The number, size and complexity of research databases continue to grow unabated. Yet, despite significant advances in information technology, scientists continue to struggle with 'data wrangling' or issues of finding, sharing, and reusing data, often for unanticipated future purposes. Most research databases are designed ad hoc, by native investigators for specific research functions and immediate needs with little to no input from database designers, informaticians, or other subject matter experts. Little thought is usually given to the implications of future data retrieval, sharing and reuse. When the needs of future investigators or research requirements change, the original design of the database can become a significant barrier to meeting unanticipated needs and can impede future scientific discovery. Designing research databases to anticipate future needs is a significant challenge given there is no universally acknowledged standard or guideline for researchers to follow when designing research databases. The New Mexico Office of the Medical Investigator (OMI) received a grant from the Department of Justice in 2010. The OMI's research sought to determine if Computed Tomography (CT) scans could supplement or supplant traditional autopsies. A by-product of this research was over 6,000 full-body, three dimensional, high resolution scans on every decedent that underwent a traditional autopsy. There were no plans to reuse this treasure trove of scans and associated health information. A Modified Delphi Method was used to create a Minimum Data Set for a research database of full-body, three dimensional cadaveric images. A Snowball Sampling Method was also performed to evaluate the quality of the metadata produced by the Delphi expert group. Fifty-nine metadata variables were recommended for inclusion in the Minimum Data Set, which only included 44% of the original ad hoc variables. As a result the Minimum Data Set is thought to be applicable and relevant to more research domains and studies than the original set of metadata variables selected by the native database designers. The Snowball Validation Method verified the 59 variables selected by the Delphi expert group and suggested 3 additional fields not included in the Delphi set. Using a larger group of experts produced 56% more metadata variables than the database designers had created ad hoc. This suggests that a modified Delphi Method that queries a broad domain of experts beyond what is typically done for immediate needs is superior. The Snowball Validation Methods can also work well to check the validity of the Delphi design process. These methods can produce a Minimum Data Set of metadata variables that is more 'future-proof' than those typically created by local, native investigators alone.
Biomedical Informatics, Metadata, Cadaveric Images
Level of Degree
Biomedical Sciences Graduate Program
First Committee Member (Chair)
Second Committee Member
Third Committee Member
Berry, Shamsi. "Metadata determination for a cadaveric collection." (2014). http://digitalrepository.unm.edu/biom_etds/91