A multiscale mathematical model integrating viral load improves reconstruction of transmission chains and epidemiological parameters in household studies

Source abstract

Abstract Household studies combined with mathematical modelling are central to inference of respiratory pathogen transmission, but they usually rely on symptom onset or follow-up, final infection status, or qualitative PCR results. Quantitative viral-load measurements, although directly derived from PCR assays, are rarely used to reconstruct individual infection trajectories and transmission links. We developed a Bayesian modeling framework integrating within-host viral dynamics and household transmission processes. Using simulated household outbreaks, we compared three observation designs: based on time of symptom onset, longitudinal qualitative PCR testing, and longitudinal quantitative viral load sampling. We evaluated their ability to recover unbiased infection times, transmission chains, generation intervals, incubation periods, transmission probabilities, and individual infectiousness profiles. Quantitative viral-load improved transmission chains reconstruction compared with symptom onset and qualitative PCR designs. Across transmission-probability settings, perfect reconstruction of transmission chains reached 87-93%, compared with 46-57% using symptom-onset data and 68-78% using qualitative PCR. The correct number of generations of transmission was recovered in 94% of households. Variability in generation intervals, incubation periods, and transmission probabilities were recovered, whereas simpler designs mainly recovered average quantities and failed to capture inter-individual heterogeneity. These findings support integrating longitudinal viral-load measurements into household transmission studies to improve inference of transmission dynamics and individual infectiousness. Author Summary Respiratory viruses often spread within households. Most household studies rely on symptom onset or simple positive or negative PCR results, which provide limited information about how infectious each individual was over time. It is therefore difficult to determine exactly when each person became infected and who infected whom. We developed a computational framework that combines individual viral-load trajectories with a household transmission model to reconstruct infection times, transmission chains, and key epidemiological quantities. Using simulated household outbreaks, we compared three types of study data: symptom onset only, repeated qualitative PCR testing, and repeated quantitative viral-load measurements. We found that quantitative viral-load data substantially improved reconstruction of transmission chains and recovered variation in infectiousness between individuals, whereas symptom-based and qualitative PCR designs mainly recovered average quantities and missed important heterogeneity. These results suggest that collecting longitudinal quantitative viral-load measurements can make household transmission studies more informative. More broadly, our framework shows how within-host biological measurements can be integrated with between-host transmission models to improve inference of infectious disease dynamics.