Kristin Rogers

Background: Health Evidence provides access to quality appraisals for >10,000 evidence syntheses on the effectiveness and cost-effectiveness of public health and health promotion interventions. Maintaining Health Evidence has become increasingly resource-intensive due to the exponential growth of published literature. Innovative screening methods using artificial intelligence (AI) can potentially improve efficiency. Objective: The objectives of this project are to: (1) assess the ability of AI-assisted screening to correctly predict nonrelevant references at the title and abstract level and investigate the consistency of this performance over time, and (2) evaluate the impact of AI-assisted screening on the overall monthly manual screening set. Methods: Training and testing were conducted using the DistillerSR AI Preview & Rank feature. A set of manually screened references (n=43,273) was uploaded and used to train the AI feature and assign probability scores to each reference to predict relevance. A minimum threshold was established where the AI feature correctly identified all manually screened relevant references. The AI feature was tested on a separate set of references (n=72,686) from the May 2019 to April 2020 monthly searches. The testing set was used to determine an optimal threshold that ensured >99% of relevant references would continue to be added to Health Evidence. The performance of AI-assisted screening at the title and abstract screening level was evaluated using recall, specificity, precision, negative predictive value, and the number of references removed by AI. The number and percentage of references removed by AI-assisted screening and the change in monthly manual screening time were estimated using an implementation reference set (n=272,253) from November 2020 to 2023. Results: The minimum threshold in the training set of references was 0.068, which correctly removed 37% (n=16,122) of nonrelevant references. Analysis of the testing set identified an optimal threshold of 0.17, which removed 51,706 (71.14%) references using AI-assisted screening. A slight decrease in recall between the 0.068 minimum threshold (99.68%) and the 0.17 optimal threshold (94.84%) was noted, resulting in four missed references included via manual screening at the full-text level. This was accompanied by an increase in specificity from 35.95% to 71.70%, doubling the proportion of references AI-assisted screening correctly predicted as not relevant. Over 3 years of implementation, the number of references requiring manual screening was reduced by 70%, reducing the time spent manually screening by an estimated 382 hours. Conclusions: Given the magnitude of newly published peer-reviewed evidence, the curation of evidence supports decision makers in making informed decisions. AI-assisted screening can be an important tool to supplement manual screening and reduce the number of references that require manual screening, ensuring that the continued availability of curated high-quality synthesis evidence in public health is possible.

The National Collaborating Centre for Methods and Tools’ (NCCMT) Rapid Evidence Service conducts rapid reviews on priority questions to respond to the needs of public health decision-makers. Given the vast quantity of literature available, a key challenge of conducting rapid evidence syntheses is the time and effort required to manually screen large search results sets to identify and include all studies relevant to the research question within an accelerated timeline. To overcome this challenge, the NCCMT investigated the integration of artificial intelligence (AI) technologies into the title and abstract screening stage of the rapid review process to expedite the identification of studies relevant to the research question. The NCCMT is funded by the Public Health Agenda of Canada and affiliated with McMaster University.

Background: The internet is commonly used by older adults to obtain health information and this trend has markedly increased in the past decade. However, studies illustrate that much of the available online health information is not informed by good quality evidence, developed in a transparent way, or easy to use. Furthermore, studies highlight that the general public lacks the skills necessary to distinguish between online products that are credible and trustworthy and those that are not. A number of tools have been developed to assess the evidence, transparency, and usability of online health information; however, many have not been assessed for reliability or ease of use. Methods: An electronic database search was conducted between 2002 and 2012 to identify published papers describing tools that assessed the evidence, transparency, and usability of online health information. Papers were retained if the tool described was assessed for reliability, assessed the quality of evidence used to create online health information, and was quick and easy to use. When no one tool met expectations, a new instrument was developed and tested for reliability. Reliability between two raters was assessed using the intraclass correlation coefficient (ICC) for each item at two time points. SPSS Statistics 22 software was used for statistical analyses and a one-way random effects model was used to report the results. The overall ICC was assessed for the instrument as a whole in July 2015. The threshold for retaining items was ICC>0.60 (ie, “good” reliability). Results: All tools identified that evaluated online health information were either too complex, took a long time to complete, had poor reliability, or had not undergone reliability assessment. A new instrument was developed and assessed for reliability in April 2014. Three items had an ICC<0.60 (ie, “good” reliability). One of these items was removed (“minimal scrolling”) and two were retained but reworded for clarity. Four new items were added that assessed the level of research evidence that informed the online health information and the tool was retested in July 2015. The total ICC score showed excellent agreement with both single measures (ICC=0.988; CI 0.982–0.992) and average measures (ICC=0.994; CI 0.991–0.996). Conclusions: The results of this study suggest that this new tool is reliable for assessing the evidence, transparency, and usability of online health information that is relevant to optimal aging.

BACKGROUND The internet is commonly used by older adults to obtain health information and this trend has markedly increased in the past decade. However, studies illustrate that much of the available online health information is not informed by good quality evidence, developed in a transparent way, or easy to use. Furthermore, studies highlight that the general public lacks the skills necessary to distinguish between online products that are credible and trustworthy and those that are not. A number of tools have been developed to assess the evidence, transparency, and usability of online health information; however, many have not been assessed for reliability or ease of use. OBJECTIVE The first objective of this study was to determine if a tool assessing the evidence, transparency, and usability of online health information exists that is easy and quick to use and has good reliability. No such tool was identified, so the second objective was to develop such a tool and assess it for reliability when used to assess online health information on topics of relevant to optimal aging. METHODS An electronic database search was conducted between 2002 and 2012 to identify published papers describing tools that assessed the evidence, transparency, and usability of online health information. Papers were retained if the tool described was assessed for reliability, assessed the quality of evidence used to create online health information, and was quick and easy to use. When no one tool met expectations, a new instrument was developed and tested for reliability. Reliability between two raters was assessed using the intraclass correlation coefficient (ICC) for each item at two time points. SPSS Statistics 22 software was used for statistical analyses and a one-way random effects model was used to report the results. The overall ICC was assessed for the instrument as a whole in July 2015. The threshold for retaining items was ICC>0.60 (ie, “good” reliability). RESULTS All tools identified that evaluated online health information were either too complex, took a long time to complete, had poor reliability, or had not undergone reliability assessment. A new instrument was developed and assessed for reliability in April 2014. Three items had an ICC<0.60 (ie, “good” reliability). One of these items was removed (“minimal scrolling”) and two were retained but reworded for clarity. Four new items were added that assessed the level of research evidence that informed the online health information and the tool was retested in July 2015. The total ICC score showed excellent agreement with both single measures (ICC=0.988; CI 0.982–0.992) and average measures (ICC=0.994; CI 0.991–0.996). CONCLUSIONS The results of this study suggest that this new tool is reliable for assessing the evidence, transparency, and usability of online health information that is relevant to optimal aging.