A recent study suggests that an AI tool called Foresight has the potential to predict future health outcomes by creating digital twins of patients. Foresight utilizes generative pre-trained transformers, which are the same type of large language models used by ChatGPT. Initially, the researchers in the UK trained the models using medical records. Subsequently, they fed new healthcare data into the tool, resulting in virtual replicas of patients. These digital twins were then used to forecast various outcomes, including disease development and medication requirements. One prominent application that excites scientists is the potential for faster diagnosis. When tested with US data, the digital twins accurately predicted the next condition with an 88% success rate.

However, they were less effective when using British data, correctly forecasting subsequent conditions 68% and 76% of the time, based on information from two National Health Trust (NHS) organizations. Nevertheless, the researchers have confidence that digital twins can assist in guiding diagnoses, treatment plans, and clinical research. Study co-author James Teo, the director of data science and AI at King’s College Hospital, considers these forecasts to represent "possible multiverses" for understanding diseases. "This patient digital twin can provide insights and 'what if' scenarios by generating AI forecasts from text in health records related to any disorder, test, medication, treatment, or complication across all disease groups into the future, " Teo mentioned. Further adjustments and tests are necessary before deploying this technology in practice, but progress is being made. The researchers are already working on an enhanced version called Foresight 2. For detailed information, you can refer to the study findings published in Lancet Digital Health.