Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The authors develop NOAH, a task-agnostic generative transformer pretrained on 559 million multimodal clinical events across 299,000 patients from the MIMIC family to represent continuous, lifetime health trajectories. While existing longitudinal clinical models are predominantly discriminative, vocabulary-constrained, and reliant on naive monotonic time steps, NOAH unifies medical imaging, time-series, categorical events, and clinical notes within a variational latent space equipped with bidirectional continuous-time integration. The model successfully executes time-controlled autoregressive forecasting and counterfactual simulation while yielding state representations that transfer effectively to time-to-event predictions, 15 ICD chapters, and 29 clinical comorbidities.
Clinical histories are fundamentally irregular and multimodal, yet foundation models have treated them as static or uniformly discrete—until a generative variational transformer trained on 559 million events unified continuous-time forecasting, multi-modal health records, and counterfactual intervention simulation.
The digitization of healthcare has generated vast, longitudinal, and multimodal patient records over a lifetime, yet fully exploiting these data to represent and predict patient state trajectories remains a critical challenge. Current AI models often struggle to capture the complex, irregular temporal dynamics and inherent stochasticity of real-world multimodal patient data. Existing AI approaches for modeling longitudinal patient records are predominantly discriminative, limited to a few modalities, constrained by closed categorical vocabularies, treating time as a monotonic inductive bias, or they are limited in forecasting future patient states. We introduce NOAH, a time-aware, task-agnostic, generative transformer model representing and forecasting the full multimodal patient journey. NOAH features a novel bidirectional time integration and a variational latent space to capture the continuous evolution of patient states and the stochasticity of clinical trajectories. Built from over 559 million clinical events from 431,000 hospital visits of 299,000 patients across the MIMIC dataset family, NOAH natively processes medical images, time-series and numeric signals, categorical events, as well as structured and unstructured clinical records. NOAH is the first truly holistic generative model in its field, enabling autoregressive forecasting with optional time control, zero-shot classification, and counterfactual intervention simulation. It generates highly informative and predictive patient state representations that demonstrate strong performance in probing for clinical outcomes, 15 ICD chapters, and 29 comorbidities, as well as in time-to-event prediction. Seamlessly handling diverse modalities and complex temporal dynamics, NOAH provides a versatile, task-agnostic, scalable foundation for intelligent predictive systems in personalized clinical care and digital medicine.