JMIR Publications

ABSTRACT

Background:

During the COVID-19 pandemic, control policies were often informed by social contact patterns rather than observed transmission pathways. Understanding who infected whom, and in which settings, is essential for modelling and designing interventions that interrupt transmission of vulnerable populations while minimizing societal harms in other groups.

Objective:

This study aimed to identify age- and proximity-specific SARS-CoV-2 transmission patterns in South-West Germany using routine contact tracing data.

Methods:

We conducted a retrospective secondary analysis of contact tracing records from the Rhine-Neckar District (also responsible for Heidelberg city) Public Health Authority from January 27, 2020, to December 26, 2021. Each observation represented a dyadic relationship between a confirmed reference case and a close contact. Dyads were categorized by reference-contact age pairs between: young people aged 0-19 years, adults aged 20-65 years, and elderly people aged 66 years and older; as well as by proximity, ranging from shared name and address to lower-intimacy categories such as shared postcode with different name and address. The primary outcome was recorded transmission following contact. Observed and expected age-pair distributions were compared under random mixing assumptions. A generalized linear mixed-effects logistic regression model with random intercepts for reference cases assessed associations between age pairing, proximity, and transmission. Unadjusted relative risks were calculated for interpretability.

Results:

The dataset included 66,505 contacts linked to 20,077 reference cases, with 19,892 recorded transmissions. Contact and transmission patterns differed significantly from random mixing (P<.05) and were concentrated within age groups and high-intimacy settings. Young-to-elderly transmission occurred at only 0.15 times the expected level (95% CI 0.12-0.19), whereas elderly-to-elderly transmission occurred 5.89 times more often than expected (95% CI 5.12-6.94). Proximity, nevertheless, was the strongest predictor of transmission. The final mixed-effects model showed strong discrimination, with Tjur R2=0.515. Fixed effects explained 15.1% of outcome variance, increasing to 62.7% after inclusion of reference-case random effects. The intraclass correlation coefficient was 0.561, indicating substantial clustering by source case. Observed risk was highest for elderly-to-elderly contacts sharing the same name and address (69.1%; RR 1.35, 95% CI 1.27-1.43, relative to adult-to-adult shared-household contacts) and lowest for young-to-young contacts sharing only a postcode (5.5%; RR 0.11, 95% CI 0.10-0.12).

Conclusions:

In this regional contact tracing dataset, SARS-CoV-2 transmission was driven primarily by domestic proximity and source-case heterogeneity rather than broad intergenerational mixing. Young people were rarely observed as sources of infection for elderly people, while elderly-to-elderly transmission in shared domestic settings represented the highest-risk pathway. To enable real-world transmission risk estimation and more targeted public health responses, future contact tracing systems need to document: a reference between index or reference cases and their contact persons, data of potential relevance to transmission, and the disease outcome of the contact person.