Comprehensive Multispecialty Medical Records Dataset

# General Overview

The table lists multiple specialties such as Allergy & Immunology, Anaesthesiology, Cardiology, Critical Care, Dentistry, Dermatology, Emergency Medicine, Endocrinology, ENT, Family Medicine, Gastroenterology, General Medicine, Geriatric Medicine, Haematology, Haematology-Oncology, Internal Medicine, Nephrology, Neurology, Neurosurgery, Obstetrics & Gynaecology, Occupational Medicine, Oncology, Ophthalmology, Oral Surgery, Orthopaedics, Osteopathic, Pain Medicine, Pathology, Paediatrics, Physiatric, Physical Therapy, Plastic Surgery, Podiatry, Preventive Medicine, Psychiatry, Psychotherapy, Pulmonology, Radiology, Rehabilitation, Rheumatology, Sleep Medicine, Speech Therapy, Sports Medicine, Surgery, Urology, Wound Care, and Unknown.

Each specialty is further categorized by work types such as Clinic Note, Consultation, Letter, Operative Report, Progress Report, Radiology Report, Transfer Summary, etc.

The table provides the sum of total character count and total audio hours for each work type within each specialty.

The Comprehensive Multispecialty Medical Records Dataset represents a vast and diverse compilation of medical records sourced from multiple specialties, making it an invaluable resource for advancing healthcare innovations. This dataset is designed to meet the growing needs of researchers, data scientists, and healthcare professionals seeking high-quality, real-world data for developing and validating machine learning models, enhancing clinical workflows, and driving data-driven decision-making in the healthcare industry.

# **Scope and Scale:**

- **Specialties Included:** The dataset spans 31 distinct medical specialties, ranging from Cardiology and Internal Medicine to more specialized fields such as Sleep Medicine, Psychiatry, and Rheumatology. Each specialty is well-represented with a diverse array of document types, providing a comprehensive view of patient care across different medical disciplines.
- **Document Types:** The dataset includes a variety of work types such as Clinic Notes, Consultation, Letters, Operative Reports, Progress Reports, Radiology Reports, and Transfer Summary. This extensive categorization allows for in-depth analysis and modelling, making it ideal for applications requiring diverse clinical documentation.
- **Volume:** With a total character count exceeding 24.7 billion and over 271,997 hours of audio recordings, this dataset offers unparalleled depth, making it suitable for large-scale machine-learning projects and comprehensive clinical research.

## **Key Highlights**

- **Total Character Count:** 24,706,408,846
- **Total Audio Hours:** 271,997.465
- **Specialties:** Physician dictation and transcripts cover 31 specialties such as Cardiology, Internal Medicine, Family Medicine, Surgery, and others.

**Dictation Audio Devices:** Audio captured from various devices:

- Telephone Dictation (54.3%)
- Digital Recorder (24.9%)
- Speech Mic (5.4%)
- Smartphone (2.7%)
- Unknown (12.7%).

**Geographical Coverage:** The dataset includes dictation audio from physicians across nearly all US states, ensuring broad regional representation and variability in medical practice styles.

**Physician Age Groups:** The dictation audio encompasses a wide age range of physicians, from 30 to over 70 years old, providing a diverse perspective on clinical practice and communication styles.

- 30-50 years (32%)
- 50-70 years (54%)
- 70+ years (13%).

# Compliance and Privacy:

All audio and text records in the dataset have been meticulously de-identified in accordance with Safe Harbor Guidelines and HIPAA compliance, ensuring that patient privacy is fully protected while maintaining the integrity and usability of the data for research and development purposes.

<p><br/></p>

# 总体概述

该表格列出了多个医学专科，包括变态反应与免疫学（Allergy & Immunology）、麻醉科（Anaesthesiology）、心脏科（Cardiology）、重症医学（Critical Care）、牙科（Dentistry）、皮肤科（Dermatology）、急诊医学（Emergency Medicine）、内分泌学（Endocrinology）、耳鼻喉科（ENT）、家庭医学（Family Medicine）、胃肠病学（Gastroenterology）、普通内科（General Medicine）、老年医学（Geriatric Medicine）、血液学（Haematology）、血液肿瘤学（Haematology-Oncology）、内科学（Internal Medicine）、肾脏病学（Nephrology）、神经病学（Neurology）、神经外科（Neurosurgery）、妇产科（Obstetrics & Gynaecology）、职业医学（Occupational Medicine）、肿瘤学（Oncology）、眼科（Ophthalmology）、口腔外科（Oral Surgery）、骨科（Orthopaedics）、整骨疗法（Osteopathic）、疼痛医学（Pain Medicine）、病理学（Pathology）、儿科学（Paediatrics）、物理医学（Physiatric）、物理治疗（Physical Therapy）、整形外科（Plastic Surgery）、足病诊疗（Podiatry）、预防医学（Preventive Medicine）、精神病学（Psychiatry）、心理治疗（Psychotherapy）、肺脏病学（Pulmonology）、放射学（Radiology）、康复医学（Rehabilitation）、风湿病学（Rheumatology）、睡眠医学（Sleep Medicine）、言语治疗（Speech Therapy）、运动医学（Sports Medicine）、外科学（Surgery）、泌尿科学（Urology）、伤口护理（Wound Care）以及未知专科（Unknown）。

每个专科还会根据工作文档类型进一步分类，包括临床笔记（Clinic Note）、会诊记录（Consultation）、信函（Letter）、手术报告（Operative Report）、病程记录（Progress Report）、放射学报告（Radiology Report）、转诊总结（Transfer Summary）等。

该表格提供了每个专科下各工作文档类型的总字符数与总音频时长之和。

本综合多专科医疗记录数据集（Comprehensive Multispecialty Medical Records Dataset）是源自多个医学专科的海量、多样化医疗记录汇编，可为医疗健康领域的创新发展提供极具价值的研究资源。本数据集旨在满足研究人员、数据科学家以及医疗从业者日益增长的需求，助力他们获取高质量的真实世界数据，以开发、验证机器学习模型，优化临床工作流程，并推动医疗健康行业的数据驱动决策。

# 范围与规模

- **涵盖专科**：本数据集涵盖31个独立的医学专科，覆盖范围从心脏科、内科学等常见专科，到睡眠医学、精神病学、风湿病学等细分专科。每个专科均配有多样化的文档类型，能够全面展现不同医学学科下的患者诊疗全貌。
- **文档类型**：本数据集包含多种工作文档类型，如临床笔记、会诊记录、信函、手术报告、病程记录、放射学报告以及转诊总结等。这种细致的分类支持深度分析与建模，非常适合需要多样化临床文档的应用场景。
- **数据体量**：数据集总字符数超过247亿，音频录制时长超过271997小时，具备无与伦比的数据深度，适用于大规模机器学习项目与全面的临床研究。

## 核心亮点

- **总字符数**：24,706,408,846
- **总音频时长**：271,997.465小时
- **覆盖专科**：医师口述与转录内容涵盖31个医学专科，包括心脏科、内科学、家庭医学、外科学等多个领域。

**口述音频设备来源**：音频采集自多种设备：
- 电话口述（54.3%）
- 数字录音笔（24.9%）
- 语音麦克风（5.4%）
- 智能手机（2.7%）
- 未知来源（12.7%）。

**地域覆盖**：本数据集收录了来自美国几乎所有州的医师口述音频，确保了广泛的区域代表性与医疗执业风格的多样性。

**医师年龄分层**：口述音频涵盖了年龄跨度广泛的医师群体，年龄从30岁至70岁以上，能够提供关于临床执业与沟通风格的多元视角。
- 30-50岁（32%）
- 50-70岁（54%）
- 70岁以上（13%）。

# 合规与隐私

数据集中的所有音频与文本记录均已按照安全港指南（Safe Harbor Guidelines）与健康保险流通与责任法案（HIPAA，Health Insurance Portability and Accountability Act）的要求进行了严格的去标识化处理，在确保患者隐私得到充分保护的同时，保留了数据用于研究与开发的完整性与可用性。

搜集汇总

背景与挑战

背景概述

该数据集涵盖31个医学专科的临床记录，包含24.7亿字符文本和27.2万小时音频，数据来自全美各州不同年龄段医师的多样化诊疗记录。所有数据均通过HIPAA合规脱敏处理，适用于医疗AI模型开发和临床研究。