15_Total_Nets_2024-2026.Rmd

---
title: "Frontier plots for each country showing person-years of ITN access vs total nets required with World Bank populations 2020-2035"
author: "Hannah Koenker"
date: "`r format(Sys.time(), '%d %B, %Y')`"
output:
  html_document:
    toc: yes
    toc_depth: 2
    toc_float: yes
    fig_width: 7
    fig_height: 6
    fig_caption: yes
    df_print: kable
editor_options:
  chunk_output_type: console
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE, message = FALSE)
library(haven)
library(tidyverse)
library(geofacet)
library(broom)
library(readxl)
library(janitor)
library(countrycode)
library(data.table)
library(stringr)
library(stringi)
library(ggrepel)

# ![](images/TH_logo.jpg){width=1.0"}

######################
# Set figure theme and colors
######################

theme_set(theme_classic())

```

These graphs show relative and total net need for the three-year period 2024-2026 (inclusive) for countries in sub-Saharan Africa where ITN retention times have been estimated. NOTE: Total populations are used to determine net need; they do NOT take into account subnational areas where populations are not targeted for ITN distribution due to low malaria risk or because the area is targeted for IRS. Thus for countries like Namibia, Botswana, Angola, Kenya, etc, the total net needs here are overestimates. However, the relative savings/increases between ITN strategy options would remain the same irrespective of population size.

Assumptions:
  1. The total population of the country receives ITNs
  1. ANC and EPI channels are active and reach 6% of the population annually in all scenarios (ANC/EPI nets are included in the totals below)
  1. Target population ITN access is 80%.
  1. Full Scale CD scenario assumes a very recent mass campaign in 2023. ITN access would be lower for countries/areas where the most recent campaign was in 2021 or 2022
  


```{r grid}
geofacet_fname <- "data/geofacet_ssa.csv"
ssa_grid <- fread(geofacet_fname) %>% 
  arrange(code)
```

```{r afpop}

## Extract Africa population data from UN WPP 2022 Demographic Indicators Medium File (https://population.un.org/wpp/Download/Standard/CSV/)
ls <- read_excel("data/MAP_retentiontime_DM_medianlifespan.xlsx", sheet="Sheet1") %>%
  clean_names() %>%
  filter(type=="MAP") %>% 
  drop_na(country) %>%
  rename(lifespan=median_retention_years) %>%
  separate(x95_percent_ci, into = c("lb", "ub"),
           sep = "\\s*(–|-)\\s*", convert = TRUE) %>% 
  select(-site)

dm <- read_excel("../DM Maps/Median lifespan DM multisite.xlsx") %>% 
  clean_names() %>% 
  rename(lifespan=medianlifespan) %>%
  separate(x95_percent_ci, into = c("lb", "ub"),
           sep = "\\s*(–|-)\\s*", convert = TRUE) %>% 
  drop_na(lifespan) %>% 
  select(country, lifespan, lb, ub) %>% 
  mutate(type="DM")

lsdm <- ls %>% 
  bind_rows(dm) %>% 
  mutate(iso3=countrycode(country, origin="country.name", destination="iso3c")) %>% 
  mutate(iso3=case_when(type=="MAP" ~ as.character(iso3)))

levels <- lsdm$country[ls$type=='MAP'][rev(order(ls$lifespan[ls$type=='MAP']))]
lsdm$country_ord <- factor(lsdm$country, levels = levels)

# save a small version of retention times for joining into results tables:
rettimes <- lsdm %>% 
  filter(type=="MAP") %>% 
  mutate(name=case_when(country=="Burkina" ~ "Burkina Faso",
                        # country=="Cote d'Ivoire" ~ "Côte d'Ivoire",
                        country=="Congo (Republic of)" ~ "Rep. Congo",
                        TRUE~as.character(country))) %>% 
  dplyr::select(-type, -country, -lb, -ub, -both, -country_ord) %>% 
  mutate(name=stri_trans_general(name,"Latin-ASCII")) %>% # remove o-hat from Cote d'Ivoire
mutate(iso_a2 = countrycode(name, origin = "country.name", destination = "iso2c"))

a3levels <- rettimes$iso3 # make a list of the isoa2 countries included in retention times for later
a2levels <- rettimes$iso_a2 # make a list of the isoa2 countries included in retention times for later

afpop <- as.data.frame(read_csv("data/WPP2022_Demographic_Indicators_Medium.csv")) %>% 
  clean_names() %>% 
  select(iso2_code, iso3_code, location, t_population1july, time) %>% 
  mutate(t_population1july=t_population1july*1000) %>% 
  rename(pop=t_population1july,
         year=time,
         iso_a2=iso2_code,
         country_code=iso3_code,
         country_name=location) %>% 
  filter(year>2019 & year<2041) %>% 
  filter(iso_a2 %in% a2levels) %>% 
  pivot_wider(names_from=year, names_prefix="x", values_from=pop)


```

```{r}
table80rtlong <- read_csv("output/table80rtlong.csv")
```

```{r totalnetsdf}
# Figure for Total Nets Needed in a 3 year timeframe (2024-2026) for different scenarios
## WHY DOESN"T ROUTINE GET COUNTED IN SCEN 2 and 3??

tnets3 <- table80rtlong %>% 
  mutate(scenario_4=na_if(scenario_4,0.1)) %>% 
  left_join(afpop, by=c("iso3"= "country_code")) %>% 
  mutate(routine=.06,
         tot0=x2024*routine+x2025*routine+x2026*routine+x2026/1.8,
         tot2=x2024*(scenario_2/100)+x2025*(scenario_2/100)+x2026*(scenario_2/100)+x2024*routine+x2025*routine+x2026*routine,
         tot3=x2024*routine+x2025*routine+x2026*routine+x2026/1.8+x2024*(scenario_3/100)+x2025*(scenario_3/100),
         tot4=x2024*routine+x2025*routine+x2026*routine+x2026/scenario_4,
         tot5=x2024*routine+x2025*routine+x2026*routine+x2024/scenario_5+x2026/scenario_5,
         tot6=x2024*routine+x2025*routine+x2026*routine+x2024/1.8+x2026/1.8,
         s0vs2=tot2/tot0,
         s0vs3=tot3/tot0,
         s0vs4=tot4/tot0,
         s0vs5=tot5/tot0,
         s0vs6=tot6/tot0)
        




totalnets <- tnets3 %>% 
  select(iso3, lifespan, tot0, tot2, tot3, tot4, tot5, tot6) %>% 
  pivot_longer(contains("tot"), names_to = "scenario", names_prefix = "tot", values_to = "totalnets")

scenpct <- tnets3 %>% 
  select(iso_a2, iso3, lifespan, contains("s0vs")) %>% # s2vs0, s2vs3, s2vs4, s2vs5
  mutate(iso3=case_when(iso_a2=="ER" ~ "ERI",
                                TRUE ~ as.character(iso3))) %>% 
  mutate(s0vs0=1,
         code=iso3) %>% 
  pivot_longer(contains("s0"), names_to="scenario", names_prefix = "s0vs", values_to = "pct") %>% 
  mutate(iso2=countrycode(code, origin="iso3c", destination="iso2c")) %>% 
  mutate(name=countrycode(code, origin="iso3c", destination="country.name"))

avgscenpct <- scenpct %>% 
  group_by(scenario) %>% 
  summarize(mean=mean(pct, na.rm = TRUE)) 
```

```{r map-3yr-totalnets, warning=FALSE}
scenpct %>%
  # filter(name!="Eritrea") %>%
  ggplot(aes(x = scenario, y = pct, fill = scenario)) +
  geom_col() +
  # geom_hline(yintercept = 0) +
  # scale_y_continuous(name = "Percent",
  #                    breaks = seq(-1, 1, 0.25),
  #                    labels = seq(-1, 1, 0.25)) +
  scale_x_discrete(labels = c("", "", "", "","")) +
  scale_y_continuous(labels = scales::percent) +
  facet_geo( ~ code, grid = ssa_grid, label = "code") +
  labs(title = "Percentage difference in total nets needed over 2024-2026\nto maintain ≥80% ITN access, relative to Status Quo\n(3 year campaigns quantified using population/1.8, with ANC/EPI ITNs)",
       x = "",
       y = "Percent",
       fill = "Scenario") +
  scale_fill_brewer(
    palette = "Dark2",
    labels = c(
      "Status Quo", 
      "Full Scale CD",
      "Campaigns + CD",
      "3-year Campaigns, \n(tailored)",
      "2-year Campaigns, \n(tailored)",
      "2-year Campaigns (pop/1.8)"
    )
  ) +
  theme_void() +
  theme(
    # legend.position="none",
    # axis.text.x = element_text(angle=45, hjust=1)
    # axis.text.x = element_blank(),
    axis.line = element_blank(),
    axis.line.x = element_blank(),
    axis.line.x.bottom=element_line(color="white"),
    axis.ticks.x = element_blank(),
    # axis.ticks.y = element_blank(),
    axis.text = element_text(size = 5),
    # panel.grid.major.x = element_line(color = "darkgrey", size = 0),
    strip.text.x = element_text(size = 6),
    strip.background = element_rect(color = "white", linewidth = .5),
    legend.key.size = unit(.25, 'cm'),
    #change legend key size
    # legend.key.height = unit(1, 'cm'), #change legend key height
    # legend.key.width = unit(1, 'cm'), #change legend key width
    legend.title = element_text(size = 10),  #change legend title font size
    legend.text = element_text(size = 8), #change legend text font size)
    # axis.text.y = element_blank(),
    # axis.ticks = element_blank(),
    axis.title.x = element_blank(),
    # axis.title.y = element_blank(),
    # legend.position="none",
    panel.background = element_blank(),
    panel.border = element_blank(),
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    plot.background = element_blank()
  )
```

```{r testplot, eval=FALSE}

mill <-  scales::unit_format(unit = "M", scale = 1e-6,accuracy = 0.1)
s2 <- tnets3 %>% 
  filter(iso3=="CMR") %>% 
  pull(scenario_2)
s3 <- tnets3 %>% 
  filter(iso3=="CMR") %>% 
  pull(scenario_3)
s4 <- tnets3 %>% 
  filter(iso3=="CMR") %>% 
  pull(scenario_4)
s5 <- tnets3 %>% 
  filter(iso3=="CMR") %>% 
  pull(scenario_5)
# scenario_6 qfactor not needed, it's just 1.8

tnets3 %>% 
  filter(iso3 == "CMR") %>% 
  select(contains("tot"),iso3, lifespan, country_name, contains("scenario")) %>% 
  pivot_longer(cols=contains("tot"), names_to = "scenario") %>% 
  ggplot(aes(
    x = scenario,
    y = value,
    fill = as.factor(scenario))) +
  geom_col() +
  geom_text(aes(label=mill(value)), hjust=.5, vjust=-.5) +
  scale_fill_brewer(
    palette = "Dark2",
    labels = c(
      "Status Quo (pop/1.8)",
      paste0("Full scale CD at pop x ",s2,"%"),
      paste0("Campaigns + CD at pop x ",s3,"%"),
      paste0("3-year campaigns at pop/",s4),
      paste0("2-year campaigns at pop/",s5),
      "2-year campaigns (pop/1.8)"
      )) +
theme(axis.text.x = element_blank()) +
  scale_y_continuous(labels = unit_format(unit = "M", scale = 1e-6)) +
    labs(x = "",
       y= "Total nets needed 2024-2026",
       fill = "Scenario",
       title = paste0("Cameroon",": "," years median retention time"))
  
```

Below are plots for each country, using the quantification factor that maintains 80% ITN access. One campaign is conducted in 2026 in the "Status Quo" and "Campaigns + CD" scenarios, while the Full Scale CD scenario assumes ITN access is relatively high due to a recent campaign in 2023.

```{r totalnets_function}
# function that plots total nets needed for GF 2024-2026 round to achieve 80% ITN access 

totalfunctionloop <- function(iso3) {
  
mill <-  scales::unit_format(unit = "M", scale = 1e-6,accuracy = 0.1)

 # ptitle <- paste0("p",{{iso3}})
lifespan <- tnets3 %>% 
   filter(iso3 == {{iso3}}) %>% 
  pull(lifespan)

name <- tnets3 %>% 
  filter(iso3 == {{iso3}}) %>% 
  pull(country_name)

s2 <- tnets3 %>% 
  filter(iso3=={{iso3}}) %>% 
  pull(scenario_2)
s3 <- tnets3 %>% 
  filter(iso3=={{iso3}}) %>% 
  pull(scenario_3)
s4 <- tnets3 %>% 
  filter(iso3=={{iso3}}) %>% 
  pull(scenario_4)
s5 <- tnets3 %>% 
  filter(iso3=={{iso3}}) %>% 
  pull(scenario_5)  
            
totplot <- tnets3 %>%
  filter(iso3 == {{iso3}}) %>%
 select(contains("tot"),iso3, lifespan, country_name, contains("scenario")) %>% 
  pivot_longer(cols=contains("tot"), names_to = "scenario") %>% 
  ggplot(aes(
    x = scenario,
    y = value,
    fill = as.factor(scenario))) +
  geom_col() +
  geom_text(aes(label=mill(value)), hjust=.5, vjust=-.5) +
  scale_fill_brewer(
    palette = "Dark2",
    labels = c(
      "Status Quo (pop/1.8)",
      paste0("Full scale CD at pop x ",s2,"%"),
      paste0("Campaigns + CD at pop x ",s3,"%"),
      paste0("3-year campaigns at pop/",s4),
      paste0("2-year campaigns at pop/",s5),
      "2-year campaigns (pop/1.8)"
      )) +
theme(axis.text.x = element_blank()) +
  scale_y_continuous(labels = scales::unit_format(unit = "M", scale = 1e-6)) +
    labs(x = "",
       y= "Total nets needed 2024-2026",
       fill = "Scenario",
       title = paste0(name,": ",lifespan," years median retention time"))
 

print(totplot)
  plotname <- paste0("figs/tnets3/t3_",iso3,".pdf")
  ggsave(plotname)
  
}

```


```{r totalloop, warning = FALSE, message=FALSE}
for(i in a3levels) {
   totalfunctionloop(i)
}
```