Evaluating the harmonisation potential of diverse cohort datasets

Area of Science:

• Neuroscience
• Data Science
• Biostatistics

Background:

• Data discovery is crucial for scientific advancement but faces challenges due to data volume and heterogeneity.
• Data harmonization offers a potential solution to improve the efficiency of data discovery across multiple datasets.
• Neurodegeneration research generates vast datasets, necessitating effective methods for data integration and analysis.

Purpose of the Study:

• To harmonize a set of 124 neurodegeneration-related variables using the C-Surv data model.
• To assess the feasibility and effectiveness of different harmonization strategies.
• To evaluate the comparability of harmonized data across diverse population cohorts.

Main Methods:

• Harmonization of 124 variables using simple calibration, algorithmic transformation, and Z-distribution standardization.
• Application of harmonization rules based on inclusive data conventions.
• Implementation of the harmonization scheme on data from four distinct population cohorts.

Main Results:

• Out of 120 variables found, 111 (93%) showed complete or near-complete correspondence with the harmonized schema.
• Harmonization was achievable for remaining variables with minor loss of granularity.
• Sufficient comparability was achieved across datasets for effective data discovery.

Conclusions:

• Data harmonization is a viable strategy for enhancing data discovery in neurodegeneration research.
• The applied methods achieved significant data comparability with minimal loss of informativeness.
• This work supports future efforts to expand harmonization to more variables and datasets, fostering data discovery tools.