New Model Shows COVID-19 Spread Is Concave: Impact of Each New Case Diminishes

Summary: A new coronavirus model finds that the spread of COVID-19 is concave: the incremental impact of each additional infected person diminishes as more people become infected.

Source: WUSTL

Key finding: Researchers at Washington University in St. Louis report that COVID-19 transmission does not scale linearly with the number of contagious individuals, as many standard models assume. Instead, transmission is concave—each additional infectious person contributes less to new infections when the number of infectious people is already high. The team attributes this pattern to overlapping social networks, where people repeatedly interact within the same households, workplaces and social circles.

The study, led by Meng Liu, Raphael Thomadsen and Song Yao of Olin Business School, adjusts the commonly used susceptible-infected-recovered (SIR) framework to allow the rate of spread to vary with the number of contagious people. This modification yields forecasts that align more closely with observed data, particularly the gradual declines in daily case counts seen in many areas after initial peaks.

Why the concave pattern matters

Conventional SIR models assume that new cases rise proportionally with the number of infectious people, which predicts sustained exponential growth in the absence of interventions. The Washington University team found that this proportional assumption does not match observed dynamics: after early exponential growth or after reopening, case counts tend to stabilize and decline more slowly than a proportional model would predict.

Thomadsen explains that, under a concave transmission function, a short period of exponential growth is followed by a longer phase of relatively stable or slowly decreasing new cases. Liu emphasizes that the principal difference between their model and the standard SIR approach is that the new model does not force transmission to be strictly proportional to the current number of contagious individuals.

Method and data

The researchers combined county-level confirmed case data with measures of social distancing derived from aggregated cellphone GPS location information. Their flexible model incorporates several factors: the share of the county population not yet confirmed infected, the daily number of newly infected individuals, and county population density. By estimating transmission flexibly rather than imposing a fixed proportional relationship, the model captures how overlapping social connections reduce the marginal impact of each additional contagious person.

They report that COVID-19 clusters frequently within households, nursing homes and workplaces—settings where many contagious individuals tend to expose many of the same susceptible people. That clustering is consistent with the observed concave relationship between infectious population size and new infections.

Other influences and policy implications

Using the combined case and mobility data, the team also evaluates the impact of non-pharmaceutical measures and weather. Their analysis shows that social distancing has a strong effect on reducing transmission rates. Humidity appears to have a modest effect, while temperature shows no statistically significant influence in their results.

The model enables scenario-based forecasts through the summer and into early fall of 2020 under different reopening strategies. Based on cellphone data at the time of the study, Americans were returning to about 60% of pre-pandemic mobility, measured as the share of people who stay home exclusively on a given day. At that level of social distancing, the model projects a fairly steady, slow plateau in daily cases—declining from just over 20,000 cases per day in early June to about 14,000 per day by the end of September.

By contrast, the researchers estimate that extending the maximum observed level of social distancing for a few additional weeks could reduce daily cases to roughly 2,000 per day by late September. A complete and immediate return to pre-pandemic behavior would likely produce an initial surge in cases lasting about two months, followed by a new long-term plateau at a significantly higher level than under current distancing patterns. The authors note, however, that widespread mask use, hand hygiene and other protective measures could moderate the effects of increased mobility and reduce the magnitude of any surge.

Conclusions

The study highlights the importance of recognizing how social network structure alters epidemic dynamics. Because people tend to interact repeatedly within overlapping circles, a concave transmission relationship is a plausible explanation for why outbreaks often slow more gradually than standard proportional models predict. Policymakers seeking to limit COVID-19 spread should weigh the substantial benefits of sustained social distancing, particularly in densely populated areas where overlapping contacts are more common. A rushed or complete reopening could be costly in terms of additional cases, while careful, measured approaches and complementary protective behaviors can help keep transmission closer to lower forecast scenarios.

About this research

Source: WUSTL

Media contact: Sara Savat – WUSTL

Note: This summary reports findings from the Washington University team and describes model-based projections and observed data. It does not provide medical advice. Public health recommendations and local policies may have evolved since this analysis.

Feel free to share this coronavirus news.