BACKGROUND: Information retrieval in primary care is becoming more difficult as the volume of medical information held in electronic databases expands. The lexical structure of this information might permit automatic indexing and improved retrieval. OBJECTIVE: To determine the possibility of identifying the key elements of clinical studies, namely Patient-Population-Problem, Exposure-Intervention, Comparison, Outcome, Duration and Results (PECODR), from abstracts of medical journals. METHODS: We used a convenience sample of 20 synopses from the journal Evidence-Based Medicine (EBM) and their matching original journal article abstracts obtained from PubMed. Three independent primary care professionals identified PECODR-related extracts of text. Rules were developed to define each PECODR element and the selection process of characters, words, phrases and sentences. From the extracts of text related to PECODR elements, potential lexical patterns that might help identify those elements were proposed and assessed using NVivo software. RESULTS: A total of 835 PECODR-related text extracts containing 41,263 individual text characters were identified from 20 EBM journal synopses. There were 759 extracts in the corresponding PubMed abstracts containing 31,947 characters. PECODR elements were found in nearly all abstracts and synopses with the exception of duration. There was agreement on 86.6% of the extracts from the 20 EBM synopses and 85.0% on the corresponding PubMed abstracts. After consensus this rose to 98.4% and 96.9% respectively. We found potential text patterns in the Comparison, Outcome and Results elements of both EBM synopses and PubMed abstracts. Some phrases and words are used frequently and are specific for these elements in both synopses and abstracts. CONCLUSIONS: Results suggest a PECODR-related structure exists in medical abstracts and that there might be lexical patterns specific to these elements. More sophisticated computer-assisted lexical-semantic analysis might refine these results, and pave the way to automating PECODR indexing, and improve information retrieval in primary care.
BACKGROUND: Information retrieval in primary care is becoming more difficult as the volume of medical information held in electronic databases expands. The lexical structure of this information might permit automatic indexing and improved retrieval. OBJECTIVE: To determine the possibility of identifying the key elements of clinical studies, namely Patient-Population-Problem, Exposure-Intervention, Comparison, Outcome, Duration and Results (PECODR), from abstracts of medical journals. METHODS: We used a convenience sample of 20 synopses from the journal Evidence-Based Medicine (EBM) and their matching original journal article abstracts obtained from PubMed. Three independent primary care professionals identified PECODR-related extracts of text. Rules were developed to define each PECODR element and the selection process of characters, words, phrases and sentences. From the extracts of text related to PECODR elements, potential lexical patterns that might help identify those elements were proposed and assessed using NVivo software. RESULTS: A total of 835 PECODR-related text extracts containing 41,263 individual text characters were identified from 20 EBM journal synopses. There were 759 extracts in the corresponding PubMed abstracts containing 31,947 characters. PECODR elements were found in nearly all abstracts and synopses with the exception of duration. There was agreement on 86.6% of the extracts from the 20 EBM synopses and 85.0% on the corresponding PubMed abstracts. After consensus this rose to 98.4% and 96.9% respectively. We found potential text patterns in the Comparison, Outcome and Results elements of both EBM synopses and PubMed abstracts. Some phrases and words are used frequently and are specific for these elements in both synopses and abstracts. CONCLUSIONS: Results suggest a PECODR-related structure exists in medical abstracts and that there might be lexical patterns specific to these elements. More sophisticated computer-assisted lexical-semantic analysis might refine these results, and pave the way to automating PECODR indexing, and improve information retrieval in primary care.
Authors: Florian Boudin; Jian-Yun Nie; Joan C Bartlett; Roland Grad; Pierre Pluye; Martin Dawes Journal: BMC Med Inform Decis Mak Date: 2010-05-15 Impact factor: 2.796
Authors: Jiantao Bian; Charlene Weir; Prasad Unni; Damian Borbolla; Thomas Reese; Yik-Ki Jacob Wan; Guilherme Del Fiol Journal: J Med Internet Res Date: 2018-06-25 Impact factor: 5.428