Early prediction of disease progression in tetanus enables timely intervention which can improve outcome. We have developed a machine learning model to predict transition to severe tetanus. The model uses continuous pulse plethysmography waveforms recorded from low-cost wearable pulse oximeters. Data from these monitors is read by the machine learning model in evaluating change over time. If a pre-specified threshold is reached, an alert is generated. To evaluate the feasibility of a clinical decision support system that incorporates our model, we describe a prospective study in adults with tetanus admitted to the intensive care unit in a tertiary hospital in Vietnam. The study aims to evaluate the frequency and accuracy of the alerts and potential clinical usefulness. For the purposes of this study, our machine-learning model runs in parallel with clinical care, and alerts are only seen by the study team who evaluate the patient status at the time of the alert.