Comparison of contact tracing methods: A modelling study.

Introduction: Contact tracing has been a key tool to contain the spread of diseases and was widely used by countries during the COVID-19 pandemic. However, evaluating the effectiveness of contact tracing has been challenging. Approaches to contact tracing were diverse and country-dependent, with operations utilizing different tracing methods under varied environments. To provide guidance on contact tracing for future preparedness, we assessed the effectiveness of contact tracing methods under varied environments using Singapore's population structure and COVID-19 as the disease model. Methods: We developed a transmission network model using Singapore's contact tracing data and the characteristics of COVID-19 disease. We explored three different tracing methods that could be employed by contact tracing operations: forward tracing, extended tracing and cluster tracing. The forward tracing method covered the period starting two days before case isolation, the extended tracing method covered the period starting 16 days before case isolation, and the cluster tracing method combined forward tracing with cluster identification. Contact tracing operations traced detected cases from surveillance and issued interventions for identified contacts, and we constructed combinations of varied scenarios to replicate variability during pandemic, namely low case-ascertainment or high case-ascertainment and either testing of contacts or quarantine of contacts. We examined the impact of varied contact tracing operations on disease transmission and provider costs. Results: Model simulations showed that the effectiveness of contact tracing methods varied under the four different scenarios. Firstly, under low case-ascertainment with testing of contacts, contact tracing reduced transmission by 12 %-22 %, with provider costs ranging between US$2943.56 to US$5226.82 per infection prevented. The most effective tracing method to control infection was cluster tracing, followed by extended tracing and forward tracing. Secondly, under low case-ascertainment with quarantine of contacts, transmission was reduced by 46 %-62 %, with provider costs below US$4000 per infection prevented. The cluster method reduced transmission by 62 %, enough to bring the reproduction number to close to unity and was the least costly. Extended tracing reduced transmission by 50 % but costed the most, while forward tracing reduced transmission by 46 %. Thirdly, under high case-ascertainment with testing of contacts, the average transmission was reduced by 20 %-26 %, with provider costs to prevent an infection ranging between US$1872.72 to US$3165.09. There was less variability between tracing methods, with cluster tracing reducing transmission the most, followed by extended tracing and forward tracing. Lastly, under high case-ascertainment and quarantine of contacts, contact tracing was the most effective, with provider costs below US$800 per infection prevented. All tracing methods were equally effective in disease containment, bringing the reproduction number below unity and stopping disease transmission early. Discussion: We conclude that contact tracing operated most effectively when implemented with high case-ascertainment rates and quarantine of contacts; disease transmission is stopped early, and the low number of contacts enable tracing operations to be more manageable and less costly. However, the pandemic situation can be dynamic, with fluctuations in resources available for case-ascertainment and quarantine adherence, which can impact the effectiveness of contact tracing. Adapting contact tracing methods to the situation can optimize disease control. Therefore, it is recommended to develop a flexible contact tracing approach that facilitates strategy switching based on resource availability and the skills of tracing operations.