OBJECTIVES: To explore the extent to which trustworthiness of treatment guidelines has changed since the introduction of Institute of Medicine (IOM) standards.
STUDY DESIGN AND SETTING: In this systematic survey, we searched Medline, Embase, PsycINFO, and Trip from January to December 2010 (pre-IOM) and 2022 (post-IOM) for guidelines that were informed by a systematic review of evidence, written in English, and included recommendations for the treatment of any health conditions in individuals of any age. Using the 10-item modified National Guideline Clearinghouse Extent of Adherence to Trustworthy Standards (NEATS) instrument, paired reviewers independently assessed trustworthiness of 70 randomly selected guidelines from 2010 and 70 from 2022. We calculated mean difference (MD) and corresponding 95% confidence interval (CI) comparing guidelines from 2010 with 2022 for mean score across four critical items, mean score across all items, and score of individual items on 5-point Likert scales in which 1 was least adherence and 5, highest adherence. We also explored change in the proportion of guidelines that achieved a mean score of 4 or higher across critical items.
RESULTS: The average of mean scores increased from 2.28 to 2.74 (MD: 0.46, 95% CI: 0.13-0.79) across critical items and from 2.20 to 2.52 (MD: 0.32, 95% CI: 0.08-0.56) across all items. The proportion of guidelines with a mean score of 4 or higher across critical items increased from 2.9% to 18.6% (absolute difference: 15.7%, 95% CI: 5.8%-25.6%). The items with the greatest increase in the average score included disclosure of funding sources and disclosure and management of financial conflict of interests, followed by a smaller increase in synthesis of evidence, study selection, and search strategy. Other items showed little to no change from 2010.
CONCLUSION: As assessed by the modified NEATS instrument, although trustworthiness of treatment guidelines has improved after the introduction of IOM standards, improvements proved modest and post-IOM guidelines trustworthiness remains suboptimal.