Objective: Development of a prediction model for 90-day in-hospital mortality in pregnant COVID-19 patients older or equal 18 years old. Design: Retrospective cohort study Setting: Admissions from Mexican emergency services from April 2020 to November 2023 Population: Pregnant patients with a diagnosis of COVID, confirmed by antigen or PCR, admitted in any Mexican health unit, and older or equal eighteen years of age. Methods: The population was divided in primary (2020-2022) and secondary cohorts (2023) (external validation). 10-fold cross-validation was performed to create training and testing subsets (internal validation). Association between mortality and variables was assessed by Cox regression analyzes and the least absolute shrinkage and selection operator (LASSO). A mortality risk scale was also constructed. Discrimination was evaluated by area under the receiver operating curve (AUC-ROC), calibration by Hosmer-Lemeshow test and calibration slopes; and overall performance by Brier scaled score. Results: 11,554 patients were included in the study, of whom 372 died. The final model included six predictors: intubation, pneumonia, diabetes, chronic kidney disease, time to care and high-risk pregnancy. Similar discrimination and calibration characteristics were observed between the training (AUC-ROC:0.873 [0.851-0.894]), Hosmer-Lemeshow test (X ²=6.38 [p=0.270]), calibration slope [R ²=0.409]) and the test subset (0.884 [0.814-0.955]), X ²=1.64 [p=0.949] and R ²=0.472). Meanwhile, the external validation showed mixed results (0.999 [0.996-1.000], X ²=0 [p=1.000], R ²=0.166). Conclusions: The developed model may aid the analysis of life-threatening cases and improve clinical decision-making to reduce the burden on the healthcare system and decrease overall mortality. Fundings: No funding sources are declared. Keywords: Predictive model, mortality, COVID, pregnancy, Mexico