---
dataset_info:
  features:
  - name: note_id
    dtype: string
  - name: client_id
    dtype: string
  - name: scenario_id
    dtype: string
  - name: date
    dtype: string
  - name: note
    dtype: string
  splits:
  - name: train
    num_bytes: 3098462
    num_examples: 14533
  download_size: 1245656
  dataset_size: 3098462
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: mit
language:
- nl
tags:
- medical
pretty_name: Fruta Records
size_categories:
- 10K<n<100K
---

# Synthetic Nursing Home Records Dataset

## Dataset Overview

This dataset contains synthetically generated records for nursing home clients. The records represent the day-to-day documentation created by nursing staff during client care. All data is completely artificial but designed to realistically represent typical nursing documentation. This dataset contains **NO REAL PATIENT DATA**.

## Dataset Creation

The records were generated using four different LLM models:
- Azure OpenAI: GPT-4o-mini
- Azure OpenAI: GPT-4o
- Anthropic: Claude-3-5-Sonnet-20240620
- Ollama: Phi4

Each record is a set of notes, derived from the client profiles and scenarios (available in companion datasets). The LLMs were instructed to create realistic daily nursing notes that reflect the detailed observations, interventions, and client status that would typically be documented by nursing home staff.

## Dataset Structure

The dataset consists of the following fields:

| Field | Description |
|-------|-------------|
| note_id | Unique identifier for each nursing note (format: n[ward_initial]_[number]) |
| client_id | Unique identifier linking to the client in the profiles dataset |
| scenario_id | Unique identifier linking to the scenario that this record relates to |
| date | Date and time when the record was written |
| note | Content of the nursing record |

## Relationship with Other Datasets

This dataset is part of a synthetic nursing home data collection:
- **Profiles**: Contains the baseline client information
- **Scenarios**: Contains weekly summaries of events
- **Records** (this dataset): Contains detailed daily nursing notes

## Usage Notes

This dataset is intended for:
- Training small language models on nursing home documentation
- Showcasing NLP applications in nursing home settings
- Creating realistic but non-sensitive datasets for healthcare NLP experiments

This synthetic dataset allows for privacy-preserving NLP development in healthcare. However, users should acknowledge that it may not capture the nuances of real-world nursing documentation.

## Author

Created by Eva Rombouts

The code for this dataset generation is available on GitHub: https://github.com/ekrombouts/GenCareAI

## Notes

- All data is in Dutch