Background: Pancreatic cancer is expected to be the third deadliest cancer in the US in 2020. Many real-world studies of pts with mPDAC are restricted to single centers, limiting the generalizability of the insights they generate. There is a need to understand prognostic factors of survival in a broader setting to aid in tailoring treatment strategies for pts. This study aimed to identify important population-based predictors related to survival among pts diagnosed with mPDAC. Methods: Data were extracted for pts diagnosed with mPDAC between Jan 2017 and Dec 2019 from the Flatiron Health database. Predictive models for overall survival from the start of each treatment were developed using multivariable Cox proportional hazards regression. Treatment specific predictive models were generated for pts treated with first line (1L) gemcitabine + nab-paclitaxel (GNP), 1L FOLFIRINOX, 1L gemcitabine monotherapy (gem-mono), and 2L liposomal irinotecan-based regimens. The holdout method was used for cross-validation, splitting the data into 70% training / 30% validation. Age at diagnosis, sex, body mass index, smoking status, and ECOG performance score (PS) were included in all models due to clinical importance. Demographic, clinical characteristics, hematological labs, liver function tests (LFTs), and serum bilirubin levels were assessed for inclusion into the models. Uno’s concordance statistic (c-statistic) was used to assess the predictive accuracy of the models. Results: Of the 3,572 pts included in the study, 44% (n = 1,557) received 1L GNP, 27% (n = 954) received 1L FOLFIRINOX, 7% (n = 265) received gem-mono, and 22% (n = 796) received other regimens. 38% (n = 1,345) pts received 2L and of those, 17% (n = 222) received liposomal irinotecan-based regimens. Among all 1L pts, the following were included in the final model: prior surgery, white blood cell (WBC) counts, serum albumin, LFTs (ALP and ALT), serum bilirubin, and ascites (c-statistic: 0.65). The model for pts treated with GNP differed from the overall model via the addition of neutrophil counts and removal of serum bilirubin and ascites (c-statistic: 0.67). Stage at initial diagnosis was included in the model only for pts treated with 1L FOLFIRINOX (c-statistic: 0.68). Among pts treated with gem-mono the LFTs were not included in the model (c-statistic: 0.78). ALP, serum albumin, and WBC counts were important predictors of survival among pts treated with 2L liposomal irinotecan-based regimens (c-statistic: 0.70). Across all regimens the strongest predictors of survival were ECOG PS, serum albumin, and ALP. Conclusions: In one of the largest contemporary real-world studies of patients with mPDAC to date, important population predictors of survival in pts receiving systemic treatment were identified. Further validation studies are needed to understand the generalizability of these results.