocessed. Ovariable amount was created to be used in health impact assessments. The preprocess code does not work at Opasnet.

You can find both codes at [https://github.com/jtuomist/pfas/blob/main/EFSA_amount_preprocess.R Github].

<rcode label="EFSA data preprocess (run on own computer)">
# This is code Op_en1836/ on page [[Domestic fish consummption of the general population in Finland]]

# This code is for archival purposes only. It cannot be run at Opasnet.

# http://www.efsa.europa.eu/sites/default/files/chronicgdaytotpop.xlsx
efsa1 <- read_xlsx("data/chronicgdaytotpop.xlsx", sheet="L1_All_subjects_g_day", skip=2) # total population
efsa1 <- efsa1[efsa1$Country=="Finland" & grepl("Fish", efsa1$`Foodex L1`),]

efsa3 <- read_xlsx("data/chronicgdaytotpop.xlsx", sheet="L3_All_subjects_g_day", range="R3C1:R131491C16")
efsa3 <- efsa3[efsa3$Country=="Finland" & grepl("Fish", efsa3$`Foodex L2`),]

#tmp <- aggregate(efsa3$Mean, by=efsa3["Foodex L3"], FUN=sum)
#tmp <- tmp[order(-tmp$x),]
#                                            Foodex L3           x
#38                      Salmon and trout (Salmo spp.) 74.26157419
#18                                          Fish meat 62.54528672
#49                                     Tuna (Thunnus) 10.16391067
#50                              Whitefish (Coregonus)  9.87886514
#30                                   Herring (Clupea)  9.22486875
#12                             Fish and potatoes meal  7.58100000
#35                                      Perch (Perca)  4.43950031
#23                                      Fish products  2.38921114
#8                        Cod and whiting (Gadus spp.)  2.00739494
#16                                         Fish balls  1.87321838
#17                                       Fish fingers  1.87172960
#24                                           Fish roe  1.09550296
#26                      Flounder (Platichthys flesus)  0.76270739
#1                                 Anchovy (Engraulis)  0.66095951
#40                        Sardinops (Sardinops sagax)  0.50412277
#42                                Seafood-based meals  0.48000000
#5                                      Bream (Charax)  0.32305245
#33                                 Mackeral (Scomber)  0.26503085
#39                     Sardine and pilchard (Sardina)  0.24428571
#41             Sea catfish and wolf-fish (Anarhichas)  0.15912625
#21                                         Fish paste  0.13883529
#28                                  Hake (Merluccius)  0.06756757
#29                        Halibut (Hippoglossus spp.)  0.03500000
#2                                      Babel (Barbus)  0.00000000
#3                                       Bass (Marone)  0.00000000
# the rest is zero

# http://www.efsa.europa.eu/sites/default/files/chronicgdayconsumers.xlsx
efsa1c <- read_xlsx("data/chronicgdayconsumers.xlsx", sheet="L1_Consumers_only_g_day", range="R3C1:R1132C15") # consumers only
#                    col_types=c(rep("text",5),rep("numeric",10)))
efsa1c <- efsa1c[efsa1c$Country=="Finland" & grepl("Fish", efsa1c$`Foodex L1`),]

efsa3c <- read_xlsx("data/chronicgdayconsumers.xlsx", sheet="L3_Consumers_only_g_day", range="R3C1:R18488C16")
efsa3c <- efsa3c[efsa3c$Country=="Finland" & grepl("Fish", efsa3c$`Foodex L2`),]
efsa3c$nr <- efsa3c$`Nr Consumer` / efsa3c$`% Consumers`
efsa3c$coeff <- efsa3c$STD / efsa3c$Mean
#mean(efsa3c$coeff[efsa3c$coeff>0])
#[1] 0.7353969
#sd(efsa3c$coeff[efsa3c$coeff>0])
#[1] 0.2678654
# It seems that the coefficient of variation is pretty close to 0.75 across all fish species and age groups.
# Therefore, we could make a model where we pool many level3 categories into e.g. imported fish and use local species separately.
# Useful groups: Salmon, fish meat, tuna, herring, whitefish, fish products (incl fish and potatoes meal), 
# perch, flounder, and other species.
# Sample from the ecdf as many as there are actual consumers and form new groups.

out <- data.frame() # This has as many rows as there were consuming individuals on the original studies.
for(i in 1:nrow(efsa3c)) {
  out <- rbind(out, data.frame(
    Row = i,
    Study = efsa3c$Survey[i],
    Participants = efsa3c$nr[i],
    Group = efsa3c$`Pop Class`[i],
    Food = efsa3c$`Foodex L3`[i],
    Result = approx(x=c(0.05, 0.1, 0.5, 0.95, 0.975, 0.99), y=unlist(efsa3c[i,11:16]), n=efsa3c$`Nr Consumer`[i])$y
  ))
}

write_csv(out, "data/EFSA_fish_consumption_consumers_fi_g_per_day.csv")

tmp <- efsa3c[c(3,5,11:16)] # quantiles saved
colnames(tmp)[colnames(tmp)=="Median"] <- "P50"
tmp <- melt(tmp, id.vars = c("Pop Class","Foodex L3"))
tmp$variable <- as.numeric(gsub("P","",tmp$variable))

ggplot(tmp, aes(x=value, y=variable, colour=`Pop Class`, group=`Pop Class`))+geom_line()+ # construct from data and quantiles
  facet_wrap(~`Foodex L3`, scales="free_x")

ggplot(out, aes(x=Result, color=Group))+stat_ecdf()+ # construct from sampled data
  facet_wrap(~Food, scales="free_x")

# The two graphs above illustrate that the consumer-only quantiles can successfully be used to approximate values and
# create a dataset for further sampling and analysis. Now food groups can be combined before sampling.
# Individuals' consumption is then created by sampling each approximation with the probability given (% Consumers)
# and then summing up to useful categories.

</rcode>

<rcode name="EFSA_amount" label="Initiate ovariable amount" embed=1>
# This is code Op_en1836/EFSA_amount on page [[Domestic fish consumption of the general population in Finland]]

library(OpasnetUtils)

amount_raw <- opasnet.csv("8/88/EFSA_fish_consumption_consumers_fi_g_per_day.csv",
                       wiki = "opasnet_en", sep=",", dec=".", header=TRUE)

amount <- Ovariable(
  "amount",
  dependencies = data.frame(Name="amount_raw"),
  formula = function(...) {
    dat <- amount_raw
    out <- data.frame()
    grouping <- unique(dat[c("Group","Food")])
    if(openv$N>0) {
      for(i in 1:nrow(grouping)) {
        tmp <- dat[dat$Group == grouping$Group[i] & dat$Food == grouping$Food[i] , ]
        out <- rbind(out, data.frame(
          Iter = 1:openv$N,
          Group = tmp$Group[1],
          Food = tmp$Food[1],
          Result = sample(tmp$Result, size=openv$N, replace=TRUE) * rbernoulli(openv$N, p=(nrow(tmp) / sum(unique(tmp$Participants)))) # There may be several studies but they never have the exact same number of patients.
        ))
      }
    } else {
      for(i in 1:nrow(grouping)) {
        tmp <- dat[dat$Group == grouping$Group[i] & dat$Food == grouping$Food[i] , ]
        out <- rbind(out, data.frame(
          Group = tmp$Group[1],
          Food = tmp$Food[1],
          Result = median(tmp$Result) * (nrow(tmp) / sum(unique(tmp$Participants)))
        ))
      }
    }
    
    # sort(unique(efsa3c$`Foodex L3`))
    fishes <- data.frame(
      Food = c(
        "Anchovy (Engraulis)",                    "Bream (Charax)",                        
        "Cod and whiting (Gadus spp.)",           "Fish and potatoes meal",                
        "Fish balls",                             "Fish fingers",                          
        "Fish meat",                              "Fish paste",                            
        "Fish products",                          "Fish roe",                              
        "Flounder (Platichthys flesus)",          "Hake (Merluccius)",                     
        "Halibut (Hippoglossus spp.)",            "Herring (Clupea)",                      
        "Mackeral (Scomber)",                     "Perch (Perca)",                         
        "Salmon and trout (Salmo spp.)",          "Sardine and pilchard (Sardina)",        
        "Sardinops (Sardinops sagax)",            "Sea catfish and wolf-fish (Anarhichas)",
        "Seafood-based meals",                    "Tuna (Thunnus)",                        
        "Whitefish (Coregonus)"),
      Fish = c("Imported fish","Bream","Imported fish",rep("Fish product",7),"Flounder",rep("Imported fish",2),
               "Herring","Imported fish", "Perch","Salmon",rep("Imported fish",4), "Tuna","Whitefish")
    )
    
    out <- merge(out, fishes, all.x=TRUE)
    out <- Ovariable(output=out, marginal = colnames(out)!="Result")
    return(out)
  },
  unit="g/d"
)

objects.store(amount, amount_raw)
cat("Ovariable amount and dataframe amount_raw stored.\n")
</rcode>

==== Fisheries statistics 2006 ====

<anacode>Farmed_salmon_table+(Fish_onsumption_tabl/Population_of_fin)+Herring_consumption_ +(Wild_salmon_consumpt/population_of_fin)</anacode>

==See also==

* [[Domestic fish consumption of the pregnant women in Finland‎‎]]
* Finravinto 2017 Study http://urn.fi/URN:ISBN:978-952-343-238-3

== References  ==

<references />