This paper analyzes differences in the grant/no-grant status of patents within patent families by country. Many companies apply for the same patent across multiple countries with the aim of acquiring rights to goods and services globally. However, it is often the case that patents within the same patent family are granted in one country but not in another. We analyze this discrepancy from two aspects: information on the patent, such as the number of claims and citations, and words used in the patent specifications, such as the usage of “seasonal” words also used in articles published around the time the patent was filed. In this study, we focus on pharmaceutical patents filed in the United States, India, and Brazil. We then build a machine learning model to predict the grant/no-grant status of patents within the same patent family, including the case where a patent is granted in one country but not in another. Experimental results show that our model predicts the grant/no-grant status of patents within a patent family with high accuracy.