Premium
P1‐557: A METHOD FOR THE DEVELOPMENT OF A DISEASE PROGRESSION COURSE USING TWO SINGLE COHORTS
Author(s) -
Kim Seonwoo,
Woo Sook-Young,
Cho Soo Hyun,
Seo Sang Won
Publication year - 2019
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2019.06.1162
Subject(s) - cohort , medicine , disease , confidence interval
Background: The DPUK cross-cohort multi-use ontology facilitates the analysis of multiple cohort datasets for the testing of emerging hypotheses. Data are curated to a common standard which simplifies the analytic challenge of working across multiple datasets; organised into 22 level-1 categories and 120 level-2 categories, machine-readable across all ontology elements. An ontology requires the curation of cohort data to common standard thereby enabling multiple heterogeneous datasets to be analysed using a single set of conventions for data structure, variable naming, and value labelling. Conventionally, ontologies are optimised for specific usecases which involve structural complexity which is rarely relevant to cohort-based analyses and although cohort datasets are complex, the conventions underlying their organisation vary widely and are often non-specific. The DPUK ontology is based on five main principles: structuring of datasets according to areas of measurement; constraining of variable name length; identification of each variable in the context of a specific cohort and wave; and machine readability of all ontology elements. Here we present the DPUK cross-cohort multi-use ontology.Methods: 1. The data are organised into 22 level-1 categories and 120 level-2 categories. 2. The variable name has five elements (fields) separated by an underscore. 3. The source cohort is identified in Field 1 by a three letter acronym and allows data source to be readily identified in pooled analyses. 4. Field 2 identifies the data type and is a two digit number and provides a natural structure for data discovery and variable selection. 5. Field 3 describes the measurement that was made, the variable names are limited to 12 characters or 17 characters dependent on complexity. 6. Field 4 describes the position of the measurement in a sequence (array) of serial measurements made on one occasion. 7. Field 5 describes the measurement wave.Results: Not applicable. Conclusions: An intuitive coding structure was adopted whereby syllable-based acronyms, word fragments, abbreviations and minimal numbers were utilised to facilitate easy interpretation. The DPUKcommon datamodel is to be applied across all DPUK cohorts, standardising naming conventions and facilitating multiple dataset analyses for dementia-focused and epidemiological research.