CRIT AI 4.0: A Locally Deployable Artificial Intelligence System for Mortality Risk Stratification and Clinical...

Intensive care unit (ICU) mortality in India remains high, with risk stratification inconsistently applied in resource-limited settings where AI-based decision support requires cloud connectivity unavailable to most Indian hospitals. We describe CRIT AI 4.0, a proof-of-concept, locally deployable, offline-capable AI-assisted decision support system for resource-constrained Indian ICUs.

A Python-based Windows application was developed, operating fully offline. A random forest (RF) classifier was trained on a synthetic dataset of 16,000 encounters across 17 disease categories with 37 input variables. Seven validated clinical scoring modules, six-shift fluid balance tracking, nursing documentation, and an on-device large language model (LLM) assistant were integrated, with no external data transmission.

On the synthetic held-out test set (n = 3,200), the model achieved an area under the receiver operating characteristic curve (AUROC) of 0.90 (95% confidence interval (CI): 0.88-0.92), sensitivity 0.86, specificity 0.77. All metrics reflect synthetic test-set performance only.

CRIT AI 4.0 demonstrates the technical feasibility of offline artificial intelligence (AI)-augmented ICU decision support in resource-constrained settings. This is a proof-of-concept study; prospective validation is underway.