ArkivX – Journal System Integrated with OpenAI ChatGPT
Background and Development
Ascendi has developed its proprietary healthcare documentation system, ArkivX, over the past two decades. The project began in 1998 under the name Journal 2000, initially as a more basic prototype.
ArkivX is a professional nursing documentation system based on the VIPS model — a well-established, research-based framework from Sweden designed for structured nursing records. The model was developed by Professor Emerita Margaretha Ehnfors and colleagues, introducing the principle of keyword-based clinical documentation.
The name ArkivX was inspired by the popular TV series The X-Files, which aired in Sweden under the title Arkiv X. Like the series’ agents Mulder and Scully, who investigated unexplained phenomena, the system’s name reflects the ambition to explore and organize complex, human-centered information.
Integration with OpenAI and ChatGPT
Today, ArkivX is fully developed and integrated with OpenAI through a secure API key for automated quality assurance of journal entries.
Users can write daily notes in their own words and preferred language. Once approved, ChatGPT automatically refines the text — restructuring it into concise, formal Norwegian journal language that aligns with professional documentation standards.
The command console is designed to be intuitive and familiar, allowing healthcare professionals to interact with the system naturally while maintaining full control of structure and terminology.
Example of the command prompt configuration:
“You are part of a healthcare journal system where users enter daily information informally.
Your task is to refine and rewrite the text concisely without losing important details.
Format the notes in formal, third-person Norwegian.• Avoid using ‘pasienten’, ‘beboer’, ‘vedkommende’, ‘individet’, or ‘bruker’.
• Write in the present tense whenever possible.
• Do not use the word ‘person’.
• Replace ‘personalet’ with ‘vi’ or ‘oss’.
• Replace ‘pasienten’ with ‘ham’.#”
Managing AI Hallucinations
It is important to note that ChatGPT within ArkivX retrieves information only from the user’s journal entries — not from the internet.
While large language models can generate hallucinations (fabricated information that appears plausible), ArkivX is purpose-built to minimize this risk. Through its command structure and verification logic, the system defines ChatGPT’s operational scope precisely, ensuring safe and traceable use of generative AI in clinical documentation.
ArkivX also applies an internal filter layer that blocks potential hallucinations with the lowest possible risk factor, maintaining consistency and reliability in every generated entry.
Relevant Nordic and International References
- Emsley, R. (2023) – “ChatGPT: these are not hallucinations – they’re fabrications.”
Explores how generative models produce fabricated rather than merely hallucinatory content. - “The Chat-Chamber Effect: Trusting the AI Hallucination.”
Describes how closed feedback loops can amplify AI-generated misinformation. - “AI Hallucination: Towards a Comprehensive Classification …”
Presents a taxonomy of distorted information types in generative models. - “Major Research into ‘Hallucinating’ Generative Models Advances Reliability of AI.”
Outlines methods for detecting and mitigating hallucinations, particularly relevant in healthcare contexts.
Command-Driven Reporting and Structure
ArkivX allows users to generate unlimited custom reports using command-based logic.
For example, to create a structured deviation report under the keyword AVVIK (incident), the system interprets the following instruction:
“%Avvik#
You are part of a healthcare journal system. When documenting a discharge summary (‘epikrise’), follow these standards:
- Language and Style
• Use formal, third-person Norwegian.
• Write in present tense.
• Do not include the writer’s name; use ‘N.N.’ for the documenting unit.
• Avoid words such as ‘pasient’, ‘personen’, or ‘beboer’.- Summary Requirements
• Indicate total days covered and number of journal entries.
• Summarize and categorize all deviations (‘avvik’).
• Merge related notes under each main keyword into one coherent text block.
• Reflect observed behavior, participation, needs, and interventions.- Visualization
• Create a diagram or timeline of all significant events across the period.
• Include deviations, activities, and condition changes.#”
This structure enables ArkivX to generate detailed, coherent summaries and visual analyses automatically — improving both documentation quality and efficiency.
Flexible and Scalable AI Integration
The system supports multiple OpenAI model configurations. Users can specify the desired model (for example, 5.oi) directly in the command console, tailoring output depth and tone to the context — whether for clinical reporting, management summaries, or research documentation.
In short, ArkivX combines 20 years of domain-specific development with modern AI, creating a secure, adaptive and professional platform for healthcare documentation.