On the Progression of COVID-19 in Portugal: A Comparative Analysis of Active Cases Using Non-linear Regression

Ana Milhinhos, Pedro M. Costa

Research output: Contribution to journalArticlepeer-review

5 Citations (Scopus)
2 Downloads (Pure)


Portugal is often portrayed as a relatively successful case in the control of COVID-19's March 2020 outbreak in Europe due to timely confinement measures, commonly referred to as the “lockdown”. As in other European Union member states, by late April, Portugal was preparing the phased loosening of such measures scheduled for the beginning of May. Despite a modest reduction in infection rates by that time, there was insufficient data to reliably forecast imminent scenarios. Using the South Korea data as scaffold, which became a paradigmatic case of recovery following a high number of infected people, we fitted the Portuguese data to biphasic models using non-linear regression and compared the two countries. The models, which yielded a good fit, showed that recovery would be slow, with over 50% active cases months after the lockdown. These findings acted at the time as a warning, showing that a high number of infected individuals, together with an unknown number of asymptomatic carriers, could increase the risk of a slow recovery, if not of new outbreaks. A month later, the models showed more favorable outcomes. However, shortly after, as the effects of leaving the lockdown became evident, the number of infections began rising again, leaving Portugal in a situation of inward and outward travel restrictions and baffling even the most conservative forecasts for the clearing of the pandemic.

Original languageEnglish
Article number495
JournalFrontiers in public health
Publication statusPublished - 11 Sep 2020


  • coronavirus
  • European Union
  • modeling
  • non-linear estimation
  • statistical forecasting


Dive into the research topics of 'On the Progression of COVID-19 in Portugal: A Comparative Analysis of Active Cases Using Non-linear Regression'. Together they form a unique fingerprint.

Cite this