BACKGROUND: A minimum dataset (MDS) can be determined ad hoc by an investigator or small team; by a metadata expert; or by using a consensus method to take advantage of the global knowledge and expertise of a large group of experts. The first method is the most commonly applied. OBJECTIVE: Here, we describe a use of the third approach using a modified Delphi method to determine the optimal MDS for a dataset of full body computed tomography scans. The scans are of decedents whose deaths were investigated at the New Mexico Office of the Medical Investigator and constitute the New Mexico Decedent Image Database (NMDID). METHODS: The authors initiated the consensus process by suggesting 50 original variables to elicit expert reactions. Experts were recruited from a variety of scientific disciplines and from around the world. Three rounds of variable selection showed high rates of consensus. RESULTS: In total, 59 variables were selected, only 52% of which the original resource authors selected. Using a snowball method, a second set of experts was recruited to validate the variables chosen in the design phase. During the validation phase, no variables were selected for deletion. CONCLUSION: NMDID is likely to remain more "future proof" than if a single metadata expert or only the original team of investigators designed the metadata. Thieme. All rights reserved.
BACKGROUND: A minimum dataset (MDS) can be determined ad hoc by an investigator or small team; by a metadata expert; or by using a consensus method to take advantage of the global knowledge and expertise of a large group of experts. The first method is the most commonly applied. OBJECTIVE: Here, we describe a use of the third approach using a modified Delphi method to determine the optimal MDS for a dataset of full body computed tomography scans. The scans are of decedents whose deaths were investigated at the New Mexico Office of the Medical Investigator and constitute the New Mexico Decedent Image Database (NMDID). METHODS: The authors initiated the consensus process by suggesting 50 original variables to elicit expert reactions. Experts were recruited from a variety of scientific disciplines and from around the world. Three rounds of variable selection showed high rates of consensus. RESULTS: In total, 59 variables were selected, only 52% of which the original resource authors selected. Using a snowball method, a second set of experts was recruited to validate the variables chosen in the design phase. During the validation phase, no variables were selected for deletion. CONCLUSION: NMDID is likely to remain more "future proof" than if a single metadata expert or only the original team of investigators designed the metadata. Thieme. All rights reserved.
Authors: Paul A Harris; Robert Taylor; Robert Thielke; Jonathon Payne; Nathaniel Gonzalez; Jose G Conde Journal: J Biomed Inform Date: 2008-09-30 Impact factor: 6.317
Authors: Liza J McCann; Jamie J Kirkham; Lucy R Wedderburn; Clarissa Pilkington; Adam M Huber; Angelo Ravelli; Duncan Appelbe; Paula R Williamson; Michael W Beresford Journal: Trials Date: 2015-06-12 Impact factor: 2.279
Authors: Anne-Fleur Domensino; Ieke Winkens; Jolanda C M van Haastregt; Coen A M van Bennekom; Caroline M van Heugten Journal: Health Qual Life Outcomes Date: 2020-02-17 Impact factor: 3.186