Global concerns have grown regarding emerging infectious diseases (EIDs) caused by previously unknown pathogens. Considering that strengthening surveillance capacity for unknown diseases is one of the core capacities for preparedness and early response to EIDs, identifying areas with poor capacity could be beneficial to prioritize regions for the improvement of surveillance. In this regard, we aimed to develop prediction models to identify high risk areas for low surveillance capacity for unknown diseases in a global scale. Unexplained death events reported between 2015 and 2019 were collected from two internet-based surveillance systems, ProMED-mail and Global Public Health Intelligence Network. From the reports, the number of reported unexplained deaths at the first report and the time gap between death and report were extracted as measures for sensitivity and timeliness of surveillance capacity, respectively. Using geographical locations of the reports and published global scale spatial data, including demographic, socioeconomic, public health and geographical variables, we fitted two boosted regression tree models to predict regions with the low sensitivity and timeliness. The performance of prediction model for the low sensitivity showed moderate validity, but in terms of the model for timeliness, the performance was unreliable. Therefore, we provided predicted risk only for low sensitivity. The mean predicted risks of low sensitivity were, respectively, 45.2%, 37.4%, 12.5%, and 3.0% in low-income, lower middle-income, upper middle-income, and high-income countries. Enhancing surveillance capacity in low-income countries is highly required, given the predicted low level of sensitivity despite the importance of early response.