Malariaverse Workshop Lightning round: Umbrella, treasure & om

Pete Winskill

Session aims:

  • A whistle-stop tour of some other malariaverse packages
  • umbrella: for fitting seasonal profiles
  • treasure: unit costing
  • om: for (budget-constrained) stratification optimisation

Umbrella

  • 📦 [malariasimulation] captures the seasonality in transmission using fourier series parameters fitted to rainfall data.
  • 📦 [umbrella] includes some simple functionality to fit these parameters.
  • 📦 [umbrella] also has some helper functions to download CHIRPS rainfall data^1.
  • site_file$seasonality has fitted seasonal parameters for the default sites.

Umbrella workflow

  1. Define your spatial data. This will be spatial polygon(s). I like to use the 📦 [sf] for spatial data. (I have downloaded and processed sf files associated with all sites: contact me for a link.)
  2. Access satellite rainfall “data” rasters for the region/period/resolution you require
  3. Extract rainfall data associated with your spatial units
  4. Use umbrella::fit_fourier() to fit parameters that can be used in a site file, or directly in 📦 [malariasimulation].

Umbrella caveats

  1. Usually use a single, representative seasonal profile for sites.

    • Needed for forwards projections
    • May miss inter- and intra-annual variation of interest

  2. The assumption that ⬆️ rain = ⬆️ mosquitoes = ⬆️ malaria does not always hold!

Umbrella info

  • Check out the umbrella website for more information and a simple introduction vignette.

treasure

  • 📦 treasure, is a simple repository for unit costing information that has been used to cost out large global/country scenarios
  • Mostly serves as a single, version controlled reference for unit costs collected from the literature, WHO and Global Fund.
  • No time- or country-specific costing information.
  • If anyone is interested in improving the database, please let me know!

Treasure info

om

  • A common challenge is finding the optimum package of interventions across a country, where interventions can be targeted sub-nationally.
  • Often this question is asked with the additional constraint of a finite budget.
  • 📦 om is here to help!

om inputs

  • 📦 om requires specific inputs to work:
  • An impact matrix giving details of the outcome measure we are optimising, for example cases averted. Each row of the matrix is a sub-unit (e.g. district) and each column a unique intervention package.
  • A cost matrix giving details of the associated cost of each option in the impact matrix
  • A budget

om and complex budgets

  • The budget specification in 📦 om can be simple: e.g. $100
  • It can also be more complex. For example we might want to optimise the intervention packages across 5 countries, where each country has a domestic budget (to be spent internally) as well as a shared regional budget.

om info

  • Check out the om website for more information, a simple introduction vignette and a guide for implementing complex budgets.