Longer Term Updates

• Simulating HIV-negative individuals

  • This will allow the model to more accurately capture the increasing difficulty of finding undiagnosed individuals.
  • However, this will require a significant re-working of the background C model (cascade)

• Prevention Intervention

  • We will simulate this by weighting beta (transmission probability) by some value between 0 and 1.
  • This will allow us to reduce the number of new infections, thereby simulating prevention efforts that avert infections
  • The model will not be specific as to what these prevention methods entail, as with other interventions, but will demonstrate their impact
  • This issue is that simulating interventions alters the number of new infections (by changing the distribution of infectiousness of individuals), so if we were to simulate 'prevention' as any other intervention, we can't make a simModel - baseModel type comparison because the number of new infections changes with each intervention permutation.
    • The direct (reducing beta) and in-direct (getting people to ART) effects are mixed, so if the prevention intervention is turned off, then still the model would still report that infections have been averted (but this was an in-direct effect of other interventions)
  • The only solution is to simulate all interventions normally, and then re-run all interventions but this time including a prevention intervention that reduces beta by say 25%, then we can illustrate the impact of this strategy, but it means that weighing up prevention vs. testing vs. retention etc. interventions is much harder
  • Of course prevention is favoured heavily so we need an effective means of capturing cost here – see HERE for details on the prevention branch

• Ensuring each "90" target is hit

  • Currently users can adjust optimisation targets by adjusting a slider for each "90"
  • The model simply calculates the resulting Suppressed / PLHIV (73% if we are aiming for 90-90-90, 0.9^3) and interpolates strategies for getting to that value
  • The issue is that there are multiple means of getting 73% viral suppression: e.g. 90-90-90, 73-100-100, 100-100-73, 100-80-91 etc.
  • The model could be adapted to only select simulations that exactly hit the other targets, instead of simply focusing on the end-goal of these targets... open to discussion

• Decreasing projection interval

  • As we edge closer to 2020, the model should be updated to take onboard 2016 estimates of the cascade, and only project 2017 -> 2020
  • Then 2017, take current cascade estimates and project 2018 -> 2020 etc. etc.