Digital Debunking: Are Social Distancing Measures Actually Effective?

Statistically predicting a return to normalcy so we can fix our “quarantine bangs.”

Countries around the world have enacted social distancing measures in an effort to combat the novel coronavirus pandemic. As people remain sheltered in their homes to reduce the spread of COVID-19, a new trend has emerged, yielding disastrous results. Without access to barbers and hairdressers, the rise of the amateur haircut has swept the globe. Self-inflicted bang blunders, coif catastrophes, and dye-job disasters are at an all-time high.

Think before you cut!

So when can we realistically expect social distancing orders to be safely lifted? Are social distancing measures actually effective?

To attempt to answer this question, we built a predictive model to forecast the spread of a country’s COVID-19 cases by modifying the famous SIR model and inputting the latest available historical data. This discrete dynamical model is represented by a system of nonlinear finite difference equations using Altair Activate™, an integration platform for system-level analysis.

With this model, we set out to answer two questions.

  • Based on the predicted spread of infection in a country, when might its government begin considering lifting social distancing measures?
  • What would have been the outcome had no stay-at-home directives been issued?

Developing the Predictive Model

Italy has been one of the countries hardest hit by the coronavirus, so we first built our model focusing on the spread of COVID-19 in the Italian population. On March 9, Italian Prime Minister Giuseppe Conte issued a national lockdown to slow the spread of the virus.

The development of the block-based predictive system model was divided into two main phases. The first phase involved system tuning using actual data related to the spread of COVID-19 in Italy, sourced from the Italian civil protection department, Protezione Civile. Next, the system needed to be adjusted by comparing the differences between the historical data and the model prediction. The resultant optimized model provided results that correlated with 95% accuracy to the most relevant data.

Phase two used the tuned model to make predictions on the potential future spread of the virus. The model plots the predicted amount of people per day that will become infected but will not have been identified through a positive test. Additionally, it plots the predicted number of people who will be found positive through testing and thus isolated to prevent infecting others.

The model was based on the following assumptions.

  • The number of people infected by the virus and identified through a positive test are considered “removed” from the model. These individuals are assumed to be isolated, so they can’t infect additional people.
  • The model adheres closely to the historical data, but we cannot be sure of the validity of its future-looking predictions.
  • The data is based on the actions undertaken in Italy, including number of tests performed, date of lockdown decrees, and the assumptions made about common Italian social habits. This is a simple model, intended to produce a broad idea of the trends and the order of magnitude of the number of infections.

The Outcome

Model of the COVID-19 spread in Italy. Note: All data is updated as of April 15, 2020

When Italy instituted its national lockdown on March 9, there were roughly 1800 new cases of COVID-19 being identified per day and rising. We estimated that, conservatively, the government will not consider lifting this order until new identified cases decline to the point that they return to March 9 levels or below.

If social distancing rules continue to be widely respected and nothing else changes, our model shows that the situation in Italy will be greatly improved over the coming weeks. The model predicts that new identified cases per day will decline to a number at or below pre-decree levels around the first of May.

The current shelter-in-place directive in Italy is scheduled to end on Sunday May 3, which correlates well with our statistical model – a positive sign that Italy’s strict social distancing efforts have been effective to reduce COVID-19 infections and that relaxing of distancing rules could be on the horizon if the reduction in cases follows the predicted pattern.

Identified COVID-19 cases and new unidentified infections per day in Italy

Estimated and actual change in number of identified COVID-19 cases per day in Italy

We then ran the model and removed the variable of the lockdown order. Without the influence of the Italian government’s decree, rather than reaching peak infection rates in early March, the peak would not be reached until April 23, ultimately infecting two thirds of the Italian population. The stark differences between these two models demonstrate the power and importance of adhering to social distancing guidelines, proving that social distancing efforts do in fact work.

Myriad factors could influence actual outcomes, including changes in government policies, advancements in testing and treatment, and changes in societal behavior, but hopefully these predictions provide some optimism in a time where we could all use some good news.

In Italy, it’s likely that you’ll have to wait until at least May 3 to get your next haircut, but our data-driven predictive modeling does show that social distancing measures are having a demonstrable effect on stemming the number of cases of COVID-19 and that there is an end in sight.

Looking good Lloyd!

If we all do our part to limit the spread of COVID-19, hopefully soon we’ll be able to go back to our favorite restaurant, visit friends and family, and maybe most importantly, finally get a haircut again from a real professional. Stay healthy, stay safe, and stay home!

To learn more about Altair Activate, download a free trial of the software. Download the predictive model used in this article, here.

For more information about the COVID-19, please visit the World Health Organization (WHO) website.