Fuzzy inference system flags sepsis risk in real time on cloud warehouse

Intensive care units generate continuous streams of physiological measurements, and interpreting those streams quickly enough to matter has long challenged hospital analytics teams. A new study published in BMC Medical Informatics and Decision Making presents an event-based framework that runs sepsis risk stratification directly on a managed cloud data warehouse, using Google BigQuery and standard SQL to build a unified, temporally aware clinical event dataset from the MIMIC-IV v3.1 database.

The researchers implemented a Mamdani fuzzy inference system designed to represent clinical reasoning with three input variables: the qSOFA score, the number of hours since suspected infection, and an infection confidence measure. Thirteen linguistic rules translate these inputs into a graded risk score, an approach the authors contrast with conventional threshold-based warnings that yield only binary alerts.

Evaluated on 38,888 ICU events drawn from 17,783 hospital admissions, the fuzzy model produced well-calibrated risk scores, with a Brier score of 0.090. Of the rule-based alerts, 79.7 percent were classified as high fuzzy risk, while the remaining 20.3 percent were assigned moderate risk based on temporal context, allowing the system to distinguish alerts that share the same threshold trigger but differ in clinical circumstances.

According to the abstract, the system showed consistency with existing clinical guidelines while providing graded, interpretable outputs. Interpretability is a central claim: rather than a black-box score, the fuzzy rules encode explicit clinical logic, so each output can be traced back to the linguistic rules and input variables that produced it.

The architectural finding may prove as consequential as the model itself. The authors report that the framework demonstrates real-time, event-based analytics can be supported by managed cloud warehouses without specialized stream processors. Building the temporal event pipeline in standard SQL on BigQuery means hospitals already using cloud data infrastructure could adopt the approach without standing up separate streaming systems.

The work is framed as a proof of concept on a retrospective research database, and the abstract does not report prospective clinical deployment or effects on patient outcomes. The published version is an early-release, peer-reviewed accepted article that remains subject to further edits before the final version of record appears. Nonetheless, the study argues that fuzzy reasoning offers scalable, interpretable risk stratification that binary thresholds cannot achieve, positioning managed cloud warehouses as a viable foundation for real-time clinical decision support.

Subject of Research: Real-time sepsis risk stratification using a fuzzy inference system on a cloud data warehouse

Article Title: Real-time sepsis risk stratification on cloud data warehouses: a fuzzy inference system on MIMIC-IV

Article References: Danda, J. M. R., Sakib, N., Haritha, I., Prasanna, Y. L., & Shareef, S. K. (2026). Real-time sepsis risk stratification on cloud data warehouses: a fuzzy inference system on MIMIC-IV. BMC Medical Informatics and Decision Making. https://doi.org/10.1186/s12911-026-03892-5

Image Credits: AI Generated

DOI: 10.1186/s12911-026-03892-5

Keywords: sepsis, fuzzy inference system, MIMIC-IV, Google BigQuery, cloud data warehouse, risk stratification, qSOFA, intensive care, clinical decision support, real-time analytics, Real-time, risk