Marburg_underreporting.Rmd

---
title: "Marburg_underreporting"
output: html_document
date: "2023-06-12"
---

## Outbreak of Marburg in Equatorial Guinea, 2023

Loading the data

```{r, include=FALSE}
library(tidyverse)
```

```{r}
MVD_linelist <- read_csv("Data/Marburg_EqGuinea_linelist.csv")
head(MVD_linelist)
```

The data from the Marburg outbreak that took place in Equatorial Guinea from December 2022 until April 2023 includes the case type (probable vs confirmed), the status of the case (recovered/dead/unknown), the week of illness onset, and, for some of the cases, information about their gender, age, occupation, and relation to other outbreak cases.

_Question: can `datadelay` use weeks as input rather than days?_

Given the lack of active surveillance and contact tracing, it is likely that only the most severe cases (which would most likely result in death) were detected and counted. Therefore, it is likely that a proportion of the outbreak cases have not been ascertained.

Because of the lack of death dates, it is not possible to estimate the time-varying CFR, but it is possible to use the information from cases with known outcomes to calculate the static CFR, and then use this to estimate the proportion of cases that are being missed.

```{r}
MVD_linelist %>% summarise(total_cases = length(Onset_week)) #40 cases
MVD_linelist %>% filter(Type == "confirmed") %>% summarise(total_cases = length(Onset_week)) #17 confirmed
MVD_linelist %>% filter(Status == "dead") %>% summarise(total_deaths = length(Onset_week)) #35 deaths
MVD_linelist %>% filter(Status == "dead" & Type == "confirmed") %>% summarise(total_deaths = length(Onset_week)) #12 deaths among confirmed cases

35/40 #CFR=87.5% among total cases
12/17 #CFR=70.5% among confirmed cases
```

The overall CFR is 87.5%. From a systematic review by [Nyakarahuka](https://bmcinfectdis.biomedcentral.com/articles/10.1186/s12879-016-2045-6), we assume that the baseline CFR is 61%. Thus, the proportion of ascertained cases is:

```{r}
61/87.5 
```

Now, adding dates for the known deaths, based on the reported delay between onset-death of 8-9 days by [WHO](https://www.who.int/news-room/fact-sheets/detail/marburg-virus-disease):

_How to distinguish recovered vs unknown status_

```{r}
MVD_linelist$Death_week <- MVD_linelist$Onset_week + 8
MVD_linelist$Death_week <- as.Date(ifelse(MVD_linelist$Status == "unknown"|MVD_linelist$Status == "recovered", NA, MVD_linelist$Death_week))
```

```{r}
library(incidence2)
MVD_cases_deaths <- incidence2::incidence(MVD_linelist,c("Onset_week","Death_week")) |> complete_dates()
MVD_cases_deaths <- pivot_wider(MVD_cases_deaths, names_from = count_variable, values_from = count)
names(MVD_cases_deaths)[2:3]=c("deaths","cases")
MVD_cases_deaths <- MVD_cases_deaths[c(1,3,2)]
MVD_cases_deaths <- as.data.frame(MVD_cases_deaths)
```

```{r}
library(epiparameter)
set.seed(1)
extract_param(type = "range", values = c(8, 2, 16), distribution = "gamma", samples = 77) #2.095 and 4.513
Marburg_onset_death <- epidist(disease = "marburg", epi_dist = "onset_to_death", prob_distribution = "gamma", prob_distribution_params = c(shape=2.095, scale=4.513))

known_outcomes_MVD <- known_outcomes(MVD_cases_deaths, Marburg_onset_death)

plot_known_outcomes(known_outcomes_MVD)

```