Beyond Binary: A Layered Clinical Decision Support Architecture

Title: Beyond Binary: A Layered Clinical Decision Support Architecture

Introduction: The Problem

Between 5% and 15% of clinical consultations involve diagnostic errors, which can lead to severe consequences such as missed cancer diagnoses, untreated sepsis, and delayed spinal infections. Current AI clinical decision support tools primarily provide binary outputs, lacking the ability to quantify uncertainty, adapt to patient changes over time, or assist clinicians in navigating complex treatment combinations.

These tools introduce new risks, such as over-reliance on black-box recommendations or static snapshots that become outdated as patient conditions evolve. Purely neural approaches excel at detecting signals in medical data but cannot explicitly model uncertainty or update confidence as new clinical data becomes available.

Proposed Solution: A Three-Layer Architecture

Layer 1 - Bayesian Diagnostic Reasoning (Static Uncertainty)

The first layer addresses diagnostic uncertainty by applying Bayesian decision theory. This approach calculates posterior probabilities of candidate diagnoses based on prior disease prevalence and likelihood ratios derived from clinical literature. By incorporating a utility function that assigns asymmetric costs to false positives and false negatives, the model can recommend actions that minimize expected clinical costs, particularly in reducing missed abnormalities.

A Bayesian network CDSS trial for assessing serious spinal pathology in patients with low back pain demonstrated that the system maintained diagnostic sensitivity even with incomplete clinical data. Another study trained a Bayesian belief network on uveitis etiology diagnosis and showed superior accuracy compared to clinicians' initial impressions, with explicit uncertainty representation through posterior distributions.

Layer 2 - TimesFM Time-Series Forecasting (Dynamic Evolution)

The second layer tackles the dynamic nature of clinical risk. Most CDSS tools assess patients at a single moment, while clinical risk is constantly evolving. Google's TimesFM foundation model, trained on large-scale time-series corpora, can predict multivariate clinical trajectories, such as vital sign trends, laboratory value evolution, and deterioration risk, directly from electronic health record (EHR) streams.

By enabling early warning systems for sepsis, COPD exacerbation, or other forms of clinical deterioration, TimesFM allows clinicians to intervene earlier when deterioration begins hours before it becomes clinically obvious. This layer addresses the temporal staticism of current CDSS tools, providing a more dynamic approach to monitoring patient conditions.

Layer 3 - QAOA Quantum Optimization (Bounded Combinatorial Advantage)

The third layer addresses the combinatorial nature of treatment selection. Purely neural approaches struggle to navigate the vast treatment decision space, especially when considering multiple drugs and their potential interactions. Quantum Approximate Optimization Algorithm (QAOA) quantum optimization can help bound the combinatorial advantage in treatment selection.

By leveraging quantum computing principles, QAOA can efficiently explore and identify optimal treatment combinations, reducing the risk of suboptimal or harmful treatment plans. However, it is essential to note that prospective clinical validation and regulatory approval under the EU AI Act 2024/1689 and FDA CDS Guidance (January 2026) are necessary before deployment.

Conclusion

The proposed three-layer CDSS architecture addresses distinct failure modes of purely neural CDSS tools. By incorporating Bayesian diagnostic reasoning, dynamic time-series forecasting, and quantum optimization for treatment selection, this architecture aims to build a decision support system that knows what it does not know, continuously monitors patient conditions over time, and helps clinicians make informed decisions in complex treatment scenarios.

However, further clinical validation and regulatory approval are crucial before implementing this layered approach in real-world clinical settings.