Bidirectional contact tracing could dramatically improve COVID-19 control

William J. Bradshaw, Ethan C. Alley, Jonathan H. Huggins, Alun L. Lloyd, Kevin M. EsveltNature Communications 2021 · DOI

TL;DR

This paper demonstrates that "bidirectional" contact tracing, which identifies both those exposed and those who may have infected others, significantly outperforms traditional "forward-only" tracing for COVID-19 control. By extending tracing windows or leveraging high-uptake digital tools, bidirectional tracing can more than double the reduction in the effective reproduction number (R<sub>eff</sub>), offering greater resilience to imperfect case detection and testing.

What Problem Does This Paper Address?

The ongoing COVID-19 pandemic highlighted the critical role of contact tracing in controlling transmission. However, existing "forward-tracing" protocols primarily focus on identifying individuals who have been exposed to a confirmed case. This approach can miss significant transmission chains, particularly those driven by asymptomatic or undiagnosed individuals, who are known to contribute substantially to the spread of SARS-CoV-2. The paper addresses the gap in understanding how to optimize contact tracing strategies to be more robust and effective, especially in the face of imperfect real-world conditions like low case ascertainment and variable test sensitivity.

Methods

The researchers employed a stochastic branching-process model, adapted from previous work on SARS-CoV-2 forward-tracing. This model simulates disease spread and the effectiveness of different tracing interventions. They explored various scenarios by varying parameters such as:

  • Tracing Strategy: Forward-only vs. Bidirectional (manual and hybrid digital/manual).
  • Tracing Window: Expanding the time frame for identifying contacts (e.g., from 2 to 6 days pre-symptom onset).
  • Digital Exposure Notification Uptake: The proportion of cases with participating smartphones and those who share their data.
  • Case Ascertainment: The percentage of symptomatic cases identified by health authorities.
  • Test Sensitivity: The accuracy of diagnostic tests.
  • Basic Reproduction Number (R<sub>0</sub>): The inherent transmissibility of the virus.

The model's output focused on the effective reproduction number (R<sub>eff</sub>) and the percentage of outbreaks controlled.

Datasets

Not discussed in the paper. The study relies on a simulation model parameterized with data from existing literature on COVID-19 epidemiology.

Key Insights and Ideas

  • Bidirectional Tracing is a Game Changer: The core insight is that looking "backward" to identify infectors, in addition to "forward" to identify the exposed, dramatically improves control. This is because infectors are statistically more likely to have infected multiple individuals, making them high-value targets for breaking transmission chains.
  • Extended Tracing Windows are Crucial: For diseases like COVID-19 with significant pre-symptomatic transmission, a short tracing window (e.g., 2 days) is insufficient. Expanding this window to 6 days pre-symptom onset offers substantial gains, even with manual tracing alone.
  • Digital Tracing's Potential and Pitfalls: While digital exposure notification (DEN) offers speed and scale, its effectiveness is highly sensitive to user uptake and data-sharing rates. High uptake (near universal smartphone participation and data sharing) is needed for DEN to rival the performance of extended manual tracing.
  • Hybrid Approaches Offer Robustness: Combining manual and digital tracing can leverage the strengths of both. Hybrid systems are particularly effective when digital uptake is high or when manual tracing windows are short, offering a more resilient approach.
  • Bidirectional Tracing is More Resilient: The study highlights that bidirectional tracing is significantly more robust to challenges like low case ascertainment and reduced test sensitivity compared to forward-only tracing. This makes it a more practical strategy in real-world scenarios where these factors are often suboptimal.

Key Results

  • Doubling of R<sub>eff</sub> Reduction: Bidirectional tracing, particularly with an extended 6-day manual tracing window, can more than double the reduction in R<sub>eff</sub> compared to forward-only tracing.
  • Significant Improvement with Extended Window: Extending the manual tracing window from 2 to 6 days pre-symptom onset resulted in an approximately 85% improvement in performance compared to a 2-day bidirectional window and a 275% improvement over forward-only tracing.
  • High-Uptake Digital Tracing Equivalence: To match the performance of a 6-day manual bidirectional tracing window, digital exposure notification requires nearly 100% of smartphone users to participate and share data.
  • Robustness to Imperfect Data: Bidirectional tracing showed significantly lower R<sub>eff</sub> values across a range of case ascertainment rates and test sensitivities compared to forward-only tracing.
  • Hybrid Performance: High-uptake hybrid tracing with a 6-day manual window achieved substantial R<sub>eff</sub> reductions, and even low-uptake hybrid tracing with a 2-day manual window outperformed manual-only approaches.

Limitations and Future Work

  • Model Simplifications: The model does not account for demographic, geospatial, or behavioral variations, nor does it distinguish between mild and severe symptoms. It also assumes only true cases are included, simplifying the trade-off between test sensitivity and specificity.
  • Real-World Implementation Challenges: The paper acknowledges that increasing the manual tracing window would require more resources and personnel, and the effectiveness of digital tracing hinges on achieving very high participation rates, which may be difficult to attain with opt-in systems.
  • Future Work Suggestions: The authors suggest that future research should explore the efficacy of bidirectional tracing using a wider variety of epidemiological models to better quantify the ratio of benefits to costs. They also propose that efficient prioritization of forward and backward tracing can minimize costs, and that digital systems could be designed to encourage self-quarantine for recent contacts while prioritizing prompt testing and tracing of earlier contacts.

Summary updated

Abstract

Contact tracing is critical to controlling COVID-19, but most protocols only "forward-trace" to notify people who were recently exposed. Using a stochastic branching-process model, we find that "bidirectional" tracing to identify infector individuals and their other infectees robustly improves outbreak control. In our model, bidirectional tracing more than doubles the reduction in effective reproduction number (R<sub>eff</sub>) achieved by forward-tracing alone, while dramatically increasing resilience to low case ascertainment and test sensitivity. The greatest gains are realised by expanding the manual tracing window from 2 to 6 days pre-symptom-onset or, alternatively, by implementing high-uptake smartphone-based exposure notification; however, to achieve the performance of the former approach, the latter requires nearly all smartphones to detect exposure events. With or without exposure notification, our results suggest that implementing bidirectional tracing could dramatically improve COVID-19 control.
BibTeX
@inproceedings{bradshaw2021bidirectional,
  author = {William J. Bradshaw and Ethan C. Alley and Jonathan H. Huggins and Alun L. Lloyd and Kevin M. Esvelt},
  title = {{Bidirectional contact tracing could dramatically improve COVID-19 control}},
  booktitle = {Nature Communications},
  year = {2021},
  doi = {10.1038/s41467-020-20325-7},
}
Figure