Big data in medical science--a biostatistical view

Area of Science:

• Medical research
• Data science
• Biostatistics

Background:

• Advancements in data acquisition and storage generate vast medical datasets (terabytes).
• Traditional analysis methods are insufficient for handling these large-scale "big data" sets.
• The utility and significance of big data in medical science remain under active investigation.

Purpose of the Study:

• To illustrate the application of big data analytical techniques in medical research.
• To review the literature on big data analysis in medicine.
• To highlight critical considerations for accurate analysis of large datasets.

Main Methods:

• Illustrative examples of big data analytical techniques.
• Selective literature review.
• Discussion of critical aspects to avoid errors in big data analysis.

Main Results:

• Machine learning facilitates pattern recognition in big data.
• Differential weighting of patient characteristics (e.g., age, sex) is essential before similarity detection.
• Conventional data analysis experience can inform causal inference from big data.

Conclusions:

• Big data techniques, including clustering, can analyze population-level observational data and identify patient subgroups.
• Big data analysis offers complementary insights to traditional clinical trials.
• Emerging statistical methods for causal analysis in big data will benefit medical science with necessary adaptations.