Methods for identifying health status from routinely collected health data: An overview

Area of Science:

• Health informatics
• Observational study methodology
• Real-world data analysis

Background:

• Routinely collected health data (RCD) are increasingly used for evaluating medical product effectiveness and safety.
• Accurate identification of health status from RCD is crucial for robust observational studies.
• Current methods, often relying on International Classification of Diseases (ICD) codes, have limitations in universality and accuracy.

Purpose of the Study:

• To outline key steps and methodological considerations for identifying health statuses using RCD in observational studies.
• To address the insufficient understanding of methodologies for health status identification from RCD.
• To promote the utilization of advanced methods, such as machine learning, in RCD research.

Main Methods:

• Review of existing methodologies for health status identification in RCD.
• Discussion of the limitations of International Classification of Diseases (ICD) codes.
• Exploration of the potential and application of machine learning techniques for health status identification.

Main Results:

• Current methods for health status identification from RCD are not fully understood and may not be universally applicable.
• Machine learning methods show promise for improved accuracy but are underutilized in RCD studies.
• Standardized approaches are needed to enhance the reliability of health status algorithms.

Conclusions:

• Improved methodologies for identifying health status from RCD are essential for advancing medical research.
• Adopting advanced techniques like machine learning can significantly enhance the accuracy and credibility of observational studies using RCD.
• This review provides a framework to boost the trustworthiness of findings derived from RCD.