Background: The early detection of COVID-19 patients with interstitial pneumonia at high risk of dismal outcome is necessary to deliver proper care and optimize management of limited resources. Objective: The aim of this study was to analyse the performance of pre-existing scores in predicting in-hospital mortality and ICU transfer at admission in an Acute Medical Unit. Methods: 106 consecutive patients with acute respiratory failure due to COVID-19 interstitial pneumoni admitted to Acute Medical Unit were enrolled. The performances of NEWS, SIRS, RAPS, REMS, qSOFA, APACHE II, CURB-65 and PSI were analysed by the Area Under the Receiver Operator Characteristic (AUROCs) and standard indices of accuracy. Results: Considering in-hospital mortality PSI and APACHE II had the higher AUROCs, 0.83 (95% CI 0.75-0.91) and 0.80 (95% CI 0.71-0.88), followed by REMS, 0.77 (95% CI 0.67-0.86), and CURB-65, 0.73 (95% CI 0.63-0.82), whereas the AUROCs of the other scores were < 0.7. PSI and APACHE II had good sensitivity (0.92 and 0.97), negative predictive value (0.96 and 0.97) and negative likelihood ratio (0.1 and 0.1), accurately identifying patients at low risk to die. However, the low specificity (0.70 and 0.47) and positive likelihood ratio (3.1 and 1.8) could limit their usefulness in predicting in-hospital mortality. Considering ICU admissions all the scores, except NEWS, SIRS and qSOFA, showed a worse performance. Conclusions: PSI and APACHE II showed good prognostic results in predicting in-hospital mortality but no pre- existing score validated for acute care settings was totally satisfactory to predict adverse outcomes in COVID-19 interstitial pneumonia after admission to Acute Medical Unit. The application setting and selected outcome criteria should always be considered to evaluate and compare scoring systems’ performance analysis.