1 of 36

Tutorial�Introduction to models and data: �HIV in Harare

Reshma Kassanjee

University of Cape Town

SACEMA�

MMED 2024

�

Based on content by John Hargrove and Brian Williams

000

2 of 36

​

Harare, Zimbabwe

​

HIV prevalence

during 1980-2012

​

Pregnant women attending antenatal clinics

​

​

2

Hargrove et al. Epidemics 2011

3 of 36

South Africa

3

https://www.nicd.ac.za/wp-content/uploads/2021/11/Antenatal-survey-2019-report_FINAL_27April21.pdf

4 of 36

​

You are going to model this data!

​

​

4

5 of 36

Outline

​

  • Recap of relevant methodological principles
  • More background on the data and setting
  • The goals of the exercise
  • Discuss the first model that you will be fitting

​

​

… leave you to play

​

5

6 of 36

6

7 of 36

7

Model world: a (often simpler) representation of the real world

8 of 36

  • Does not need to exactly capture every mechanism of the real world

​

complex simple

  • Difficult to:�implement model,�interpret results,�obtain inputs, …

​

  • Aim: find the simplest model that can adequately answer your question

8

Model world: a (often simpler) representation of the real world

  • May not be able to accurately answer

9 of 36

Want your model to ‘fit’ the data

  • Outputs from your model should be consistent with available data

​

​

9

10 of 36

Want your model to ‘fit’ the data

  • Outputs from your model should be consistent with available data

​

  • E.g. HIV prevalence survey estimates � should be ‘close’ to � the modelled prevalence � in the population at those times

​

  • Not aligned 🡪 model world is not adequately describing your real world

10

11 of 36

11

Want your model to ‘fit’ the data

12 of 36

  • If there is misfit, this could be caused by an incorrect structure of model or incorrect inputs

12

Want your model to ‘fit’ the data

13 of 36

Want your model to ‘fit’ the data

  • Various methods for trying to ensure (or check whether) your model fits the data
  • Today, we will be checking ‘by eye’

13

Prevalence (y) over time (x)

Model output

Data

14 of 36

​

You are going to model this data!

​

​

14

15 of 36

John Hargrove and Brian Williams

15

16 of 36

The data: Prevalence

​

​

​

16

​

Surveys of pregnant women attending nine antenatal clinics in different years

​

n = 22 684

​

​

17 of 36

The data: disease-related mortality

​

​

​

17

Total mortality from records provided by the City of Harare

​

From 1980 to before the HIV epidemic: declining mortality

​

Extracted trends for disease-related mortality

​

18 of 36

The data: incidence

​

​

​

18

  • Two estimates from two cohort studies of pregnant women (conducted in 1991-1993 and 1997-2000)

​

19 of 36

The data – can we find a model?

​

​

​

19

20 of 36

What you will be doing…

  • You will start with the simplest possible model, �with the smallest number of parameters
  • You will not be surprised to find that you do not get a good fit to the data
  • You should try to reflect on why
  • Then try a slightly more complex model �(adding one feature)
  • Reassess

​

�

20

21 of 36

What you will be doing…

  • Use prevalence data, checking the fit ‘by eye’
  • Shiny app (by Carl Pearson and Juliet Pulliam) – �you do not need to the do the model implementation!

​

​

21

22 of 36

22

The first �model

23 of 36

23

The first �model

  • Understand the model
  • Try out different values for the inputs
  • Compare model outputs to data points

24 of 36

24

The first �model

  • Understand the model
  • Try out different values for the inputs
  • Compare model outputs to data points
  • What may be ‘wrong’ with this model?

25 of 36

25

26 of 36

26

27 of 36

27

28 of 36

28

29 of 36

29

30 of 36

30

31 of 36

31

32 of 36

32

33 of 36

33

34 of 36

34

35 of 36

Instructions

  • Groups of <4 people
  • Think about what it all means, answer Qs, and discuss
  • When done with Harare, try the other two datasets�(South Africa, Uganda)
  • Can also look at the model code
  • We will pause discuss ‘box cars’, and end with a discussion

​

​

​

​

​

​

35

remotes::install_github('ICI3D/ici3d-pkg')

ICI3D::hivTutorial()

You should have already done this:

36 of 36

This presentation is made available through a Creative Commons Attribution-Noncommercial license. Details of the license and permitted uses are available at� http://creativecommons.org/licenses/by-nc/3.0/

Introduction to modelling changes in HIV prevalence and incidence in Harare, Zimbabwe

�Attribution: J. Hargrove, B. Williams, R. Kassanjee

Clinic on the Meaningful Modeling of Epidemiological Data

​

For further information please contact figshare@ici3d.org.

000

000

© 2014-2024 International Clinics on Infectious Disease Dynamics and Data

000

000