Background: Early identification and timely therapeutic strategics for potential critical patients with coronavirus disease 2019 (COVID-19) are of crucial importance to reduce mortality. We aimed to develop and validate a prediction tool for 30-day mortality for these patients on admission. Methods: Consecutive hospitalized patients admitted to Tongji Hospital and Hubei Xinhua Hospital from January 1 to March 10, 2020, were retrospective analyzed. They were grouped as derivation and external validation set. Multivariate Cox regression was applied to identify the risk factors associated with death, and a nomogram was developed and externally validated by calibration plots, C-index, Kaplan-Meier curves and decision curve. Results: Data from 1,717 patients at the Tongji Hospital and 188 cases at the Hubei Xinhua Hospital were included in our study. Using multivariate Cox regression with backward stepwise selection of variables in the derivation cohort, age, sex, chronic obstructive pulmonary disease (CON)), as well as seven biomarkers (aspartate aminotransferase, high-sensitivity C-reactive protein, high-sensitivity troponin I, white blood cell count, lymphocyte count, D-dimer, and procalcitonin) were incorporated in the model. An age, biomarkers, clinical history, sex (ABCS)-mortality score was developed, which yielded a higher C-index than the conventional CURB-65 score for predicting 30-day mortality in both the derivation cohort 10.888 [95% confidence interval (CI), 0.869-0.907] vs. 0.696 (95% CI, 0.660-0.731)] and validation cohort [0.838 (95% CI, 0.777-0.899) vs. 0.619 (95% CI, 0.519-0.720)], respectively. Furthermore, risk stratified Kaplan-Meier curves showed good discriminatory capacity of the model for classifying patients into distinct mortality risk groups for both derivation and validation cohorts. Conclusions: The ABCS-mortality score might be offered to clinicians to strengthen the prognosis-based clinical decision-making, which would be helpful for reducing mortality of COVID-19 patients.