JMIR Publications

ABSTRACT

Background:

Clinical artificial intelligence (AI) is often developed around patient-level prediction, automation, and documentation assistance, while nursing work remains underrepresented in dominant AI paradigms. Nursing care is continuous, surveillance-based, documentation-intensive, and shaped by workflow conditions that are often invisible to patient-centered prediction models.

Objective:

This Viewpoint proposes a documentation-native and workflow-aware framework for nursing informatics, clarifying how routine electronic health record (EHR) behavior and audit-log metadata may be used as contextual signals of documentation burden, shift-level strain, surveillance intensity, workflow congestion, and AI governance conditions.

Methods:

The manuscript synthesizes recent literature on nursing documentation burden, nurse-EHR interaction modeling, audit-log analytics, nursing workload prediction, human-centered AI, implementation science, and sociotechnical digital health implementation. It positions documentation-derived signals as a contextual governance layer rather than as autonomous clinical decision-making inputs.

Results:

The framework defines documentation-derived signals, distinguishes them from existing audit-log analytics and workload prediction, proposes an operational taxonomy of measurable feature families, and describes how these signals could integrate with traditional clinical data streams. It also outlines technical feasibility requirements, including temporal fidelity, EHR audit-log heterogeneity, interoperability, privacy, transportability, and data governance, and proposes a 5-step translational roadmap for silent-mode validation.

Conclusions:

Documentation-native AI should not be used for punitive surveillance or autonomous decision-making. When validated, governed, and interpreted with nursing leadership, documentation-derived signals may help health systems make workload, workflow strain, AI safety, and subgroup-specific implementation conditions more visible. The next phase of this work requires multicenter validation, nurse-reported outcome linkage, implementation science methods, fairness auditing, and clear safeguards against surveillance misuse.