template_design_report.Rmd

---
title: "*pfhrp2/3* Planner: Design report"
date: "*Downloaded on: `r Sys.Date()` | DRpower interactive app v1.0.1*"
output: 
  html_document:
    # theme: flatly
      version: 5
      bg: "#F9F9F9"
      fg: "#5e5ea3"
      primary: "#5e5ea3"
      base_font: !expr bslib::font_google("Lato")
      code_font: !expr bslib::font_google("JetBrains Mono")
params:
  design_ss_icc: NA
  design_ss_prev: NA
  design_final_sizes: NA
  design_nclusters: NA
  design_paramprev: NA
  design_paramicc: NA
  design_paramsims: NA
  design_poweroutput: NA
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)

library(tidyverse)
library(kableExtra)
library(janitor)

set.seed(10)
```

```{css, echo=F}
.title{
  color: #605ca3;
}

.date{
  color: #605ca3;
  text-align: right; 
  font-size: 14px;
}

/* Styles for the h3 headers */
h3 {
  color: #605ca3 !important;
}

hr {
  opacity: 1; 
  border-top: 2.5px solid #605ca3;
}
```

```{r create objects from params, echo=F}
df_final_sizes <- params$design_final_sizes
n_clusters <- params$design_nclusters
param_prev <- params$design_paramprev
param_icc <- params$design_paramicc
param_sims <- params$design_paramsims
power_output <- params$design_poweroutput
```
-----

### Background
This report presents the results generated by using the [*pfhrp2/3* planner web application](https://shiny.dide.ic.ac.uk/DRpower-app/). This application was used to help guide the appropriate number of health facilities (or sites) and a target sample size per health facility for the design of a *pfhrp2/3* deletion prevalence study. For more information on the statistical method used, see the [DRpower R package website](https://mrc-ide.github.io/DRpower).

### Final sample sizes

The adjusted sample sizes based on the expected drop-out proportion in each of the `r params$design_nclusters` health facilities are shown below and should be considered for study design. 

```{r echo=F}
kable(df_final_sizes %>% 
        # add total samples row
        adorn_totals(name = "Total samples", where = "row", ,,, -c(cluster, percent_dropout)) %>% 
        # annoyingly need to change dash to endash because rmd changes it to a bullet point thinking its a list!
        mutate(percent_dropout = ifelse(row_number() == nrow(df_final_sizes)+1, "--", percent_dropout)),
      format = "html",
      digits = 2,
      col.names = c("Health facility", "Target sample size", "Dropout (%)", "Adjusted sample size")) %>%
  kable_styling(
    bootstrap_options = "striped",
    position = "left",
    full_width = F) %>% 
  row_spec(nrow(df_final_sizes) + 1, italic = T)
```

### Estimated power

To estimate the power of the study assuming the target sample sizes as per above, we used the [DRpower R package](https://mrc-ide.github.io/DRpower) with the following parameters: 

<!-- - The sample sizes for the `r params$design_nclusters` clusters (shown above) -->
- Prevalence of `r params$design_paramprev`%
- Intra-cluster correlation of `r params$design_paramicc` 
- `r params$design_paramsims` simulations

The estimated power is `r power_output$power`% (95%CI: `r ceiling(power_output$lower)` - `r ceiling(power_output$upper)`%).

```{r fig.height=3, fig.width=3, echo=F}
power_output %>% 
  ggplot() +
  geom_segment(aes(x = " ", xend = " ",y = lower, yend = upper), color = "black", linewidth = 1) +
  geom_point(aes(x = " ", y = power),
             size = 2,
             shape = 21,
             fill = "mediumpurple") +
  geom_hline(yintercept = 80, color = "darkgrey", linetype = "dashed") +
  geom_text(aes(x= " ", y = 83, label = "80% threshold"), color = "darkgrey", size = 3) +
  scale_y_continuous(labels = scales::percent_format(1, scale = 1), limits = c(0,100)) +
  labs(x = "",
       y = "Estimated power") +
  theme_light() +
  theme(text = element_text(size = 12))
```