Bone marrow metastasis (BMM) in patients with solid cancers is indicative of advanced-stage disease with a poor prognosis. The clinical features and outcomes remain unclear. We aimed to develop a scoring system to predict survival in these patients to help with clinical decision making. A total of 165 adult patients diagnosed with solid cancers and BMM between 2000 and 2014 were selected as the derivation cohort. A risk model was developed using multivariate logistic regression from the derivation cohort and a marrow metastases prognostic score (MMPS) was generated. An independent cohort of 156 patients from 3 other hospitals was selected using the same recruiting criteria to validate the MMPS as a predictor of survival. The MMPS was calculated based on 4 independent prognostic variables: the Eastern Cooperative Oncology Group performance scale, site of cancer, platelet count, and neutrophil-to-lymphocyte ratio. Patients in both the derivation and validation cohorts were stratified into good, intermediate, and poor prognostic groups based on their MMPS. The median survival in each risk group of the derivation cohort was 241, 58, and 11 days for the good, intermediate, and poor prognostic groups, respectively, and 305, 65, and 9 days, respectively, in the validation cohort. The c-statistic values for prediction of mortality at 3, 6, and 12 months were significantly higher for the MMPS than for the Eastern Cooperative Oncology Group performance scale in both cohorts. We developed a risk model that accurately predicted survival in adult patients with solid cancers and BMM. This scoring system may help patients and clinicians with treatment decisions.