In Novartis/patprofile: Patient profile module

library(knitr)
opts_chunk$set(echo = TRUE, eval = FALSE, message = FALSE)

Adding the patient profile to an app

The patient profile has been packaged as a shiny module to facilitate deployment. It can be added to any app that uses standard ADaM datasets with only a couple of lines.

Server

Server side, ADaM datasets are passed in the module call.

callModule(patientProfile_mod, id = "pp_module1",
           ADSL = myADSL, ADAE = myADAE, ADCM = myADCM, ADLB = myADLB,
           ADEG = myADEG, ADEX = myADEX, ADPC = myADPC)

Aside from ADSL, all other datasets can be left NULL. Missing datasets simply disable some of the functionalities of the module. Moreover, whenever informative columns are missing, the submodules will attempt to find alternative sources within the given data.

UI

UI side, patientProfile_modUI can be used to

patientProfile_modUI(id = "pp_module1")

Submodules

While the patient profile is meant to be used as a single top level module that will adapt to the data available, submodules are exported by the package and can be used individually. They are currently impplemented as tabPanel, so their use is restricted to compatible containers (navbarPage/avaNavbarPage, tabsetPanel/avaTabPanel, ...).

For example, to include the Exposure & Lab values module alone

# Server
callModule(module = exlb_mod, id = "exlb_auto", uid = uid,
           ADSL = ADSL, ADLB = ADLB, ADEX = ADEX, ADPC = ADPC)
# UI (within a tabset or menubar)
exlb_UI(ns("exlb_auto"))

When used in this manner, it is the developer's responsibility to add the appropritate data checks to ensure correct behaviour.

Refer to the individual submodules documentation for details on the required inputs.

Linking patient profile with other modules

A uid argument can be passed to the patient profile. By creating a new input binding in the server code and linking it to a Shiny.onInputChange event, we can set the participant profile to automatically select a participant of interest targeted from another plot.

# Server code
uid <- reactiveVal()
observe({
  ns <- session$ns
  if(!is.null(input$patient_js)){
    print("Updating selected patient")
    uid(input$patient_selector)
  }
})
# Module call
callModule(patientProfile_mod, id = "pp_module1", uid = uid, ADSL = myADSL)

Existing implementations

The AVA safety explorer is linked with the patient profile, such that when a participant is clicked in the timelines or safety outlier explorer, the subject is pre-selected in the patient profile module.

Likewise, if the eDISH plot is also part of the app, it will have the same interaction with the participant profile.

Custom plots

An example is available in the pp-template shiny app embeded within the package.

Using plotly, we can setup a plot with patient ids as key.

p <- plot_ly(data = data, x = ~ADT, y = ~AVAL,
             key = ~USUBJID,
             type = "scatter", mode = "markers",
             text = ~paste("uid:", USUBJID))

With the onRender function of the htmlwidgets package, javascript code can be added to plot events. We can use this to link a custom plot with the patient profile

# p is a plotly object
p <- p %>% onRender("
  function(el){
    el.on('plotly_click', function(d){
      selsub = d.points[0].data.key[d.points[0].pointNumber];
      console.log('Click: ', selsub);
      Shiny.onInputChange('pp_module1-patient_js', selsub);
    })
  }
")

The chunk of code assume that the module id is "pp_module1". If the module is named differently, the render string can be built

module_id <- "pp_module2"
render <- paste0("
  function(el){
    el.on('plotly_click', function(d){
      selsub = d.points[0].data.key[d.points[0].pointNumber];
      console.log('Click: ', selsub);
      Shiny.onInputChange('", module_id, "-patient_js', selsub);
    })
  }
")

If the plot function is part of its own module, then the namespace must be passed instead of the module name.

target <- ns("patient_js")
render <- paste0("
  function(el){
    el.on('plotly_click', function(d){
      selsub = d.points[0].data.key[d.points[0].pointNumber];
      console.log('Click: ', selsub);
      Shiny.onInputChange('", target, "', selsub);
    })
  }
")

Creating submodules for patprofile

All visualization tools were created as self contained submodules, making it trivial to extend the functionalities of the patient profile by adding new submodules.

Developers are encouraged to submit merge requests with their own shiny modules or open issues to suggest new visualization on the package's GitLab page.

Novartis/patprofile documentation built on April 24, 2020, 7:22 a.m.