The COVID-19 pandemic necessitated contact tracing strategies to control the spread of the virus following lockdown relaxations. Technology-based solutions, such as contact tracing smartphone apps, were proposed to automate this process. These apps leverage Bluetooth technology to track close contacts, notifying users of potential exposure upon infection. However, debates arose regarding their effectiveness and privacy implications. Existing studies using various methodologies explored optimal adoption rates and necessary complementary measures (social distancing and testing). While some analytical models suggested optimism, others highlighted potential negative consequences, like overwhelming the testing system if adoption rates were high but testing capacity low. This study aimed to address these complexities using an agent-based model to simulate the interplay of app usage, testing, and behavioral factors.

The paper reviews several studies on COVID-19 mitigation strategies and the use of contact tracing apps. Some analytical mathematical models demonstrated that instantaneous contact tracing, as enabled by apps, could lead to epidemic control with a sufficient adoption rate (at least 60%). However, other models highlighted potential issues such as unrealistic isolation numbers if testing capacity was insufficient. A developing model even suggested that the app might be ineffective compared to alternative measures like random testing. The authors acknowledge the range of approaches and the conflicting conclusions presented in the literature, setting the stage for their own investigation using a more detailed agent-based model.

The researchers developed an agent-based model (ABM) to simulate COVID-19 transmission and mitigation strategies within a synthetic population representing Glasgow, Scotland. The model incorporated a multi-layered social network representing households, friendships, workplaces, schools, and random contacts. Each agent had individual characteristics (age, gender) and behaviors. Daily contacts were simulated based on social network structure and contact frequencies. Viral transmission probability varied across contact types (household contacts having a higher transmission probability than random contacts). Disease progression followed an age- and gender-dependent SEIR model with parameters derived from COVID-19 patient data. Mitigation strategies included contact tracing apps (CTA), COVID-19 testing, and agent compliance with self-isolation. The model simulated various scenarios combining different levels of CTA adoption, testing capacity, compliance with self-isolation, and testing policies (prioritizing symptomatic cases vs. no priority). The model was calibrated to match the basic reproduction number (R0) estimated for the UK. The model's contact patterns were validated by comparison to data from a UK population contact survey. A total of 140 scenarios were simulated, each repeated 20 times to account for stochasticity.

The study's key findings revolve around the interaction between contact tracing app adoption, testing capacity, and testing policy. * **Testing without tracing:** Increasing testing capacity significantly reduced infections, plateauing around 3% of the population being tested weekly, which was sufficient to test all symptomatic cases and a portion of those notified through other means (relatives, classmates, etc.). * **CTA with priority to symptomatic cases:** Introducing the contact tracing app (CTA) while prioritizing symptomatic cases for testing showed a strong synergistic effect. Higher CTA adoption rates consistently led to lower infection rates, irrespective of testing capacity. With 80% CTA adoption, infections dropped substantially. * **CTA without priority to symptomatic cases:** When testing capacity was limited and symptomatic cases were not prioritized, increasing CTA adoption did not always reduce infections. High CTA adoption rates, coupled with limited testing capacity, led to more infections than scenarios without the CTA. This counterintuitive effect occurred because the increased demand for tests from CTA notifications overwhelmed the system, preventing many symptomatic individuals from getting tested. * **CTA user compliance:** High compliance with self-isolation among CTA users led to slightly lower infection rates, especially at lower testing capacities and higher CTA adoption rates. However, even with low compliance, the CTA still showed a significant infection reduction. In essence, the model showed that the CTA's effectiveness hinges on adequate testing capacity and a testing policy prioritizing symptomatic cases. Without this, the increased demand for tests from the app could be detrimental.

The results highlight the intricate interplay between technological interventions and public health infrastructure. The contact tracing app, while offering a promising tool for mitigating COVID-19, relies heavily on sufficient testing capacity and effective resource allocation. The model's finding that prioritizing symptomatic cases is crucial underscores the need for well-defined testing strategies. The study also sheds light on the importance of considering behavioral factors, showing that even with imperfect compliance, the CTA remains effective. The findings provide important policy implications, emphasizing the need for governments to invest in both technological solutions and robust testing systems to maximize the benefits of contact tracing apps. The observed synergistic effect between testing and the CTA suggests that technological interventions should be integrated into a comprehensive public health response.

The study concludes that smartphone-based contact tracing is a viable mitigation strategy, but its success depends on sufficient testing capacity and a well-managed testing policy prioritizing symptomatic individuals. Higher CTA adoption leads to lower infection rates, provided that the testing system can handle the increased demand. The model suggests the importance of a coordinated approach, integrating technological advancements with strengthened public health infrastructure. Further research could investigate the impact of heterogeneous adoption rates across demographics, the effects of technical limitations of the CTA, and the development of more sophisticated testing and resource allocation strategies.

The model simplifies reality, relying on assumptions and parameters estimated from emerging evidence. The contact patterns were validated, but the individual-level transmission probabilities were calibrated using aggregated R0 data, which might not uniquely represent the actual distribution of contacts and transmission rates across various social settings. The model also assumes a perfect functioning CTA and accurate, timely test results, potentially overestimating the app's effectiveness in real-world scenarios. Finally, the model doesn't consider demographic variations in app adoption, potentially affecting the generalizability of results.