Search papers, labs, and topics across Lattice.
This paper introduces DRIFT, a hybrid forecasting framework that combines a direct model for primary predictions with a recursive, action-conditioned model for constrained corrections, specifically targeting ICU physiological trajectories influenced by treatment interventions. Evaluated on extensive datasets from MIMIC-IV and eICU-CRD, DRIFT demonstrates a significant reduction in mean absolute error for mean arterial pressure (MAP) compared to an action-conditioned Temporal Fusion Transformer (TFT-action). While the overall accuracy improvement is modest, DRIFT shows superior performance in scenarios where treatment sequences are altered, highlighting its robustness in dynamic clinical settings.
DRIFT outperforms existing models by effectively integrating treatment interventions into ICU forecasting, achieving lower error rates even when treatment sequences change.
Many time-series forecasts depend not only on prior observations but also on actions specified during the forecast period. In intensive care units (ICUs), future vital signs and laboratory values are influenced by treatments such as vasopressors. However, models that predict the full future sequence all at once make little use of these treatments, whereas autoregressive models can accumulate errors. We introduce DRIFT, a hybrid framework in which a direct model produces the primary forecast and a recursive, action-conditioned model contributes constrained corrections. We evaluate DRIFT on 6,046 admissions from MIMIC-IV and 8,345 admissions from eICU-CRD. Averaged across the 8-, 24-, and 48-hour forecast endpoints, DRIFT reduces mean absolute error for mean arterial pressure (MAP) by 0.673% relative to an action-conditioned Temporal Fusion Transformer (TFT-action) on MIMIC-IV and achieves the lowest corresponding error among the compared models on eICU-CRD. Although the overall accuracy improvement is modest, a MIMIC-IV audit restricted to windows in which the supplied treatment sequence was altered showed that DRIFT achieved lower observed-target MAP error than TFT-action at 8 and 24 hours. Treatment-sequence alteration increased DRIFT's MAP error by 0.21-0.26 mmHg more than it increased TFT-action's error, with prediction changes occurring primarily after the supplied paths diverged. In a separate robustness experiment, the MAP advantage persisted under three shared checkpoint-selection rules emphasizing overall endpoint error, MAP error, or both equally.
