Zeidyen/epimodelhiv-aimee-agyw

Modelled population HIV impact of an AI health companion (Aimee) on HIV testing and PrEP uptake among South African AGYW, using EpiModelHIV (SSA heterosexual module). # epimodelhiv-aimee-agyw **Modelled population HIV impact of an AI health companion (Aimee) on testing and PrEP uptake among South African AGYW.** A network-based HIV transmission modelling study using EpiModelHIV (SSA heterosexual module) to estimate the population-level HIV-incidence impact of an AI chatbot that increases HIV-testing frequency and PrEP uptake among adolescent girls and young women (AGYW, 15–24) in South Africa. > **Status:** scoping / scaffold. Not yet calibrated. No results are valid yet. --- ## The idea in one diagram ``` Clover Field Study (observational) → chatbot effect on testing & PrEP uptake │ │ │ ▼ │ EpiModelHIV (SSA heterosexual, AGYW) │ │ ▼ ▼ effect-size inputs ───────────────► counterfactual scenarios │ ▼ HIV infections averted by the chatbot ``` The empirical cohort study estimates **who engages → who tests / starts PrEP**. This model propagates those behavioural changes to **population HIV incidence**. ## Key design decisions - **Population:** AGYW 15–24, heterosexual, generalized epidemic (NOT MSM). Age-disparate partnerships (AGYW ↔ older male partners) are retained — they are the dominant incidence driver. - **The chatbot is not modelled mechanistically.** Its *effect* is applied as a perturbation to existing model parameters: HIV-testing rate, PrEP initiation/coverage, and (if supported by data) PrEP persistence. - **Associational → causal discipline.** Cohort effects are confounded by self-selection, so they enter as a **scenario range** (conservative / central / optimistic), never a single causal point estimate. ## Model foundation The study builds on the **EpiModel Gallery "HIV Transmission with Care Cascade and PrEP"** model (MIT licensed, runs on current **EpiModel 2.6.x**) — vendored in `R/base_model/`. It already provid …