What is ABM all about?
Agent-based modeling (ABM) is used increasingly and unconstrained by disciplinary boundaries in many scientific fields such as engineering, biology, social science and economics for the analysis of organized, complex systems. (Heath et al. 2009) The basic idea is, to model independent entities which interact with their environment based on certain behavior rules. Thus, dynamic non-linear effects can be modeled.
Why should I care? Aren’t there more than enough modeling methods in health economics?
As all models are simplifications of reality, there are a few major questions surrounding the most adequate way to depict the reality of diseases and interventions by the right choice of the modeling technique itself as well as the construction of the model. Both steps inherit the risk of omitting important factors biasing he results of the model. For example it may be inappropriate to model a disease with recurrent disease states using a decision tree which only allows for ne-directional progression through the model.
Also, patient characteristis (e.g. age, socio-economic status) which strongly influence the effectiveness of interventions may not always be appropriately considered certain types of models. The difficulties described above especially apply for the modeling of infectious diseases. The spread of infections not only depends on the course of the disease, but also on the population structure and on behavioural aspects, e.g. mobility or social networks. Furthermore, through the heterogeneity of patients or patient groups, effects caused by certain individuals may not be covered if only expected values of a population are incorporated in the model. A famous example to demonstrate the influence of a single individual is the spread of human immunodeficiency virus (HIV) in the United States (US) via a flight attendant in the 1980’s. (see Auerbach et al., 1984) Modeling the spread of HIV using only expected values regarding mobility and number of sex-partners could have failed to include such events and therefore the spread of the disease would have been underestimated.
For the above reason, our project will focus on modeling syphilis using an agent-based model. As the natural history of syphilis is well understood and because of its status as a reportable disease data availability is a minor concern. Furthermore the treatment is rather boring because simple penicillin injections work fine as there are no know resistances.
However there are some interesting facts about this sexual transmitted disease:
1. The fact is that its a STD. As human mating habits are rather complex and diverse there is the possibility for huge gains in precision when explicitly modeling this behavior. Furthermore it is understood that syphilis has a higher prevalence in some subpopulations e.g. men-having-sex-with-men. In those subpopulations the prevalence is rising. Modeling those patterns is a challenge for traditional methods but may be addressed adequately with ABM
2. Long-term consequences if untreated. Especially neurosyphilis possibly leading to dementia is a rather severe late effect associated with high costs.
3. Debate on universal antenatal screening. As the prevalence in the general public is rather low there is an ongoing debate if universal screening of pregnant women is still economically justified.