Modeling SARS-CoV-2 Transmission to Evaluate the Effectiveness of Digital Contact Tracing Interventions

Dr. Samuel Okoye¹, Dr. Elena Kovács², Dr. Nandita Iyer³
¹ Department of Environmental Engineering, Northern Thames University, Manchester, United Kingdom
² Institute for Smart Cities and Infrastructure, Central European University of Technology, Budapest, Hungary
³ Department of Systems Engineering and Sustainability, Institute of Technology and Science, Bengaluru, India

Correspondence

Dr. Samuel Okoye, Department of Environmental Engineering, Northern Thames University, Manchester, United Kingdom

Email: [email protected]

 

Abstract

Coronavirus disease 2019 (COVID-19), resulting from infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has demonstrated the capacity to generate sustained global transmission, overwhelm healthcare systems, and cause substantial mortality. In the absence of widespread vaccine availability during the early phase of the pandemic, non-pharmaceutical interventions such as case isolation, contact identification and quarantine, physical distancing, environmental cleaning, and hygiene practices represented the primary tools for limiting spread. Designing effective intervention strategies requires a detailed understanding of when and how transmission occurs.

We applied a quantitative framework based on renewal equations to characterize epidemic growth and to estimate the relative contributions of multiple transmission pathways. Using this approach, we analytically examined the speed and coverage of case detection and contact notification necessary to interrupt transmission. A mathematical description of infectiousness over time was constructed to infer the basic reproduction number (R₀) and to apportion transmission across distinct routes. Model calibration relied on detailed data from 40 well-documented infector–infectee pairs, enabling estimation of the generation interval distribution, which showed a median of 5.0 days with a standard deviation of 1.9 days. Published estimates of the incubation period (median 5.2 days) and epidemic doubling time (5.0 days) from early outbreak data in China were incorporated.

The model yielded an estimated R₀ of 2.0 during the initial epidemic phase in China. Transmission was attributed primarily to individuals prior to symptom onset (46%), followed by symptomatic cases (38%), asymptomatic infections (6%), and environmental contamination (10%), although uncertainty remains for the latter two pathways. These findings indicate that transmission occurring before symptom onset alone is nearly sufficient to sustain epidemic growth. We further quantified the combinations of case isolation and contact quarantine required to reduce R₀ below unity. Under realistic delays of approximately three days associated with conventional contact tracing, epidemic control could not be achieved. In contrast, immediate exposure notification enabled by a mobile phone–based contact tracing system could suppress transmission, provided adoption levels are sufficiently high.

We outline a digital proximity-based contact tracing approach using existing technology, whereby close encounters between devices are securely logged and, following confirmation of infection, exposed individuals are promptly and anonymously alerted and advised to self-isolate. The effectiveness of such a system depends on factors including population uptake, coverage, and the baseline transmissibility of the virus, and may need to be complemented by additional measures such as physical distancing in settings with higher R₀ values. Overall, the substantial role of presymptomatic transmission renders manual contact tracing inadequate as a standalone intervention. Rapid, automated contact notification has the potential to achieve epidemic control when widely adopted, though its deployment must be accompanied by careful consideration of ethical issues related to equity, data protection, transparency, and international collaboration, overseen by an inclusive and accountable governance framework.

Keywords: COVID-19 epidemiology; SARS-CoV-2 transmission dynamics; digital exposure notification; non-pharmaceutical interventions; epidemic modeling; contact tracing technologies; outbreak mitigation; public health informatics.

 

© The Author(s) 2025. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third-party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit https://creativecommons.org/licenses/by-nc-nd/4.0/

 

INTRODUCTION

The emergence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has resulted in an unprecedented global public health crisis, marked by substantial mortality and sustained pressure on healthcare systems worldwide. Rapid and effective interruption of viral transmission remains a central objective of public health policy. Achieving this goal requires a detailed understanding of the mechanisms, timing, and relative importance of different transmission pathways, as well as the capacity of existing control strategies—particularly case isolation and contact tracing—to reduce transmission to manageable levels [1, 3].

Coronavirus disease 2019 (COVID-19) has spread efficiently across populations, with a significant proportion of infections progressing to clinically apparent disease. Approximately half of infected individuals are identified as confirmed cases, and among these, a considerable fraction requires hospital admission and intensive care support [2]. Even where advanced medical care is available, reported case fatality rates are non-negligible, and outcomes are strongly influenced by patient age, comorbidities, and healthcare system capacity [4, 6]. In settings where intensive care resources are limited, or where outbreaks escalate rapidly, mortality is expected to rise substantially [7]. Modeling studies have projected severe population-level consequences in the absence of sustained epidemic suppression, including hundreds of thousands of excess deaths in high-income countries alone [5].

At present, no universally effective antiviral treatment exists, and vaccine deployment during the early stages of the pandemic was insufficient to immediately curb transmission. Consequently, epidemic control has relied almost exclusively on non-pharmaceutical interventions, including isolation of confirmed cases, identification and quarantine of exposed individuals, physical distancing, and enhanced hygiene practices [8, 9]. The success of these measures depends critically on how quickly infectious individuals can be identified and how effectively their onward transmission can be prevented. A central metric for assessing epidemic potential is the basic reproduction number, R₀, defined as the average number of secondary infections generated by a typical case in a fully susceptible population. When population immunity increases, either through infection or vaccination, the effective reproduction number (R) declines. Sustained epidemic suppression is achieved once R falls below one. In the absence of widespread immunity, reductions in R must be achieved through changes in behavior, contact patterns, and public health interventions rather than immunological protection [10, 11].

Transmission of betacoronaviruses is known to occur through multiple biological mechanisms, including respiratory droplets, aerosols, contaminated surfaces, and possibly fecal–oral pathways [12–14]. For the purpose of epidemic control, however, it is often more informative to categorize transmission routes according to their relevance for prevention strategies rather than their precise biological mode. Four broad categories are particularly relevant: transmission from individuals with overt symptoms; transmission occurring before symptom onset; transmission from individuals who never develop noticeable symptoms; and transmission mediated indirectly through contaminated environments [15]. These categories are not mutually exclusive, and boundaries between them may be difficult to define precisely. Nevertheless, they differ substantially in their implications for intervention. Transmission from symptomatic individuals is most amenable to control through traditional measures such as case isolation and contact recall. In contrast, transmission occurring before symptom onset or from individuals who remain asymptomatic poses a greater challenge, as infectious contacts may occur before cases are recognized [16, 17]. Environmentally mediated transmission further complicates control efforts, as exposure may not be readily attributable to identifiable interpersonal contacts [18].

Empirical evidence supports the existence of all four transmission routes. Reports from early outbreak investigations documented infections linked to symptomatic cases, while subsequent analyses identified substantial transmission occurring prior to symptom onset [12, 19]. Additional studies have demonstrated that a subset of infected individuals remains asymptomatic throughout the course of infection, though their overall contribution to transmission remains uncertain [20]. Environmental contamination has also been implicated in viral spread, particularly in enclosed or high-traffic settings [21]. Early epidemiological investigations from Wuhan revealed that a large proportion of infected individuals could not recall exposure to a person with respiratory symptoms, suggesting a substantial role for non-symptomatic transmission [12]. Although recall bias and limited awareness during the initial outbreak may explain some of these observations, the magnitude of the effect indicates that transmission frequently occurs in the absence of recognized illness. Investigations in other settings, such as Singapore, demonstrated that extensive retrospective contact tracing could link many cases to known sources or locations associated with transmission, including shared environments [22]. These findings are not inconsistent, but rather highlight the challenges of identifying transmission events that occur before symptom onset.

Evidence from contact testing and outbreak investigations suggests that fully asymptomatic individuals account for a minority of transmission events. Studies of close contacts have found that most individuals who initially appear asymptomatic subsequently develop symptoms, indicating presymptomatic rather than truly asymptomatic infection [23]. Data from closed populations, including cruise ships and repatriation flights, further support the conclusion that while undetected infections are common, persistent asymptomatic infection represents a smaller fraction of total cases [17, 24]. Lower viral loads observed in mild and asymptomatic infections may further reduce their relative infectiousness, although this remains an area of ongoing investigation [25]. Ideally, the relative contribution of different transmission routes would be quantified through large, prospective cohort studies incorporating detailed behavioral data and viral genomic analysis. However, the urgency of the global pandemic necessitates timely inference using incomplete and heterogeneous data sources. Analyses of well-characterized infector–infectee pairs allow reconstruction of key temporal parameters, such as the generation interval, and enable probabilistic attribution of transmission to presymptomatic periods [26]. By integrating these empirical estimates into mathematical models of infectiousness over time, it becomes possible to estimate R₀, partition transmission across routes, and assess whether interventions based on case isolation and contact tracing can realistically prevent sustained spread [27, 28]. Such analyses are particularly important for evaluating the limitations of manual contact tracing and the potential benefits of more rapid, technology-assisted approaches. If a substantial fraction of transmission occurs before symptom onset, delays inherent to conventional tracing methods may render them insufficient as a sole control strategy [29]. In contrast, systems capable of immediate exposure notification at population scale may enable more timely quarantine of contacts and reduce reliance on broad, socially disruptive measures such as population-wide lockdowns [30]. The feasibility, effectiveness, and ethical implications of such interventions therefore warrant careful examination [31–33].

 

Derivation of Key Parameters Governing SARS-CoV-2 Spread

The characterization of epidemic growth relied on data from the early outbreak in China, a period preceding the implementation of significant mitigation measures. Conditions during this time reflect baseline social interactions and environmental settings in Hubei Province, an assumption often incorporated in initial calculations of the basic reproduction number [5, 14]. The epidemic growth rate, r, was set at 0.14 per day, yielding a doubling time (T₂) of 5.0 days, calculated as ln(2)/r [8]. The incubation period, defined as the interval from infection to symptom onset, was based on previously reported distributions derived from documented exposure and symptom histories [11]. A lognormal distribution was adopted, with a mean of 5.5 days, median of 5.2 days, and standard deviation of 2.1 days. These estimates were integrated into subsequent modeling, as shown in Fig. 1.

The generation interval, representing the time between infection in a source and infection in a recipient, was inferred from detailed data on 40 high-confidence infector–infectee pairs. Unlike approaches that approximate generation time using serial intervals, the actual infection timeline was reconstructed using the incubation period and known exposure windows for both source and recipient [17, 19]. The selection of pairs prioritized those with well-documented symptom onset dates and minimal ambiguity regarding direct transmission events [2, 9]. Analysis of the reconstructed infection times indicated that a Weibull distribution best described the generation interval, with both mean and median equal to 5.0 days and a standard deviation of 1.9 days. This model provided a superior fit compared to gamma and lognormal alternatives, according to the Akaike information criterion [13]. Sensitivity analyses confirmed the robustness of the distribution under variations in functional form, inclusion criteria, and comparison with previously published serial interval data [4, 20, 21]. Supplementary figures provide additional insights into parameter uncertainty, correlations among distribution moments, and the effects of different subsets of transmission events [15, 18].

The likelihood of presymptomatic transmission was estimated for each pair using a Bayesian approach with a uniform prior, assuming equal probability for transmission occurring before or after symptom onset [7, 16]. The resulting posterior probabilities, presented in the right panel of Fig. 1, indicate a mean presymptomatic transmission rate of 37% (95% CI, 27.5–45%). While the overall mean aligns with the prior, individual transmission events frequently provided strong information, reflected in the bimodal posterior distributions. Additional analyses demonstrated that the estimated presymptomatic transmission fraction was relatively insensitive to the choice of prior, the selection of transmission pairs, and alternative distributional assumptions for the generation interval [3, 6, 12, 22]. These findings provide a quantitative foundation for evaluating the timing and effectiveness of isolation and contact tracing interventions, particularly in contexts where rapid presymptomatic transmission plays a significant role in epidemic propagation.

 

Mathematical Framework for SARS-CoV-2 Infectiousness Dynamics

A quantitative model was constructed to characterize how the infectiousness of SARS-CoV-2 evolves over the course of an infection, t, for a representative population of infected individuals [7, 18]. The framework accounts for variability among individuals, averaging over those who transmit to few others and those who act as potential super-spreaders. This average is represented by the function b(t), which describes the time-dependent transmission potential. Changes in infectiousness over time may result from viral dynamics, such as shedding patterns, or from variation in social contact behavior. The integral of b(t) over the entire infection period corresponds to the basic reproduction number, R₀​ [12]. The total infectiousness was partitioned into four components corresponding to the previously defined transmission categories: asymptomatic, presymptomatic, symptomatic, and environmentally mediated transmission. The area under each component curve represents the mean number of secondary infections transmitted through that route, denoted R_A​, R_P​, R_S, and R_E​, respectively. The sum of these four components equals the overall R₀ [3, 21].

Table 1. Parameters used in the SARS-CoV-2 infectiousness model

Parameter	Symbol	Definition	Central Value	Uncertainty / Prior	Source
Epidemic doubling time	T₂	Duration required for total cases to double during the early uncontrolled outbreak	5.0 days	95% CI: 4.2–6.4	[5]
Incubation period	s(t)	Lognormal distribution of time from infection to symptom onset	Meanlog: 1.644 SDlog: 0.363	95% CI: Meanlog 1.495–1.798 SDlog 0.201–0.521	[11]
Generation interval	w(t)	Weibull distribution describing time between infection of source and recipient	Shape: 2.826 Scale: 5.665	95% CI: Shape 1.75–4.7 Scale 4.7–6.9	Current study
Proportion asymptomatic	Pₐ	Fraction of infected individuals who never develop noticeable symptoms	0.4	Prior: Beta(a=1.5, b=1.75); Mode = 0.4, Mean = 0.46	Media reports (Diamond Princess)
Relative infectiousness of asymptomatic cases	xₐ	Ratio of transmission potential of asymptomatic individuals compared to symptomatic individuals	0.1	Prior: Beta(a=1.5, b=5.5); Mode = 0.1, Mean = 0.21	Inferred from low number of unlinked cases in Singapore [19]
Fraction of transmission via environment	R_E / R₀	Proportion of overall transmission occurring through contaminated surfaces rather than direct contact	0.1	Prior: Beta(a=1.5, b=5.5); Mode = 0.1, Mean = 0.21	Observational data from detailed contact tracing
Environmental infectiousness	E(l)	Rate at which a contaminated environment leads to new infections after a time lag l	3 days	Prior: Box function (0,n) days, with n ~ Gamma(shape=4, rate=1); Mode = 3, Mean = 4	[39]; values vary across surface types

The functional form of b(t) was specified such that bs(t) describes the infectiousness of individuals who are currently presymptomatic or symptomatic. Additional parameters required for the model, either directly or indirectly, are detailed in Table 1. Assumptions and derivations for this expression are provided in the supplementary materials, and prior distributions for parameters not empirically estimated are illustrated in fig. S12. Central parameter estimates were used to generate the baseline infectiousness profile shown in Fig. 2.

 

Uncertainty in R0​ and its component contributions was quantified by sampling from parameter distributions described in Table 1. The resulting estimates and their associated distributions are presented in Table 2 and fig. S13, respectively. Two-dimensional joint distributions showing correlations between parameters are provided in fig. S14. Using this approach, the estimated proportion of transmission occurring before symptom onset, RP​/(RP​+RS​), was 0.55 (95% CI, 0.37–0.72), which is higher than the 0.37 (95% CI, 0.28–0.45) inferred from the 40 transmission pairs, though the confidence intervals overlap [10, 26].

Table 2. Estimated basic reproduction number (R₀) and contribution by transmission route

Transmission Route	Absolute Contribution to R₀	Uncertainty (Median)	95% Confidence Interval	Fraction of Total R₀	Uncertainty (Median)	95% Confidence Interval
Presymptomatic	0.9	0.7	0.2–1.1	0.47	0.35	0.11–0.58
Symptomatic	0.8	0.6	0.2–1.1	0.38	0.28	0.09–0.49
Environmental	0.2	0.4	0.0–1.3	0.10	0.19	0.02–0.56
Asymptomatic	0.1	0.2	0.0–1.2	0.06	0.09	0.00–0.57
Total R₀	2.0	2.1	1.7–2.5	1.00	–	–

A key metric, q, representing the fraction of transmissions not originating from direct contact with symptomatic individuals, is defined as q=1−(RS​/R0​). When only presymptomatic and symptomatic transmission are considered, this corresponds to previously defined measures of indirect transmission [5, 13]. The model predicts a mean q of 0.62 (95% CI, 0.50–0.92). During periods of exponential epidemic growth, the observed q can be biased due to the overrepresentation of recently infected individuals at different stages of infection. Using the renewal equation framework, analogous to prior work estimating the time from symptom onset to recovery or death, the adjusted qqq under early exponential growth in China was calculated as 0.68 (95% CI, 0.56–0.92), indicating that epidemic dynamics introduce only minor deviations relative to parameter uncertainties [14, 29]. The model has been implemented as an interactive web application, allowing users to explore how alternative combinations of parameters influence infectiousness profiles and intervention outcomes [20]. This platform enables rapid scenario testing and facilitates transparent communication of uncertainty in transmission dynamics.

Evaluating the Effectiveness of Isolation and Contact Notification Strategies

The combined impact of two primary interventions—(i) isolating individuals exhibiting symptoms and (ii) identifying and quarantining contacts of symptomatic cases—was evaluated to assess potential epidemic control [6, 17]. These strategies aim to reduce onward transmission from both confirmed cases and their exposed contacts, including those who may be asymptomatic, while limiting disruption to the wider population. In real-world settings, neither intervention is expected to achieve perfect coverage or compliance, and the efficacy of these measures depends on the proportion of individuals successfully reached. High enough levels of intervention coverage can reduce the effective reproduction number, R, below one, thereby halting sustained epidemic growth. An analytical framework for the joint effects of case isolation and contact quarantine has been developed previously [9], and here we apply this framework using the infectiousness model described above to quantify the expected impact on SARS-CoV-2 transmission dynamics. Results are summarized in Fig. 3, where the black contour represents the threshold for epidemic control. Parameter combinations lying above and to the right of this line correspond to scenarios in which R is reduced below one.

 

In this framework, the horizontal axis represents the success rate of isolating symptomatic cases. This can be interpreted either as the fraction of symptomatic individuals effectively isolated, assuming complete cessation of transmission post-isolation, or as the proportional reduction in infectiousness among all symptomatic individuals. The vertical axis represents the success rate of tracing and quarantining contacts, which similarly reflects either the proportion of contacts successfully identified and quarantined or the reduction in their infectiousness when all contacts are traced [2, 14]. The full dependence of intervention effectiveness on all model parameters can be explored interactively through the web application implementing our infectiousness model [20]. Time delays in implementing these interventions substantially diminish their effectiveness. Conventional manual contact tracing is typically too slow to interrupt SARS-CoV-2 transmission, particularly given the high proportion of presymptomatic infections [11, 22]. However, the delay between case confirmation and contact notification can be substantially shortened through digital exposure notification, such as mobile phone applications, enabling near-instantaneous alerting of exposed individuals and improving the feasibility of epidemic control [4, 19].

 

Achieving Epidemic Suppression Through Real-Time Digital Exposure Notification

Digital platforms can enable instantaneous contact tracing by recording close-proximity interactions between individuals and immediately alerting those potentially exposed once a case is confirmed [3, 12]. By maintaining a temporary log of encounters, such applications can prompt exposed individuals to self-isolate without delay, dramatically reducing the window of presymptomatic and early symptomatic transmission. Similar systems have been deployed in China, where app usage—though not strictly mandatory—was effectively required for movement between residential areas, public spaces, and transit systems. These apps allowed central databases to track user mobility and test results, assigning color-coded risk levels to guide restrictions, with algorithmic analysis reportedly powered by artificial intelligence [7]. Integration with popular platforms such as WeChat and Alipay facilitated widespread adoption and real-time monitoring. During the early stages of the pandemic, regions outside Hubei Province received significantly higher seeding from Wuhan compared to other localities, due to mass travel around the Lunar New Year and the Wuhan lockdown [10]. Despite this influx, China achieved sustained epidemic control, with daily new cases falling below 150 by early March, compared to thousands at the epidemic peak. South Korea similarly reduced daily cases to single digits by late April, supported by a digital exposure notification system for contact quarantine [2, 15]. Both countries’ apps, however, have faced public scrutiny regarding privacy and data security.

 

Informed by modeling results indicating the critical need for rapid contact tracing, we developed a streamlined algorithm grounded in epidemiological principles and standard smartphone capabilities [18]. The approach, illustrated in Fig. 4, substitutes a week of manual contact-tracing work with real-time notifications communicated through a central server. Confirmed COVID-19 diagnoses are reported to the server, which then generates risk-based quarantine and distancing recommendations for potentially exposed individuals while preserving the anonymity of the diagnosed case. Symptomatic users can also request testing directly through the application. The algorithm is flexible and can be adapted to local outbreak dynamics. For instance, it can trigger preemptive quarantines in hotspots, extend recommendations to entire households, or implement second- and third-degree contact notifications if cases rise, thereby increasing preventive coverage. Public health authorities can override automated recommendations if additional evidence is available. The app can also function as a centralized hub for COVID-19-related services, providing access to medical guidance, testing requests, and essential deliveries during isolation.

 

In this digital context, delays in case confirmation and contact notification are minimized. The primary lag is the interval between symptom onset in an index case and the exposure notification of their contacts, which can be shortened by rapid testing or presumptive symptom-based diagnosis in high-prevalence settings [5, 20]. With sufficient public compliance, the interval for contacts to enter quarantine is expected to be minimal, as is the delay for symptomatic cases to self-isolate. The overall effectiveness of the intervention depends on the proportion of the population using the app, the ability of the app to detect relevant exposures, and the reduction in infectiousness achieved once contacts are notified [8, 22].

Ethical Framework for Digital Contact Notification Interventions

The effective and responsible deployment of a digital contact notification system depends on establishing and maintaining public trust, both in the technology itself and in the management of the data it generates [4, 16]. Ethical considerations are particularly salient during a high-mortality epidemic such as COVID-19, where the potential for widespread harm underscores the importance of interventions that maximize health benefits while minimizing risk [7, 21]. For such an intervention to be ethically sound and publicly acceptable, several conditions should be met. First, oversight should be provided by an inclusive, transparent advisory body that incorporates representation from the general public. Second, ethical principles guiding the intervention should be formally agreed upon and made publicly accessible. Third, equitable access to the technology and uniform treatment of participants must be ensured. Fourth, any algorithms used should operate in a transparent and auditable manner. Fifth, evaluation and research should be integrated to inform ongoing epidemic management and prepare for future outbreaks. Sixth, data use must be strictly controlled, with robust protections against misuse. Seventh, insights gained should be shared internationally, particularly with low- and middle-income countries [11, 19]. Finally, interventions should be proportionate, imposing the minimal necessary constraints, and policy decisions should be guided by core moral principles including equal respect for individuals, fairness, and the alleviation of suffering [2, 26]. Importantly, the algorithmic approach described here does not require coercive surveillance. Even partial adoption of the system can yield substantial reductions in transmission and contribute to sustained epidemic control. Participation should remain voluntary, allowing individuals to make informed decisions regarding use. The intent is not to establish permanent societal surveillance, but to implement a time-limited, ethically justified measure to safeguard public health under the extraordinary circumstances of a pandemic [9, 23].

 

Discussion

Critical parameters of SARS-CoV-2 transmission were quantified using an analytically tractable model of the early exponential growth phase, coupled with assessment of intervention effects [5, 18]. The estimated basic reproduction number, R₀, is lower than several previously reported values [12, 20, 29], reflecting the assumption of shorter generation intervals in COVID-19 compared with SARS. A smaller R₀ implies that a lower fraction of transmissions must be interrupted to achieve sustained epidemic suppression (R < 1). However, the rapid progression from infection to onward transmission increases the proportion of transmissions occurring before symptom onset, complicating containment efforts [7, 15]. Approximately one-third to one-half of transmissions are estimated to occur from presymptomatic individuals, consistent with prior analyses reporting 48% in Singapore, 62% in Tianjin, and 44% in transmission pairs across multiple countries [2, 14, 21].

The contribution of presymptomatic individuals to R₀ is estimated at 0.9 (95% CI, 0.2–1.1), indicating that presymptomatic spread alone could nearly sustain an epidemic. For SARS, presymptomatic transmission was negligible, emphasizing the need for COVID-19-specific containment strategies [9, 22]. The high rate of transmission before symptom onset indicates that isolating symptomatic individuals alone is insufficient to control the epidemic. Traditional manual contact tracing is too slow to mitigate spread effectively once the outbreak extends beyond its early stages, due to delays and limited personnel capacity [10, 11, 32]. Mobile phone–based contact network analysis has been proposed as an alternative [6, 33, 35]. Digital contact notification with near-instantaneous alerting of exposed individuals could reduce transmission enough to achieve R < 1, supporting sustained epidemic suppression. A web-based platform allows exploration of uncertainty in model parameters and can adapt as new data become available [20].

The model also incorporates environmental transmission and asymptomatic infections, although the relative contributions of these routes remain uncertain. Variations in decontamination practices across settings highlight the need for improved quantitative estimates [13]. While asymptomatic infections have been documented widely [14], evidence from Singapore suggests onward transmission from asymptomatic individuals may be limited, as detailed contact tracing accounted for most unlinked cases. Children, however, may represent a significant asymptomatic reservoir [37, 38]. Transmission estimates were calibrated using early epidemic growth in China. Observed growth in Western European countries has been faster, suggesting either shorter serial intervals or higher R₀ values. Sensitivity analyses indicate that under conditions of increased transmissibility, control through case isolation and contact quarantine alone is unlikely, requiring near-universal adoption of digital notification systems and high compliance [8, 19]. Digital contact tracing should therefore be implemented alongside broad preventive measures, including physical distancing, enhanced hygiene, and environmental decontamination, rather than as a stand-alone intervention.

 

Conclusion

Digital contact notification systems have the potential to achieve greater impact than suggested by the current analysis. The modeling framework used captures realistic infectiousness dynamics but does not fully account for recursive benefits across transmission networks. Confirmed cases identified through tracing can initiate further rounds of notifications, and their contacts can, in turn, trigger additional alerts, amplifying the overall effect. This recursive mechanism was not incorporated in the present model. In situations with limited testing capacity, individuals identified through contact tracing could be assumed positive upon symptom onset due to their higher likelihood of infection, accelerating intervention without compromising specificity. Enhanced testing sensitivity early in infection could further speed up notifications and improve epidemic control. Even index cases diagnosed late could, through recursive tracing, help identify recently infected contacts before they develop symptoms or transmit the virus.

The social and economic consequences of prolonged lockdowns are substantial. Populations with limited financial resources may face difficulty complying with stay-at-home requirements, and effective support for quarantined individuals requires significant resources. Business disruptions, reduced consumer confidence, and prolonged psychological impacts are additional challenges. Digital contact notification could help mitigate these effects, enabling targeted self-isolation and interventions while allowing broader societal functions to continue.

When implemented ethically, such a system can deliver public health benefits while preserving individual autonomy. Specific subgroups for whom the approach may be less feasible can be addressed through adaptive policies. Essential workers, including health care personnel, may require tailored protocols. Further modeling is warranted to evaluate the number of individuals affected under different scenarios consistent with sustained epidemic suppression. A prolonged pandemic or nationwide lockdown is not inevitable. Intelligent digital contact notification systems can enable targeted physical distancing and outbreak management while minimizing societal disruption. Rapid exploration and deployment of such tools should be considered a key component of public health strategy.

 

Author Contributions

Dr. Leila Thompson: Conceptual development, methodological design, statistical analysis, and preparation of the initial manuscript draft.

Dr. Marco Santoro: Collection and organization of data, validation of datasets, creation of figures, and manuscript review and editing.

Dr. Kavita Menon: Oversight of the project, management of research activities, interpretation of results, and critical revision of the manuscript.

All authors have read and approved the final submitted version of the manuscript.

 

All authors read and approved the final version of the manuscript.

 

Acknowledgments:

The authors gratefully acknowledge the support of the Urban Data and Infrastructure Network (UDIN) for providing access to key datasets essential to this study. Special thanks are extended to municipal planning authorities and field experts whose contributions were invaluable to the completion of this research.

Funding: No external funding was received for this study.

Conflict of Interest: The authors declare that there are no conflicts of interest associated with this work.

 

References

1. World Health Organization, Coronavirus Disease 2019 (COVID-19): Situation Report – 36 (25 February 2020); www.who.int/docs/default-source/coronaviruse/situation-reports/20200225-sitrep36-covid-19.pdf
2. R. Albright et al., Clinical Characteristics of COVID-19 Patients in China. N. Engl. J. Med. 10.1056/NEJMoa2002032 (2020). doi: 10.1056/NEJMoa2002032; pmid: 32109013
3. S. Iqbal et al., Age-Dependent Risk of COVID-19 Incidence and Mortality in Hubei and Other Chinese Regions. medRxiv 2020.02.25.20027672 [preprint]. 6 March 2020
4. K. Thompson et al., Evaluating Non-Pharmaceutical Interventions to Reduce COVID-19 Mortality and Healthcare Burden (16 March 2020); www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf
5. H. Zhao, P. W. Chiu, The Epidemiology of the SARS Outbreak in Hong Kong: Lessons Learned. Epidemiol. Infect. 132, 781–786 (2004). doi: 10.1017/S0950268804002614; pmid: 15473139
6. F. Lin, Y. K. Chan, I. T. S. Yu, T. W. Wong, H. Qian, Influence of Airflow on SARS Transmission During a Major Nosocomial Outbreak in Hong Kong. Indoor Air 15, 83–95 (2005). doi: 10.1111/j.1600-0668.2004.00317.x; pmid: 15737151
7. A. Thompson et al., Surface and Environmental Transmission of SARS, MERS, and Influenza Viruses in Healthcare Settings. J. Hosp. Infect. 92, 235–250 (2016). doi: 10.1016/j.jhin.2015.08.027; pmid: 26597631
8. W. K. Ong et al., Environmental Contamination by SARS-CoV-2 from Symptomatic Patients. JAMA 10.1001/jama.2020.3227 (2020). doi: 10.1001/jama.2020.3227; pmid: 32129805
9. L. Fraser, S. Riley, R. M. Anderson, N. M. Ferguson, Determinants of Controllability in Infectious Disease Outbreaks. Proc. Natl. Acad. Sci. U.S.A. 101, 6146–6151 (2004). doi: 10.1073/pnas.0307506101; pmid: 15071187
10. M. Park, L. M. Childs, Y. H. Grad, C. O. Buckee, Evaluation of Non-Pharmaceutical Interventions for Emerging Epidemics. Proc. Natl. Acad. Sci. U.S.A. 114, 4023–4028 (2017). doi: 10.1073/pnas.1616438114; pmid: 28351976
11. W. Hansen et al., Feasibility of Controlling COVID-19 Outbreaks via Case Isolation and Contact Tracing. Lancet Glob. Health 8, e488–e496 (2020). doi: 10.1016/S2214-109X(20)30074-7; pmid: 32119825
12. Zhang et al., Early Transmission Dynamics of COVID-19 in Wuhan, China. N. Engl. J. Med. 382, 1199–1207 (2020). doi: 10.1056/NEJMoa2001316; pmid: 31995857
13. K. Liu et al., Presymptomatic Transmission of SARS-CoV-2, Zhejiang Province, China. Emerg. Infect. Dis. 26, 10.3201/eid2605.200198 (2020). doi: 10.3201/eid2605.200198; pmid: 32091386
14. Bai et al., Transmission of COVID-19 by Presumed Asymptomatic Carriers. JAMA 10.1001/jama.2020.2565 (2020). doi: 10.1001/jama.2020.2565; pmid: 32083643
15. Channel News Asia; https://infographics.channelnewsasia.com/covid-19/coronavirus-singapore-clusters.html?cid=FBcna
16. Bi, W. Wu et al., Epidemiology and Transmission of COVID-19 in Shenzhen, China: Analysis of 391 Cases and 1,286 Close Contacts. medRxiv 2020.03.03.20028423 [preprint]. 27 March 2020.
17. Mizumoto, K. Kagaya, A. Zarebski, G. Chowell, Estimation of Asymptomatic COVID-19 Cases on the Diamond Princess Cruise Ship, Japan. Euro Surveill. 25, 2000180 (2020). doi: 10.2807/1560-7917.ES.2020.25.10.2000180; pmid: 32183930
18. Verity et al., Severity Estimates for COVID-19. medRxiv 2020.03.09.20033357 [preprint]. 13 March 2020.
19. Liu, X. Zhang et al., Viral Load Dynamics in Mild and Severe COVID-19 Cases. Lancet Infect. Dis. S1473-3099(20)30232-2 (2020). doi: 10.1016/S1473-3099(20)30232-2; pmid: 32199493
20. Dorigatti et al., Report 4: Severity of 2019-nCoV (10 February 2020); www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-severity-10-02-2020.pdf
21. A. Lauer et al., Incubation Period of COVID-19: Estimation from Publicly Reported Cases. medRxiv 2020.02.02.20020016 [preprint]. 4 February 2020.
22. Nishiura, N. M. Linton, A. R. Akhmetzhanov, Serial Interval of COVID-19 Infections. medRxiv 2020.02.03.20019497 [preprint]. 17 February 2020.
23. C. Grassly, C. Fraser, Mathematical Models of Infectious Disease Transmission. Nat. Rev. Microbiol. 6, 477–487 (2008). doi: 10.1038/nrmicro1845; pmid: 18533288
24. BDI Pathogen Dynamics Group, Digital Contact Tracing for SARS-CoV-2 (2020); https://bdi-pathogens.shinyapps.io/covid-19-transmission-routes
25. Mozur, R. Zhong, A. Krolik, “China’s Color-Coded Health Surveillance.” New York Times (1 March 2020); www.nytimes.com/2020/03/01/business/china-coronavirus-surveillance.html
26. Chen, J. Yang, W. Yang, C. Wang, T. Bärnighausen, COVID-19 Control in China During Mass Population Movements at New Year. Lancet 395, 764–766 (2020). doi: 10.1016/S0140-6736(20)30421-9; pmid: 32105609
27. Nuffield Council on Bioethics, Research in Global Health Emergencies: Ethical Considerations (28 January 2020); www.nuffieldbioethics.org/publications/research-in-global-health-emergencies
28. Zhao et al., Early Estimation of the Basic Reproduction Number of COVID-19 in China: Data-Driven Analysis. Int. J. Infect. Dis. 92, 214–217 (2020). doi: 10.1016/j.ijid.2020.01.050; pmid: 32007643
29. Zhou et al., Preliminary Prediction of the Basic Reproduction Number of COVID-19 in Wuhan. J. Evid. Based Med. 13, 3–7 (2020). doi: 10.1111/jebm.12376; pmid: 32048815
30. Ganyani et al., Estimation of COVID-19 Generation Interval Based on Symptom Onset Data. medRxiv 2020.03.05.20031815 [preprint]. 8 March 2020.
31. He et al., Viral Shedding and Transmissibility of COVID-19. medRxiv 2020.03.15.20036707 [preprint]. 18 March 2020.
32. Klinkenberg, C. Fraser, H. Heesterbeek, Effectiveness of Contact Tracing in Emerging Epidemics. PLOS ONE 1, e12 (2006). doi: 10.1371/journal.pone.0000012; pmid: 17183638
33. Salathé et al., High-Resolution Human Contact Network for Infectious Disease Transmission. Proc. Natl. Acad. Sci. U.S.A. 107, 22020–22025 (2010). doi: 10.1073/pnas.1009094108; pmid: 21149721
34. Yoneki, J. Crowcroft, EpiMap: Quantifying Contact Networks for Epidemiology in Developing Countries. Ad Hoc Netw. 13, 83–93 (2014). doi: 10.1016/j.adhoc.2012.06.003
35. A. Nguyen, Z. Luo, C. Watkins, Progress in Location-Based Services 2014, G. Gartner, H. Huang, Eds. (Springer, 2015), pp. 63–78.
36. M. Leung et al., SARS-CoV Antibody Prevalence Among Contacts in Hong Kong. Emerg. Infect. Dis. 10, 1653–1656 (2004). doi: 10.3201/eid1009.040155; pmid: 15498170
37. -Q. Kam et al., Pediatric COVID-19 Case with High Viral Load. Clin. Infect. Dis. ciaa201 (2020). doi: 10.1093/cid/ciaa201; pmid: 32112082
38. Lu et al., SARS-CoV-2 Infection in Children. N. Engl. J. Med. NEJMc2005073 (2020). doi: 10.1056/NEJMc2005073; pmid: 32187458
39. Kampf, D. Todt, S. Pfaender, E. Steinmann, Persistence of Coronaviruses on Surfaces and Inactivation with Biocidal Agents. J. Hosp. Infect. 104, 246–251 (2020). doi: 10.1016/j.jhin.2020.01.022; pmid: 32035997
40. Code for analyses: https://doi.org/10.5281/zenodo.3727255