The researchers built their analysis around a mathematical branching-process model, sometimes described as an epidemic tree, that tracks who infected whom through successive generations of a growing outbreak. This transmission genealogy is essential for simulating contact tracing, because tracing only works if the model knows which secondary cases belong to which confirmed positive. Infected individuals in the model pass through a presymptomatic and infectious phase, develop symptoms after an incubation time drawn from a lognormal distribution, and then remain infectious and symptomatic for a period matching the length of their presymptomatic phase, an assumption motivated by the nearly symmetrical infectiousness profile reported for SARS-CoV-2. A fraction of cases remain asymptomatic throughout and can only be caught through contact tracing.

Three strategy parameters govern the model’s mitigation machinery. The test turnaround time measures how long elapses between ordering a test and receiving the result, a delay that includes both waiting for the test itself and waiting for the analysis. Test sensitivity is the probability that a truly infected person tests positive. Tracing efficiency is the fraction of secondary cases successfully found and isolated after a positive result. These parameters are neither fixed nor independent: surges in case numbers can stretch turnaround times, which slows tracing and allows more secondary infections, creating a vicious cycle, while faster tests such as rapid antigen tests generally sacrifice sensitivity compared with slower PCR-based analysis.

To compare strategies, the team computed the effective reproduction number under each combination of parameters, averaging results over hundreds of simulations with offspring drawn from a negative binomial distribution that captures superspreading, with a basic reproduction number of two and parameters otherwise calibrated to SARS-CoV-2. For each choice of tracing efficiency and turnaround time for the slow test, they identified a breaking point: a rapid-test sensitivity below which the accurate test wins and above which the fast test wins. With a tracing efficiency of 80 percent, a rapid test with a one-day turnaround remained preferable even as its sensitivity dropped considerably, whereas when the accurate test’s results took six days to arrive, a rapid test with sensitivity as low as 0.76 still outperformed it.

The broader picture emerged when the researchers varied tracing efficiency across the full range. The breaking-point curves all originate in the lower-right corner of the parameter space, where tests are perfectly accurate but no tracing occurs, and sweep leftward as tracing improves. In other words, the faster test only earns its advantage when contact tracing is there to exploit its speed. For a disease with an infectiousness profile similar to COVID-19, highly accurate tests with turnaround times under six days generally reduce transmission better than less accurate rapid tests for most parameter choices, but highly effective tracing flips the balance in favor of speed.

The shape of the infectiousness profile itself proved decisive. When the team simulated a disease that was twice as infectious during the presymptomatic phase as in the later half of infection, all breaking-point lines shifted left, meaning fast turnaround became more valuable. The logic is intuitive: when most transmission happens early and in people who feel perfectly healthy, identifying infected individuals quickly is paramount, and speed can compensate for a lower probability of detection. Delays, by contrast, waste the narrow window in which intervention does the most good.

The study then relaxed a key simplifying assumption. Real rapid antigen tests do not have a constant sensitivity; their performance tracks the viral load of the infected person. Drawing on a meta-analysis by Brümmer and colleagues that reported an average sensitivity of 76.3 percent for rapid antigen tests, rising above 90 percent in cases with high viral loads, the researchers modeled sensitivity as a weighted function of the infectiousness profile, keeping the average over the course of disease fixed while raising sensitivity on the days when infected people are most contagious. Under this more realistic assumption, the rapid test fared markedly better: for empirically plausible sensitivity values and nearly all levels of contact tracing, the rapid test beat the PCR-style test whenever the slower test’s turnaround exceeded four days. Notably, increasing tracing efficiency from zero to 20 percent substantially boosted the rapid test’s performance, while beyond that threshold both test types benefited similarly, producing nearly vertical breaking-point lines that signal diminishing marginal returns.

Perhaps the most striking result concerns what happens when contact tracing is absent altogether. The researchers proved analytically that in the no-tracing limit, a slow test with perfect sensitivity always reduces infections at least as well as any faster, less accurate alternative, even one with zero turnaround time. The reason lies in the model’s assumption that people isolate while awaiting results. Perfect self-isolation before and during testing does much of the mitigation work, and a false negative releases an infected person early, causing more onward transmission than the delay of an accurate test ever would. Curiously, this means that if everyone adheres strictly to isolation upon symptoms, a strategy of self-isolation without any testing performs at least as well as testing with perfect sensitivity and no delay, although the authors note the practical economic cost of isolating people whose similar symptoms stem from other illnesses.

The analytical framework also clarified how the pieces fit together. The team derived a closed-form expression for the effective reproduction number in which the mitigation policy appears as a subtraction term from the basic reproduction number, and they showed that the contribution of test sensitivity is linear in the no-tracing case. Extending the analysis to contact tracing introduced genuine complexity, because the probability of being traced depends recursively on whether one’s infector tested positive, which depends on whether the infector was traced in turn, and tracing only succeeds if the traced person is still actively infectious. The researchers resolved this with Monte Carlo estimates of the conditioned distributions sampled from their simulations, achieving close agreement between theory and simulation across most of the parameter space, with divergence appearing only near the epidemic threshold where simulated outbreaks die out quickly.

The authors are careful to frame their conclusions conservatively. Several model assumptions, including perfect isolation compliance, no isolation after a false negative, a relatively low basic reproduction number, and a perfectly sensitive slow test, all tilt the comparison in favor of the accurate test, meaning their findings likely underestimate the benefits of rapid testing. The model considers only forward tracing of descendants of confirmed cases, excludes mass screening and two-way tracing, and does not model explicit tracing delays, each of which could shift the balance further. Still, the practical message is clear: when testing infrastructure comes under severe stress and waiting times balloon, switching to rapid, less sensitive tests can be a viable and even preferable strategy for curbing transmission, provided that contact tracing capacity stands ready to convert fast diagnoses into fast isolation of secondary cases.

Subject of Research: Trade-offs between test sensitivity, turnaround time, and contact tracing efficiency in test-trace-isolate epidemic mitigation strategies

Article Title: Are fast test results preferable to high test sensitivity in contact-tracing strategies?

Article References: Boilesen, M. R., Græsbøll, K., & Juul, J. L. (2026). Are fast test results preferable to high test sensitivity in contact-tracing strategies?. PLOS Complex Systems, 3(10), e0000134. https://doi.org/10.1371/journal.pcsy.0000134

Image Credits: AI Generated

DOI: 10.1371/journal.pcsy.0000134

Keywords: epidemiology, contact tracing, rapid antigen tests, test sensitivity, turnaround time, branching process model, SARS-CoV-2, test-trace-isolate, effective reproduction number, viral load, mathematical modeling, public health