Article type
Year
Abstract
Background: A major challenge in updating CPGs is to efficiently identify new, relevant evidence. As part of a non-communicable diseases (NCDs) clinical practice guidelines implementation project in Colombia, we evaluated the efficiency and feasibility of three approaches for identifying up to date literature. This abstract presents the preliminary results of an updating process followed for the selected recommendations from the Type 2 Diabetes Mellitus (T2DM) Colombian CPG.
Objectives: To identify, among three different approaches, which resulted in (a) less time consuming and (b) more accurate results (i.e. identifying relevant up-to date evidence) using fewer steps.
Methods: We compared three approaches: (A) based on PubMed for MEDLINE: a restrictive search strategy using the minimum number of Medical Subject Headings (MeSH) terms and text words required from the original search strategies published in the original CPG, plus a narrow filter for systematic review (SR) identification; (B) Based on Epistemonikos database: a broad search strategy using only population and intervention key words and (C) based on the recently launched new Living overview of evidence (L.OVE) platform which uses artificial intelligence searching: a revision of SR references included in the correspondent PICO-specific T2DM-L.OVE platform. Two people independently ran searches and applied predefined selection criteria following the A and B strategies, a third person performed approach C. We compared the number of references retrieved and the number of steps required to find out relevant up-to date evidence. In case of no identifying up-to-date SRs relevant for the recommendation updating, we plan to complement searches in CENTRAL for Randomized Control Trial (RCT) identification and compare these results with the "primary studies" references included in the T2DM-L.OVE platform.
Results: We updated searches for a total of 8 recommendations; 4 recommendations had original searches published in the CPG. Approach B retrieved fewer number of references than A. Additionally, B identified the majority of key references for the recommendation updating. Compared to approach C, all key new SR identified with A or B were included in the corresponding PICO-specific T2DM-L.OVE. Approach C allowed fewer steps as it was not necessary to run searches. However, the new L.OVE platform is still completing the reference classification process, therefore several references included under each PICO set were not specifically related to the PICO. In these preliminary results we did not need to search for new RCTs, as we found key SR in all cases.
Patient or healthcare consumer involvement: The L.OVE platform-based approach promises to be a very efficient way to update CPG recommendations. Healthcare consumer involvement is essential to support this collaborative project that could greatly facilitate the task of updating CPGs.
Objectives: To identify, among three different approaches, which resulted in (a) less time consuming and (b) more accurate results (i.e. identifying relevant up-to date evidence) using fewer steps.
Methods: We compared three approaches: (A) based on PubMed for MEDLINE: a restrictive search strategy using the minimum number of Medical Subject Headings (MeSH) terms and text words required from the original search strategies published in the original CPG, plus a narrow filter for systematic review (SR) identification; (B) Based on Epistemonikos database: a broad search strategy using only population and intervention key words and (C) based on the recently launched new Living overview of evidence (L.OVE) platform which uses artificial intelligence searching: a revision of SR references included in the correspondent PICO-specific T2DM-L.OVE platform. Two people independently ran searches and applied predefined selection criteria following the A and B strategies, a third person performed approach C. We compared the number of references retrieved and the number of steps required to find out relevant up-to date evidence. In case of no identifying up-to-date SRs relevant for the recommendation updating, we plan to complement searches in CENTRAL for Randomized Control Trial (RCT) identification and compare these results with the "primary studies" references included in the T2DM-L.OVE platform.
Results: We updated searches for a total of 8 recommendations; 4 recommendations had original searches published in the CPG. Approach B retrieved fewer number of references than A. Additionally, B identified the majority of key references for the recommendation updating. Compared to approach C, all key new SR identified with A or B were included in the corresponding PICO-specific T2DM-L.OVE. Approach C allowed fewer steps as it was not necessary to run searches. However, the new L.OVE platform is still completing the reference classification process, therefore several references included under each PICO set were not specifically related to the PICO. In these preliminary results we did not need to search for new RCTs, as we found key SR in all cases.
Patient or healthcare consumer involvement: The L.OVE platform-based approach promises to be a very efficient way to update CPG recommendations. Healthcare consumer involvement is essential to support this collaborative project that could greatly facilitate the task of updating CPGs.