In 2020, the Italian city of Milan became one of the first epicenters of COVID-19 outside of Asia. Its decision-makers were ill-prepared in managing the outbreak. This research conducted content analysis of geotagged tweets in Milan during the government enforced lockdown in March 2020 and after the lockdown in May 2020. It provides a temporal and translational comparison of over 545,000 Italian language tweets and their English translation. The Italian tweets were harvested using Italian and English keywords. The translation tool deepL was used to translate all Italian tweets into English. Three methods were used for the temporal and translational comparisons: word clouds and frequency tables, pointwise mutual information (PMI) score, and Latent Dirichlet Allocation (LDA) topic models. The temporal comparison reveals that tweets during the lockdown in March 2020 (Phase I) focused more on the containment of the virus and the disruption on daily routine, like soccer events and gatherings. When the lockdown was lifted in May 2020 (Phase II), the themes included testing, origin, vaccines, possible treatments for COVID-19, political criticism on pandemic management, and unsafe public health behavior. The translation comparison revealed that translation mistakes can make or break communication and understanding during high-stakes situations like the pandemic. It showed contextual mistranslations, mistranslations due to encoding, gender errors, and possible machine learning model errors that can be improved over time. By combining both the temporal and translational analysis, this work hopes to help local leaders in managing future pandemics and other healthcare crises from a multilingual perspective.