Healthcare Data Basics

Healthcare Data, Information, and Knowledge

• Data are symbols or observations reflecting differences in the world. An example is 250.00. (Note: Data is the plural of datum)
• Information is data with meaning. For example, ICD-9 code 250.00 means type 2 diabetes.
• Knowledge is information that is justifiably believed to be true. For example, obese patients are more likely to develop type 2 diabetes.

Introduction to Computers and Data Types

• Computers process binary information—zero (off) and one (on). Each zero or one is a bit; 8 bits equal a byte. Bits and bytes don't inherently have meaning.
• Bits assemble into various data types: integers (e.g., 345, 669988), floating-point numbers (e.g., 14.1, -1.23), characters (e.g., a, z), and character strings (e.g., "hello", "goodbye").
• Data can be organized into formats like image files (JPG, GIF, PNG), text files, audio (WAV, MP3), and video (WMV, MP4). These formats don't define the information contained within.
• Data is the focus of computer scientists, while information is the focus of informatics and informaticians.

Information Retrieval

• Information retrieval involves computer science (data) and informatics (information). A diagram shows the overlap of these fields concerning data & information retrieval, databases, searching, sorting, vocabularies, and ontologies.

Data and Information

• Computer data often lacks inherent meaning and needs additional information (dates, qualifiers) for understanding (e.g., blood glucose = 127. Was this reading in mg/dL? Was the sample taken fasting?).
• Standardization in data exchange is crucial for interoperability between different computer systems.

Information to Knowledge

• Clinical data warehouses (CDWs) are used to transform medical information into knowledge.
• Electronic Health Records (EHRs) now contain structured (coded, e.g., ICD-9 codes) and unstructured data (e.g., free text or natural language).
• Free text interpretation requires natural language processing (NLP).

Clinical Data Warehouse (CDW)

• Data from various sources (EHRs, Radiology, Pathology) is copied into a staging database, cleaned, and loaded into a common database.
• These databases contain meta-data (which describes data). An example is ICD-type data.
• Tools like descriptive analytics are used to analyze CDW data (e.g., number of patients with breast cancer, their age, and menopausal status).
• CDWs aggregate healthcare data better and analyze trends for public health, as well as for research and cost estimates.

i2b2 Platform

• i2b2 (Informatics for Integrating Biology and the Bedside) is an open-source platform for research.
• This Harvard project is used across a number of institutions in the United States.
• The platform aggregates genomic and clinical information, organized in a star schema.
  • Facts (diagnoses, lab results)
  • Dimensions that detail facts
  • Data from multiple hospitals can be aggregated.

Concept Extraction

• Several systems extract concepts from free text in EHRs or CDWs. Examples given include:
  • CTAKES (with precision 0.80, recall 0.65)
  • MetaMap (with precision 0.32, recall 0.53)
  • MEDLEE (with precision 0.86, recall 0.77)

Challenges of Informatics

• Other industries (e.g., banking) have less ambiguity in data, compared to healthcare, where subjective factors ("I feel sick") are difficult to quantify and vary widely between patients and physicians.
• EHR data aims to be precise, but sometimes needs flexibility over time.
• Complexity of healthcare data. An example is the HL7 RIM model, highlighting the detailed structure required to represent all aspects of healthcare.

Conclusions

• Computer scientists focus on data, while informaticians focus on information.
• A gap exists between unstructured healthcare data and usable information.
• Transforming information into knowledge is crucial for informatics.
• Clinical data warehouses are pivotal in aggregate analysis and clinical research.