Calibration of individual-based models to epidemiological data: A systematic review

Area of Science:

• Epidemiological modeling
• Computational epidemiology
• Public health informatics

Background:

• Individual-based models (IBMs) are vital for informing public health policy on infectious disease spread.
• Effective calibration of IBMs requires robust parameter-search strategies and goodness-of-fit (GOF) measures.
• Current reporting standards for IBM calibration methods in infectious disease epidemiology are inconsistent.

Purpose of the Study:

• To review and analyze calibration methods used in individual-based models for infectious disease spread.
• To assess the reporting quality of parameter-search strategies and GOF measures in published epidemiological studies.
• To identify gaps and recommend improvements in the calibration of IBMs for public health policy.

Main Methods:

• Systematic literature review of simulation-based calibration methods for IBMs.
• Search of PubMed for studies on HIV, tuberculosis, and malaria epidemiology (2013-2018).
• Inclusion criteria: models storing individual-specific data and calibration against population targets.

Main Results:

• Eighty-four articles met the review criteria; 48% used quantitative GOF with algorithmic search strategies.
• The remaining 52% had unidentified or informal search strategies, with limited quantitative GOF assessment.
• Only 17% provided a rationale for their calibration method choice, and 37% reported model validation.

Conclusions:

• Reporting on calibration methods in epidemiological modeling studies is suboptimal.
• Adoption of well-documented, algorithmic calibration methods can enhance reproducibility and inference quality.
• Further research is needed to compare the performance of different calibration strategies and GOF measures.