Home

Overview of the MIGHTI Simulation Framework

What is MIGHTI?

MIGHTI (Model of Inter-Generational Health, Transmission, and Interventions) is a Python-based agent-based modeling framework designed to simulate the dynamics of health and disease across a population. Built upon the Starsim simulation engine, MIGHTI enables complex interactions between diseases, interventions, social determinants, and individual agent characteristics.

MIGHTI simulates a wide range of health conditions (HCs), including both noncommunicable diseases (NCDs) such as diabetes, cardiovascular disease, and depression, and infectious diseases (IDs) such as HIV, tuberculosis, and HPV.

Overview of the MIGHTI Simulation Framework

---

Core Inputs

MIGHTI requires the following input datasets:

• Demographics: Age distribution, fertility, and mortality (via life tables)
• Disease Parameters: Acquisition, remission, mortality, relative risks, condition-specific attributes
• Intervention Coverage: Time-varying data (e.g., ART scale-up, T2D treatment programs)
• Social Determinants (SDoH): Individual-level states like housing stability, education, and income

Prevalence-matching calibration (e.g., for p_acquire) is used to ensure realistic simulations. See Calibration.

---

What Makes MIGHTI Unique?

Structured Disease Interactions

• HIV-HC and HC-HC interactions are supported via connectors
• Interactions adjust acquisition susceptibility (rel_sus) dynamically
• Input via interpretable CSV files (data/processed/rel_sus.csv and parameter rel_sus column)

Intervention Logic

• Supports screening, treatment, and prevention
• Coverage can be static or time-varying
• Interventions can modify care delivery, disease outcomes, and SDoH states
• Flexible enough to include SDoH-targeting interventions (e.g., housing support)

Social Determinants of Health (SDoH)

• Modules like NeighbourhoodSituation, EconomicSituation, and EducationSituation
• Per-agent binary states with optional maternal inheritance
• Primary default pathway: adherence penalties and intervention targeting

CASM-Based Adherence Disruption

• CASM stands for constellation of alcohol, substance, and mood-related
• Adherence affected by binary flags for 7 CASM conditions:
  • Alcohol use, depression, anxiety, pain, tobacco, opioid, stimulant use
• Multiplicative adherence model with SDoH factors
• Treatment of CASM conditions can improve adherence via spillover effects when conditions remit

Pathways from Interventions to Health Outcomes

Pathway diagram: Attach your intervention → adherence → outcomes figure to this wiki page.

---

Model Capabilities

• Track incidence, prevalence, and mortality for each disease
• Simulate intervention rollouts and counterfactuals (via scenario scripts)
• Quantify effect of SDoH and adherence disruptors
• Life expectancy and YLL utilities via mighti/analysis/life_expectancy.py and demography analyzers (optional; not shown in the default mighti_main.py example)
• Prevalence calibration via Optuna (mighti/calibration/calibration_diseases_ver2.py)

Research-specific analyses (e.g., multi-condition life expectancy attribution) live in companion project repositories that import MIGHTI as a library.

---

Wiki Pages for MIGHTI

• Getting Started — Installation, dependencies, and how to run your first simulation.

• Parameter Preparation — Guide to preparing and formatting inputs such as p_death, dur_condition, and demographic files.

• Health Conditions — Overview of how diseases (e.g., HIV, diabetes, cancer) are modeled in mighti/diseases/.

• Disease Interactions — HC-HC and HIV-HC connectors in mighti/interactions/connectors.py.

• Adherence and CASM — Adherence pipeline in mighti/interventions/adherence.py.

• Social Determinants of Health — SDoH modules in mighti/sdoh/.

• Interventions — HIV wrappers, SDoH programs, and disease-specific treatments in mighti/interventions/.

• Analyzers — Outcome tracking modules in mighti/analyzers/.

• Calibration — Fitting p_acquire (and optionally p_death) to observed prevalence with Optuna.

• Simulation Examples — End-to-end run of mighti_main.py and plotting helpers.

• Microcosting and Cost-Effectiveness Analysis — MicrocostingAnalyzer and CEA workflow.