HIV indicators - natural course of an HIV epidemic or results of sociobehavioral changes?
Difficulty in choosing a basic HIV indicator to compare HIV epidemics across countries
What HIV indicators should we use to compare the epidemics across countries & what HIV indicators give us a sense of where the epidemic is?
HIV incidence:
- Expectation: Lower numbers of new HIV infections per year is a good sign
- Contextualization: A lower number of new HIV infections is obviously better than a higher number of new HIV infections, HIV incidence depends on multiple factors and cannot be directly compared across countries
- Example: Country A has a slightly higher HIV incidence than country B, but country A has 100% higher HIV prevalence than country B, which is doing better?
HIV prevalence:
- Expectation: Lower HIV prevalence means less people living with HIV and should be a good sign
- Contextualization: A lower number of people living with HIV is obviously better than a higher number, but HIV prevalence depends on new HIV infections and mortality of HIV/AIDS patients making it difficult for direct comparison
- Example: Take again country A with 100% higher HIV prevalence than country B, but country A has very little HIV/AIDS mortality thanks to widespread use of ART while country B has very little use of ART and HIV mortality lags HIV incidence by ~10 years, which is doing better?
HIV mortality:
- Expectation: Lower HIV mortality means less people are dying of HIV/AIDS
- Contextualization: A lower number of deaths due to HIV/AIDS is obviously better than a higher number, but this calue obviously also depends on the HIV prevalence
- Example: Country A above has little HIV/AIDS mortality compared to overall HIV prevalence, but country B has similar HIV mortality simply because HIV prevalence is low...
A very simple compartmental model for HIV
SID model:
Of course many many assumptions (that are not correct in real life) need to be made to keep it very simple:
- No births (but we focus on 15-49 population anyways)
- No deaths other than HIV/AIDS related deaths
- No CD4 levels of infectiousness
- Basically anything not specifically discussed below is assumed not to exist..
S --> Susceptible state:
An S individual is simply someone susceptible to the disease, meaning anyone in the population who is not immune to the disease.
I --> Infectious state:
Once an individual is exposed to the disease, he will develop this disease and become infectious (unless suppressed thanks to ART).
D --> Death state:
HIV infections eventually become AIDS and leads to death.
SID compartmental model disease dynamics
S → I
Individual-level
Going from S to I for an indivudal in a particular year depends on three things:
- the proportion of infectious people in the population that year:
i(t)=PrevalenceAlthough with ART the equation may become:
i(t)=(1−ARTcoverage)∗Prevalence
- the number of exposure events the individual has per year (this means number of drug injections, number of sexual partners, etc etc):
- the chance for an S to contract the disease after such an exposure event :
which combines many factors into one, including:
- Male circumcision
- Use of condoms
- Other STDs
- Use of PrEP
- It can also incluse ARTcoverage (see later points)
- and so on..
We can combine the last two into
β=ρ∗rPopulation level
On a population-level however, the number of S that will become I also depends on the proportion of S itself in the population (obviously if there are no S in the first place, no one will can become I).
So we add the following requirement:
- the proportion of susceptible people in the population that year:
s(t)=1−HIVPrevalenceSo the change in the number of S in a population in a given year is :
−β∗i(t)∗s(t)and so HIV incidence can be given by:
HIVIncidence=β∗i(t)∗s(t)⟺HIVIncidence=β∗HIVPrevalence∗(1−HIVPrevalence)or
⟺HIVincidence=β∗(1−ARTcoverage)∗HIVPrevalence∗(1−HIVPrevalence)A new HIV indicator to compare across countries
As we have data from UNAIDS on ARTcoverage, HIVincidence, and HIVprevalence from 1990 on, we can easily calculate β as follows:
β=HIVPrevalence−HIVPrevalence2HIVincidenceAnd as seen above, β can be considered as a factor representing all socio-behavioral characteristics.